Kaufman - Piston Ring Dry Running

Chair of Mechanics
Wear of dry-running piston rod sealing rings:

modelling and experiments
November 2018
Dipl.-Ing. Andreas Kaufmann, BSc

Affidavit
I declare on oath that I wrote this thesis independently, did not use other than the
specified sources and aids, and did not otherwise use any unauthorized aids.
I declare that I have read, understood, and complied with the guidelines of the senate of
the Montanuniversität Leoben for “Good Scientific Practice”.
Furthermore, I declare that the electronic and printed version of the submitted thesis
are identical, both, formally and with regard to content.
Andreas Kaufmann
Matriculation Number: 00835288
Date, 15.11.2018
Abstract
This work presents a predictive calculation tool for the wear of dry-running piston rod
sealing elements, called “packing rings”, which are placed into the pressure packing of
reciprocating compressors. Packing rings are performance determining components in
two ways: i) they are key to an efficient operation during their lifetime and ii) they cause
very expensive downtime and loss of production if failing unexpectedly. The rings’ duty
is to seal the compression chamber towards the ambient. “Dry-running” applications are
tribologically most challenging because lubrication must not be used.
Due to their working principle the packing rings are inevitably subject to wear. Proper
sealing performance over the rings’ lifetime is only achieved by geometrically compensating
wear.
Time and cost intensive laboratory testing is important and industry standard, but several
quantities of interest (stress, strain, contact pressure, wear pattern, etc.) are hardly or not
at all accessible via experiments. Finite element (FE) calculations allow to gain insight
into these quantities and their evolution with time, which is key to better understanding
and further improving packing rings.
The modelling and the simulation procedure are compared against analytical models and
purpose-developed experiments proving the reliability and predictive capabilities of the
numerical approach. This FE tool allows for the first time to bridge the gap between
laboratory testing and real-world applications.
i
Kurzfassung
Diese Arbeit präsentiert eine Berechnungsmethode zur Verschleißvorhersage für troc-

kenlaufende Kolbenstangendichtelemente, genannt “Packungsringe”. Packungsringe sind
leistungs- und effizienzbestimmende Bauteile in reziprokierenden Kompressoren: zum ei-
nen müssen die Ringe Dichtheit gewährleisten, um einen Kompressor effizient betreiben zu
können, zum anderen führt ein plötzliches Versagen der Dichtringe zu ungeplanten, kos-
tenintensiven Stillständen und Produktionsausfällen. Die Ringe haben die Aufgabe, die
Kompressionskammer von der Umgebung abzudichten. Am tribologisch anspruchsvoll-
sten sind “trockenlaufende” Anwendungen. Hier ist die Verwendung von Schmierstoffen
prozessbedingt untersagt.
Durch ihre Funktionsweise verschleißen Packungsringe zwangsläufig. Dieser Verschleiß
muss geometrisch kompensiert werden, um über die gesamte Lebenszeit entsprechende
Dichtwirkung zu zeigen.
Zeit- und kostenintensive Laborversuche sind wichtig und derzeit Industriestandard. Man-
che auslegungstechnisch relevanten Größen (Spannung, Dehnung, Kontaktdruck, Versch-
leißbild, etc.) sind jedoch kaum oder nicht experimentell zugänglich. Finite Element
Rechnungen erlauben Einblick in diese Größen und deren Evolution mit der Zeit. Da-
durch wird tiefgehenderes Verständnis erlangt und eine effiziente Weiterenwicklung der
Packungsringe ermöglicht.
Die Verschleißberechnung und die dahinterliegenden Modelle werden mit analytischen
Modellen und Experimenten verglichen, somit kann die Zuverlässigkeit und die Vorhers-
agekraft dieses Berechnungswerkzeuges gezeigt und erstmals eine Verbindung zwischen
Labortests und echter Anwendung geschaffen werden.
ii
Acknowledgments
First I thank HOERBIGER for the opportunity to conduct my PhD thesis. I was (and
still am) allowed to work in an interesting, challenging field of research that connects ther-
modynamics, analytical and mechanical modelling, numerical simulation, test rig design,
material testing and component testing.
Heartfelt thanks goes to my supervisor Prof. Thomas Antretter Head of the Institute of
Mechanics. He turned out to be not only a supervisor, but a true mentor by combining
his great in-depth simulation and modelling knowledge with the ability to find time for
fruitful and directive discussions whenever necessary. I also thank all colleagues from the
Institute of Mechanics for their help, especially Dr. Michael Toifl, Aleksandr Sannikov and
Dr. Robert Kaiser. Special thanks goes to Karl Flicker, who helped occasionally setting
up calculations on the cluster.
On the side of HOERBIGER I thank Dr. Bernhard Spiegl and Dr. Tino Lindner-Silwester
who initiated and enabled this project. Special thanks to Dr. Lindner-Silwester and
Dr. Matthias Kornfeld for creating a working environment, which allowed me to focus, to
work without unnecessary (management) disturbances and to grow with the project. I
thank them for many open, directive, fruitful and productive discussions and their effort
especially with the analytical modelling. Due to their positive, encouraging critique du-
ring the recent years and while writing papers they enabled me to perform this work.
I thank the whole HOERBIGER R&D team for their support in gaining insight into com-
pressor technology. I want to express my gratitude to the R&D workshop team (Martin
Lagler, Alexander Jandl, Christopher Habetler, Martin Molnar, Dominik Ratschka) for
their support preparing test specimen and packing rings, building test rigs, maintaining
them and executing tests. Here special thanks goes to conscientious Alexander Jandl, who
- despite occasional honest disagreements - simply made things work when they initially
did not.
Heartfelt thanks goes to Dr. Marian Janko-Grasslober. He helped me a great number of
times to re-focus and brainstorm. Thank you for your unbroken positive spirit and for
encouraging me.
I thank Dr. Thomas Dick from tortuetec GmbH. His expertise in modelling and material
testing improved this work.
At last I thank my family: my parents Johann and Christa for their huge support through
all this years of studying. My brother Martin for his hospitality (at the beginning of this
adventure) and the occasionally given input on programming. My beloved wife Margarita
for her incredible tolerance and patience and my daughter, Sophia, who shows me - again
and again - that the world is full of smaller and bigger wonders.
iii
Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
Kurzfassung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
1 Introduction 1
2 State of the art 3

2.1 Piston rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Pressure packing rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Wiper rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4.1 Pure PTFE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.2 Fillers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5 Gas dryness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3 Mechanisms governing wear 23

3.1 Contact pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Gas pressure distribution in a sealing surface . . . . . . . . . . . . . . . . . 26
3.3 Consideration of time scales . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 Background and idea of the model . . . . . . . . . . . . . . . . . . 28
3.3.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3.2.1 Wear equation . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.2.2 Pressure load . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.2.3 Time scales . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.3 Example: non-linear wear law . . . . . . . . . . . . . . . . . . . . . 38
3.3.4 Error estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.5 Proof of concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4 Material 46
4.1 Tribological modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.1.1 Frictional force tester . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.2 Rod surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.3 Coefficient of friction . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1.3.1 Normal force between rod and segmented ring . . . . . . . 54
4.1.3.2 Garter springs . . . . . . . . . . . . . . . . . . . . . . . . 56
4.1.3.3 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1.4 Wear coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Mechanical modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
iv
4.2.1 Basic considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.2 Analytical beam model . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3 HY54 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3.1 Material composition . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3.2 Material characterisation . . . . . . . . . . . . . . . . . . . . . . . . 74
5 Modelling 76
5.1 Wear algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.2 Implementation of wear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 Numerical implementation of the gas pressure distribution in the dynamic
sealing surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Analytical model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4.2 Numerical implementation and analytical results . . . . . . . . . . . 90
5.4.3 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.5.1 Model set-up and parameter variation . . . . . . . . . . . . . . . . . 96
5.5.2 Evaluation procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.5.3 Calculations with incompressible gas . . . . . . . . . . . . . . . . . 106
5.5.4 Calculations with compressible gas . . . . . . . . . . . . . . . . . . 117
5.5.5 Calculations with different elastic moduli . . . . . . . . . . . . . . . 122
6 Experimental validation 125

6.1 Test rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1.1 Packing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.2 Test procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.2.1 Wear measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.2.2 Measurement reproducibility . . . . . . . . . . . . . . . . . . . . . . 140
6.3 Test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.3.1 Test A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.3.2 Test B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.4 Comparison of experiment and calculation . . . . . . . . . . . . . . . . . . 156
6.4.1 Frictional force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.4.1.1 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.4.1.2 Comparison with calculation . . . . . . . . . . . . . . . . 160
6.4.2 Wear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7 Conclusion and Outlook 164
Appendix 165
Developed code 170

Start calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Backbone of the calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Initial FORTRAN subroutine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
FORTRAN subroutines to calculate wear . . . . . . . . . . . . . . . . . . . . . . 177
v
Obtain information about sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Find the neighbour nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Input generator for the wear calculation . . . . . . . . . . . . . . . . . . . . . . 221
Input generator gas pressure sub-calculation . . . . . . . . . . . . . . . . . . . . 224
Acronyms and Symbols 236

Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
List of Figures 241
List of Tables 252
Bibliography 254
vi
Chapter 1
Introduction
Sealing elements are performance determining components in reciprocating compressors

in two ways: i) they are key to an efficient operation during their lifetime and ii) they
cause very expensive downtime if they fail unexpectedly.
The sealing elements of a pressure packing, i.e., the so called “packing rings” are inevitably
subject to wear, while sealing the compression chamber towards the ambient. In this
work the packing rings will also be simply referred to as “rings”. Their working principle
is “self-energizing”, i.e. their seal effect arises from the differential gas pressure to be
sealed [44]. The wear has to be compensated by the ring geometry to ensure proper
sealing performance over the rings’ lifetime.
In many applications packing rings are lubricated, which decreases wear and increases
the rings’ lifetime. But there are applications where no gas contamination (e.g. by a
lubricant) is tolerated, e.g. the food industry, when the gas is of highest purity or in
chemical processes. Such non-lubricated applications are called “dry-running” or “non-
lube” applications. They are tribologically very challenging, since the lack of lubrication
poses greater demands on the rings’ material and design.
The tribology of each and every compressor packing is unique, even for packings in what
is considered the “same” application. Field experience shows that sometimes out of
two redundant machines one is a trouble maker, whereas the other one performs well.
Nonetheless the lifetime of a dry-running packing ring is roughly between 2000 and 8000 h,
which makes the rings often the maintenance interval determining component. Obviously
there is a strong demand for longer lifetime of dry-running packing rings in the industry.
Common industry practise is to perform experiments on modified test compressors with

purpose-designed highly instrumented packings. This allows to benchmark ring designs as
well as materials to a certain extent. But by the nature of any tribological problem even
very small occurring differences in the test set-up lead to (to a certain degree) different
results. For example the contact surface preparation, the batch-to-batch difference of the
same material grade or the manual finishing (deburring, lapping) of a packing ring have
unknown influence on the tribological system. These experiments are important, but time
and cost intensive. Furthermore, most experiments are a black-box to a certain extent.
Not all quantities of interest can be measured during the runtime, like contact pressures
1
or the wear pattern. The latter can only be determined at the end of a wear experiment.
This work presents a simulation tool, which bridges the gap between laboratory testing
and real-world applications. Careful modelling combined with FE calculations allows to
gain insight into many quantities which are hardly or not at all accessible via experiments:
contact pressures, the wear pattern, stresses and strains. Additionally the simulation al-
lows to track these quantities over time.
An analytical model with a simplified ring configuration validates the correct numerical
implementation of the wear algorithm. Furthermore, the model is experimentally vali-
dated with the intention to prove the reliability and predictive power of the calculation
tool.
2
Chapter 2
State of the art
This chapter briefly presents the state of the art of sealing elements in a reciprocating
compressor with the focus on the packing rings, their geometry and their materials. Furt-
her detailed information can be found online on the homepage of any notable rings and
packings producer or compressor packager, e.g. Ariel Corporation, Burckhardt Compres-
sion, Castanet-SA, Cook Compression, Compressor Products International (CPI), HOER-
BIGER, Howden-Thomassen, Kranz - Packungen Kuhrmeier GmbH, Stasskol GmbH.
Over time different compressor types have been developed, which can be split into two
basic groups: positive displacement and dynamic compressors, see Figure 2.1. Both groups
have subgroups with different working principles, advantages and problems [21, 27, 81].
This work focuses on reciprocating compressors.
compressors
positive displacement dynamic
reciprocating rotatory axial centrifugal
single double diaphragm scroll ring liquid screw vane lobe

acting acting
Figure 2.1. Compressor types
In reciprocating compressors the following dynamic sealing elements are used: valves,
piston rings, packing rings and wiper rings. Figure 2.2 shows a compressor-cut. The
valves, the piston with the piston rings and the rider rings, the rod with the pressure
packing including the packing rings, the intermediate packing, and the wiper rings sealing
are visible.
3
suction valve piston
with quantity
regulation system rider ring suction
valve
rod
head crank
end end
discharge discharge intermediate

valve
piston
valve pressure packing wiper
sealing packing packing
rings
Figure 2.2. Cut of a compressor including the piston rings, the rider rings, the pressure
packing, the intermediate packing and the wiper packing. All packings consist
of multiple rings. The photo taken is from [63].
The term dynamic is valid for valves and rings, but the meaning is different. Valves move
dynamically between seat and guard but seal statically. The valve kinematic is characte-
rized by high accelerations and subsequent impacts. Only in one position, when the valve
contacts the seat, the valve seals.
For rings dynamic refers to the relative motion between ring and rod (packing & wiper
rings) or ring and liner (piston rings). Due to the manifold application areas of compres-
sors the configurations of the rings within compressors vary with the application. The
ring design and the number of rings vary depending on the gas type, the compression
ratio, the end pressure, the materials used, the lubrication, and the gas delivery quantity.
Wiper rings seal the crank case and prevent oil leakage. These rings work in lubricated
or semi-lubricated mode and are mentioned for completeness only, since this work focuses
on the wear of dry-running packing rings.
Piston and packing rings face a very similar situation during application. Both ring types
have to seal the compression chamber. They have the same self-energized working prin-
ciple inevitably leading to wear. Further both seal types work under difficult conditions:
The gas pressure to be sealed varies periodically between suction and discharge pressure,
the relative velocity changes in magnitude and direction and with it the contact situation
changes permanently.
One of the major differences between packing and piston rings is that for non-vertical
4
machines the piston rings have to carry the load of the piston, whereas the packing rings
are pressed against the rod by the gas pressure and garter spring only. To achieve higher
sealing performance and longer service time the piston rings consist of two groups: the
piston sealing rings and the rider rings. The latter should only carry the piston weight
without sealing. If a rider ring seals, it wears faster due to the higher contact pressure.
When designing a rider ring, the axial width is determined by the contact pressure and
the (limited) construction space. The contact pressure should be as low as possible to
increase lifetime. Piston sealing rings seal at their outer diameter, whereas the packing
rings seal at their inner diameter. Thus the design of rings is adapted to their function.
Another major difference between piston and packing rings is the possibility of cooling
the contact partners during operation. A cooling circuit can be included into the cylinder
surrounding the liner. This leads to an effective cooling system for the compression
chamber and the piston rings. Cooling the packing housing is possible but by far not that
effective. Packing rings, which are usually made out of polymeric materials, insulate the
dynamic contact surface well [22]. Some attempts have been made to include a cooling
system into the rod [74], but these systems are not commonly used or commercially
available.
Leakage in the piston sealing system is in double-acting compressors less critical than in
the packing sealing system, because the gas still remains inside the compressor, i.e. only
efficiency is lost. Conversely, leakage in the packing sealing system leads to a loss of gas.
2.1 Piston rings
Piston sealing rings are cut into one, two or three pieces. A cut can be manufactured
straight, under a certain angle, or with an overlap, Figure 2.3. The latter version allows
to manufacture gas-tight piston rings. They are oversized in order to compensate wear.
Bores on the inner diameter and a groove on the outer diameter lead to a pressure-balanced
design with lower wear. Such a design is used for high compression ratios.
5
a) b) c1)
c2) c3) d)
e) f) g)
Figure 2.3. Piston ring designs: a) uncut ring, b) one piece radially straight-cut ring,
c) angle-cut rings: c1) one piece, c2) two piece or c3) three piece design, d)
one piece overlap/step-cut ring, e) two piece gas-tight cut ring, f) radially
straight-cut ring with pressure-balancing groove, g) radially straight-cut ring
pair.
The simplest (and cheapest) rider rings are solid rings, which are shrink-fitted onto the
piston. To ensure easy installation the rider rings are cut either straight or with an angle.
Radial and axial pressure relieve grooves avoid a too high pressure difference across the
axial ring’s width. Figure 2.4 presents an overview of typical single-piece rider rings. The
rider rings are also available as two piece rings, radially or angularly cut. To prevent
rotation of the ring an additional hole is provided for the peg on the piston.
6
a) b) c)
d) e) f)
Figure 2.4. Rider ring designs for single piece rings: a) uncut ring, b) radially straight-
cut ring, c) angle-cut ring, d) angular-cut ring with radial pressure-relieve
grooves, e) angular-cut ring with axial pressure relieve grooves, f) angular-cut
ring with radial and axial pressure relieve grooves. Similar versions of d) to
f) can be found with the cut and the pressure relieve grooves in straight axial
direction. Two-piece rings have another cut shifted by 180◦ .
2.2 Pressure packing rings
In the pressure packing several rings and/or ring pairs seal the compression chamber to
the environment. The number of sealing elements within a packing is usually 3 to 8 [33],
but even packing cases with more than 10 sealing elements are in the field. Additionally,
depending on the pressure level, the first ring on the cylinder side is a pressure breaker with
defined leakage gaps. Its name originates from its function to break down the dynamic
pressure peaks, see Figure 2.2. Pressure breakers can be designed as radially contacting
or non-contacting labyrinth seals depending on the application. The contacting designs
are for light and the non-contacting design for heavy gases. They are either single-piece
uncut rings or (mostly three-piece) radially-cut rings. The latter have either no bore
clearance and axial gaps or they are designed as zero-gap with a bore clearance, i.e. the
inner diameter of the pressure breaker is slightly bigger than the rod diameter. Pressure
breakers can be single or double-acting. The former have radial slits on their high pressure
side to relieve any pressure from the packing back into the cylinder. The shape and the
number of the radial slits vary dependent on the diameter, operating conditions and
7
manufacturer. The segments of a multi-piece pressure breaker are held together by a
garter spring.
a) b) c) d)
Figure 2.5. Most common pressure breakers: a) single piece, double acting b) single piece,
single acting, and c) three piece, double acting d) three piece, single acting.
The designs can have either i) bore clearance and no gap between the segments
or ii) no bore clearance and gap between the segments for three piece rings.
The bore clearance or gap are small compared to the ring’s other dimensions.
The geometry of seal rings is often more complex. A single seal ring usually consists of
several pieces held together by a garter spring, Figure 2.6. Standard ring groups are e.g.
the single-acting radial-tangential ring pair (a+b or a+c in Figure 2.6) or the double-acting
tangential-tangential ring pair (b+b or c+c). In contrast to these ring pairs the balanced
cap design (BCD) ring (f) is a single ring, which allows to build shorter packings. Further
designs and combinations of rings or ring pairs can be found in the literature [22, 27].
The geometric diversity and complexity of packing rings arises from the need of two basic
features: Firstly, every cut or gap in the design has to be sealed, either by a second ring or
by another ring segment. Secondly, the occurring wear has to be compensated to maintain
the sealing performance [17, 23]. Most seal rings have an initial “wear gap”, which closes
during operation due to wear.
This work presents the wear of packing rings on the example of the BCD. Figure 2.7
shows the basic geometric features of a BCD ring, including the wear gap, the garter
spring groove and the pressure balancing groove. The wear gap and the pressure balancing
groove are designed such that when the wear stop is reached the pressure balancing groove
has worn off entirely. The idea is that the ring is pressed adequately to the rod without
losing sealing efficiency even when the ring’s circumferential stiffness sharply increases
due to closing of the wear gaps. If the pressure balancing groove has disappeared the ring
is pressed more strongly against the rod.
8
a) b) c) d) e)
f1) f2)
Figure 2.6. Overview of seal rings in a packing: a) 3 piece radial cut ring, b) 3 piece
tangential cut ring with wear stop, c) 3 piece tangential to rod cut ring without
wear stop, d) 6 piece tangential cut ring (“bridge”), e) 4 piece ring design f)
balanced cap design (BCD) ring (f1 shows the pressure side, f2 the sealing
side).
Current ring designs are mostly made from high-performance polymer-based composites,
with highly temperature dependent material properties. The arising contact temperature
in non-lubricated compressors has great influence on the packing ring stiffness. Between
rod and packing housing is a clearance, which prevents contact and damage of these parts,
but allows the rings to creep in with increasing temperature. This deformation can disable
the ring’s ability to compensate wear. A backup ring is used to reduce this clearance.
This prevents creeping in most cases and increases the thermal coupling of the ring to
the cooler packing housing. A backup ring is usually made from metal, but also high
performance polymers are used.
Backup rings are usually either solid, i.e. single piece, uncut rings or three piece radially
cut rings, see Figure 2.8. The rod-ring clearance is a few tenth of a millimetre. The radial
cuts have to be finished precisely to prevent leakage. Generally the leakage of a packing
ring (pair) is dominated by the leakage through manufacturing imperfections.
9
BALANCED CAP DESIGN (BCD) RING
cap angle garter spring groove
A-A
pressure
sealing segment balancing
groove
wear gap
A A
inner diameter
outer diameter
marked with ’P’
cap segment
axial thickness
SEALING SEGMENT
pressure balancing groove
B-B axial groove width
B
C (3 : 1)
C
groove
height
radial
axial groove position
CAP SEGMENT D-D D
garter spring groove D
Figure 2.7. The BCD consists of two cap and two sealing segments. The wear gap, the
pressure balancing groove and the garter spring groove are marked. A “P”
marks the pressure side to prevent reverse installation.
10
a) b)
Figure 2.8. Overview of backup rings in a packing: a) solid (i.e. single piece, uncut)
backup ring, b) 3 piece radial cut backup ring.
2.3 Wiper rings
Wiper rings (or scraper rings) have the task to seal the crank case. They prevent the oil
from leaking along the rod into the pressure packing housing. These rings are working
in lubricated or semi-lubricated mode only. Well designed wiper rings scrape the oil off
the rod, which is then returned to the crank case. A bigger outer diameter allows to stop
splashing oil from the crank case, Figure 2.9a), whereas a smaller radial dimension allows
higher contact pressures on the scraping edges, Figure 2.9b).
The necessary contact pressure results from the garter springs only, since wiper rings are
not loaded by any gas pressure. The contact pressure has to be high enough to scrape
off the oil but not too high to prevent wear [27]. Since no gas pressure should act on a
wiper ring, it may move axially inside its cup. This movement, induced by the rod, is
called “shuttling”. To prevent damage from shuttling some wiper rings or ring groups are
side-loaded with springs.
a) b) c) d)
Figure 2.9. Oil wiper designs: a) oil wiper with bigger outer diameter to block splashing
oil, b) oil wiper with smaller outer diameter to increase scraping performance,
c) and d) profiled oil wiper ring with two scraping edges each.
11
2.4 Materials
Rings are made of various materials (metals, polymers and polymer compounds), depen-
ding on the operation conditions and their function. Metallic rings can be found in i)
lubricated service in older compressors as cylinder rings, ii) in both lubricated and non-
lubricated service as backup rings and iii) in rare instances even sealing rings are partly
made of metal.
Today most compressors are equipped with polymer or polymer compound sealing rings.
The change from metal to polymer based rings has the benefit of protecting the counter
part (liner or rod) from wear. Liner and rod are much harder than a polymeric ring and
thus hardly subject to wear during operation, which saves maintenance costs.
Polytetrafluoroethylene (PTFE) is an excellent tribological material. Its wear properties

can be improved by orders of magnitude if used as a compound. Current compounds
consist of a polymeric matrix like poylether ether ketone (PEEK), polyphenylene sulfide
(PPS) or PTFE or a compound of these and fillers like bronze, carbon, carbon fibres
(CFs), coke, graphite, glass fibres (GFs), molybdenum disulfide (MoS2 ), etc. Often the
compounds are highly filled, with a filler content of more than 40 wt%.
The literature review below gives an overview of relevant mechanical and tribological pro-
perties of pure PTFE, several PTFE compounds and the effect of fillers on a compound’s
tribology.
Many papers report on the mechanical and/or tribological behaviour of PTFE and/or
PTFE compounds. The mechanical properties have been determined in tension and com-
pression mode, with different specimen types and with different test parameters. The
variety in the published tests makes comparison difficult, but helps to gain insight into
the material’s behaviour. Tribological investigations are even harder to compare, since
the number of different parameters is greater. The tribological papers differ not only in
the reported test methods, the specimen geometries but also in the test parameters, e.g.
relative velocities ranging from µm s−1 [9] to m s−1 [71].
The literature can be grouped into several categories:
• The type of motion:
◦ reciprocating (either circumferentially or linearly)

◦ unidirectional (pin-on-disk or block-on-ring)
◦ torsional.
• The counter surface is either
◦ the same material as the specimen or

◦ a different material, mostly steel.
12
The counter surface characteristics are hardly found to be equal. Although the
surface preparation and finishing methods seem similar, the result is most of the
time different e.g. surface roughness, surface hardness, surface profile.
• The specimen type varies with the test set-up. Bhushan [5, chapter 11] gives a
general overview of wear testing set-ups including a short description. The test
methods listed below are the set-ups used in the reviewed papers.
◦ The pin-on-disk set-up seems the most common test method. Here the pin
is placed one one side of the rotating or reciprocating disk. In this category
different specimen types and set-ups are used:
− Mostly the pin is made of the (polymeric) specimen material [3,6,7,38,42,
46, 71, 72, 76, 78, 84],
− sometimes the disk is ,made of the investigated polymeric material [37,69,
83] and
− rarely both, pin and disk, are from the investigated material [34].
A further sub-categorisation is based on the pin form:
− usually the pin is cylindrical [6, 7, 38, 42, 46, 69, 71, 72, 76, 78, 84],
− sometimes the contacting end is spherical [3, 37, 83].
◦ In the block-on-ring (also pin-on-cylinder [5, chapter 11]) set-up a block is
pressed against the outer diameter of the ring [39, 41].
◦ Other less popular set-ups are closer to an application:
− the disk-on-disk set-up [34] (also called thrust washer [5, chapter 11])
− ring-on-ring set-up [68]
− bearing tester [73]
− self-developed set-ups, which could be called “ring-on-rod” set-ups [12, 47,
84].
• Test parameters used:
◦ relative speed
◦ applied load/contact pressure
◦ temperature, which is hardly controlled but often measured
◦ test atmosphere:
− humidity
− fluid (mainly gas) type
− temperature
− static gas pressure
The mechanical and tribological properties of pure PTFE are for many applications (also
for rings and packings) not sufficiently stable. The low strength, high tendency to creep
and high wear rate are problematic. These properties can be improved with fillers. Before
addressing compounds and fillers several knowledgeable facts concerning pure PTFE are
highlighted.
13
2.4.1 Pure PTFE
Domininghaus [20, pages 543-565] gives a good overview over the historical development,
the mechanical and tribological properties and the different applications of PTFE. The
reviewed papers are divided into two partly overlapping groups: firstly, papers which
focus on the mechanical properties [11, 15, 34, 58, 59, 64, 79] and the effects of morphology
and crystallinity [7, 15, 16, 38, 58, 59, 64] and secondly, papers which study friction [3, 6, 12,
39, 56, 69, 70, 77, 86] and wear [6, 12, 38, 39, 66, 69, 70, 76–78, 86]. In some references pure
PTFE is used as a reference for the tested compounds.
In comparison to other commodity and engineering polymers, PTFE is a very heavy

polymer. The theoretical values for the density of fully amorphous and fully crystalline
PTFE are 2040 and 2300 kg m−3 , respectively [59]. Usual density values of PTFE are
between 2150 and 2200 kg m−3 [20, 59].
The transition temperatures of PTFE at atmospheric pressure are [20, 37, 59]
• −97 ◦C - glass transition temperature of amorphous PTFE

• 19 ◦C - crystal transformation: triclinic crystals below and hexagonal crystals above
• 30 ◦C - crystal transformation: hexagonal crystals below and pseudohexagonal cry-
stals above
• 127 ◦C - glass transition temperature of amorphous PTFE from solid to under-cooled
liquid state
• 313 ◦C - melting temperature for the amorphous phase
• 326 - 331 ◦C - melting temperature for the crystalline phase
The crystal transformation temperatures around room temperature have to be kept in

mind when performing material testing.
The following sheds light on the mechanical properties of pure PTFE, considering the
influence of hydrostatic pressure, the load-mode dependency of the mechanical properties,
the effect of the strain rate and the effect of temperature. The end of this section presents
some tribological aspects of PTFE.
The influence of hydrostatic pressure on the material properties was already investigated
in the early 1970’s [15, 64]. Christiansen et al. [15] investigated the effect of hydrostatic
pressure on the stress-strain behaviour of PTFE. The effect on the yield stress is repre-
sented by the factor −0.07, which gives the gradient of the yield stress as a function of
the hydrostatic pressure.
Utilizing this factor the influence of a hydrostatic pressure load on a sealing ring in a com-
pressor can be estimated. The yield stress is 2.76 MPa at 54 ◦C, whereas a hydrostatic
pressure of 5 MPa would add only 0.24 MPa (10 %).
The yield stress and the bulk modulus change over the tested hydrostatic pressure range
due to a morphological change. A schematic phase diagram of PTFE can be found in
14
either [59, Fig. 1] which shows the four known phases of PTFE between 10 and 120 ◦C
or [20, Bild 2-331] where only three solid phases are distinguished over a temperature
range of -100 to 500 ◦C. Both show pressures up to 800 MPa. The triple point is at 70 ◦C
and 450 MPa [20].
Sauer, Mears and Pae [64] found that the yield strength and the elastic modulus increases
while the strain at fracture decreases with increasing hydrostatic pressure up to 517 MPa.
The compressive yield stress increases linearly over the pressure range. The compressive
elastic modulus increases linearly, but has a discontinuity near 550 MPa due to a phase
transition. These increases are explained with the free volume theory: with increasing
pressure PTFE loses free volume and becomes more brittle. The amorphous regions
become more glass-like and less rubbery. The degree of cold drawing is reduced and
strain hardening increases. The strain at fracture decreases. If a crack forms, no plastic
tearing occurs and rupture follows. Sauer et al. stated that the Coulomb yield criterion
predicts that the yield stress should be greater in compression than in tension, which fits
to the presented measurements.
Rae and Dattelbaum [59] investigated PTFE’s compressive properties and found that the
Poisson’s ratio varies between 0.4 and 0.5. A trend towards 0.5 was found for strains
greater than 5 %, which agrees with earlier reported data by Sauer et al. [64]. Later Rae
and Brown [58] published a comparison of compression and tension data in a low strain
regime (max. 2 % strain). PTFE shows an asymmetric behaviour. Under compression the
material is stiffer than under tension. The Poisson’s ratio is inexplicably mode dependent,
ν = 0.36 under tension and ν = 0.46 under compression.
The compressive tangent modulus is highly temperature dependent and considerably lower
if measured at larger strains, approximately 500 MPa for large strains [59]. This lower
compressive modulus is possibly better suitable for engineering purposes. Further Rae and
Dattelbaum found that [59, (page 7624)] “During deformation, PTFE stores much more
of the work done as structural change (≈ 30 %) than typical metals (< 10 %). This energy
is presumably used in crystallographic and amorphous chain rearrangement resulting in
less sample heating than a comparable metal and therefore lowering the thermal softening
effect of ’self’ heating.”
Rae and Brown [58] published that the tensile properties of PTFE are highly strain
rate and temperature dependent, whereas crystallinity plays a minor role. Higher loa-
ding rates and lower temperatures result in a stiffer material response. Rae and Brown
give a detailed discussion on the temperature failure mechanisms of PTFE and their
micro-structure dependency. They see the micro-structure of PTFE as “three component
compound” consisting of crystalline, amorphous and quasi-ordered materials. The latter
are either close to a crystalline region in the material or an amorphous region which rear-
ranged under load. At low loading rates (0.001 s−1 ) the strain hardening happens at the
same rate in compression and tension (approximately 0.25 MPa %−1 above a strain level
of 10 %). The true stress level in compression is approximately 3 MPa higher than in ten-
sion above 10 % strain. The effect of hydrostatic pressure published by Christiansen [15]
is approximately 1/10 of the difference in yield stress at this load level. Thus there has
15
to be another reason for this asymmetry.
From the data presented by Rae and Brown [58] (two PTFE types and a strain rate of
0.005 s−1 ) the dependence of the stress level on the temperature is estimated. At 10 %
strain the true stress at 25 ◦C is 11 MPa, at 150 ◦C it is 3 MPa. This gives a temperature
dependency of 0.06 MPa ◦C−1 . The rate decreases with increasing temperature. A tem-
perature change of 1 ◦C has approximately the same impact on the mechanical properties
as 1 MPa hydrostatic pressure. This shows how sensitive PTFE reacts to temperature
changes.
Cyclic compression tests [11] showed similar effects. The compression modulus depends
on the loading rate and the ratcheting strain depends on the mean stress. The linear in-
crease of the compression modulus as well as the higher ratcheting strains at low loading
rates indicate that, if possible, PTFE is subject to creep and/or relaxation.
Some tribological aspects of pure PTFE, like the ability to form a transfer film, are discus-
sed in the following section together with the tribological aspects of PTFE compounds.
The following sheds some light on frictional behaviour µ of PTFE and the high depen-
dency on the test parameters. The static and the dynamic coefficients of friction are
nearly equal for PTFE, no stick-slip occurs [20, 73]. The temperature dependence of the
coefficient of friction is presented by Pleskachevsky and Smurugov [56] for a PTFE-steel
contact pair. Starting with µ = 0.23 at room temperature the coefficient of friction drops
to a minimum of µ = 0.13 around 130 ◦C and then increases again to µ = 0.16 at 170 ◦C
(block-on-ring, load 400 N, speed 0.5 m s−1 ).
Babuska et al. [3] presented the temperature dependency of PTFE against PTFE and
against MoS2 in bone-dry atmosphere under 1 N and 3 mm s−1 in a ball-on-flat set-up.
The coefficient of friction increases from 0.15 at −150 ◦C to a maximum of 0.25 at −50 ◦C
and then decreases to 0.07 at room temperature. Between 50 ◦C and 125 ◦C µ is approx-
imately 0.05, then it increases again to 0.08 at 150 ◦C.
Blanchet and Kennedy [6] showed that PTFE against 316 stainless steel has a velocity de-
pendent coefficient of friction, when loaded with 6.55 MPa. At approximately 0.01 m s−1
µ is 0.02, increasing v to 0.1 m s−1 gives µ=0.1 and at 0.2 m s−1 µ = 0.18. Although the
coefficient of friction does not significantly change at 0.1 m s−1 , the wear rate increases by
two orders of magnitude.
Briscoe et al. [9] found that the PTFE-steel contact pair has a nearly constant coefficient
of friction of 0.06 between 25 ◦C and 250 ◦C (load 10 N, speed 0.45 mm s−1 ). At 0.16 m s−1
µ decreases from 0.35 to 0.15 when increasing the load from 25 to 75 N.
A decrease of µ from 0.23 to 0.17 was detected between room temperature and 90 ◦C by
Smurugov et al. [70], (steel counterface, load 0.25 MPa, speed 0.125 m s−1 ).
Shibo et al. [69] showed the motion dependency of the coefficient of friction with a steel
ball on a PTFE disk. In ascending order µ is 0.1, 0.118 and 0.12 for unidirectional motion,
reciprocating motion and torsion, respectively.
Tanaka and Kawakami [72] found that µ increases from 0.2 to 0.23 if the sliding speed is
increased from 0.1 to 2.5 m s−1 for a load of 10 and 50 N.
So dependent on the test parameters the coefficient of friction varies for PTFE over an
order of magnitude between 0.02 and 0.35 This shows the sensitivity of a PTFE based
tribological system.
16
2.4.2 Fillers
Fillers and their effect on PTFE compounds have been studied over decades by many
groups worldwide. The compounds can be divided into three categories, for each a list
containing the filler type, the filler content and the literature source are noted. Table 2.1
gives an overview of the studied compounds with a single inorganic filler. Table 2.2 lists
compounds with 2 or more inorganic fillers. Table 2.3 considers compounds of PTFE
and another polymer. Polymeric compounds often use PTFE as a filler to improve the
tribological behaviour. The presented compounds usually have a wear rate which is at
least 1-2 orders of magnitude below the wear rate (10−3 mm3 N−1 m−1 ) of pure PTFE.
Preferably, wear rates should be below 10−8 mm3 N−1 m−1 (e.g. [10]).
The compounds presented in the literature usually have a lower coefficient of wear than
pure PTFE. It is difficult to compare the presented compounds, since they differ in the
filler type and content, the manufacturing processes as well as the test equipment and the
test parameters. In the following only some key aspects are highlighted.
The fact that PTFE forms a transfer film on the counter surface makes this polymer at-
tractive for tribological applications. The transfer film reduces friction and wear, but its
formation and appearance is influenced by many parameters. It depends on the filler type
and shape, the test parameters (load, velocity, type of motion) and the counter surface
properties [6, 9, 12, 16, 37–39, 42, 56, 68, 69, 72, 77, 82, 83, 86].
The formation of the transfer film takes some time and is commonly referred to as
“running-in” period. In this period the wear rate is usually higher than in the following
steady-state [38, 75]. The running-in can take from several minutes [75] to a day [73].
Metals as filler promote the film formation [86]. The influence of the type of motion was
studied by Shibo et al. [69], comparing unidirectional rotation, linear reciprocating and
torsional motion. Under torsional motion the transfer film is thicker and fractured, here
debris is not expelled as in the other two types of motion, where the pin does not stay at
the same location.
Under water no transfer film is formed, thus the wear rate is significantly higher [38].
Close investigations on the formation and the effect of sliding speed and contact pressure
have been carried out by Li et al. [42]. They found that a thinner film has a higher
wear resistance. A thin film can be obtained by lower speeds and higher pressures, but
the surface quality and wear resistance of the transfer film is higher at lower pressures.
Furthermore, a certain surface roughness is conducive to a strongly bonded transfer film.
Tevrüz [73] found that the film thickness increases over time and with increasing load.
Furthermore, a thick, continuous and smooth film gives a low coefficient of friction, which
is in agreement with Conte and Igartua [16].
An investigation by Uruena et al. [77] showed the transfer film alone is not the reason
for ultra low wear, it is a combination of polymer and transfer film. The wear of transfer
films was tested with a micro-tribometer and is orders of magnitudes greater than the
wear of the polymer/counter surface pair. The influence of the testing direction and the
number of cycles forming the transfer films varied. The testing direction has no significant
influence. After a certain number of over-travels a proper transfer film establishes. Its
wear resistance does not show significant improvement with more frequent over-travels.
17
Table 2.1. PTFE compounds with a single inorganic filler
filler type amount (wt%) source annotation

glass (fibre) 25 [6] k increases with increasing velocity, GF
prevents surface failure
25 [16]
25 [38] GFs perform better than glass beads (hol-
low or solid) and glass flakes
15, 25 [37]
25 [72] lowest k at high load
40 [65] significant counter surface damage by the
fibres
graphite 15 [6] flakes form, k is velocity independent
15 [16]
0-50 [83] above 40 % filler k increases sharply, lo-
west k at 20 %
15 [72]
carbon or 30 [84] CF reinforced carbon matrix compounds,
as well as CF compounds
carbon fibre 35 [73]
25 [16]
bronze 40 [6] flakes, k decreases with increasing velocity
40 [72] low k if the load is low
60 [16]
MoS2 20 [72]
0-50 [83] above 40 % filler k increases sharply, lo-
west k at 10 %
miscellaneous
Al 5 [77] aluminium particles
AlN 1-5 [79]
CaAlSi3 O5 n.a. [75] amount of man-made calcium aluminosi-
licate fibres not specified
Cu 30 [86]
45 [42]
K2 Ti4 O9 1-25 [68] potassium titanate whiskers
K2 Ti6 O13 1-25 [68] potassium titanate whiskers
Ni 30 [86]
(OH)8 Si4 Al4 O10 0-15 [82] ultra-fine kaolin particles
Pb 30 [86]
Si3 N4 1-5 [79]
TiO2 1-25 [68] whiskers performed better than particles
20 [72]
UFD 1-10 [39] ultra fine diamond (UFD)
ZrO2 40 [72]
18
Table 2.2. PTFE compounds with multiple inorganic fillers
filler type and amount (wt%) source

carbon (18) + graphite (7) [37]
CF (6) + graphite (8) + MoS2 (2) [47]
GF (5) + MoS2 (5) [71]
GF (15) + MoS2 (3) [16]
GF (20) + MoS2 (5) [37]
GF (20) + graphite (5) [16]
modified GF (5) + Pb (60) [12]
Table 2.3. PTFE compounds with another polymer (compound)
second polymer amount (wt%) source annotation

FEP 35, 50 [7] fluorinated ethylene propylene (FEP)
PA 70 [34] polyamide (PA); mainly investigated glass
fibre filled PA
80 [60] PA66
PEEK 0 - 100 [9] poylether ether ketone (PEEK), 10-15 %
PTFE in PEEK performs best
0 - 100 [10] optimum at 15-25 % PEEK in PTFE
80 [65] commercial grade: PEEK is already fil-
led with 10 % carbon fibres, the 20 % are
PTFE and graphite
0 - 100 [46] optimum at 10-20 % PTFE in PEEK, lo-
west wear rate at 5 % and lowest µ at 15 %
PTFE in PEEK
PPDT fibres 2.5, 10 [37] poly-p-phenyleneterephtalamide (PPDT)
Pleskachevsky and Smurugov [56] suggested that (pure) PTFE (against steel) has an
operation point where most efficient self-lubrication occurs. At this operation point (tem-
perature) the load bearing ability and tribological characteristics of the transfer film are
at the optimum. This is in agreement with the findings of Blanchet and Kennedy [6]: the
wear regime of pure PTFE can change abruptly from mild to severe wear, when increasing
the sliding speed at a given contact pressure.
Smurugov et al. [70] presented an explanation for the thermally activated wear mecha-
nisms of PTFE.
Invesigating different filler types, Tanaka and Kawakami [72] showed that the aspect ratio
(l/r) of fillers is important. The higher the aspect ratio the higher is the filler’s load-
supporting action. This was confirmed by Khedkar et al. [37], adding that a higher volume
19
fraction of (glass) fibres increases the thermal stability and conductivity and effectively
decreases the wear rate. Shi et al. [68] also reported that whisker-filled PTFE composites
show a lower wear coefficient than compounds which are filled with particles of lower
aspect ratio. Furthermore, they added that their high aspect ratio fillers prevents the
dominant wear mechanism for pure PTFE: adhesion.
On the other hand Tanaka and Kawakami [72] reported that lamellar fillers as graphite
or MoS2 and particle like fillers like metal oxides have an extremely small aspect ratio
or an aspect ratio about unity, respectively. These fillers do not have a significant load-
supporting action. (This was confirmed by Blanchet and Kennedy [6].)
Additionally, the filler size is important [72]. Too small particles are not well incorporated
in the matrix material and can easily be transferred to the transfer film on the counter
part. Bigger fillers and fibres are firmly embedded into the matrix and can rub against
the counter surface without being pulled from the matrix and thus reduce wear. On
such “bigger” fillers like bronze or GF a PTFE film was found. This film reduces the
wear of the fillers and thus the wear of the whole compound. Tanaka and Kawakami [72]
concluded that the wear rate of PTFE compounds depends only weakly on the type and
shape of the filler, if the filler has suitable wear resistance and size (i.e. several microns
e.g. 30 µm).
Zhang et al. [86] used Cu, Pb and Ni powder as filler and reported an increase of the
load-supporting action. It can be concluded that grain size of the metal powder is “big
enough” to increase the wear resistance.
Klaas et al. [38] compared the effects of solid and hollow glass beads, glass flakes and GF
as filler. They found that the compounds with solid glass beads, hollow glass beads or
glass flakes produce a thicker transfer film and experience a higher wear rate than the
compound with GF. They reported that the GFs polish the counter surface and produce
a uniform transfer film at a low wear rate. The wear rate decreases further when adding a
(not specified) solid lubricant. It can be guessed that the solid lubricant is either MoS2 or
graphite. This corresponds to the earlier findings of Cheng et al. [12] that the (modified)
GFs increase the ability to form a transfer film and the transfer film’s adhesion to the
countersurface.
Xiang and Gu [82] confirmed Tanaka’s statement concerning too small particles. They
found that too many ultra-fine kaolin particles result in a non-uniform transfer film. This
is likely due to abrasion by agglomerated particles.
Khedkar et. al [37] found that carbon particles as filler can plough the PTFE matrix,
destroy the transfer film leading to three-body abrasion. Graphite does not hinder this
process.
Tanaka and Kawakami [72] considered that compared to MoS2 , graphite adheres more
strongly to the matrix and thus a PTFE+graphite compound exhibits less wear than a
PTFE+MoS2 compound. A benefit of MoS2 was reported by Khedkar et al. [37] and
Conte et al. [16]: MoS2 may prevent the direct contact of GF and the transfer film and
thus the abrasive effect of GF is reduced. So MoS2 helps to reduce wear rate and maintain
a low coefficient of friction. Conte et al. compared different compounds and found that
the substitution of 5 % out of 25 % GF by the same amount of graphite results in lower
wear. The substitution of 10 % GF by only 3 % MoS2 outperforms the compound with
20
25 % GF.
Friedrich et al. [26] reported that carbon fibres perform better than glass fibres.
The coefficient of friction of PTFE based composites is generally independent on the filler
type and similar to unfilled PTFE. This fact is ascribed to the formation of a transfer
film [72]. The independence of the filler shape was confirmed by Shi et al. [68], adding
that the crystal structure of the filler has a high influence on the coefficient of friction.
Yang et al. [84] presented an interesting study on packing ring materials. They claim
that the coefficient of friction does not automatically mean lower frictional force and
power consumption at a given constant load. They suggested that the coefficient of
friction depends on the contact pressure, the sliding velocity, the temperature, the surface
hardness and possibly other factors. Tzanakis et al. [75] studied the coefficient of friction
of a PTFE compound filled with calcium aluminosilicate in detail. The speed was varied
between 0.25 and 0.5 m s−1 , the load between 0.4 and 1.6 N in 4 steps and the roughness
of a high carbon steel between Ra 0.125 and 0.7 µm in 4 steps. The influence of these
factors on the contact temperature is: i) Doubling the load increases the temperature
by about 35 % of the initial value (in ◦C). ii) Doubling the sliding velocity increases
the temperature by about 50 %. iii) The increase of temperature with roughness is more
complicated: increasing the roughness profile from Ra 0.125 to 0.25 µm adds 17 % to the
contact temperature regime, but from Ra 0.25 to 0.5 µm it adds 30 %, whereas an increase
from 0.5 to 0.7 µm just adds 3 %.
Blanchet and Kennedy [6] found that fillers also govern the size and shape of wear frag-
ments. Larger particles can turn a subsurface crack’s trajectory to the sliding surface,
leading to finer debris.
In contrast unfilled PTFE exhibits mild or severe wear, depending on the test parame-
ters. In the severe wear regime debris stripes are found around the wear tracks. The
morphology of the transferred material turns to fragmented sheets. These are removed
from the contact zone by the oscillatory motion of the disk [6]. Similar results were found
later [37].
Mamaev et al. [47] investigated the compressive behaviour of a PTFE compound after
exposing it to air or oil for 3 or 6 months at 20 or 100 ◦C temperature and uniform com-
pression of 4 or 16 MPa. None of the three independent factors (environment, temperature
and pressure) has a dominating influence on the wear rate, but the combination of these
factors has a significant impact. The elastic modulus, elongation at break and wear rate
does not change significantly even after an exposure of 6 months (approximately 10 %).
They concluded that their compound (c.f. Table 2.2) is fit for the service as a long-term
piston sealing.
In combination with other polymers PTFE is used as a filler to decrease a compound’s

coefficient of friction and wear by forming a transfer film. In combination with the high-
temperature polymer PEEK several studies have been published [9,10,46,65]. Each group
reported that the compound performs tribologically better than the original materials.
The lowest wear rate can be found at 5 % PTFE in PEEK, but the lowest coefficient of
friction (0.25) is found at 15 % PTFE in PEEK, Lu et al. [46]. In comparison Briscoe
21
et al. [9] found the lowest coefficient of friction to be about 0.1 (over a broad range of
mixtures between 10 and 80 %) and showed that µ depends on the load and the operating
conditions. They stated (and agreed with Lu et al.) that the optimum between mechanical
and tribological properties is between 10 and 15 % PTFE in PEEK. The latter has the
lower coefficient of friction, but is mechanically weaker. In contradiction Burris and
Sawyer [10] found the optimum wear rate at 20 % PEEK in PTFE. The coefficient of
friction varied between 0.135 and 0.111 for the compounds. They also presented the
development of the wear rate and coefficient of friction over the reciprocating cycles: The
former decreases, the latter increases.
2.5 Gas dryness
The choice of material for a non-lubricated application also depends on the dryness of the
compressed gas. The moisture within a gas can work as lubricant or lead to significantly
higher wear rates. Three different types of dryness are differentiated by the gas’ dew
point: wet, dry and bone-dry in Table 2.4.
Table 2.4. Gas dryness and typical examples [32]
name dew point example compression process

◦C
wet > -20 air, chemical industry (CO2 )

air separation (O2 ), chemical industry (CO2 ),
dry -20 > and > -40
natural gas (turbine fuel gas)
air separation (Ar, He, N2 ),
bone-dry < -40
natural gas (liquid natural gas boil off)
22
Chapter 3
Mechanisms governing wear
Archard’s wear equation (3.1) is most frequently used in the literature when modelling
wear. It assumes that the wear rate dw dt is proportional to the contact pressure pc and the
relative velocity v of the contact partners. The proportionality factor, the wear coefficient
k, has to be determined experimentally.
dw
= kpc v (3.1)
dt
We assume that Archard’s equation is valid for the wear of packing rings. The next section
elaborates how to obtain the true, wear-relevant solid-solid contact pressure.
The relative velocity can directly be calculated from the crank mechanism, stroke and
the running speed of the crank drive. For the wear calculation we assume that a mean
velocity acts over time. Section 3.3 presents the considerations and thoughts on the time
scales behind this assumption.
3.1 Contact pressure
The gas pressure on the cylinder side pcyl and on the crank side pcrk load a packing ring,
we assume pcyl > pcrk . The pressure differential leads to contact between ring and rod and
ring and cup. Due to this pressure differential the ring seals pcyl from pcrk . This working
principle is called “self-energizing”, which is defined by Lindner-Silwester and Hold [44]
as “the seal effect arises from the differential gas pressure to be sealed”.
Figure 3.1a shows ring, cup and rod and additionally the gas pressure loads on a sealing
ring. The contact pressure distribution pc,flat at the sealing surfaces can be calculated
using a FE solver, if the loads, the ring geometry, the material properties and the contact
situation are known. This can be done in a straightforward way assuming that all contact
surfaces are not only nominally but also microscopically perfectly flat. Hence, no gas can
creep into the sealing surfaces, see Figure 3.1b.
In reality the surfaces of the contact partners are nominally, but not microscopically flat.
Thus gas can creep into the gaps between the contact partners and a gas pressure drop
23
arises from pcyl to pcrk . Figure 3.1c shows a gas pressure distribution pg in the sealing
surfaces assuming that surface roughness exists. The considerations concerning this gas
pressure distribution are detailed in the following section 3.2.
The true solid-solid contact pressure pc,rough can be calculated by taking all gas loads on
the free and the sealing surfaces into account. The gas pressure distribution in the sealing
surfaces pg reduces the previously calculated contact pressure pc,flat
pc,rough = pc,flat − pg . (3.2)
The resulting true solid-solid contact pressure pc,rough is highlighted green in Figure 3.1d.
Proportionally to pc,rough the ring’s dynamic sealing surface wears. No wear occurs bet-
ween ring and cup, since the relative velocity of these parts is zero. Figure 3.1e shows the
wear relevant contact pressure distribution pc,rough acting on the ring.
For later use, convenience and the convention used in equation (3.1) the true contact
pressure is renamed to
pc := pc,rough . (3.3)
24
a) b)
pcyl cup pcyl cup
cylinder side
cylinder side
pc,flat
crank side
crank side
ring ring
pcrk pcrk
rod rod
pc,flat
c) d)
pcyl cup pcyl cup
pg pg
cylinder side
pc,flat cylinder side pc,flat

crank side
crank side
ring ring
pcrk pcrk pc,rough

rod rod
pg pg pc,rough
pc,flat pc,flat
e)
cup
cylinder side
pc,rough
crank side
ring
rod pc,rough
Figure 3.1. Mechanisms governing ring wear. a) Gas pressure load on ring pcyl and pcrk .
b) Contact pressure pc,flat assuming perfectly flat contact surfaces. c) Gas
pressure pg in the sealing surfaces is added, which results from surface micro-
roughness. d) The true solid-solid contact pressure pc,rough (highlighted green)
takes the gas pressure drop pg into account. e) pc,rough to which wear is
proportional, if relative motion occurs.
25
3.2 Gas pressure distribution in a sealing surface
Here the investigation of Lindner-Silwester [43, section 2.3.1] concerning the gas pressure
distribution in the sealing surfaces of packing rings is summarised:
On a microscopic level the surfaces of ring and rod are rough. The surface roughness
results in a gap between the parts and allows gas to creep in. The gap is then replaced
by an equivalent constant gap independent of x, y, see Figure 3.2. With a gap height of a
few microns, an order of magnitude estimation reveals that inertia effects are negligible.
Hence, the flow, commonly referred to as Hele-Shaw flow [80] in the incompressible case,
is governed by the balance between frictional forces and pressure forces. In this limiting
case the absolute gas pressure pg is uniform over the gap height so that pg = pg (x, y).
pcyl −pcrk
When density variations are negligible which holds true for pcyl << 1, the pressure
distribution is governed by
∇2 pg = 0. (3.4)
When the pressure differential pcyl − pcrk across the ring can no longer be neglected
compared to pcyl , gas compressibility effects have to be taken into account. In the limiting
case of negligible inertia, i.e. kinetic energy, the adiabatic flow under consideration is to
a good approximation an isothermal one. For an ideal gas the density is then linearly
proportional to the absolute pressure pg , which together with the continuity equation
leads to
∇2 pg 2 = 0. (3.5)
For the numerical implementation the fact will be taken advantage of that “The equations
for many different physical situations have exactly the same appearance.” as is described
in Volume II, chapter 12, section 1 “The same equations have the same solutions” [24].
Steady heat flow problems in isotropic, homogeneous solid bodies of constant thermal
conductivity are also governed by Laplace’s equation
∇2 Tg = 0. (3.6)
Having the thermal conductivity vary in a linear way with the temperature Tg results in
∇2 Tg 2 = 0. (3.7)
This allows to use standard FE code to calculate the pressure distribution in the sealing
surfaces by treating it as heat transfer problem.
26
cup
pcyl
pg
cylinder side pc,flat
crank side
ring
pcrk
rod
pg pc,flat
Hele-Shaw configuration
z
smooth, soft (ring) material
gap area
y
asperity contact x
gap height
hard, rough (rod) material
Figure 3.2. Zoom into the dynamic sealing surface. The softer ring contacts the harder
rod, the surface roughness results in a gap. Here all surface roughness is
depicted for the rod. The gap is assumed to be of constant height and allows
to consider the gas flow in a Hele-Shaw configuration.
3.3 Consideration of time scales
This section presents the two physical time scales on which the wear process of rings
occurs. A fast time scale t which is of the magnitude of a crank revolution and a slow
time scale T which is of greater magnitude representing the wear process. The time scales
are matched and two examples show that the wear law on the slow time scale is valid and
allows to calculate wear over time correctly with highly reduced computational effort.
Note: In this section 3.3 and all its subsections the naming convention is slightly changed
to increase readability: the tilde denotes dimensional quantities, all non-tilde quantities
are non-dimensional.
Acknowledgement: This section was elaborated in cooperation with Dr. Matthias Korn-
feld.
27
3.3.1 Background and idea of the model
While maintaining sealing performance the packing rings wear over millions of strokes.
The wear is compensated geometrically and thus the sealing element geometry changes
slowly over time. These changes are slow compared to the time scale in which the piston
moves. Every single stroke involves a highly complex wear process due to the changes
in velocity and pressure. But the wear due to a single stroke is barely detectable and a
geometric change can hardly or not at all be measured.
The problem must be tackled with two different time scales: One is defined by the service
interval of the sealing elements, the slow time scale. This slow time scale is important when
considering long-term material effects (e.g. creep) and the wear of the sealing element.
Furthermore, reciprocating compressors are running with several hundred (approximately
200-2000, application dependent) strokes per minute and with every stroke the sealing
element wears. This process of wearing is physically occurring on the fast time scale of a
single stroke. Nevertheless, in this work regardless of the time scale the wear problem is
dealt with on a macroscopic level, similar to Archard’s considerations [1].
The computational costs for modelling wear over the sealing element’s lifetime are enor-
mous, if every single stroke has to be resolved. Even with massive computational resources,
such a calculation will take longer than the real-world wear process. Furthermore mo-
delling contact and wear in FE software is complex and may lead to convergence issues.
Thus it is beneficial or even necessary to reduce the complexity of the contact problem.
The analytical model presented shows, how to calculate the wear process on the (slow)
numerical time scale without resolving all real-time cycles occurring on the (fast) physical
time scale.
For sealing elements the load on the outer diameter varies between two pressure levels
(the suction p̃s and the discharge pressure p̃d , dependent on the compressors working
point), compression ratio, and delivery quantity. Additionally the gas pressure level might
change over time. It is assumed that the difference between suction pressure and discharge
pressure is constant even if the pressure level changes, Figure 3.3.
28
Figure 3.3. Schematic pressure function over time, for some cycles a zoom is shown on
the right.
The load-pressure function p̃L (t̃) generally can be developed into p̃cycle (t̃) from the com-
pression cycle and p̃p (t̃) depending on the process and application, e.g. filling a tank with
gas. The latter can be expanded into a Taylor series, with the initial pressure p̃p (t̃=t̃0 )
and the pressure change over time
p̃L (t̃) = p̃cycle (t̃) + p̃p (t̃)

∂p̃p (t̃) (3.8)
= p̃cycle (t̃) + p̃p (t̃0 ) + t̃=t̃0
(t̃ − t̃0 ) + h.o.t.
∂t̃
where t̃0 is usually set at t̃ = 0.
Figure 3.4a shows schematically the gas-pressure for the compression chamber, which is
equal to the gas pressure in the first cup. Additionally a possible gas pressure for the
second cup is drawn, which depends on the sealing efficiency of the previous sealing ele-
ment and is generally unknown [17,22]. The contact pressure p̃c (t̃) in the dynamic sealing
surface depends on the gas pressure on the outer diameter and on the ring geometry,
which changes due to wear. Evidently the sliding velocity is also a periodic function on
the same time scale, Figure 3.4b.
29
cylinder pressure (pressure 1st cup)
a)
pressure 2nd cup
p̃d
pressure
p̃s
0 90 180 270 360

crank angle (◦ )
b)
speed
0 90 180 270 360

crank angle (◦ )
Figure 3.4. a) Pressure over crank angle for the first and second sealing element. b) Rod
speed over crank angle. For both diagrams 0◦ is the head end centre point.
At the piston’s turning points the contact conditions change from sliding to sticking and
back to sliding into the other direction. This transition is not accounted for. Many
different wear equations have been developed in the recent decades, most of them only
for very special cases [50]. The oldest, best-known and still most frequently used wear
equation is Archard’s equation (3.9) [1]. (This equation is redefined here to be compliant
to the naming convention.) It states that the wear rate dw̃( dt̃
t̃)
is proportional to the
contact pressure p̃c (t̃) and the sliding velocity ṽ(t̃). The proportionality constant is the
wear coefficient k̃.
dw̃(t̃)
= k̃ p̃c (t̃)|ṽ(t̃)| (3.9)
dt̃
Archard’s equation performs well under constant conditions. Thus the wear coefficient is
30
often determined in a pin-on-disk experiment, with constant contact pressure and sliding
velocity or as a parameter of these values e.g. [6, 7]. To account for reciprocating motion
the wear coefficient can now be determined either as a parameter of velocity and pressure
or for a given motion and different contact pressures.
Some authors have worked with similar approaches, like for quasi-steady-state sliding
contacts by Lengiewicz et al. [40]. Lengiewicz et al. showed a concept for pin-on-flat
configurations, namely for the ball-on-flat configuration, how to cope with the two inher-
ited time scales of the wear-controlled problem under constant load, with both contact
partners changing their shape due to wear. Peigney [55] analytically investigated a wear
on half-plane problem with a rigid indenter on a solid continuum. He showed the asymp-
totically reached wear profile, if it exists for the investigated load. With his presented
minimization approach wear problems with cyclic loadings are considered. Another pro-
cedure how to analytically tackle steady-state wear problems for elastic bodies in relative
motion was presented by Páczelt and Mróz [52]. Páczelt and Mróz used a wear law where
the influence of contact pressure and relative sliding velocity is non-linear. After calcu-
lating the contact pressure distribution and the wear velocity vector with the variation
principle the shape of the contact surface is determined iteratively by the iterative op-
timization procedure. As extension of their method, Páczelt and Mróz mentioned that
it can be expanded to i) oscillating sliding motion ii) a change in the contact area due
to wear. These two poinst were addressed in a later work [53], where Páczelt and Mróz
presented the implementation of reciprocating sliding motion (pin-on-flat configuration)
using a variational principle and the p-version of finite elements.
Here a solution for wear problems using Archard’s wear equation for periodic pressure
and periodic sliding is presented. This enables to tackle long-term wear problems, like
the wear of dynamic sealing elements, numerically efficiently.
3.3.2 Model
The wear process of sealing rings of a compressor packing is considered. The sealing
elements are exposed to a reciprocating motion of the piston rod at the contact surface
between the rings and the piston rod with a periodicity given by the speed of the com-
pressor ω̃. Moreover, the rings are pressed onto the piston rod through a gas pressure
load p̃L acting on the ring surfaces and eventually yields a time-varying contact pressure
p̃c (t̃).
The specific wear coefficient k̃, reflecting the abrasion of sealing ring material from a ma-
croscopic point of view, i.e. a height reduction per unit of time, is expected to be small
compared to the radial ring height h̃ suggesting the introduction of a new time scale as-
sociated with the wear process.
The wear process and the pertaining suitable time scales for its description shall be descri-
bed by means of matched asymptotic expansions. To this end non-dimensional quantities
and formal asymptotic expansions are provided first. Subsequently the governing equati-
ons are discussed and an asymptotic matching of the time scales is performed.
31
3.3.2.1 Wear equation
The non-dimensional quantities, tilde denotes dimensional quantities, are defined as fol-
lows,
t = t̃ω̃/2π (3.10)
pc = p̃c (t̃)/p̃s (t̃0 ) (3.11)
w = w̃(t̃)/h̃(t̃0 ) (3.12)
Z 2π/ω̃
v = ṽ(t̃)/ṽM , with ṽM = |ṽ(t̃)|dt̃ (3.13)
0
Here t̃ denotes the time, ω̃ the angular speed of the compressor, p̃c (t̃) the contact pressure
between the sealing ring and piston rod surface, p̃s the suction pressure, w̃(t̃) the local
abrasion of sealing ring material due to wear, h̃ the radial ring height, ṽ(t̃) the piston rod
velocity and ṽM the mean piston rod velocity.
Archard’s equation (3.9) can be written in the following form
dw(t) 2π k̃ p̃s ṽM

= pc (t)|v(t)| = εt pc (t)|v(t)| (3.14)
dt h̃ω̃
with the non-dimensional parameter εt = (2π k̃ p̃s ṽM )/(h̃ω̃) accounting for all wear as-
sociated material properties from a macroscopic point of view. In the following, it will
be assumed that εt 1 which can be justified for typical values found for compressor
applications as shown in Table 3.1. Here, values for εt are in the range of 10−14 to 10−4 .
Table 3.1. Estimation of εt
name value range dimension

suction pressure 0.1 ≤ p̃s ≤ 10 MPa
mean rod velocity 0.1 ≤ ṽM ≤ 10 m s−1
wear coefficient 10−9 ≤ k̃ ≤ 10−5 mm3 N−1 m−1
angular velocity 5 ≤ ω̃ ≤ 30 s−1
(radial) ring height 5 ≤ h̃ ≤ 50 mm
32
3.3.2.2 Pressure load
The dimensionless pressure load pL (t), which can be considered as a given function, is
defined as follows:
pcycle (t) = p̃cycle (t̃)/p̃s (t̃0 ) (3.15)

pp (t0 ) = p̃p (t̃0 )/p̃s (t̃0 ) (3.16)

∂pp (t) ∂p̃p (t̃) ∂p̃p (t̃)
= /max (3.17)
∂t ∂t̃ ∂t̃

∂p̃p (t̃) 2π ∂pp (t)
pL (t) = pcycle (t) + pp (t0 ) + max t
∂t̃ ω̃ p̃s (t̃0 ) ∂t (3.18)
∂pp (t)
= pcycle (t) + pp (t0 ) + εL t
∂t
The pressure from the process is a function with two degrees of freedom: i) the initial
pressure level p̃p (t̃0 ) and ii) the pressure gradient ∂p̃p (t̃)/ ∂t̃. The initial pressure level is
related to the suction pressure level at p̃s (t̃0 ). To keep the information of the magnitude
of the dimensional pressure gradient, the pressure gradient is made dimensionless with a
characteristic gradient of the function, the maximum gradient: max( ∂p̃p (t̃)/ ∂t̃). Thus the
magnitude of the dimensionless gradient ∂pp (t)/ ∂t is of order 1 and the non-dimensional
∂p̃ (t̃)
parameter εL = max( p∂t̃ ) ω̃2π
p̃s accounts for the change of the process pressure load.
From Table 3.2 values of εL can be determined, ranging between 10−9 and 10−2 .
Table 3.2. Estimation of εL
name value range dimension

suction pressure 0.1 ≤ p̃s ≤ 10 MPa
∂p̃p (t̃)
pressure change rate 10−7 ≤ ∂t̃
≤ 10−3 MPa s−1
angular velocity 5 ≤ ω̃ ≤ 30 s−1
While the motion of the piston rod is exactly given by the crank mechanism and hence
the piston rod velocity v(t) is known a priori, the contact pressure pc (t) depends on the
periodically varying gas pressure load as well as on geometrical changes due to wear and
therefore is not known in advance.
Thus, for subsequent analysis the contact pressure pc (t) is decomposed into a periodic part
pper (t), associated with the gas pressure, and a non-periodic part pnon per (t), associated
with the change of the contact pressure with time due to wear, as follows
pc (t) = pper (t) + pnon per (t) (3.19)
With the inspection analysis performed in the previous section a formal asymptotic ex-
pansion for the contact pressure and the wear can be formulated. For these two quantities
33
the inspection analysis suggests the following scaling
pc = p0 (t) + εt p1 (t) + O(εt 2 ) (3.20)

w = w0 (t) + εt w1 (t) + εt 2 w2 (t) + O(εt 3 ) (3.21)
Inserting the asymptotic expansions (3.20) and (3.21) into Archard’s equation (3.14) gives
d
(w0 (t) + εt w1 (t) + εt 2 w2 (t) + . . .) = εt (p0 (t) + εt p1 (t) + . . .)|v(t)| (3.22)
dt
which finally yields in leading order
d
O(1) : (w0 (t)) = 0. (3.23)
dt
Evaluating also the second and third order approximations yields
d
O(εt ) : (w1 (t)) = p0 (t)|v(t)| (3.24)
dt
d
O(εt 2 ) : (w2 (t)) = p1 (t)|v(t)| (3.25)
dt
For later use expanding the non-periodic part of the decomposition pnon per (t) being a
function of the contact pressure change due to i) wear of the contact area Ac (determined
by the stiffness and the wear of the ring geometry) and ii) an external pressure load change
over time pL (t) gives
Z t Z t
∂pc ∂Ac ∂w ∂pc ∂pL
pnon per (t) = 0 + t=0
dτ t + t=0
dτ + h.o.t. , (3.26)
∂Ac ∂w 0 ∂τ ∂pL 0 ∂τ
Rt Rt
where higher order terms can be neglected due to ( 0 ∂∂wτ dτ )2 0 ∂∂wτ dτ . The pressure
load function pL (t) must be a smooth function. Inserting equation (3.24) and the last
term of (3.18) (the first two terms are considered in the periodic pressure pper (t)) into
(3.26) finally yields
Z t Z t
∂pc ∂Ac ∂pc ∂pp
pnon per (t) = εt t=0
p0 (τ )|v(τ )|dτ + εL t=0
dτ . (3.27)
∂Ac ∂w 0 ∂pL 0 ∂τ
Now there are three different cases to be considered case A) εL ∼ O(1), case B) εL ∼ O(εt )
and case C) εL εt :
• Case A) εL ∼ O(1) indicates a significant pressure change within few compression

cycles, e.g. an emergency shut-down of a compressor. Here the solution can be
obtained piece-wise before and after the pressure change, Figure 3.5a.
• Case B) εL ∼ O(εt ) could be the filling of a cavern with natural gas, a pipeline or
a (big) gas storage, Figure 3.5b.
• Case C) εL εt means the pressure on the ring is constant, Figure 3.5c.
34
(a) Case A: εL ∼ O(1)
(b) Case B: εL ∼ O(εt )
(c) Case C: εL εt
Figure 3.5. Schematic sketch of the different load changing situations.
35
The cases A) and C) are generic limiting cases of B), thus only case B) will be considered
further: Inserting equation (3.27) into the decomposition (3.19) and comparison of the
order of magnitudes of the expansion of the contact pressure (3.20) shows
p0 = pper (t),
Z t Z t (3.28)
∂pc ∂Ac ∂pc ∂pp
p1 = t=0
p0 (τ )|v(τ )|dτ + t=0
dτ .
∂Ac ∂w 0 ∂pL 0 ∂τ
Hence, equation (3.25) yields
Z t
dw2 ∂pc ∂Ac ∂w
= t=0
dτ
dt ∂Ac ∂w 0 ∂τ
Z t
(3.29)
∂pc ∂Ac
= t=0
p0 (τ )|v(τ )|dτ t|v(t)|
∂Ac ∂w 0
3.3.2.3 Time scales
Wear analysis solely based on the leading order approximation based on equation (3.23)
obviously does not predict any wear. Visible effects only appear if higher order terms are
taken into account at longer times giving rise to introducing a second time scale, further
referred to as slow time scale T , where the wear effects become significant. Here, the
wear, to be precise, wear of the order of magnitude of the ring geometry, is zero on the
fast time scale t
T = εt t. (3.30)
Consequently, the asymptotic expansions for the contact pressure and wear on the time
scale T can be written as
P (T ) = P0 (T ) + εt P1 (T ) + O(εt 2 ) (3.31)
W (T ) = W0 (T ) + εt W1 (T ) + O(εt 2 ) (3.32)
Here, the asymptotic expansion is denoted via capital letters to distinguish the expansion
of the slow time scale quantities from the fast time scale quantities. Inserting the expan-
sions (3.31) and (3.32) into Archard’s equation (3.14), which has to hold true even on the
slow time scale gives
d
O(1) : (W0 (T )) = P0 (T )V (T ) (3.33)
dT
d
O(εt ) : (W1 (T )) = P1 (T )V (T ) (3.34)
dT
With the equations (3.23) to (3.25) and equation (3.33) and (3.34) the problem is properly
posed. Here, the quantities on the t time scale contain the boundary conditions, namely
the contact pressure and the rod velocity, while the wear is governed by the leading order
equation (3.33) on the T time scale. Hence, to close the problem formulation the two
time scales need to be matched properly.
36
To this end, the following matching condition is introduced, i.e. wear on the real (fast)
time scale must accumulate to the same value as in the virtual (slow) time scale,
1
lim w(t = O( )) = W (T = O(1)), (3.35)
εt →0 εt
which matches the wear w(t) on the fast time scale t for a sufficiently large time with the
wear W (T ) on the slow time scale T where the time is an order one quantity. Inserting the
equations (3.24) and (3.33) into the matching condition and integrating over an arbitrary
time step ∆T finally yields
Z ∆T Z ∆T
εt
W0 (T = ∆T ) = lim εt p0 (t)|v(t)|dt = P0 (T )V (T )dT (3.36)
εt →0 0 0
Note that only the leading order is non-zero due to εt 2 , εt 3 , ... ∼ 0. Expanding the
periodic part of the contact pressure and the piston rod velocity into a Fourier series
∞
X
p0 (t) = S0 + Sn,1 cos(2πtn) + Sn,2 sin(2πtn) (3.37)
n=1
∞
X
|v(t)| = V0 + Vn,1 cos(2πtn) + Vn,2 sin(2πtn) (3.38)
n=1
and introducing these series into the matching condition (3.36) gives
Z ∆T
εt C0
W0 (∆T ) = εt p0 (t)|v(t)dt = (S0 V0 + C0 )∆T = S0 (V0 + )∆T , (3.39)
0 S0
where only terms of O(1) are considered. Here S0 and V0 denote the mean quantities of
the contact pressure and the
R absolute value of Rpiston velocity, respectively. C0 denotes a
2 2
constant arising from the cosR (2πtn)dt andR sin (2πtn)dt terms
R of the Fourier series,
whereas the other terms like cos(2πtn)dt, sin(2πtn)dt or cos(2πtn) sin(2πtn)dt are
negligible.
Comparing equation (3.39) with the leading order approximation (3.33) on the slow time
scale shows that the contact pressure P0 as well as the piston velocity V can be considered
as constant on the slow time scale T and has to fulfil the following condition for proper
matching of the time scales
P0 = S 0 (3.40)
C0
V = V0 + (3.41)
S0
Thus, the leading order approximation of Archard’s equation for reciprocating contact
surface motion and periodically varying contact pressure on a slow time scale, where wear
effects become significant, is found to be
dW
= S0 K(S0 , V0 ), (3.42)
dT
37
where K(S0 , V0 ) is a virtual wear parameter which can be determined directly by evalu-
R1
ating the terms in equation (3.41) and S0 = 0 p0 (t)dt is the averaged contact pressure
seen on the fast time scale. The approximation of Archard’s law stated in equation (3.42)
reflects the wear behaviour on a slow time scale derived from the governing equations for-
mulated on the fast physical time scale. Thus, equation (3.42) is the proper formulation
of a wear law for the implementation in numerical codes studying wear effects where two
time scales are present.
3.3.3 Example: non-linear wear law
The asymptotic concept was presented with Archard’s (linear) wear equation (3.9) as
basis. Since the general non-linear wear equation form (e.g. [62]) can be written as
∆w̃ = k̃2 p̃c a ṽ b t̃c , (3.43)
where k̃2 is the wear coefficient, p̃c the contact pressure, ṽ the relative velocity, t̃ the time
of sliding and a, b and c are a set of parameters, which have to be (experimentally) derived
for a contact pair.
To show the validity of the presented concept a non-linear wear equation (3.43) will be
considered with the parameters a = 2, b = c = 1:
dw̃(t̃)
= k̃2 p̃c 2 (t̃)|ṽ(t̃)| (3.44)
dt̃
This choice of parameters makes an analytical solution amenable without enormous ef-
forts.
The non-dimensional quantities are defined as shown above in the equations (3.10) to
(3.13). Thus the wear equation can be rewritten into dimensionless quantities as follows
dw(t) 2π k̃2 p̃s 2 ṽM 2

= pc (t)|v(t)| = εt pc 2 (t)|v(t)| (3.45)
dt h̃ω̃
Inserting the asymptotic expansions (3.20) and (3.21) into the non-linear wear equa-
tion (3.45) gives
d
(w0 (t)+εt w1 (t) + εt 2 w2 (t) + ...) =
dt
= εt (p0 (t) + εt p1 (t) + ...)2 |v(t)| (3.46)
= εt (p0 2 (t) + 2εt p0 (t)p1 (t) + εt 2 p1 2 (t) + ...)|v(t)|
which finally yields in leading order

d
O(1) : (w0 (t)) = 0. (3.47)
dt
38
Evaluating also the second and third order approximations yields
d
O(εt ) : (t) = p0 2 (t)|v(t)| (3.48)
dw1
d
O(εt 2 ) : (t) = 2p0 (t)p1 (t)|v(t)| (3.49)
dw2
Consequently the slow time scale is derived as described in the equations (3.30) to (3.32).
Leading to
d
O(1) : (W0 (T )) = P0 2 (T )V (T ) (3.50)
dT
d
O(εt ) : (W1 (T )) = 2P0 (T )P1 (T )V (T ) (3.51)
dT
By using the matching condition (3.35) and integrating over an arbitrary time increment
∆T yields
Z ∆T Z ∆T
εt
2
W0 (t = ∆T ) = lim εt p0 (t)|v(t)|dt = P0 2 (T )V (T )dT (3.52)
εt →0 0 0
Next the velocity term is expanded into a Fourier series as described in equation (3.38).
The square of the periodic part of the contact pressure is expanded as follows
∞
X
p0 2 (t) = S0,nl + (Sn,1 cos(2πtn) + Sn,2 sin(2πtn)) (3.53)
n=1
Thus the matching condition yields

Z ∆T
εt
W0 (∆T ) = εt p0 2 (t)|v(t)|dt
0
= (S0,nl V0 + C0 )∆T (3.54)
C0
= S0,nl (V0 + )∆T
S0,nl
where the result is similar to equation (3.39). Note that P0 2 (T ) = S0,nl . For the linear and
the non-linear wear equation the result looks formally equal, but the derived constants S0
R ∆T
ε
and S0,nl are different. For the first case, Archard’s wear equation (3.9), S0 = 0 t p0 (t)dt,
R ∆T
ε
whereas for the non-linear case S0 = 0 t p0 2 (t)dt. Note that in general S0,nl 6= S0 2 .
Figure (3.14) now shows the difference between the linear and the non-linear wear law. It
must be emphasized that in the non-linear wear case it is not valid to simply insert the
mean value into Archard’s equation.
39
p0
pressure p0 2
S0
S0,nl
time
Figure 3.6. Comparison of the cyclic pressure for the linear (red) and non-linear (blue)
wear equation including the resulting “mean” values for both models, S0 and
S0,nl respectively.
3.3.4 Error estimation
The presented concept works well in a closed vicinity of the evolution point T = 0. The
question arises at which time increment ∆T the approximation errors become no longer
negligible.
The truncation error of this approach can be estimated by evaluating the higher order
terms. Since the physical behaviour governed by the equations on the fast time scale must
be reflected also on the slow time scale the O(εt )-terms stated in equation (3.34) have to
exhibit a similar form as the terms stated in equation (3.29)
∂Ac T ∂W ∂pc T
Z Z
dW1 ∂pc ∂pL
= dT V0 + dT
dT ∂Ac ∂W 0 ∂T ∂pL 0 ∂T
(3.55)
∂pc T
Z
∂pc ∂Ac ∂pL
= S0 K(S0 , V0 )T + dT .
∂Ac ∂W ∂pL 0 ∂T
Rt∂pL R εTt ∂pL RT ∂pL ∂T 1 RT ∂pL ∂T

Note: 0 ∂t dt = 0 ∂t
dt = 0 ∂T ∂t εt
dT = 0 ∂T
dT , where ∂t
= εt for case B.
Assuming piecewise constant terms ∂∂Apcc , ∂∂A

W
c and ∂pc
∂pL
and integrating equation (3.55)
over an arbitrary time step ∆T finally yields
∆T T
∆T 2
Z Z
∂pc ∂Ac ∂pc ∂pL
W1 = S0 K(S0 , V0 ) + dτ dT . (3.56)
∂Ac ∂W 2 ∂pL 0 0 ∂τ
The inner integration over τ is equivalent to the integration in (3.55), but for distinction of
integrand and integration limit T was substituted by τ . Thus, the asymptotic series (3.32)
40
can be written as
Z ∆T Z T
∂pc ∂Ac ∆T ∂pc ∂pL
W = S0 K(S0 V0 )∆T (1 + εt ) + εt dτ dT . (3.57)
∂Ac ∂W 2 ∂pL 0 0 ∂τ
Comparing this result to the numerical implementation of Archard’s equation as stated
in equation (3.42) yields the following condition for the time step size
∂pc ∆T T ∂pL
Z Z
∂pc ∂Ac ∆T
εt 1 and εt dτ dT 1. (3.58)
∂Ac ∂W 2 ∂pL 0 0 ∂τ
Here, the terms ∂pc / ∂Ac and ∂Ac / ∂W are assumed to be of order O(1). This assumption
is violated for configurations where two bodies are in contact in a single point or line, e.g.
the wear of a spherical pin, since the term ∂Ac / ∂W |(T =0) O(1/εt ) becomes very large.
For such geometrical configurations the wear time scale overlaps with the physical time
scale associated with the surface motion. Thus, the problem has to be treated first on the
fast physical time scale until a finite and sufficiently large contact area is present, which
allows switching to the slower wear time scale.
Furthermore, ∂pL / ∂t ≤ O(εt ) is fulfilled for the considered cases B and C.
For the problem formulation of packing rings the term ∂∂Apcc can be further expanded into
∂p ∂pf
a term ∂AAc resulting from the change in contact area and a term ∂Ac
resulting from the
change in gas pressure acting on the ring surfaces.
∂pc ∂pA ∂pf
= + (3.59)
∂Ac ∂Ac ∂Ac
∂pA ∂pf
Since for typical packing ring geometries the term ∂Ac
is larger than the term ∂Ac
, the
∂pA
term ∂Ac
determines the size of the time increment for updating the geometry. The
∂p ∂p
integer of the ratio ∂AAc / ∂Afc represents the calculation cycle after which the gas pressure
distribution in the dynamic sealing surface has to be updated. Hence, condition (3.58)
finally yields the following condition for the maximum time step size
errthreshold
∆Tmax ≤ ∂p ∂p
, (3.60)
εt ∂AAc ∂∂A
W
c + εt ∂Afc ∂∂A
W
c
where the error threshold errthreshold is chosen sufficiently small.
3.3.5 Proof of concept
With a numerical experiment it can be shown, that the developed concept is valid. A
calculation on the fast time scale is compared to a calculation on the slow time scale.
Here a simplified sealing element configuration is considered, Figure 5.8, namely an uncut
single piece packing ring of rectangular cross section [36]. This is closely elaborated in
section 5.4. Note that the conclusions drawn here are valid for sealing elements of higher
geometrical complexity as well. In equation (5.7) (from [36]) the change of the inner ring
41
radius due to wear over time is dependent of i) geometrical features: the outer ring radius,
the rod diameter; ii) material parameters: the shear modulus and Poisson’s ratio; iii) the
pressure acting on the outer radius of the ring; iv) the wear coefficient and v) the relative
velocity between the contacting surfaces.
For both time scales an Euler forward scheme is implemented to calculate the radial ring
wear over time. On the fast time scale the piston rod movement is implemented as a
sine function, which is a good approximation of the crank shaft motion, Figure 3.4b,
and is resolved in 10◦ steps. On the slow time scale the wear-model relevant velocity is
implemented as mean value of the absolute value of the rod velocity. The pressure on the
outer diameter is held constant in both cases. This case is similar to the load situation
of the last (crank end) packing ring of a piston rod sealing system [22].
wi+1 = wi + εt v(t)pc ∆t (3.61)

W i+1 = W i + vpc ∆T (3.62)
The results resolving the problem on the fast time scale as given in equation (3.29) as
well as the numerical implementation as given by equation (3.62) are shown in Figure 3.7.
The parameters for the validation are listed in Table 3.3. The rod diameter is denoted
with D̃.
Table 3.3. Parameters for the validation
parameter value and dimension non-dimensional value

R̃(o) 33.4 mm
25.4 mm 1.315
R̃(i)
D̃ 50.8 mm
h̃ = 8 mm 6.35
R̃(o) −R̃(i)
p̃L 10 MPa
500 MPa 0.02
G̃
ν 0.4 0.4
D̃
R̃(i) 0 = 2 25.4 mm -
ṽM 5 m s−1 -
For a more complex load case on the same ring, with a given, changing pressure load
function the dimensionless pressure load and resulting contact pressure are depicted in
Figure 3.8 and the resulting dimensionless wear in Figure 3.9.
42
0.12
0.1
4 ·10−9
0.08 w(t)
t0
3
w − w(t0 )
wear (-)
0.06 2
w(t0 )
1
0.04
0
0 0.5 1
t − t0
0.02
w(t)
W (T )
0
0 1 2 3 4 5 6 7 8 9
revolutions ·107
Figure 3.7. Validation of the model: The diagram shows the accumulated wear (per ring
height) over the number of revolutions. The calculation of the fast time scale
t is drawn as red line, every tenth calculated point on the slow time scale T
is marked with a blue circle. The smaller plot shows the wear over the 107 th
revolution, here each of the 36 calculated points is marked with a red plus.
43
7
−7
0 ·10
6 pc (t)
pc -pc (t0 )
-4
5
-8
pressure (-)
-12
3 0 0.5 1
t − t0
2
t0 pc (t)
Pc (T )
pc (t0 ) 1 p̃L /p̃s
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
revolutions ·107
Figure 3.8. The change in the given pressure load (dotted line) and the resulting contact
pressure over time, calculated on both time scales. The zoom shows the
contact pressure of the 107 th revolution on the fast time scale. A plus marks
every calculated point in time.
44
·10−2
7
5 1.5 ·10−9
t0 w(t)
w − w(t0 )
4 w(t0 ) 1
wear (-)
3 0.5
2 0
0 0.5 1
t − t0
1 w(t)
W (T )
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
revolutions ·107
Figure 3.9. Time scale model comparison for the accumulated wear (per ring height) over
the number of revolutions for the changing load presented in Figure 3.8. The
calculation of the fast time scale t is drawn as red line, every tenth calculated
point on the slow time scale T is marked with a blue circle. The smaller plot
shows the wear over the 107 th revolution, here each of the 36 calculated points
is marked with a red plus.
Although contact pressure, sliding velocity and wear rates change within a single stroke
on a fast time scale, it could be shown that by using an asymptotic approach wear can
be calculated quantitatively correct on a slow time scale without considering every single
stroke. It is possible to numerically implement wear only on the slow time scale without
resolving the reciprocating motion of the contacting surfaces, because these quantities are
analytically included into the wear coefficient on the slow time scale.
Through the introduction of the slow time scale the calculation times are reduced signi-
ficantly even for the simplified one dimensional analytical wear consideration presented
above. The wear calculation for example 1, Figure 3.7, on the fast time scale evaluates
3.24 · 109 points whereas on the slow time scale only 330 points are evaluated.
45
Chapter 4
Material
The goal to model packing ring wear efficiently can only be achieved by making necessary
model restrictions and assumptions. This chapter presents the tribological modelling of
a mounted packing ring as well as the thoughts and considerations for the mechanical
modelling of the used packing ring material including all necessary test methods.
4.1 Tribological modelling
The fact that tribology is a system property is probably the greatest challenge in de-
veloping rings. Many interacting parameters influence how a ring (or packing) behaves.
Some of the parameters can be controlled more easily (e.g. rod velocity, gas pressure, gas
type), some with more effort and even then only to a certain extent (gas dryness, surface
characteristics like roughness or profile) and some are nearly uncontrollable e.g. the rod
temperature or leakage.
The contact layer where the tribo-chemical and mechanical interactions of asperities take
place is inevitably part of the model, but not resolved in detail. It is important to under-
stand that the calculated results in the contact layer, e.g. structural stresses and strains
are not reliable. The high deformation gradients and the flash temperatures occurring
during the wear process can currently not be evaluated in the same calculation where the
ring wears in the magnitude of millimetres. An example for nano-scale modelling is the
work of Barry et al. [4]. They studied only few PTFE molecule chains and found that the
coefficient of friction increases with decreasing temperature (investigated between 25 and
300 K) for the different investigated molecular chain orientations.
The fact that the contact layer is not modelled in detail does not mean that it is disregar-
ded or not of importance. The micro structure of ring and rod are not modelled, but the
gas pressure distribution arising from that micro structure is a crucial ingredient and part
of the model. The typical PTFE transfer film or its formation is not modelled explicitly,
but the coefficient of friction or the wear coefficient is derived from configurations where
a transfer film is present. This is valid since the formation of a quasi steady-state transfer
46
film is a short process compared to a ring’s lifetime.
The frictional force of a packing ring design under given boundary conditions is a bench-
mark criterion. The lower the frictional force the more likely the ring runs cooler, wears
more slowly and has a longer lifetime.
The frictional force can be derived from the contact pressure distribution in the dynamic
sealing surface assuming Coulomb friction. To compare simulation and experiment of a
loaded packing ring the coefficient of friction has to be determined as well as the gas
pressure dependent frictional force. Both is possible with the testing device presented in
the following.
4.1.1 Frictional force tester
A purpose-designed test rig allows to measure the frictional force of a packing ring moun-
ted on a rod in a T-cup, which can be pressurized, see Figure 4.1. The static part of
the frictional force tester is mounted on a standard testing machine of the type Zwick
Roell Z010, company Zwick GmbH & Co. KG, Germany. Bolts prevent rotation of the
frictional force tester, which are secured with nuts.
A rod is fixed to a 500 N load cell, precision class 0.05 %, which pulls or pushes the rod
through the packing. The radial clearance of 1.6 mm between the 50.8 mm (2”) rod and
the cup flanges is similar to the clearances in a compressor. The rods running surface
is 350 mm long. Figure 4.2 names all parts of the frictional force tester. The tests are
performed in a climate controlled laboratory at 20 ◦C.
47
load cell
top fixture
rod
top flange
T-cup
pressure inlet
bottom flange
columns
bottom fixture
Figure 4.1. Friction force tester. Left: Sketch. Right: Photograph of the mounted friction
force tester on the standard testing machine.
48
A-A
1
8 8
3
11a 10
4 9
11b 10
5
6 6
A
7
8 8
1
Figure 4.2. Side and top view of the friction testing device, including a cut, where all parts
are named. (1) pin hole, fixture to the force measurement device, (2) rod, (3)
top flange, (4) T-cup, (5) bottom flange, (6) columns, (7) bottom fixture to
the testing machine, (8) screw, (9) pressure inlet, (10) O-ring groove, (11a)
upper ring, (11b) lower ring.
49
4.1.2 Rod surface
The counter-face materials which are used most frequently in the industry are X20Cr13
steel (cylinder liner) and tungsten carbide (WC) coated steel. The rods where turned at
the HOERBIGER Wien (HW) tool shop from X20Cr13 steel. The surface finishing of the
steel rod was done at the HW. The high velocity oxygen fuel spraying (HVOF) coating
and surface finishing of the WC coated rod was performed by TLBS GmbH, Vienna. A
WC coating is commonly used for rods in non-lubricated service, consisting of WC/Co/Cr
with 86/10/4 [31].The coated rod will further be referred to as ’WC rod’.
The surface of the rods was characterized with a Hommel-Etamic T8000 and a sensing
arm of type TS1 KE5/90GD T1.9 D4/30 testing device. The roughness was determined
on six (0-5) axial positions and at each on four circumferential positions for the steel rod.
The WC rod is characterised on five (1-5) axial and four circumferential positions, see
Figure 4.3. Position 0 is not WC coated.
initial position upper ring

5 24
initial position lower ring
3
330
260
2
190
120
1
10
50
180
270 90
Figure 4.3. The blue dashed lines mark the roughness measurement positions. For the
steel rod on all 6 positions a reading is taken. The WC rod is measured on
position 1 to 5, since position 0 is not coated.
50
The roughness was determined according to the DIN EN ISO 4287 standard. On each
position over a total measurement length of 4.8 mm five measurements were performed
with 0.8 mm each. Usually the surface roughness for rods is specified with Ra = 0.3 ±
0.05µm [31]. The steel rod does not meet this standard, with a maximum of Ra = 0.8 µm,
Figure 4.4, but allows to check the influence of the roughness on the frictional force. The
WC rod meets the standard very well, Figure 4.5.
1
0◦
90◦
0.8 180◦
270◦
0.6
Ra (µm)
0.4
0.2
0
10 50 120 190 260 330
axial position (mm)
Figure 4.4. Ra measurements on the steel rod at the locations marked in Figure 4.3.
1
0◦
90◦
0.8 180◦
270◦
0.6
Ra (µm)
0.4
0.2
0
50 120 190 260 330
axial position (mm)
Figure 4.5. Ra measurements on the WC rod at the locations marked in Figure 4.3.
51
4.1.3 Coefficient of friction
To determine the coefficient of friction a special 20-piece ring design, named “segmented
ring”, is used, see Figure 4.6. For a sufficiently narrow ring segment the bending stiff-
ness becomes negligible. That allows to measure the frictional force uninfluenced by the
material’s elastic properties. The segments are equally spaced around the rod, which is
ensured by a mounting device. The segmented ring has two circumferential grooves on the
outer diameter, in which garter springs can be placed to press the ring against the rod.
This ring has no sealing properties, since it is used for the determination of the coefficient
of friction only.
The following procedure is used when assembling the segmented ring and mounting it
on a rod. The ring pieces are placed on a two-piece mounting device, consisting of a
sleeve and a spacer, see Figure 4.7 (garter springs not depicted). Then the garter springs
are mounted in the ring’s grooves. The mounting device is pulled over the rod and the
sleeve is removed, leaving the ring and the spacer on the rod, Figure 4.8. After carefully
removing the spacer the ring pieces are mounted equally spaced on the rod, see Figure 4.9.
20x18 ◦ 3.3 3.3

2
2
R2
R
12
50.8
67.8
69
Figure 4.6. The 20 piece ring sketched in 3D and its drawing, including a cross section of
a ring segment.
52
Figure 4.7. The mounting device used for the segmented ring. From left to right: Sleeve,
spacer and segmented ring.
Figure 4.8. The mounting device with the assembled ring on a rod. The garter springs
are not depicted.
53
rod
segmented ring
garter spring
ring groove
T-cup
Figure 4.9. The segmented ring is mounted on a rod, held by two garter springs. In the
background the T-cup with the upper ring groove is visible.
4.1.3.1 Normal force between rod and segmented ring
The coefficient of friction µ is calculated from the rod’s measured pull-off force and the
knowledge of the contact pressure applied by the garter springs. The friction is assumed
to be of Coulomb type, equation (4.1). The contact force is referred to as FC and the
frictional force as FF .
FF = µFC (4.1)
The contact force resulting from the garter spring force is determined from the geometry
of a single ring segment. The angle between the spring forces FS is α. A force triangle
and the law of cosines gives the contact force FC .
q p
FC = FS 2 + FS 2 − 2FS 2 cos(α) = FS 2(1 − cos(α)) (4.2)
To get the total contact force FC t between ring and rod the number of springs nS and
the number of ring segments nseg has to be taken into account. We assume that the
ring segments are all equal, thus a multiplication accounts for their number. Since garter
springs sometimes differ in their offset value at the beginning of the elongation, the forces
54
of the springs are summed up.
nS
p X
FC t = nseg 2(1 − cos(α)) FSi (4.3)
i=1
Figure 4.10 shows a sketch of a single 18◦ ring segment from the segmented ring and
the force triangle. A 2 mm milling cutter separates the segments. Thus the wrap of the
garter spring (and the angle between the spring forces in the force triangle) is 14.62◦ at
a diameter of 67.8 mm. This is the diameter of the centre of the garter springs.
18◦
1 1
FS FS
FC
50.8
67.8
0
69.
14.62◦
α
force triangle:
FS
FC α
FS
Figure 4.10. Sketch of a single ring segment and all acting forces. The force triangle for
the determination of the contact force FS is shown on the bottom.
55
4.1.3.2 Garter springs
The length of the garter springs is determined with a calliper. Spring 1 has a total length
of 195.3 mm and an effective spring length of 188.3 mm. The effective spring length is the
total spring length reduced by the length of the hooks on each spring end, see Figure 4.11.
This effective spring length is used for the calculation of the spring elongation. Spring 2
has a total length of 195.6 mm with an effective spring length of 188.6 mm. The mounted
springs are elongated to a total length of 213 mm, i.e. an elongation of 10 % of the effective
spring length.
hook 1
effective spring length
total spring length
hook 2
Figure 4.11. Sketch of a garter spring: total and effective spring length.
The spring force at an elongation of 18.7 and 19 mm is of interest for spring 1 and 2,
respectively. The garter springs are tested on a standard testing machine. The springs
are mounted loosely (no initial tension) on pins at the bottom fixture and at the load cell.
The springs are elongated by 22 mm, i.e, 11.6 % of the initial spring length.
The force-displacement curves are evaluated above 0.5 N to remove the offset resulting
from the loose mounting. Figure 4.12 shows the force-displacement curve for each garter
spring. The marked point represents the mounted situation. The springs have the same
slope after an initial offset. The offset itself is different for each spring.
56
25
garter spring force (N) 20
15
10
5
spring 1
spring 2
0
0 2 4 6 8 10 12 14 16 18 20
displacement (mm)
Figure 4.12. Force-displacement diagram of the two springs mounted on the segmented
ring. The marked point represents the spring load in the mounted configu-
ration.
When mounted on the segmented ring spring 1 acts with 23.1 N and spring 2 with 18.1 N on
the ring. From the measurement the contact force FC can be evaluated with equation (4.2)
for a single ring segment
p
FC = (18.1 + 23.1) 2(1 − cos(14.62)) = 10.5 N
and with equation (4.3) the total contact force for 20 piece ring is
FC t = 209 N.
4.1.3.3 Measurement
Cleaning with ethanol removes any oil, grease or other contamination from ring and rod
and ensures non-lubricated measurement conditions. The segmented ring is placed on the
rod with the mounting device. The rod’s and the ring’s weight are considered by setting
the load cell to zero with the ring mounted on the rod, both freely hanging from the load
cell. The ring is located inside the upper recess of the T-cup.
The coefficient of friction is determined dependent on velocity. The rod speed is varied
from 10 to 1000 mm min−1 . The latter is the testing machine’s maximum speed. The
total displacement of the rod depends on the relative velocity. For effective testing the
57
total displacement is lower at lower speeds.
Each measurement consists of at least 5 cycles. Within a cycle the rod is first pulled
upwards, then pushed downwards. The higher the relative velocity the higher the number
of cycles, see Figure 4.13 and 4.14 and Table 4.1.
The coefficient of friction is determined from the mean contact force over the measurement
distance, disregarding the first and last 10 mm of the travelled distance. Thus the result
is not influenced by acceleration or deceleration of the rod or changes in contact mode
(sliding/sticking) at the turning or the starting event.
50
frictional force (N)
-50
0 50 100 150 200 250
position (mm)
10 mm min−1 5 cycles
Figure 4.13. Frictional force measurement on the steel rod.
For each cycle the mean force in upward and downward direction is calculated and the
coefficient of friction (µup and µdown ) is determined with equation (4.1). In table 4.1 the
coefficients of friction is shown including the standard deviation. The measurement is
highly reproducible over the performed measurement cycles, with a maximum standard
deviation of 2.5 % for the steel rod and 0.42 % for the WC rod, when considering the only
one direction of motion. We assume that the coefficient of friction is independent from
the direction because both surfaces are smooth and have no orientation or surface profile.
58
Figure 4.15 presents the mean value calculated from both directions of motion including
the standard deviation.
The measurements presented in Figure 4.13 show no significant influence of the roughness
Ra (Figure 4.4) on the frictional force. The frictional force variation along the axial rod
position is not coincident with the roughness profile measured. For the measurements
on the WC rod with constant roughness, Figure 4.14, the frictional force shows more
pronounced slip-stick over the whole measurement length than on the steel rod. In both
measurements the axial clearance between ring and cup can be seen at the turning points.
The coefficient of friction varies more strongly on the WC rod. It changes from 0.16 to 0.20
between 10 and 1000 mm min−1 , respectively. On the steel rod the coefficient increases
from 0.19 to 0.21 over the same relative velocity range. Figure 4.15 and Table 4.1 show
the detailed measurement data including the number of cycles and the standard deviation
for both rods.
50
-50
0 50 100 150 200 250
position (mm)
10 mm min−1
100 mm min−1
500 mm min−1
750 mm min−1
1000 mm min−1
Figure 4.14. Frictional force measurement on the WC rod, for every rod speed 10 cycles
were run.
59
Table 4.1. Detailed velocity dependent measurement data from the determination of the coefficient of friction for HY54 running
on a steel rod or on a WC rod. The mean value and its standard deviation is given for the rod movement upwards,
downwards and independent from the movement direction.
material relative velocity displacement cycles µup µdown µ

(mm/min) (mm)
10 100 5 0.1900 ± 0.0029 0.1933 ± 0.0019 0.1916 ± 0.0030
60
100 200 10 0.1973 ± 0.0050 0.2061 ± 0.0042 0.2017 ± 0.0064
steel
500 250 10 0.2002 ± 0.0023 0.2094 ± 0.0022 0.2048 ± 0.0051
HY54
1000 250 20 0.2048 ± 0.0005 0.2138 ± 0.0009 0.2093 ± 0.0045
10 100 5 0.1547 ± 0.0005 0.1623 ± 0.0003 0.1585 ± 0.0038
100 200 10 0.1701 ± 0.0002 0.1778 ± 0.0003 0.1739 ± 0.0039
500 250 10 0.1861 ± 0.0004 0.1942 ± 0.0003 0.1901 ± 0.0041
HY54
750 250 10 0.1891 ± 0.0003 0.1988 ± 0.0003 0.1940 ± 0.0049
WC/Co/Cr
1000 250 10 0.1931 ± 0.0008 0.2023 ± 0.0006 0.1977 ± 0.0046
0.22
steel rod
0.21 WC rod
coefficient of friction (-)
0.2
0.19
0.18
0.17
0.16
0.15
101 102 103
relative velocity (mm/min)
Figure 4.15. The coefficient of friction and its standard deviation over rod speed.
4.1.4 Wear coefficient
A purpose designed test rig, supplied by Phoenix Tribology LTD, Newbury, England,
allows to determine the coefficient of wear. The test rig is of pin-on-flat type for linearly
reciprocating motion. The test rig has four test stations on two sides. On each side the
specimens are opposing each other, being separated by a heatable, moving block. On this
block the exchangeable counter face is fixed. The cartridge heater in the block allows to
set a minimum temperature for the counter face. The block temperature is controlled
and tracked using an infrared sensor. If too much frictional heat is generated during a
test, the block temperature becomes higher than the selected set value. Figure 4.16 gives
an overview of one of the two equal sides of the test rig. Figure 4.17 shows more details.
In this work the counter face is a plate of 1.4021 steel, 4 mm thick and coated with WC.
The same coating as typically used as rod coating (WC/Co/Cr, 86/10/4 [31]).
The stroke during a test is fixed. For the presented results it is set to 50 mm. The
frequency of the reciprocating motion can be set as high as 20 Hz. The stations are
placed in two chambers. This allows tests under controlled gas and moisture atmosphere
and easier temperature control. The stationary specimens are pressed against the counter
face with a defined load up to 200 N. The range of contact pressures is increased by
using a smaller and a bigger specimen type. The bigger (18 mm) specimen directly fits
into the hemispherical specimen holder. This specimen geometry is optimised to fit into
61
the specimen holder without getting stuck during a test run, Figure 4.18a. The smaller
specimen type has a steel adapter into which a cylindrical specimen of 9 mm fits into,
Figure 4.18b. The distance between specimen holder and reciprocating counter face is
permanently tracked with a capacitive sensor to determine wear.
arm with test chamber

specimen holder
heatable block with
frame for counter face fixture
load distribution
displacement sensor
gas inlet specimen holder
load cell contact force

pneumatic actuation
lid
temperature
sensor
Figure 4.16. Test rig front view.
62
reciprocating,
heatable block
counter face fixture
specimen fixture
displacement sensor
counter face
(with runmark)
Figure 4.17. Test chamber showing the counter face fixed on the reciprocating, heatable
block, the displacement sensor and the specimen fixture.
a) b)
steel adapter with

specimen 18 mm specimen 9 mm
Figure 4.18. Specimen types: a) shows the bigger 18 mm specimen, b) shows the smaller
9 mm specimen in a steel adapter.
With a single test run four specimens can be tested simultaneously. Figure 4.19 shows four
counter faces with run marks and the used specimen after a test. A test run of a material
with high wear resistance takes several days to produce trustworthy results. During this
time the two specimen holders of one side frequently shift their position symmetrically.
Figure 4.20 shows the displacement data of all four stations. These shifts are compensated
in the post processing by using the mean value of the displacement sensors of two opposing
specimens of one side of the test rig, see Figure 4.21. From the mean values of the sides
the wear coefficient is calculated with equation (4.4). The specimen’s contacting area is
As , pc the contact pressure, vM the mean velocity of the reciprocating motion. ∆t is the
time period over which the specimen height has worn by ∆hs . The wear coefficient for
63
both sides and as mean value over all stations is given in Table 4.3. Table 4.2 lists the
test parameters.
As ∆hs
k= (4.4)
pc vM ∆t
Table 4.2. Test parameters.
parameter value
block temperature 80 ◦C
contact pressure 1 MPa
mean relative velocity 1 m s−1
gas nitrogen
humidity bond-dry
specimen specimen holder

transfer layer
counter surface
with cross cut
1 2 3 4
Figure 4.19. Post test: The specimens (black) are still in the steel adapter. On the counter
surfaces a transfer layer is visible. The cross cut on the counter face is still
visible around the transfer layer. The numbers below the picture are the
labels of the stations.
64
Figure 4.20. The displacement (wear) over time of 4 specimens occurring during the same
run. Station 1 and 2 are on one side of the test rig, station 3 and 4 on the
other side. The specimen holders on one side tend to symmetrically shift
their position due to vibrations.
Figure 4.21. The mean displacement (wear) over time is given for each test side on the
tribo rig.
65
Table 4.3. Resulting coefficient of wear.
stations wear coefficient

(mm N−1 m−1 )
1+2 7.48 · 10−7
3+4 6.90 · 10−7
all 7.2 · 10−7
4.2 Mechanical modelling
This section motivates and explains the choice of the material model and discusses its
features. Subsequently, the material characteristics are briefly described.
4.2.1 Basic considerations
The list below poses several questions, which help to choose an appropriate material
model. Close consideration of the problem’s boundary conditions, experiments and (field)
experience help to find answers.
• What material characteristics have to be described by this model?
◦ As long as a sealing element is operating constantly, i.e. a compressor running

without stops, a material model is sufficient that correctly describes the loading
process over time. It is then not necessary to differentiate between (visco-
)elastic or plastic material behaviour, only the time dependency has to be
modelled correctly.
◦ If the material model has to describe starts and stops too, a true visco-elasto-
plastic material model has to be used.
• In which ranges will the temperature, time, stresses and strains be?
◦ The temperature boundary conditions (rod, packing and gas temperature) are
known from field experience and laboratory testing (e.g. [43], section 6.3).
− The rod temperature ranges between -150 and 180 ◦C. Higher tempera-
tures - up to 350 ◦C - occur sometimes, often resulting in a failure of the
packing rings and rod damage.
− The packing temperature depends on the gas temperature and the surroun-
ding (outside or room) temperature and if a packing cooling is installed.
Another complication is that a packing can have a high temperature gra-
dient from the inner to the outer diameter. The relevant temperature for
66
the sealing element in the packing is the inner wall temperature, which is
assumed to be between 20 and 180 ◦C.
− The gas temperature is assumed to be between suction temperature and
discharge temperature, thus ranging between the extreme values −180 ◦C
and 250 ◦C.
◦ The lifetime of piston sealing rings is currently about one year for non-lube
service and up to three years in lubricated service [57]. Understandably packing
rings have to have equal or higher lifetimes.
◦ The assumption of a constant loading situation and the known geometry allows
to efficiently determine the stress and strain limits for the material model.
The problem is strain controlled, since the sealing elements are wearing, thus
deforming, until their end of life. The local strains and (bending) stresses in
the ring segments can be estimated by a beam model.
• What other parameters influence the material behaviour?
◦ The gas type and the gas moisture can have significant influence on the wear
behaviour.
◦ A material can be non-isotropic, e.g. stress-strain asymmetry between the
tension and the compression regime.
• Which of the afore mentioned parameters have the biggest impact in the current
problem?
◦ Temperature concerning the stress-strain behaviour.

◦ Moisture concerning the non-lubricated wear behaviour.
• Which material model is the fittest for the purpose?
◦ This question is difficult to answer, since several aspects have to be considered

and weighted:
− the necessary complexity and accuracy of the material model,
− the effort to obtain sufficient material data to calibrate the model
− the numerical stability.
Choice: A linear elastic material model is far from the optimum fit, but it is numerically
stable and a good choice for first estimations, analytical considerations and verification
purposes. Furthermore, when the ring is in a thermally quasi-steady state and the ring
wear is high, a linear elastic material model gives a reasonable response.
4.2.2 Analytical beam model
The packing ring’s deformation is now reduced to a bending beam problem. A straight
beam with a constant load over the beam length is assumed. Figure 4.22 sketches the
67
model in the undeformed and deformed state, further the position of the neutral fibre and
the boundary fibres. The typical conventions of the beam theory are applicable:
• The beam’s cross section dimensions are small compared with the beam length.
• The cross section stays plane and is not deformed by shear.
• The beam is a Hookean body, i.e. linear elastic and isotropic.
Additionally, it is assumed that:
• The cross section is constant over the beam length.

• The load is constant over time.
Note that for the sake of simplicity the BCD’s geometry is approximated with a straight
beam. Using the theory of curved beams yields only slightly different results while adding
much more complexity. The desired estimation of (strain) magnitudes can be derived
from the theory of straight beams in a straightforward way.
undeformed z
p
h
x y
L
b
boundary layer
zmax
neutral axis
boundary layer
deformed
uB
tension
compression
Figure 4.22. Beam model in the undeformed (top) and deformed (bottom) state. The
beam dimensions L, b and h are derived from the BCD’s geometry.
68
The beam dimensions are derived from the BCD ring geometry. Figure 4.23 presents the
relevant dimensions of a BCD ring to develop the beam model: the outer diameter D(o) ,
the inner diameter D(i) , the radial ring height h and the two characteristic angles α and
β.
Using a quarter model takes advantage of the BCD’s symmetry and explains the assump-
tion of a fixation on one side of the beam. The complex BCD geometry is reduced to a
beam with rectangular, constant cross section. The medium ring diameter and α and β
define the beam length, see equation (4.6). The beams cross section is equal to the full
cross section of a BCD ring: The axial ring width b is constant and 8 mm, the radial ring
height h depends on the inner ring diameter D(i) (rod size).
D(o) − D(i)
h= (4.5)
2
D (i) h β
( + )(α + )π
L= 2 2 2 (4.6)
180
Figure 4.24 presents the influence of the rod size (equal to D(i) of a virgin ring) on the
BCD’s other geometric features.
wear gap
β
2
L β
D (o)
D (i)
α
h
2
h
Figure 4.23. Geometric values needed to determine the size of the beam. L is calculated
from the mean diameter and the angles α and β.
69
160
D(o) outer diameter
140 b axial ring width
h radial ring height
geometric data (mm) or (◦ )
120 wear gap

α
β
100
α + β/2
L
80
60
40
20
0
0 20 40 60 80 100 120 140
inner diameter D(i) (mm)
Figure 4.24. Dependence of the geometric BCD ring features on the rod size (equals D(i) ):
the outer diameter D(o) , the radial ring height h and the wear gap in (mm)
and the angles α, β and their combination α + β/2 in (◦ ). A representative
beam length L is calculated with equation (4.6).
The following is basic mechanics. The equations of the bending momentum Mb (x) (4.7),
the deflection curve d(x) (4.8), and the (maximum) deflection at the beam’s free end uB
(4.9) are dependent on the pressure load p and the beam geometry. The wear and uB
are assumed to be of the same magnitude. The second momentum of area J is given for
completeness in equation (4.10).
(L − x)2
Mb (x) = −p (4.7)
2 2 2
Lx3 x4

p L x
d(x) = − + (4.8)
2EJ 2 3 12
2 2 3 x4

p L x Lx
uB = d(x = L) = − + (4.9)
2EJ 2 3 12
bh3
J= (4.10)
12
Since the modelled problem is strain controlled, the deflection at the beam end uB is the
only input parameter to evaluate the outer fibre elongation εb . Equation (4.11) allows to
evaluate the needed pressure p to get a certain deflection uB . The stress in the outer fibres
σb can be calculated either from the bending momentum Mb and geometry parameter or
from Hooke’s law, when assuming elastic behaviour, equation (4.12).
70
Inserting equation (4.7) into (4.12) and substituting the pressure p as given in (4.11)
results in the purely kinematic relationship between the outer fibre elongation εb and the
deflection at the beam end uB . The outer fibre strain is independent of the material. The
beam height h has a linear relationship with the fibre strain via the distance between the
neutral fibre and the outer fibre zmax = h/2, see equation (4.13).
8uB EJ
p= (4.11)
L4
Mb
σb = zmax = Eεb (4.12)
J
4uB (L − x)2
εb = ± zmax (4.13)
L4
The assumption in beam theory concerning the beam dimensions is violated, when consi-
dering small rod diameters. The full beam length for one sealing segment is 2L. The ratio
2L/h gives an estimation how badly the assumption is violated. The full beam length
(2L) is less than 5 times greater than the beam height h for a rod diameter smaller than
45 mm. In the worst case the ratio is 3.6 for a 30 mm rod.
Nonetheless, the evaluation of the maximum outer fibre strain at x = 0, equation (4.13),
is performed for rod diameters between 30 and 127 mm. The ring dimensions are taken
from existing rings. Between 30 and 60 mm rod diameter the evaluation is done in steps
of 5 mm, beyond 60 mm in steps of 10 mm or bigger, see Figure 4.25. This allows to draw
several conclusions:
• The strain range the material model has to perform accurately is up to 10 %.
• For bigger rings the strains are smaller by one magnitude, i.e. 1-2 % strain.
• Calculating wear of a few tenth of a millimetre will result in strains below 1 %. This
means that the material model has to perform well in the range of small strains.
• A viable material model has to be well calibrated between 0.1 and 5 % strain, but
has to be able to handle higher strain values as well (up to 10 or 12 %). Higher
strain values are important when considering i) locally calculated strain maxima
and ii) smaller rings.
• If wear is expected in the magnitude of millimetres sealing performance can only be

ensured if the structure is compliant enough to adapt the ring’s shape to the rod.
Overly stiff materials might stop to seal or break.
• The BCD design is not scaling with the rod diameter. The axial width is constant
and the radial height increases only slightly with the rod diameter. It is expected
that the wear pattern of smaller and bigger rings looks differently because bigger
rings have less bending stiffness.
• Equation (4.13) contains no material parameter because the problem is strain con-
trolled. The conclusions drawn give already rough criteria what a sealing element
material has to satisfy:
71
◦ The above stated strain levels have to be reached without rupture.
◦ The necessary pressure load to achieve the stated strains has to be compatible
with the operating conditions. If the pressure in operation is by far bigger
than necessary for the deformation the ring is pressed firmly to the rod and
might wear away too quickly. If the pressure is smaller than needed for the
deformation the ring might become leaky, if the contact pressure between rod
and ring becomes too low.
◦ If stiff materials are to be used the ring design alone has to compensate wear.
72
0.20
0.18
30 mm
0.16 35 mm
40 mm
0.14 45 mm
50 mm
0.12 55 mm
εb (-)
0.10 60 mm
70 mm
0.08 80 mm
90 mm
0.06 100 mm
110 mm
0.04 127 mm
0.02
0.00
0 0.5 1 1.5 2 2.5 3 3.5 4
uB (mm)
0.06
30 mm
0.05 35 mm
40 mm
45 mm
0.04
50 mm
55 mm
εb (-)
0.03 60 mm
70 mm
80 mm
0.02 90 mm
100 mm
110 mm
0.01 127 mm
0.00
0 0.5 1 1.5 2 2.5 3 3.5 4
uB (mm)
Figure 4.25. The outer fibre elongation over the deflection of the beam end for different
inner (rod) diameters. The deflection represents the ring wear. Top: For
ring diameters less than 40 mm the bending strains are bigger than 10 %, if a
deflection (wear) of 3 mm is applied. Bottom: Zooms of the results between
0 and 6 % strain - the bigger the rod, the smaller the strains.
73
4.3 HY54
4.3.1 Material composition
The material used for the presented studies is a PTFE based compound with glass fibres
and graphite. The trade name of this HOERBIGER grade is “HY54”. The material data
sheet is included in the appendix 7. The grade is used in non-lubricated applications,
except if the gas is bone-dry. If it is used in a non-lubricated bone-dry application the
wear coefficient is significantly higher, which allows accelerated wear testing.
4.3.2 Material characterisation
The PTFE based compound is a rather soft material. The elastic modulus depends on
the temperature, the load direction (compression or tension) and the strain rate. At room
temperature (23 ◦C), the influence of the strain rate in tension is already significant. For
the strain rates of 0.05, 0.5 and 5 % s−1 the elastic moduli are 1425, 1651 and 1915 MPa,
respectively.
Figure 4.26 shows the decay of the elastic modulus with increasing temperature for three
strain rates. The average strain rate levels are 0.06, 0.6 and 6 % s−1 . The experiments were
carefully carried out by Dr. Thomas Dick, tortutec GmbH, Vienna. Over all performed
tests the mean Poisson’s ratio is ν = 0.4.
Short-term 5 h creep tests show the material’s reaction to different temperature levels for
compression and tension. Figure 4.27 shows the load mode and temperature dependency.
From this data a non-linear Marlow model can be derived and implemented in the simula-
tion model. This model includes the experimental time scale, which has to be considered,
when interpreting the results. The implementation of this material model is still pending,
but the data allows to get a first impression how the material behaves.
74
2 000
1 800 0.06 % s−1
0.6 % s−1
elastic modulus (MPa) 1 600 6 % s−1
1 400
1 200
1 000
800
600
400
200
0
0 20 40 60 80 100 120 140 160
temperature (◦C)
Figure 4.26. Elastic modulus dependent on strain rate and temperature. The lines con-
nect the mean values of multiple data points for each temperature and strain
rate.
lin. el. 1400 MPa

12
23 ◦C lin. el. 700 MPa
lin. el. 350 MPa
engineering stress (MPa)
10 tension 23 ◦C
55 ◦C tension 55 ◦C
8 tension 80 ◦C
6
180 ◦C comp. 23 ◦C
4 comp. 55 ◦C
comp. 80 ◦C
2 comp. 110 ◦C
comp. 140 ◦C
0 comp. 180 ◦C
0 0.04 0.08 0.12 0.16 0.2
engineering strain (−)
Figure 4.27. Non-linear elastic, temperature dependent material data derived from 5 h
tension or compression creep tests. Elastic moduli of E = 350, 700 and
1400 MPa are added to the graph.
75
Chapter 5
Modelling
This chapter describes the choice of the wear algorithm, its implementation and all ne-
cessary additional calculations with or in connection with the commercial FE software
package Abaqus, which is available at HW.
Then an analytical wear model of an uncut ring is derived to validate the numerical finite
element implementation of the wear procedure.
5.1 Wear algorithm
Literature describes different methods how to implement wear: element deletion, re-
meshing-techniques or nodal displacement. Each of them has advantages and disadvan-
tages.
Considering the following aspects allows to choose an appropriate implementation appro-
ach:
• The goal is to model wear of several millimetres, which may be more than half of
the radial ring thickness.
• Packing rings contact with a rod of circular cross section. (This allows a highly
regular mesh in the ring partitions near the rod.)
• The ring’s sealing performance is only ensured if there is no leakage path from
cylinder to crank side.
• The pressure drop from cylinder to crank side has to be considered. Therefore the
contact surface has to be piecewise smooth, but may consist of several incoherent
patches, which expand or contract over time due to wear.
• The computation time must be observed.
76
Element deletion was discarded, because by definition this method deletes elements,
which creates non-physical leakage paths or holes in the contact surface. This type of
wear modelling technique seems to be more applicable to erosive wear like in rock cut-
ting/breaking [13, 14, 45, 85].
The procedure of re-meshing the geometry after every nth wear cycle can be used in two
ways. First, the wear is calculated in post-processing and then the whole geometry is
re-meshed. Second, the re-meshing is used only for the interior of the wearing solid body
to obtain a better mesh (similar to mesh-smoothing, explained later in this section). This
method includes mapping all results from the configuration prior to the re-meshing to the
post-meshing configuration. Re-meshing is computationally expensive and by default only
applicable for a limited number of element types in Abaqus, namely planar continuum
triangles and quadrilaterals, shell triangles and shell quadrilaterals and 3D tetrahedrals
from the Abaqus library. 4-node tetrahedral elements are overly stiff and not applicable
in problems where bending and/or contact stresses are of importance. So this method is
discarded as well.
The method chosen here is nodal displacement. The Fortran-coded user-defined Abaqus
subroutine UMESHMOTION allows to move nodes without causing stresses or strains. In
this subroutine the displacement of a node within an increment is user-defined, keeping
the flux densities constant. The constant strain energy density results in the fact that
an element can deform without stresses being caused. If a surface node moves “inside”
the geometry, the body loses material, i.e. it wears. (The opposite works as well: If a
surface node moves “outside” direction, material is generated.) Figure 5.1 shows how a
single element wears proportional to the nodal contact forces, assuming constant velocity
for the contacting nodes.
contact force worn volume

nodal wear (nodal displacement
without causing stress)
a) b) c)
Figure 5.1. Working principle within a single wear increment: a) loaded and deformed
configuration, b) displacement of nodes due to wear (proportional to contact
force), c) bottom: worn configuration, top: worn material volume.
77
Wear routines following this principle were already presented in different applications
using different FE codes: The software package Ansys was used by [2, 51]. The UMESH-
MOTION in Abaqus was used by [8, 48, 61, 67]. Hegadekatte et. al [28–30] used Abaqus
for their wear calculations. They implemented a “global-incremental wear model“ as well
as a “wear processor” (calculates wear in post-processing) and compared both methods
to the Abaqus UMESHMOTION.
A natural wear limitation of this method is the element size. After several wear cycles and
sufficient wear an element might be inverted by the algorithm. This causes the calculation
to abort and has to be prevented.
Abaqus allows to use a built-in mesh-smoothing algorithm to prevent element inversion.
This smoothing algorithm moves the nodes inside a meshed solid body and tries to keep
the elements next to each other at similar size, see further [18]. The number of runs for
the mesh-smoothing can be set manually.
b) a) c) d)
pre-wear
developed configuration
smoothing loaded and Abaqus’ first
algorithm deformed no smoothing smoothing loop
25%
wear box
50%
smoothing
75%
100% 100%
contact node contact pressure

node influenced by smoothing original configuration
node in the wear box
Figure 5.2. Smoothing algorithms: a) Starting from a loaded and deformed configuration,
b) the developed smoothing algorithm moves all the nodes inside a wear box
to prevent element distortion. c) Wear of a single element without smoothing.
d) Effect of the first run of the Abaqus smoothing algorithm to reduce element
distortion.
Abaqus’ smoothing algorithm is too inefficient for this wear problem: With a low number
78
of smoothing runs the contacting elements wear away and start to invert, whereas the
elements in the inner part of the mesh are not yet effected by the smoothing, see Figure 5.2.
This is caused by the natural delay, which occurs when trying to keep the elements next
to each other at similar size. With a higher number of loops of the smoothing algorithm
turned out to be too time consuming.
The key to model wear of several element lengths (i.e. several mm) is to introduce a wear-
box. The idea of a wear-box has already been exploited by Rezaei et al. [61] and Fouvry,
Paulin et al. [25, 54]. Earlier Sui et al. [71] described this method, calling it “rezoning”.
Figure 5.2 shows a comparison between the smoothing algorithms.
5.2 Implementation of wear
This section explains how the wear simulation is set-up, how the chosen wear algorithm
is implemented, and how the wear-box is defined. Figure 5.3 presents the quarter model
of the BCD ring with the coordinate system, which is used in the following. Figure 5.4
shows a flowchart containing the simulation procedure.
axial sealing surface
ϕend
r ϕ
sealing segment cap segement

z
Figure 5.3. BCD quarter model: sealing segment (red) and cap segment (blue) viewed
from the crank side. The used cylinder coordinate system originates in the
rod axis and the ring’s axial sealing surface. ϕend gives the position of the
sealing segment’s tip, which changes due to wear.
Many packing ring designs allow to build a radially expanding wear-box, which fully
surrounds the rod. The size of the wear-box is defined as to achieve a maximum wear of
3 mm without re-meshing. The chosen approach requires:
• a regular mesh,
79
◦ where the nodes are aligned in (nearly) radial direction.
◦ where the specified number of neighbour nodes is constant for all nodes in the
dynamic sealing surface.
• the knowledge of the node labels in the dynamic sealing surface.
• the knowledge of the node labels in the dynamic sealing surface are located on an
edge or corner.
• the knowledge of the node labels of the neighbours of each node in the dynamic
sealing surface.
A python script finds the neighbour nodes in a post-processing step after the ring has
initially been loaded, see appendix 7. The script iterates over all nodes in the dynamic
sealing surface. For each contact node and for each following neighbour node the script
determines the elements, which are connected to the currently considered node. The
correct neighbour node has to be geometrically determined from all nodes of the connected
elements. The tolerances depend on the smallest distance between the current node and
all surrounding neighbour nodes. The correct neighbour node is found if the following
conditions are met:
1. The next neighbour node and the current node have the same axial coordinate (i.e.
z-coordinate) (tolerance = 0.1 · smallest distance).
2. The next neighbour node has have a greaterr-coordinate than the current node
(tolerance = 0.3 · smallest distance).
3. The next neighbour node is the potential neighbour closest to the current node.
An additionally implemented sanity check assures that exactly one neighbour is found.
If a neighbour node cannot be found or more than one neighbour is found, a warning is
issued and the neighbour search has to be solved manually. Figure 5.5 presents a simplified
flowchart of the script’s logic. The algorithm works very well for regular meshes.
This geometric procedure is computationally expensive. An advanced guess is implemen-

ted additionally to reduce the search time for the next neighbour node. The geometry
of a complex ring design like the BCD’s is partitioned to be accessible to meshing. The
regular mesh inside the wear-box is exploited by the algorithm because the node num-
bering within a partition is regular. Thus if the difference in node numbers between the
neighbour nodes is known, the next neighbour can be guessed.
After finding all neighbours of the first few (usually 3) contacting nodes the two most
frequently occurring node number increments are used to provide a suggestion of the next
neighbour node number. The most frequently occurring node number increments are up-
dated through the whole procedure.
If the first guess is wrong, a second guess is made. If it turns out wrong again, the geome-
tric algorithm is used again. For each guess geometric checks ensure that the suggested
80
node is the correct neighbour node, according to given tolerances. The tolerances depend
on the median of the distance between two neighbour nodes, the axial distance between
the nodes and the fact that the neighbour has to be radially aligned. The dependence
on the median of the distance gives the advantage that the algorithm can be used with
different mesh sizes without further changes.
Finally, after all neighbour nodes have been determined, a sanity check provides the in-
formation if any node number occurs more than once. If so, a warning message and the
surface nodes with duplicates are printed to a log file. The reason for a duplicate has
to be investigated manually. (Wrong tolerances set? Mesh quality not high enough? =
Nodes not aligned radially? Highly differing element sizes inside the wear-box?)
As a result of this algorithm all surface nodes and their neighbour nodes are written to
a comma-separated values (csv) file for further processing. (The first column contains
the contact nodes and in every subsequent row the contact node’s neighbours are listed
from the innermost to the outermost. The use of a text file is beneficial in two ways:
it is human-readable and it can easily be read by any programming language (Fortran,
Python, Bash).
With the ability to model wear over several element lengths the user-defined Fortran-coded
subroutines become more complex. Implementing a MODULE (introduced in Fortran 95)
instead of a COMMON BLOCK (used in Fortran 77) allows the independent subroutines
to share data. The used subroutines and their main tasks are:
• UEXTERNALDB: This subroutine interacts with external files and is called at the
beginning and at the end of every analysis as well as at the beginning and at the
end of every increment. It is used to:
◦ initialize the calculation at the beginning of the analysis:

− allocate arrays,
− read node sets,
− read the file containing the neighbour nodes of each contact node,
− read the pressure distribution in the dynamic sealing surface,
− read the displacement (=wear) value of the nodes within the wear-box if
the analysis is restarted after updating the pressure distribution in the
dynamic sealing surface. This allows to directly track the total wear of the
sealing element and not only the wear within the calculation between two
updates of the dynamic sealing surface.
◦ update the accumulated wear variable, to allow displaying the wear after every
increment.
◦ write the accumulated wear variable to a text file at the end of every increment.
◦ write the accumulated wear variable to a text file which can be used for the
restart of the analysis at the end of an analysis.
• URDFIL: This subroutine is used to gain access to the results file (every increment)
and read the current coordinates of the nodes in contact with the rod.
81
• UMESHMOTION is the core subroutine, which allows to move nodes (every incre-
ment) independently from their underlying material. The movement does not create
stresses or strains and gives the possibility to model wear. The subroutine is called
for every node in ascending order. The contact nodes should have lower numbers
than their neighbour nodes in the bulk, to guarantee that their position and wear
status is already known and can be used as input for the calculation of the required
displacement of the nodes in the wear-box.
Internal routines provide access to necessary data to calculate wear. Then the no-
dal displacement is calculated dependent on a node’s position: i.e. whether if it
is a contact node, a neighbour node or not even within the wear-box. For nodes
in the dynamic sealing surface the displacement results in material loss (wear) for
the model, whereas the displacement of the bulk nodes prevents element distor-
tion. Figure 5.6 shows the flowchart of the subroutine UMESHMOTION and can
be summarized as follows:
◦ GETNODETOELEMCONN returns all elements connected to the considered

node and provides this information for the next internal routine.
◦ GETVRMAVGATNODE returns the value of a chosen output. Here the con-
tact stress ’CSTRESS’ is used to obtain the nodal contact force for the consi-
dered node.
◦ If the node is in contact with the rod, calculate the nodal wear dependent on
Archard’s equation (3.1). Dependent on the absolute radial wear of the contact
node and the number of neighbours in the wear-box the displacement for each
neighbour node is calculated and stored.
◦ If the node is a neighbour node the stored displacement value is applied.
◦ If the node is neither in contact nor a neighbour node the algorithm skips this
node.
• DLOAD: This subroutine is used to specify the non-uniform gas-pressure distri-

bution in the dynamic sealing surface, which has been determined by a separate
preceding FE calculation. This is of paramount importance to obtain the correct
solid-solid contact pressure for the wear calculation, as explained in chapter 3. The
pressure values, which were determined at the nodal position, are mapped to the
integration points of each element face in the dynamic sealing surface. This map-
ping is done in the isoparametric space where the position of each integration point
is known in advance. A transformation from the isoparametric space to the model
space is not necessary, since the position of the nodes or integration points does not
change within this routine.
• UFIELD can be used to specify a pre-defined field variable or to display the nodal
movement/wear in “Abaqus Viewer”. Note that the subroutine displays the values
from one increment prior to the current one.
82
create geometry model
specify
parameters
and names prepare the wear-box †
find the neigh-

load the structure
bour nodes
create an input file

neighbour
for the pressure
node list
distribution calculation
run the pressure extract pressure

distribution calculation values for each node
create an input file for pressure

the wear calculation value list
run the wear

calculation ‡
total wear ABAQUS

no time reached calculation
PYTHON
yes scripts
post-processing
† detailed in Figure 5.5

‡ detailed in Figure 5.6
Figure 5.4. Flow chart of a wear calculation including the calculation of the pressure
distribution in the dynamic sealing surface and setting up a wear-box [35].
83
read loaded
start
parameters model
yes all contact set considered node read nodes in

nodes = next contact dynamic sealing
considered node, i = i + 1 surface, i = 0
no
no
i < threshold† suggestion 1
yes
pass
get elements no
geometric
connected to checks
current node
suggestion 2
get all nodes
(potential neigh-
bours) from yes
try next connected elements pass
potential geometric
no
neighbour checks
pass
no yes
geometric
checks
yes
neighbour
node found
write found neigh-

bour node to list
neighbour node yes
counter‡ < max
write contact
and neighbour stop no
node list to file
† threshold is the number of runs required before suggesting

‡ each contact node has the same number of neighbour nodes in the whole wear-box
Figure 5.5. Flowchart of the algorithm to find all neighbour nodes in the wear-box. The
counter i iterates over all nodes in the dynamic sealing surface.
84
Abaqus calls get node get
UMESHMOTION to element contact
for a node connectivity† pressure†
contact pressure is the node

no no
> 0 AND node is in a neighbour
contact with the rod node
yes
do not move node,
yes set WGLOBAL = 0
calculate radial
wear (absolute) radial displace- no
ment from the
global array > 0
wear < tole- no
rance value
reduce wear yes
to maximum
yes
acceptable value‡
calculate the
wear vector
is the node above
no
a edge or corner of
the contact patch
is the node inside no
the contact patch
recalculate the radial direction
wear vector toward yes = displacement
yes
next neighbour direction
set WGLOBAL
equal wear vector calculate displa-
cement vector to
next neighbour
check if node
no
is in contact
with the rod • write displacement of
node to global array
reset WGLOBAL = 0 • calculate displacement
yes in local coordinate system
• write displacement of node to global array
• calculate displacement for neighbour nodes and
return to Abaqus
write to global array
• calculate displacement in local coordinate system
† provided by internal routines within UMESHMOTION
‡ so that the node does not leave the contact surface
Figure 5.6. Flow chart for the calculation of the nodal displacement (wear) with the user-
defined subroutine UMESHMOTION. The subroutine is called for each node
in the model in ascending order.
85
5.3 Numerical implementation of the gas pressure
distribution in the dynamic sealing surface
The gas pressure drop across the dynamic sealing surface has significant impact on the
ring wear. The gas pressure drop supports the structure against the outer pressure load.
This reduces the solid-solid contact pressure and thus the wear, as described in chapter 3.
The gas pressure distribution is modelled in an Abaqus sub-calculation before the wear
calculation starts. The gas pressure distribution is updated periodically, since the dynamic
sealing surface changes due to wear over time. The sub-calculation is derived from the
loaded main model and consists of the dynamic sealing surface only. The gas pressure
distribution in the axial sealing surface is not modelled, since no wear occurs here.
A Python script reads the nodes from the Abaqus output database file, which are in the
dynamic sealing surface, and rebuilds the dynamic sealing surface with shell elements of
the type DS4. Next the script applies the boundary conditions to node sets specified in
the wear calculation. The boundary conditions are the high and low pressure values for
the cylinder and crank side, respectively, see Figure 5.7.
a)
z
b)
ϕ r
c)
Figure 5.7. Pressure distribution calculation: a) 3D model of the loaded ring geometry
viewed from cylinder side. The red elements are extracted and the nodes
on the inner diameter are used to build the dynamic sealing surface model.
b) Dynamic sealing surface model derived from a). On the highlighted nodes
boundary conditions are applied: red - high value, blue - low value. c) Re-
sulting (incompressible) gas pressure distribution.
86
The formal similarity of the gas pressure drop equation and the heat transfer equation is
exploited, as explained in section 3.2. For the incompressible gas pressure distribution a
steady-state heat transfer calculation is set-up. For the compressible gas pressure distri-
bution the coefficient of thermal conductivity varies linearly with temperature.
The heat transfer analysis is implemented in the Abaqus software package for different
element types (e.g. DS4). Calculating a temperature field with given boundary conditi-
ons (high and low temperature) is done quickly and returns a temperature distribution
over the modelled dynamic sealing surface. UEXTERNALDB reads the temperature
distribution and the DLOAD applies the nodal (temperature) values as nodal loads.
5.4 Analytical model
This section presents an analytical model, which allows to verify the developed wear
algorithm and its numerical implementation. The content of this section and most of the
figures are published in [36].
The following four independent aspects are considered when modelling wear:
• Ring geometry: The garter spring is ignored, since the force it applies is much
smaller than the loading due to the gas pressure. For simplicity, the following
discussion focuses on a simpler geometry: an uncut (i.e., single-piece) packing ring
of rectangular cross section. Axial symmetry allows easier and quicker numerical
modelling. It also makes the problem amenable to an analytical solution.
• Wear: according to Archard’s equation (3.1).
• Gas pressure drop across the dynamic sealing surface, see section 3, is ignored in
the analytical model. We assume perfectly smooth surfaces.
• Material behaviour: Since the primary focus is on verifying the numerical wear
algorithm, we use the simplest possible material model, Hooke’s law, disregarding
any material ageing effects. Stress-strain measurements at 23 ◦C give the values for
the elastic modulus E =1400 MPa and the Poisson’s ratio ν = 0.4.
5.4.1 Model
To make the problem amenable to an analytical solution, we consider a simplified con-

figuration where a constant pressure po acts on the outer diameter D(o) = 2R(o) of a
linear-elastic axisymmetric ring of rectangular cross-section (axial width b, radial height
h), i.e. the problem is axisymmetric in the r-z-plane. This pressure pushes the ring
against a rod sliding at constant speed (Figure 5.8).
Friction is ignored, as are gas pressure effects over the contact surface. Furthermore, the
ring is assumed to be in a state of plane strain in the r-ϕ-plane, so that stresses and
87
strains are functions only of the radial coordinate r. Under these assumptions, ring wear
does not vary along the axial direction z.
po b
a) b)
R(o)
h(t1 )
ŵ(t1 )
R(i) (t1 )
r pc (t1 ) D/2
Figure 5.8. Worn axisymmetric ring at t = t1 . a) Loaded worn ring (pressure po acting
on the outer diameter, contact pressure pc acting on the inner diameter),
assumed to be in state of plane strain. b) Dimensions of the worn ring in the
undeformed state.
In the initial state t = 0, the inner diameter of the undeformed ring D(i) = 2R(i) is equal
to the rod diameter D. The pressure po gives rise to a contact pressure pc which in turn
leads to wear in accordance with equation (3.1). Due to wear, R(i) grows with time and so
does the radial displacement u(i) = u(r = R(i) ) required to satisfy the contact condition
D/2 = R(i) + u(i) . Linear elasticity theory [87] provides the radial displacement field:
 !2 !2 
R (o) R (i)
 1 − 2ν + 1 − 2ν + 
r  r r 
u(r) = pc − po (5.1)

!2 !2 
2G 
R(o) R(i)

−1 1−
 
R(i) R(o)
Evaluated at r = R(i) , equation (5.1) gives the elastic radial displacement u(i) of the ring’s
inner diameter when loaded by po and pc on its outer and inner diameter, respectively. G
denotes the shear modulus and ν is the Poisson’s ratio.
At the first instant of time, the rod diameter is equal to the undeformed inner diameter
of the ring, so that u(i) (t = 0) = 0. Together with equation (5.1), this gives the initial
contact pressure pc (t = 0) and thus the initial wear rate dwdt (t = 0). However, the rate
(i) dw
at which R grows is not given by dt , since equation (3.1) gives the rate at which
88
deformed material is worn away. The following consideration shows how to derive the
dR(i) dw
proper relationship between dt and dt .
During a time increment dt, an infinitesimal layer of material of thickness dw is worn

away according to equation (3.1), this infinitesimal cylindrical shell being under stress.
The thickness dŵ of this infinitesimal shell in its undeformed state is related to dw via
the radial strain εrr according to:
dw
ŵ = (5.2)
1 + εrr (r = R(i) )
where the radial strain of the infinitesimal shell follows from equation (5.1) according to:
∂u
εrr (r = R(i) ) = (5.3)
∂r r=R(i)
Noting that the incremental change dR(i) of the inner radius is equal to dŵ gives:
dR(i) kνpc
= (5.4)
dt ∂u
1+ (i)
∂r r=R
which, with when we express pc in terms of R(i) via equation (5.1),
D 2 − (1 − ν)
G (i) − 2 + po 1 2
R
1−
ρ
pc = 2 (5.5)
1 − 2ν + ρ
ρ2 − 1
and the derivation of ∂u ,
∂r
 !2 !2 
R(o) R(i)
 1 − 2ν − 1 − 2ν − 
∂u(r) 1  r r 
= p c − po (5.6)
 
2
∂r 2G  ρ2 − 1

1 
 1− 
ρ
finally leads to the non-linear first-order differential equation:
!
kν(ρ2 − 1)

D 2(1 − ν)
G − 2 + po
dR(i) 1 − 2ν + ρ2 R(i) 1 − ( ρ1 )2
= ! (5.7)
dt
 
D 2(1 − ν) 2
G − 2 + po (1 − 2ν − ρ )
R(i) 1 − ( ρ1 )2
 
1  2ν 
1+  + po 
2G 
 1 − 2ν + ρ2 1 − ( ρ1 )2 

89
Equation (5.7) governs how the inner radius grows with time due to wear, ρ denoting the
R(o)
ratio . It can be solved incrementally by means of an Euler forward scheme.
R(i)
5.4.2 Numerical implementation and analytical results
The numerical implementation of the Euler forward scheme for the analytical model was
done within the commercially available software Matlab. The flowchart Figure 5.9 shows
the calculation steps as well as the parameters, which are chosen for the comparison with
the finite element models. Using the Matlab code gives the accumulated wear in the
deformed state w and the contact pressure pc over time.
The ring is pressed against the rod during the wear process, if the pressure po is set to
zero after the ring is already worn, a gap between the ring and the rod can be noticed.
This gap (R(i) − D/2) can be evaluated as well and turns out to be smaller than the
originally worn distance w. This can be seen in Figure 5.10 and Figure 5.11.
The numerical implementation of the analytical model enables to test quickly the effect of
different geometry parameters on the wear process. Exemplary two cases are presented.
Case 1 is a case where the ring stiffness is high enough that a wear limit is reached.
Case 2 is the case where the ring wears fully away. The parameters for those cases are
in Table 5.1, where the parameters equal for both cases are written only for case 1. The
cases differ only in the radial dimension of the rings (case 1: 8 mm, case 2: 5 mm).
Table 5.1. Exemplary cases for ring wear.
parameter case 1 case 2

R(i) 28.58 mm
R(o) 36.58 mm 33.58 mm
D 28.58 mm
b 8 mm
G 500 MPa
ν 0.4
po 10 MPa
kv 2 × 10−7 mm3 /Ns
In case 1 the wear rate of the ring decreases over time proportional to the decreasing
contact pressure. The wear stops after approximately 3 × 106 s. The accumulated wear in
the deformed state will be referred to as ultimate wear limit w∞ . An ultimate wear limit
can only be reached if the contact pressure drops to zero and there is still some radial
ring height left.
In case 2 no ultimate wear limit is reached, the ring wears away fully. Although the
pressure drops to a minimum value (less than 5 % of the initial value), it never reaches
90
zero. The ring wears further, with a low wear rate and constantly loses stiffness. At a
certain point the loss in stiffness gives rise to an increasing contact pressure. With the
contact pressure the wear rate increases and the ring wears faster until it is worn away.
start
parameter definition:
E, ν, D, R(i) , R(o) ,
po , k, ∆t, Tmax
calculate initial contact

pressure pc , j = 1
calculate:
du
dr r=R(i) (j),
pc (j + 1),
R(i) (j + 1),
t(j + 1) = t(j) + ∆t,
u r=R(i) (j),
w(j + 1),
j =j+1
R(i) (j + 1) ≥ R(o)
no or t(j + 1) > Tmax
yes
plot data
end
Figure 5.9. Numerical implementation of the analytical model to calculate the wear on
the inner surface of a plane strain uncut ring.
91
w∞
10
1
contact pressure (MPa)

0.8 8
wear gap (mm)
wear (mm)
0.6 R(i) − D/2 6

w
pc
0.4 4
0.2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5
time (s) ·106
Figure 5.10. Case 1: Change of the wear gap (R(i) − D/2), the accumulated wear w and
the contact pressure pc over time for a ring with h0 = 8 mm on a rod of
diameter D = 57.16 mm. The ring wear stops before the ring is fully worn
away.
5 10
R(i) − D/2
w
pc
4 8
wear gap (mm)
wear (mm)
3 6
2 4
1 2
0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
time (s) ·107
Figure 5.11. Case 2: Change of the wear gap (R(i) − D/2), the accumulated wear w and
the contact pressure pc over time for a ring with h0 = 5 mm on a rod of
diameter D = 57.16 mm. The ring wear continues until the ring is fully worn
away.
92
Equation (5.7) was solved via numerical integration. For the parameters given in Table 5.1,
Rt
Figure 5.10 shows how the accumulated wear w(t) = 0 dw(τ )/dτ dτ , the wear gap (R(i) −
D/2) in the undeformed state, and the contact pressure pc change with time. For each
value of the pressure loading po there is a certain value hcrit of the initial radial ring height
h0 = h(t = 0) that determines whether the wear process comes to an end before the ring
is fully worn away (Figure 5.10) h0 > hcrit or continues until the ring is fully worn away
(Figure 5.11) h0 < hcrit . Note that Figure 5.11 has a smaller initial ring height h0 h=5 mm,
with R(o) =33.58 mm, the other parameters are equal to those used in Figure 5.10.
The wear process comes to an end when either the contact pressure pc becomes zero or
when the ring has completely worn away. For the configuration under consideration where
R(i) (t = 0) = D/2, there is always a finite initial value pc (t = 0) of the contact pressure
so that the ring initially starts wearing away with R(i) increasing correspondingly.
From equation (5.1) the following relationship between the contact pressure pc and the
inner ring radius R(i) can be derived:
  !2
R(o)
  −1
R(i)

 D
(i) 2G 2po (1 − ν) 
pc =  −R + (5.8)
 
!2  !2
 2 R(i) R (i) 
R (o)
1− 1 − 2ν +
 
R(o) R(i)
This relationship is depicted in Figure 5.12 for the geometrical and material parameters
given in Table 5.1 case 1 for different values of po . Case A shows how the contact pressure
pc varies with R(i) for po = 30 MPa. As can be seen, the contact pressure first decreases
with increasing R(i) until the further loss in structural ring stiffness requires smaller and
smaller pressure differentials |po − pc | to radially displace the inner ring diameter by
the amount R(i) − D/2. Since the contact pressure stays positive over the entire range
D/2 < R(i) < R(o) , the ring fully wears away in case A. Applying a ring loading of
po = 10 MPa, which is the configuration considered for Figure 5.10, leads to case C,
where the ring wear stops at R(i) = 29.6 mm. These two regimes A and C are separated
by the limiting case B (po = 21.2 MPa) where the contact pressure only vanishes at a
single value of R(i) .
Figure 5.13 shows in dimensionless form how the ultimate accumulated wear w∞ =
lim w(t), referred to the initial ring height h0 , varies with dimensionless pressure loading
t→∞
po /G for rings of different aspect ratios h0 /D. The critical ring height hcrit for a certain
pressure loading po /G is given by the highest value of h0 /D that leads to w∞ /h0 = 1.
Values for h0 /D between 0.01 and 1 are evaluated.
93
30
case A
case B
20 case C
pc (MPa)
10
-10
30 32 34 36
D/2 R(o)
R(i) (mm)
Figure 5.12. Variation of contact pressure pc with inner radius over D/2 < R(i) < R(o) for
different pressure loadings po (case A: po = 30 MPa, case B: po = 21.2 MPa
and case C: po = 10 MPa)
100
po /G
10−1 increasing load
0.01
0.005
0.002
10−2 0.001
w∞ /h0
0.0005
0.0002
10−3 0.0001
5e-05
2e-05
10−4 1e-05
10−5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
h0 /D
Figure 5.13. Ultimate wear w∞ referred to initial ring height h0 for rings of different
aspect ratios h0 /D that are loaded by the non-dimensional pressure po /G.
94
5.4.3 Validation
Table 5.1, case 1 lists the parameters for the verification. To verify the user-defined wear
algorithms developed in FORTRAN, the analytical model is now compared with two FE
models created according to the clamping situation given in Figure 5.8. In both FE models
the nodes on the axial surfaces of the rings are constrained such that they are fixed in
axial direction.
a) A 2D axisymmetric FE model. The axisymmetric cross section is modelled with

50x50 CAX4 elements.The wear-box expands radially over 40 elements. Note that
for validation purposes a single row of elements would be sufficient (due to the plane
strain condition in r-ϕ). However, the same model has already been set up for some
more general, non plane-strain cases.
b) A quarter symmetrical 3D FE model of an uncut single piece ring. The cross section
is modelled with 16x11 C3D8 elements. The radial dimension of the outer 5 elements
is twice the radial dimension of the inner 6 elements. The smaller, inner elements
are within the wear-box.
Figure 5.14 shows the mesh of the ring cross section. The wear-box is highlighted red.
Note: The size of the wear-box does not influence the result.
a) b)
Figure 5.14. Cross section of the algorithm verification models: a) the 2D axialsymmetric
model meshed with 50x50 CAX4 elements and b) the cross section of the 3D
model meshed with C3D8 elements. The wear-box is shaded red.
Figure 5.15 shows how the contact pressure and the accumulated wear evolve over time.
The data from the analytical model is drawn as solid line (as presented in Figure 5.10).
The results of the numerical 2D and 3D FE calculation are marked with symbols. The 2D
95
calculation and the analytical model are in excellent agreement. For the 3D calculation
the agreement is good. A deviation from the analytic model can be seen only for the
wear with progressing time. This may be due to the fact that the analytical model is
formulated in geometrically linear theory, whereas the FE implementation takes geometric
non-linearities into account.
1 10
accumulated wear (mm)

0.8 8
w analytical
w numerical 2D
0.6 w numerical 3D 6
pc analytical
pc numerical 2D
0.4 pc numerical 3D 4
0.2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5
time (s) ·106
Figure 5.15. Verification of the wear algorithm of the numerical FE model against the
analytical (numerically integrated) model for the wear in the deformed con-
figuration w and the contact pressure pc for an uncut ring.
5.5 Results
In this section the wear of BCD rings is investigated for different rod sizes and wear
gaps. The main geometric features are given in Figure 2.7, with a detailed description in
section 2.2. The fact that the geometry of the BCD rings are not scaling with the rod
diameter was discussed as well, see section 4.2.2 and Figure 4.24, which already leads to
interesting effects even when assuming linear elastic material behaviour.
5.5.1 Model set-up and parameter variation
The BCD ring is quarter-symmetrical. This symmetry is utilized to reduce the model size.
The BCD geometry was adapted slightly to facilitate meshing and remove non-physical,
96
numerically problematic stress peaks. The following measures were taken:
• The round garter spring groove on the outer diameter has been changed to rectan-
gular shape, see Figure 5.16.
• The chamfer on the sealing segment’s tips has been removed.
• Two edges have been trimmed on the cap segment, to remove non-physical stress
peaks:
◦ The edge between the surface, which contacts with the sealing segment, and
the garter spring groove.
◦ The edge between the surface, which contacts with the sealing element, and
the surface of the outer diameter of the cap segment, see Figure 5.16.
wear stop (analytical rigid)
edge trimmed
cup (analytical rigid)
cap segment
rectangular
garter spring ϕ
groove
r
z
rod (analytical rigid)

sealing segment
Figure 5.16. BCD ring model set-up for a 50 mm rod. Analytical rigid bodies are displayed
gray, the sealing segment red and the cap segment blue. Edges of the cap
segment are trimmed to avoid non-physical high contact pressures, which
can lead to numerical instability. The garter spring groove is modelled as
rectangular groove to allow meshing without tetrahedrons or distortion.
Relative motion of the sealing segment and the cap segment occurs due to wear. During
the simulation the gas pressure is not applied to the newly created free surface, thus
leading to non-physical locally high bending stresses, if the two forementioned edges jut
97
out.1 Trimming the edges reduces this unintended problematic behaviour.
The garter spring is not modelled because its effect on the contact pressure (usually
0.03 MPa) is negligible compared to the contact pressure resulting from the gas pressure.
Figure 5.16 shows the quarter-model of a BCD for a 50 mm rod, the rigid bodies are shaded
grey. Rod and cup are modelled as analytical-rigid parts, the radial clearance between
rod and cup is 1 mm. Additionally another rigid part is used as wear stop for the sealing
element. The wear stop is fixed in space and only contacts with the sealing segment after
the ring has worn sufficiently. It is assumed that the BCD’s quarter-symmetry is not
broken while wearing. Figure 5.7 shows the meshed model.
The BCD geometry is manually partitioned to make it accessible to the Abaqus meshing
algorithm, to avoid mesh distortion at the beginning (and through) the analysis as far as
possible and to allow to use a wear-box. The mesh consists of C3D8 and wedge elements
of the type C3D6. The wedges are used only outside the wear-box. Figure 5.17a shows the
wear-box coloured yellow. The C3D8 elements inside have an edge length of approximately
0.5 mm. The wear-box is designed such that a maximum wear of 3 mm can be calculated.
Figure 5.17b shows the developed view of the dynamic sealing surface of the virgin BCD
the quarter model. The pressure balancing groove and the wear gap are marked.
1
Abaqus offers a pressure penetration algorithm, which was not yet mature and could not be used to
map the pressure physically correctly in the wear algorithm.
98
a) sealing segment cap segment
r
ϕ
wear box
z
z cylinder side
b)
cap segment
pressure balancing groove
sealing segment
ϕ crank side 1
2 wear gap
Figure 5.17. BCD ring for the 50 mm rod, quarter-model: a) The wear-box is coloured
yellow for the sealing (red) and the cap (blue) segment. b) The developed
view of the dynamic sealing surface with the view direction being the negative
radial direction. The pressure balancing groove area is coloured green. On
the lower right the clearance between the sealing segment and the wear stop,
half the wear gap, is marked.
The wear simulation of the BCD rings allows to investigate the effect of the rod diameter,
the pressure balancing groove and the size of the wear gap. The range of rod sizes is
represented by three rod diameters: 28.58 mm, 50 mm and 100 mm. Not all ring dimen-
sions scale with the rod diameter, cf. Figure 4.24. Adding the pressure balancing groove
unloads the ring partially and thus reduces the contact pressure and the wear rate. For
each rod size a model without a pressure balancing groove is investigated. The size of
the wear gap is 6 mm in the current configuration of a standard BCD ring. The wear gap
is varied in three steps of 6 mm, 4 mm and 2 mm for the configuration with the pressure
balancing groove. The calculations of the BCD without the pressure balancing groove
has a wear gap of 6 mm. Table 5.2 lists the 12 considered cases.
99
Table 5.2. Cases considered while screening the wear behaviour of the BCD ring.
case name rod diameter pressure balancing groove wear gap

A1 yes 6 mm
A2 yes 4 mm
28.58 mm
A3 yes 2 mm
A4 no 6 mm
B1 yes 6 mm
B2 yes 4 mm
50 mm
B3 yes 2 mm
B4 no 6 mm
C1 yes 6 mm
C2 yes 4 mm
100 mm
C3 yes 2 mm
C4 no 6 mm
The linear elastic calculations are time independent. The material properties are repre-
sented by E = 1400 MPa and ν = 0.4. The gas pressure load is set to 10 MPa, which
is very high. The gas is assumed incompressible. The rod velocity is 1 m s−1 , the wear
coefficient is 2.8 × 10−7 mm3 m−1 MPa−1 .
Note: In these linear elastic calculations the point in time when a certain wear pattern
occurs solely depends on the chosen wear coefficient. The resulting times can be used to
compare different configurations, but for a lifetime prediction the coefficient of wear of an
analysed tribo-system has to be known.
Here the maximum time increment is 600 s. Every 1.2 × 105 s (equals at least 200 wear
increments) the gas pressure distribution in the dynamic sealing surface is evaluated and
updated.
5.5.2 Evaluation procedure
After a wear calculation finishes post-processing gives the following results, which are
explained in detail below for a BCD ring mounted on a 50 mm rod:
• the wear pattern at a certain point in time,

• the evolution of wear over time,
• the contact force over time and
• the closing of the wear gap over time.
The wear pattern of the whole dynamic sealing surface can be plotted as 3D surface.
Below the 3D wear pattern a black patch represents the dynamic sealing surface. These
100
patches also indicate the remaining wear gap. If a ring reaches the wear stop, the black
patch of the sealing segment expands from 0 to 90◦ . Figure 5.18 represents the end of a
wear calculation procedure, when a maximum wear of more than 3.2 mm is achieved and
the wear-box is locally fully consumed. The wear stop has been reached and the wear gap
is gone.
3
wear (mm)
3 2.5
2
1 2
0 1.5
8
6 1
90
z( 4 75
m 60 0.5
m) 2 45
30 ◦
0 15 ϕ( )
0 0
Figure 5.18. Wear pattern of a (linear elastic) wear calculation of BCD ring on a 50 mm
rod after 23.6 d of continuous operation.
These surfaces can be produced for different points in time and give insight how the whole
ring wears.
The wear history is presented in Figure 5.19a to f, where six wear patterns are depicted at
different times. The time increment between the graphs is 6 × 105 sec (166.67 h). Between
these points in time the pressure drop is updated 5 times. It is evident, that the axial
wear gradient is small compared to the gradients in the circumferential direction. Thus
the wear data of one of the three patches of the surface plot can be represented well by
the data along a node path, z = const..
101
a) t = 6 × 105 s b) t = 1.2 × 106 s
wear (mm)
wear (mm)
3 3
2 2
1 1
0 0
8 8
6 6
90 90
z(4 75 z(4 75
mm 60 mm 60
45 45
) 2 30 ◦ ) 2 30 ◦
0 15 ϕ( ) 0 15 ϕ( )
0 0 3
6 6
c) t = 1.8 × 10 s d) t = 2.4 × 10 s
2.5
wear (mm)
wear (mm)
3 3
2 2 2
1 1
0 0 1.5
8 8
6 6 1
90 90
z(4 75 z(4 75
mm 60 mm 60
45 45
) 2 30 ◦ ) 2 30 ◦ 0.5
0 15 ϕ( ) 0 15 ϕ( )
0 0
0
e) t = 3 × 106 s f) t = 3.6 × 106 s
wear (mm)
wear (mm)
3 3
2 2
1 1
0 0
8 8
6 6
90 90
z(4 75 z(4 75
mm 60 mm 60
45 45
) 2 30 ◦ ) 2 30 ◦
0 15 ϕ( ) 0 15 ϕ( )
0 0
Figure 5.19. Wear pattern of the BCD ring quarter model on a 50 mm rod after 6 points
in time. Where a) to f) marks advancing time from 6 × 105 s (166.67 h) to
3.6 × 106 s (1000 h).
Next the wear evolution for z = 2 mm is considered, starting at the middle of a sealing
segment with ϕ = 0◦ and running to the tip of the sealing segment (where ϕend changes
over time due to the wear). Figure 5.20 gives a graphical explanation: a) The wear pattern
at a certain point in time. The pink plane at z = 2 mm represents the position of the
node path, which is depicted schematically in b). The cut between the pink plane and
the 3D wear surface gives the wear of this node path, which can be tracked over time, see
c).
This representation allows to compare with other configurations. A line is drawn from
6 × 104 s every 1.2 × 105 s, for easier visualisation and comparison every 10th line (∆t =
1.2 × 106 s) is coloured red. Within c) there is a miniature sketch, similar to b), of the
virgin-state dynamic sealing surface, where the position of the node path is drawn as a
red line. The sketch also gives the size of the wear gap and shows the existence of the
102
pressure balancing groove. Additionally the rod diameter is given.
The wear gap closes as the sealing segment wears, this can be seen on the right end of the
abscissa in Figure 5.20c. The distance between the wear contour line and the right side
of the plot (90◦ ) represents the size of the remaining wear gap.
The simulation also provides the contact force of every contacting node. The contact force
for the calculated ring is obtained by integrating over all nodes in the dynamic sealing
surface, without regard of the surface normal. This integrated contact force is displayed
as a result. Note that the contact force for a whole ring is four times this value. The rod
pull-off force or the frictional force of a ring is proportional to that contact force. The
proportionality factor is µ, when assuming Coulomb friction.
Comparing the direction independent, sum of all nodal contact forces has a big advantage:
The result is valid independent of the tribological system.
The closing of the wear gap can be monitored, using the Abaqus variable “COPEN”.
It gives the contact opening distance, the distance between two potentially contacting
partners. The distance between a node on the surface of the sealing segment’s tip and
the rigid stop is evaluated. Here the node, which is in contact with the rod and at ϕend
and z = 2 mm is chosen, see Figure 5.21. Note that the real wear gap is twice as big as
the calculated (and plotted) distance from the node to the wear stop.
Figure 5.22 shows the closing of the wear gap over time, considering the distance of the
tip of one sealing segment to the wear stop. Further it shows the development of the
contact force of the quarter model over time. When the wear gap closes (0 mm gap)
the ring’s structural stiffness increases instantly and significantly. The ring then nearly
behaves like an uncut, solid ring. This is the reason for the discontinuity in the contact
force curve. The contact force decreases faster after the closing event because then the
ring’s structural stiffness takes on more and more of the gas pressure load.
103
a) 3
wear (mm)
3 2.5
2
1 2
0 1.5
8
6 1
90
4 75
z( 60 0.5
mm 2 45
) 30 ◦
0 15 ϕ( )
0 0
z cylinder side
b) sealing segment cap segment
node path for

evaluation
ϕ crank side
c) 3.5
8 Ø 50 mm
4
3 0
0 30 60 90
2.5
wear (mm)
1.5
0.5
0
0 15 30 45 60 75 90
ϕ (◦ )
Figure 5.20. a) Wear pattern of a linear elastic wear calculation of BCD ring quarter
model on a 50 mm rod after 23.6 d. The pink plane represents the position
of a node path in tangential direction 2 mm from the cup face, schematically
shown in b). c) presents the evolution of the wear of the picked node path.
Blue lines are drawn for time increments of ∆t = 1.2 × 106 s. The red lines
indicate every 10th line.
104
a) b)
z
ϕ
wear
considered node stop
Figure 5.21. Position of the node, which is used to measure the wear gap. a) shows the
dynamic sealing surface, the sealing segment’s contact area is coloured red
and the cap’s area blue. In b) the black framed detail of a) is magnified. The
wear stop is indicated by a dotted line. The distance from the considered
node to the wear stop is monitored over time.
1 500
3
2.5
distance to wear stop (mm)

contact force (N)
1 000 2
1.5
500 1
0.5
0 0
0 5 10 15 20 25
time (day)
Figure 5.22. The evolution of the contact force (blue) and the wear gap (green) are shown
over time. When the wear gap closes (0 mm gap) after 17.1 days, the contact
force shows a discontinuity. After gap closing the contact force decreases
faster because from then on the ring has a higher structural stiffness in
circumferential direction which allows it to carry a greater part of the external
pressure load.
105
5.5.3 Calculations with incompressible gas
The wear pattern and then the contact force and the wear gap are analysed in the follo-
wing for the variations given in Table 5.2. The plots comparing these configurations are
arranged equally for each rod diameter as shown in Table 5.3.
Table 5.3. Plot arrangement in the consecutive figures. The symbol ’*’ represents a wil-
dcard character to be substituted by the letters A, B or C. The number in the
case name can be cross-referenced to Table 5.2. Case *4 is placed below case
*1 which allows to better see the effects of the pressure balancing groove.
case name *1 case name *2

configuration standard configuration modified
pressure balancing groove yes pressure balancing groove yes
wear gap 6 mm wear gap 4 mm
case name *4 case name *3
configuration modified configuration modified
pressure balancing groove no pressure balancing groove yes
wear gap 6 mm wear gap 2 mm
Figure 5.23 shows that the wear pattern of the 28.58 mm rod reaches an ultimate wear
limit. The structure is stiff enough to carry the whole gas pressure load, which gives high
lifetime. This limit is reached earlier in time and the total amount of wear is lower if the
wear gap is smaller, i.e. the wear values are arranged in the following order A3<A2<A1.
Interestingly the sealing segments of case A1 and A2 do not stop wearing at ϕ = 0◦ . The
effect of the pressure balancing groove is visible when comparing wear patterns at equal
times. For example the second red line of A4 starts at 1.59 mm at 0◦ and has a maximum
of 1.84 mm at 58◦ . The equivalent line for A1 starts at 1.35 mm at 0◦ and has a maximum
of 1.79 mm at 59◦ . The form of the wear pattern also differs. The ring without pressure
balancing groove has a more uniform pattern than the ring with the pressure balancing
groove.
The ring without pressure balancing groove (A4) shows the highest total wear not at
approximately 60◦ but at 0◦ . This is unique. The rings from cases A1 and A2 show
similar trends, but not that pronounced. Their wear pattern evens out when reaching
their saturation pattern. This indicates a change in structural stiffness that shifts the
loading situation such that the tips of the sealing segment are unloaded.
Figure 5.24 shows the evolution of the contact force and the closing of the wear gap for
the 28.58 mm rod. Any discontinuities in the curves of the contact force are either due
to the update of the gas pressure distribution in the dynamic sealing surface or the load
rearrangement after the wear gap closes. The proof that an ultimate wear level has been
reached for the cases A2 and A3 is found in the contact force. In these cases the contact
force nearly drops to zero. (Note: No information can be given if the rings would still be
sealing.) The closing speed of the wear gap is not constant over time. It slows down with
increasing time. This is more pronounced when the wear gap is bigger. The closing speed
106
with pressure balancing groove is 15 % lower than without pressure balancing groove.
The contact force of the ring with pressure balancing groove is 15 % lower than without
pressure balancing groove at the beginning of the calculation. This accordance is used as
a sanity check: Lower contact pressure results in slower wear.
3.5 3.5
8 Ø 28.58 mm 8 Ø 28.58 mm
4 4
3 0 3 0
0 30 60 90 A1 0 30 60 90 A2
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
◦
ϕ( ) ◦
ϕ( )
3.5 3.5
8 Ø 28.58 mm 8 Ø 28.58 mm
4 4
3 0 3 0
0 30 60 90 A4 0 30 60 90 A3
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
◦
ϕ( ) ◦
ϕ( )
Figure 5.23. Wear pattern evolution. The tangential node path at z = 2 mm. The rod di-
ameter = 28.58 mm. The case names are A1, A2, A3 and A4 given clockwise
starting at the upper left plot.
107
3.5 3 3.5 3
8 8
4 4

3 A1 0 3 A2 0
0 30 60 90 2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)

2.5 Ø 28.58 mm 2.5 Ø 28.58 mm
2 2
2 2
1.5 1.5
1.5 1.5
1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
time (day) time (day)
3.5 3 3.5 3
8 8
4 4

3 A4 0 3 A3 0
0 30 60 90 2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)
2.5 Ø 28.58 mm 2.5 Ø 28.58 mm

2 2
2 2
1.5 1.5
1.5 1.5
1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 5.24. Evolution of the integral contact force (blue) and the closing of the wear gap
(green) for the four investigated configurations A1 to A4, which are placed
clockwise starting at the top left, respectively. Rod diameter 28.58 mm.
Figure 5.25 shows the wear patterns of BCD calculation for the 50 mm rod. Locally the
ring wears more than 3 mm, if the wear gap is 6 mm. This is independent of the pressure
balancing groove. In these cases the wear-box is fully consumed locally. Cases B2 and
B3, with 4 and 2 mm wear gap, respectively, reach an ultimate wear level in the central
part of the sealing segment from ϕ = 0 to approximately 30◦ . The tips of the sealing
segment continue wearing. In the cases B1 and B4 the the wear rate drops towards the
end of the calculation.
The effect of the pressure balancing groove is similar to the BCDs on the smaller rod.
108
Without the groove the ring wears more uniformly. The second red line of B4 starts at
1.52 mm at 0◦ and has a maximum of 2.11 mm at 67◦ . The equivalent line for B1 starts
at 0.99 mm at 0◦ and has a maximum of 2.03 mm at 70◦ . The shape of the wear pattern
also differs.
3.5 3.5
8 Ø 50 mm B1 8 Ø 50 mm B2
4 4
3 0 3 0
0 30 60 90 0 30 60 90
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
ϕ (◦ ) ϕ (◦ )
3.5 3.5
8 Ø 50 mm B4 8 Ø 50 mm B3
4 4
3 0 3 0
0 30 60 90 0 30 60 90
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
◦
ϕ( ) ◦
ϕ( )
Figure 5.25. Wear pattern evolution. The tangential node path is at z = 2 mm. The rod
diameter = 50 mm. The case names are B1 to B4 given clockwise starting
at the top left plot.
Figure 5.26 presents the wear gap size and the contact force over time of the quarter
model of the 50 mm rod. The discontinuity after 1.4 days results from the update of the
109
gas pressure distribution in the dynamic sealing surface. The calculation of A1 stops after
23.6 days, because the wear-box is fully consumed. As mentioned, a quasi steady-state is
reached for B2 and B3, when these rings reach an ultimate wear level. Then the contact
force reaches a nearly constant level after the wear gap has closed and the structure has
worn to some extent.
3.5 3 3.5 3
8 8
4 4

3 B1 0 3 B2 0
0 30 60 90 2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)

2.5 Ø 50 mm 2.5 Ø 50 mm
2 2
2 2
1.5 1.5
1.5 1.5
1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
3.5 3 3.5 3
8 8
4 4

3 B4 0 3 B3 0
0 30 60 90 2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)
2.5 Ø 50 mm 2.5 Ø 50 mm
2 2
2 2
1.5 1.5
1.5 1.5
1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 5.26. Evolution of the integral contact force of the BCD quarter-model and the
rod (blue) and the closing of the wear gap (green) for the four investigated
BCD configurations B1 to B4. Rod diameter 50 mm.
The closing speed of the wear gap is nearly constant over time. The closing speed with
pressure balancing groove is 20 % slower than with pressure balancing groove. The contact
110
force of the ring without the pressure balancing groove is 23 % lower than without pressure
balancing groove at the beginning of the calculation.
The BCD rings on the 100 mm rod wear until the wear-box is consumed locally. Fi-
gure 5.27 shows that only B3, with 2 mm wear gap shows signs of reaching an ultimate
wear level at ϕ = 0◦ .
3.5 3.5
8 Ø 100 mm C1 8 Ø 100 mm C2
4 4
3 0 3 0
0 30 60 90 0 30 60 90
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
◦
ϕ( ) ◦
ϕ( )
3.5 3.5
8 Ø 100 mm C4 8 Ø 100 mm C3
4 4
3 0 3 0
0 30 60 90 0 30 60 90
2.5 2.5
wear (mm)
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 15 30 45 60 75 90 0 15 30 45 60 75 90
◦
ϕ( ) ◦
ϕ( )
Figure 5.27. Wear pattern evolution. The tangential node path is at z = 2 mm. The rod
diameter = 100 mm. The case names are C1 to C4 given clockwise starting
at the top left plot.
111
Figure 5.28 shows that all rings were worn to the limits of the wear-box within 30 days of
wear time. The closing speed of the wear gap is nearly constant over time. The closing
speed with the pressure balancing groove is 30 % slower than without pressure balancing
groove. The contact force of the ring with pressure balancing groove is 28 % lower than
without pressure balancing groove at the beginning of the calculation.
3.5 3 3.5 3
8 8
4 4

3 C1 0 3 C2 0
0 30 60 90 2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)

2.5 Ø 100 mm 2.5 Ø 100 mm
2 2
2 2
1.5 1.5
1.5 1.5
1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
3.5 3 3.5 3
8

4
3 C4 3 C3 0
2.5 0 30 60 90 2.5
contact force (kN)
contact force (kN)
2.5 2.5 Ø 100 mm

2 2
2 8 2
4 1.5 1.5
1.5 0 1.5
0 30 60 90
Ø 100 mm 1 1
1 1
0.5 0.5 0.5 0.5
0 0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Figure 5.28. Evolution of the integral contact force of the BCD quarter-model and the
rod (blue) and the closing of the wear gap (green) for the four investigated
BCD configurations C1 to C4. The rod diameter = 100 mm.
112
Comparison of different diameters
A comparison of different diameters shows that with decreasing initial wear gap the posi-
tion of the maximum wear ϕmax shifts towards 90◦ . If the wear gap is smaller, the sealing
segments are longer. Thus the cap segment is supported over a greater length and the
position of maximum wear shifts. Table 5.4 presents the maximum position ϕmax of wear
found in the plots of Figure 5.23, 5.25 and 5.27. For case A4 from Figure 5.23 two values
are given, since the wear maximum shifts from 50◦ to 0◦ when the wear gap is closed
and the structure wears further. A significant increase between the small and medium
diameter occurs. The difference between the medium and big rod are smaller. This is
mainly because the BCD does not scale with the rod diameter: For small rods the cap
segment seals a bigger circumferential section than for a bigger rod, see Figure 4.24.
Table 5.4. Circumferential position ϕmax of the maximum wear. For A4 two values are
given, since the maximum wear at the calculation’s end occurs at 0◦ . For
comparability the maximum position of the 3rd red line is given. Case *4 has
no pressure balancing groove. The symbol ’*’ represents a wildcard character
to be substituted by the letters A, B or C.
case name wear gap (mm) circumferential position of ϕmax (◦ )

for the rod diameter (case)
28.58 mm (A) 50 mm (B) 100 mm (C)
*3 2 65.3 77.0 77.5
*2 4 58.3 71.0 74.3
*1 6 53.5 69.6 72.5
*4 6 (0), 50 66.0 70.0
Figure 5.29 compares how the wear gap closes for all variations. The time of the wear gap
closing depends on the existance of the pressure balancing groove and the initial wear gap
size, but not significantly on the rod size. The closing speed is constant for the biggest
rod size, but for the small and medium rod it decreases over time. This effect is more
pronounced for the small rod. Table 5.5 gives the times when the wear gap closes for all
12 compared calculations. Increasing the wear gap from 2 to 4 to 6 mm increases the time
until the wear gap closes roughly in the same proportion, e.g. for the 100 mm rod from 6
to 12 to 19 days.
The pressure balancing groove increases the time until the wear gap closes for increasing
rod diameters by 15, 20 and 30 %, respectively. This is in accordance to the increase of
the contact force. Table 5.6 gives the contact force of the loaded virgin rings. Since the
contact area between ring and rod increases linearly with the diameter, the contact forces
grow roughly with the rod diameter.
A bigger wear gap results in slightly higher contact forces. The reason is that the dyn-
amic sealing surface becomes smaller and thus some of the support of the gas pressure
distribution is lost as the wear gap increases.
The effect of the pressure balancing groove is much more pronounced than changing the
113
size of the wear gap. As the rod diameter increases the groove becomes circumferentially
longer and has a higher impact. Thus the introduction of a pressure balancing groove
decreases the contact force by 14, 22.5 and 27.5 % for the small, medium and large rod,
respectively.
The decrease of the contact force over time corresponds to the closing speed of the wear
gap. On the small rod the contact force decreases faster at the beginning. This effect is
slightly visible on the medium rod, but cannot be seen at the big rod.
Table 5.5. Time in days until the wear gap is closed, for all investigated variations. Case
*4 has no pressure balancing groove.
case name wear gap (mm) wear gap closing time (days)
28.58 mm (A) 50 mm (B) 100 mm (C)
*3 2 5.4 5.6 5.9
*2 4 12.0 11.5 12.3
*1 6 18.1 17.3 18.7
*4 6 15.7 14.4 14.4
Table 5.6. Contact force of the ring variations. Case *4 has no pressure balancing groove.
case name wear gap (mm) contact force of virgin ring (N)
28.58 mm (A) 50 mm (B) 100 mm (C)
*3 2 787 1304 2359
*2 4 807 1312 2380
*1 6 815 1309 2400
*4 6 930 1690 3313
114
6 mm wear gap, no pressure balancing groove
6 mm wear gap, pressure balancing groove
3
a) Ø 28.58 mm
0
0 5 10 15 20 25 30
time (day)
b) 3
Ø 50 mm
0
0 5 10 15 20 25 30
time (day)
c) 3
Ø 100 mm
0
0 5 10 15 20 25 30
time (day)
Figure 5.29. Evolution of the wear gap for the four quarter-models and the three rod
diameters: a) 28.58 mm, b) 50 mm and c) 100 mm.
115
6 mm wear gap, no pressure balancing groove
a)
contact force (kN) 3 Ø 28.58 mm
0
0 5 10 15 20 25 30
time (day)
b)
3 Ø 50 mm
contact force (kN)
0
0 5 10 15 20 25 30
time (day)
c)
3 Ø 100 mm
contact force (kN)
0
0 5 10 15 20 25 30
time (day)
Figure 5.30. Evolution of the integral direction independent contact force over time for
the four quarter-models on the three rod diameters a) 28.58 mm, b) 50 mm
and c) 100 mm.
116
5.5.4 Calculations with compressible gas
This section presents the effect of the gas compressibility on the frictional force and the
wear of the BCD ring mounted on a 50.8 mm (2”) rod. The compressibility effects of
the gas in the dynamic sealing surface is becoming more pronounced when the pressure
differential becomes larger and the absolute pressure on the crank side smaller.
A series of numerical simulations investigates the wear pattern dependent on the gas
pressure. The material is assumed to be linear elastic with E = 1400 MPa, ν = 0.4. The
gas pressure load varies between 11 and 101 bar in steps of 10 bar. The pressure on the
crank side is ambient pressure (1 bar). Additionally a calculation with 6 bar is performed
for the comparison with the experiments. Each calculation works with a quarter model
using the available symmetry boundary conditions. The gas pressure drop in the dynamic
sealing surface is calculated for each gas pressure level with and without taking into
account compressibility effects. Figure 5.31 shows the axial gas pressure drop in the
dynamic sealing surface at ϕ = 10◦ and ϕ = 70◦ for different pressure levels.
The calculations are evaluated according to section 5.5.2. The integral contact pressure
is calculated i) after loading the structure, thus without the gas pressure distribution in
the dynamic sealing surface and ii) after adding the gas pressure in the dynamic sealing
surface and starting the wear algorithm. For the latter case Figure 5.32 shows the contact
force over time assuming a) incompressible gas and b) compressible gas.
For comparison with the experiment the results of the quarter model are scaled up to a
full BCD model, see Table 5.7 and Figure 5.33. The dependence of the frictional force on
the gas pressure pg is linear. The modelling of the gas pressure distribution is crucial to
the correct implementation of a wear calculation, since it drastically reduces the effective
contact pressure leading to wear. The implementation of the compressible gas pressure
distribution results in lower effective contact pressure, i.e. 19.8 % for the calculation with
5 bar differential pressure, 24.7 % with 10 bar differential pressure, 27.4 % with 20 bar and
then slowly increasing to 30.6 % with 100 bar differential pressure.
Figure 5.34 shows the effect of the gas compressibility on wear for three loads: a) 3 MPa,
b) 6 MPa and c) 10 MPa. The crank end pressure is at ambient pressure (0.1 MPa). As
expected, the rings wear more slowly if the compressible gas properties are taken into
account in the pressure distribution of the dynamic sealing surface. Additionally the
wear pattern becomes more uniform, since the pressure distribution is effected more in
the region where cap segment touches the sealing segment.
117
a) 10
8
9 4
0
8 0 30 60 90
incompressible
7
gas pressure (MPa)
compressible
6
0
0 2 4 6 8
z (mm)
b) 10
8
9 4
0
8 0 30 60 90
incompressible
7
gas pressure (MPa)
compressible
6
0
0 2 4 6 8
z (mm)
Figure 5.31. Axial gas pressure distribution in the dynamic sealing surface at two cir-
cumferential positions calculated with and without taking into account com-
pressibility effects, drawn as solid and dashed lines, respectively. The lines
represent different pressure levels ranging from 6 to 101 bar. a) At ϕ = 10◦
the pressure balancing groove allows the high gas pressure act between 4 and
8 mm from the cup face. b) At ϕ = 70◦ the gas pressure decreases over cap
and sealing surface.
118
a) 1 400
MPa
1 200 0.5
1
2
1 000 3
contact force (N)
4
5
800 6
7
600 8
9
10
400
200
0
0 10 20 30 40 50 60 70 80 90 100
time (day)
b) 1 400
MPa
1 200 0.5
1
2
1 000 3
contact force (N)
4
5
800 6
7
600 8
9
10
400
200
0
0 10 20 30 40 50 60 70 80 90 100
time (day)
Figure 5.32. Evolution of the contact force over time for different differential pressures
(MPa). The values given are valid for the quarter model. The crank end
pressure is ambient, the pressure load varies from 0.5 to 10 MPa. a) shows
the contact force assuming an incompressible gas pressure distribution in
the dynamic sealing surface. b) gives the results assuming compressible gas
properties.
119
Table 5.7. Contact force FC for a single (full) BCD ring on a 50.8 mm rod, depending
on the head-end gas pressure. The contact pressure is listed for i) perfectly
flat surfaces assuming no gas pressure in the sealing surface FC, flat , ii) an
incompressible gas pressure distribution FC, incompressible and iii) a compressible
gas pressure distribution FC, compressible acting in the dynamic sealing surface.
pg (head end) pg (crank end) FC, flat FC, incompressible FC, compressible
(bar) (bar) (N) (N) (N)
6 1 584 272 227
11 1 1167 544 436
21 1 2334 1096 860
31 1 3499 1640 1276
41 1 4664 2184 1688
51 1 5829 2728 2104
61 1 6992 3272 2516
71 1 8155 3812 2928
81 1 9317 4356 3340
91 1 10478 4896 3748
101 1 11638 5432 4160
·104
1.2
FC, flat
FC, incompressible
1
FC, compressible
0.8
0.6
0.4
0.2
0
0 20 40 60 80 100
gas pressure differential (bar)
Figure 5.33. Contact force FC of a single BCD ring on a 50.8 mm rod, depending on
the sealed gas pressure differential and the type of gas pressure distribution
calculation. In the dynamic sealing surface the gas is assumed to be i) non
existent, ii) incompressible and iii) compressible.
120
3
a) 8
4 delta time = 6.9 (day)
0
0 30 60 90 incompressible
2 ∆p = 3 MPa compressible
wear (mm)
0
0 15 30 45 60 75 90
ϕ (◦ )
3
b) 8
4
0
0 30 60 90
2 ∆p = 6 MPa
wear (mm)
0
0 15 30 45 60 75 90
ϕ (◦ )
3
c) 8
4
0
0 30 60 90
2 ∆p = 10 MPa
wear (mm)
0
0 15 30 45 60 75 90
ϕ (◦ )
Figure 5.34. Wear pattern evolution. Black dashed lines represent calculations with in-
compressible gas properties. Blue solid lines represent the calculations with
compressible gas properties. The plots show three different pressure diffe-
rentials: a) 3 MPa, b) 6 MPa and c) 10 MPa. The crank end pressure is at
ambient pressure (0.1 MPa).
121
5.5.5 Calculations with different elastic moduli
This section presents calculations with different elastic moduli, which represent different
temperature levels. Figure 4.26 in section 4.3 presents the dependency of the elastic
modulus on the temperature. The elastic modulus ranges from 350 MPa to 1400 MPa
between 140 ◦C and 23 ◦C. The three elastic moduli chosen for comparison are 350, 700
and 1400 MPa representing 140, 80 and 23 ◦C, respectively.
The considered BCD is mounted on a 50.8 mm (2”) rod. The gas pressure distribution
takes the compressibility of the gas into account. Two different pressure loads are used
for this evaluation: 3 and 6 MPa. The pressure on the crank side is ambient pressure.
The code pertaining to the 6 MPa calculation is provided in the appendix 7.
Figure 5.35 presents the effect of the elastic modulus on the contact forces over time for
two load levels: 3 and 6 MPa. At the beginning the high modulus rings show slightly
higher contact forces. The contact pressure decreases with increasing wear. A higher
elastic modulus results in lower contact pressure, since the ring segments carry a larger
portion of the load. This is independent of the load level, but becomes more pronounced
with increasing load. The closing of the wear gap is only slightly effected, see Figure 5.35b.
Higher elastic modulus means longer time until the wear gap closes. Figure 5.36 shows
the evolution of the wear pattern of the sealing segments. Sealing segments with lower
elastic modulus wear faster at the segment tip because the segment carries less load, the
wear distribution becomes less uniform.
This observation allows the following notion: A hot-running ring will wear faster at the
sealing segment’s tips. If the tips are consumed too fast, the segments are very thin
when the wear gap closes. The tips might be too weak to guarantee for enough structural
stability of the ring or they might even get stuck in the clearance between backup and
rod. Both cases likely lead to ring failure. This underlines that lower temperatures are
crucial for a ring to perform well over a long time.
122
a) 350 MPa
600
700 MPa
1400 MPa
500 ∆p = 3 MPa
contact force (N)
400
300
200
100
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
time (day)
b) 350 MPa
600
700 MPa
1400 MPa
500
∆p = 6 MPa
contact force (N)
400
300
200
100
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
time (day)
Figure 5.35. The evolution of the contact force over time and its dependency on the elastic
modulus is shown for two different load levels a) 3 MPa and b) 6 MPa. The
crank end pressure is ambient pressure (0.1 MPa). The elastic modulus is
given in the legend.
123
3
a) 8 350 MPa
4 700 MPa
2.5 0 1400 MPa
0 30 60 90
delta time = 6.9 day
2 ∆p = 3 MPa
wear (mm)
1.5
0.5
0
0 15 30 45 60 75 90
ϕ (◦ )
3
b) 8
4
2.5 0
0 30 60 90
delta time = 6.9 day
2 ∆p = 6 MPa
wear (mm)
1.5
0.5 350 MPa

700 MPa
1400 MPa
0
0 15 30 45 60 75 90
ϕ (◦ )
Figure 5.36. The wear pattern evolution and its dependency on the elastic modulus is
shown for two different load levels a) 3 MPa and b) 6 MPa. The crank end
pressure is ambient pressure (0.1 MPa). The elastic modulus is given in the
legend.
124
Chapter 6
Experimental validation
In a compressor the packing consists of several cups each filled with a sealing element. It
is unknown how the pressure is broken down by the sealing elements and how the pressure
in the cups changes with time [17, 22]. Field experience has shown that sometimes two
compressors of the same type, at the same site, next to each other (redundant machines)
behave unexplainably. One runs smoothly, the other one is a trouble-maker. Finding the
difference is often a tedious procedure and sometimes the reason for the problems is not
detected.
To avoid uncontrollable irregularities as much as possible a test packing has been develo-
ped to compare ring designs, to evaluate ring features and to examine the effect of different
operating conditions. Additionally a precise wear measurement procedure is presented for
the BCD ring.
Parts of the content and some of the figures of this chapter are published in [35].
6.1 Test rig
A crank drive named “short block BS102”, manufactured by LMF (Leobersdorfer Maschi-
nenfabrik, Leobersdorf, Austria), has been prepared to work with different rod diameters
and test packings. It is possible to work between 365 and 911 rpm. The drive works like a
boxer (engine) with horizontally opposing rods. The stroke is 98 mm. Since different rods
might be used on the opposite sides at the same time, compensation weights are place on
the lighter rod. The motor is placed below the crank case on the frame. Motor and drive
are connected via a belt. Figure 6.1 shows a schematic sketch of the test rig.
In comparison to real compressors no cylinders and packing boxes are mounted directly
to the crank drive as standard. With the presented set-up it is not possible to compress
gas. There is no need for compression since the last ring in a pressure packing is usually
loaded by static (suction) pressure, which is represented by the applied static pressure on
the test rig.
On the test rig different gases can be used:
125
• Compressed air, 13 bar from the production plant (HW). This pressurised air has a
dew point of 3 ◦C (5.95 g m−3 ). This gas is condsidered wet.
• Compressed air, 30 bar from the production plant (HW) with a dew point of −40 ◦C
(0.117 g m−3 ). This gas is bone-dry.
• Bone-dry nitrogen from the production plant’s 25 m3 liquid nitrogen storage tank,
with 30 bar. The gas purity is nitrogen 5.0 (≥ 99.999 % nitrogen).
• Any other non-corrosive, non-toxic gas which is supplied by a bundle of pressure

bottles.
The test rig is placed indoors and the test room’s oxygen content is constantly monitored
to prevent asphyxia.
The short block is mounted on a steel frame which has a bench with guidings on each side
i.e. for each throw. The rod is fixed at the crank head and passes a standard set of oil
wipers. Outside the crank case and the wiper packing a compensation weight is placed
onto the rod if necessary. Then there is approximately 400 mm rod length as test space
for a test packing and instead of a piston a guiding guides the rod’s free end.
6.1.1 Packing
The test packing consists of two massive steel plates between which a T-cup and flanges
are placed. The T-cup has two recesses each with a width of 12.2 mm, where a sealing
element is mounted on the rod. The hollow rod has an outer diameter of 50.8 mm and an
inner diameter of 35 mm. The hollow design reduces the rod weight and so the accelerated
masses on the test rig. Figure 6.1 shows a schematic sketch. Although no cylinder is used
the sides of the packing are referred to as “cylinder” and “crank” side. Figure 6.2 shows
a cut of the test packing.
The temperature is measured on 8 positions during a test. The packing temperature is

monitored on four positions with thermocouples of type K. 1) and 2) are placed at a radial
position of 31.9 and 35.9 mm at the packing’s cylinder end side, respectively. Sensor 3)
is placed in the middle of the T-cup. The tip of the sensor is in contact with the gas.
Position 4) is in the crank end flange. 5) the room temperature is tracked on the cylinder
side of the packing. 6) A pyrometer (impac 140, LumaSense Technologies GmbH, Frank-
furt, Germany) directly measures the rod temperature through a quartz glass in the axial
middle of the T-cup. The pyrometer works above 75 ◦C. Below this value the from the
rod emitted energy is too low for this measurement principle. 7) and 8) measure the tem-
peratures of the packing cooling water, the inlet and outlet, respectively. A heat meter
measures the effect of the packing cooling using the temperatures 7) and 8). The cooling
water from the packing cooling (secondary cooling circuit) is cooled via a heat exchanger
with the plant’s primary cooling circuit. The primary circuit is cooled outdoors, thus the
cooling water temperature of the primary circuit depends on the outside temperature.
126
First tests showed that the packing cooling is not sufficient to regulate the rod tempera-
ture especially at elevated loads and/or rod speeds. Further improvement of the cooling
was necessary to achieve rod temperatures similar to field and testing experience. So
additionally the hollow rod is cooled via pressurised air on the inside. The air is guided
with a lance into the hollow rod and blown off into ambient. This allows to work with
higher pressures and rod speeds.
flow meter pressure inlet

rod cooling pyrometer
pressure sensor
rod guiding weight compensation
crank
1 6
rod 3 4 case
2
5
rings
dewetron: 8 y x
signal processing force measurement
7 z
signal displaying
save data heat exchanger
secondary circuit
and heat meter
temperature
heat exchanger detection
primary circuit position
Figure 6.1. Shematic drawing of the test rig, the mounted packing and the 8 temperature
measurement positions. [35]
127
pyrometer temperature
position plate detection
T-cup
position
plate
flange
hollow rod
packing
ring
bolt
6
cylinder side
crank side
flange
4
3
1 2
packing
cooling
circuit
connection
bottom fixture load cell temperature sensor position

Figure 6.2. Cut through the packing. The temperature measurement positions are marked
and labeled according to Figure 6.1.
The packing ring is loaded by static gas pressure. Since the pipe connecting the nitrogen
or pressurized air tank with the test rig is longer than 25 m, it is assumed that the gas
has ambient temperature at the inlet. Before the pressure inlet in the axial centre of
the T-cup a mass flow meter and a pressure sensor are installed. The mass flow meter
monitors the flow into the packing, i.e., the leakage of both rings. The rings’ leakage can
hardly be separated, since it leaks into ambient.
The whole packing is mounted on a steel plate which is fixed to a 3D force measurement
device. It allows to measure not only the frictional force occurring between rod an ring,
but also any of the perpendicular forces.
A Dewetron (DEWE-5000) processes and displays all measured data. Every minute the
signals of 300 ms are saved. After conversion using the Software DeweSoft [19], Matlab [49]
post-processes the data.
128
6.2 Test procedure
Prior to each test the rod, the backup rings and the packing rings are prepared as follows:
The rod used for the packing ring tests is made of steel (1.4021) and coated with WC.
The coating is according to industrial standard and consists of WC/Co/Cr with a ratio
of 86/10/4 [31]. The rod’s surface is cleaned and finished to a roughness of Ra =0.3 µm.
A static leakage test is performed at 6 bar prior to mounting a ring on the rod. This
test is a standard production control test performed for all BCD rings. The leakage
limit is 16 L min−1 for a BCD ring made from HY54. No significant changes could be
found between the measurement after production and before mounting. Between ring
production and a test a maximum of half a year passed.
A 3D tactile coordinate measurement machine precisely scans the ring geometry before
and after a test. The comparison of the measurements gives the ring wear.
The sealing elements are placed in the T-cup and their initial position is noted to detect
sealing element rotation. In none of more than 30 tests rotation was detected. A bronze
backup ring is used. The axial thickness of the backup ring varies with the packing ring
thickness, such that the axial clearance is never below 0.2 mm. Prior to each test the
backup ring is lapped (industrial standard procedure).
6.2.1 Wear measurement
The BCD ring’s geometry is measured before and after the packing ring test. The initial
configuration of the virgin ring segments is compared to the worn configuration. Wear
inevitably leads to a distortion of the unmounted packing ring due to residual stresses. In
order to separate wear related shape changes from residual stress related shape changes
a sophisticated measurement procedure has been devised.
The ring is disassembled and each ring segment is separately mounted on a steel fixture,
Figure 6.3. The measurement procedure measures always a sealing and a cap segment
at the same time. Double sided adhesive tape fixes the segments at the fixture during
measurement.
To distinguish between the two sealing segments and between the two cap segments one
of each is marked with “P” before the test on the pressure side. Figure 6.4 shows the
chosen coordinate system and the segment’s reference points. All wear results displayed
are using this coordinate system.
A 3D coordinate measurement machine (CMM) (DeMeet 404, manufacturer Schut Ge-

ometrical Metrology, Groningen, The Netherlands) provides a detailed 3D scan of the
whole ring geometry. The density of the data points is highest on the inner diameter,
where the ring wears. Figure 4.9 shows two mounted segments and the xy, yz and xz
planes used to set the origin for the measurement. The measurement routine compensates
axial misalignment by setting the coordinate system on the ring’s top plane.
Matlab post-processes the exported raw data. The measured configuration before the
129
wear test is referred to as “pre” and after the test as “post” configuration. Virtual re-
orientation of each segment compensates minor misalignment resulting from the manual
mounting. Figure 6.5 shows the raw data of a pre measurement.
As next post-processing step the data is shifted and rotated. The origin is set between the
edges, which are calculated by cutting the radial planes and the outer diameter for the
cap segment or by cutting the planes, which are parallel to the y = 0 plane, and the outer
diameter for the sealing segment. Then the data is rotated so that these edges are in the
x = 0 plane. After aligning the data, the inner and outer diameter are fitted with circles
for each measured axial (z) level and the planes’ positions are recalculated. In the pre
configuration the centre points of the inner and outer diameter are equal (concentric), see
Figure 6.6. After the test run, due to wear and deformation, the centre points differ such,
that the centre points of the inner diameter are closer to the ring segment, Figure 6.7.
The significant wear on the sealing segment’s ends is visible, when comparing the pre and
the post configuration.
The cap segment deforms plastically during the test. The cap segment’s surface con-
tacting the sealing segments is bent, Figure 6.8. The sealing segment’s deformation of
the originally parallel planes is visible, they are deformed towards the centre points, Fi-
gure 6.9.
Then the radial ring dimension is evaluated dependent on the axial and circumferential
position. From the measurement points a mean outer diameter is calculated. The dis-
tance from the measurement points of the inner diameter to the mean outer diameter
is calculated to obtain the local radial ring thickness. The post-test local radial ring
thickness is subtracted from the pre-test local radial ring thickness, giving the local wear.
Figure 6.10 shows the wear pattern of the ring segments over the axial and angular po-
sition. To determine if a gradient over the axial height exists, the data is plotted as line
plot, Figure 6.11, where for each z-coordinate a line is drawn. The wear measurement and
analysis procedure is repeated twice for a single BCD ring, so the full wear is determined
and can be displayed within one plot, see Figure 6.12.
The total wear volume of a ring is derived from the measurement. The total wear volume
is hard to measure directly with contacting measurement because edges and corners are
hardly accessible and not well reproducibly measurable, especially if a geometry changes
unpredictably due to wear and deformation. Thus the measured wear volume is extrapo-
lated by a factor created from the ratio of the measurement area and the real contacting
area. The extrapolation in axial direction is reasonable, since the change in the wear
pattern in axial direction is small (Figures 6.10 and 6.11). The extrapolation in circum-
ferential direction is small since this direction is scanned with high resolution and nearly
to the segment ends. The up-scaling factor is approximately 1.4. Table 6.1 presents the
measured and calculated wear volume data for the full BCD ring, from which the sealing
segment and the cap segment without “P” are presented below.
130
x = 0 (yz plane) y z
z x
y = 0 (xz plane) y x
z = 0 (xy plane)
b)
a)
Figure 6.3. a) a cap segment and b) a sealing segment ready for 3D scanning. The seg-
ments are mounted on a steel fixture on the CMM. For both segments the
origin is set at the intersection of the three planes drawn in red (yz plane),
green (xz plane) and blue (xy plane).
v reference point
v v v v
90◦ 90◦
180◦ 0◦ 0◦ 180◦
270◦ 270◦
v v v v
Figure 6.4. Sketch of the BCD ring geometry. Left: One cap and one sealing segment
are marked with “P” on the high pressure (cylinder) side. The location of
the reference points is shown. Right: Low pressure (crank) side of the BCD,
the two marked segments are on the upper right. Both sketches show the
coordinate system used for displaying results.
131
z (mm)
m )
y( x (m
mm)
(a) Cap segment.

z (mm)
y( )
mm m
) x (m
(b) Sealing segment.
Figure 6.5. Raw measurement data from BCD ring segments. The origin is from the
CMM.
132
z (mm)
)
m
(m
y
x (m
m)
(a) Cap segment.

z (mm)
)
m
(m
x (m
m)
y
Figure 6.6. “Pre” configuration: Shifted and rotated data from 3D scans. The origins are
reset between the calculated reference points. The fitted diameters are drawn
as black line, their centre points as yellow filled circles, the fitted planes are
shaded blue. Each calculated reference point is represented by a filled red
circle and results from the intersection of a fitted outer diameter and a fitted
plane.
133
z (mm)
)
m
(m
y
x (m
m)
(a) Cap segment.

z (mm)
)
m
(m
x (m
y
m)
Figure 6.7. “Post” configuration: Shifted and rotated data from 3D scans with the orig-
ins are reset between the calculated reference points on the edges. The fitted
diameters are drawn as black line, their centre points as yellow filled circles,
the fitted planes are shaded blue. Each calculated reference point is repre-
sented by a filled red circle and results from the intersection of a fitted outer
diameter and a fitted plane.
The Figures 6.8 and 6.9 present the difference resulting from wear and plastic deformation
between pre and post configuration for the cap and sealing segment, respectively. The
134
pre configuration is drawn in green, the post configuration in black. The dots represent
measurement points.
pre
10
post
(b)
y (mm)
-10
-20 -10 0 10 20
x (mm)
(a)
1
y (mm)
0.75
0.5
-20 -10 0 10 20
x (mm)
(b)
Figure 6.8. (a) Overview of the pre and post test data of a BCD cap segment. (b) The
deformation of the cap segment facing the contact area contacting the sealing
segments.
135
20
(b) pre
post
10
y (mm)
(c)
-10
-30 -15 0 15 30
x (mm)
(a)
20
15
y (mm)
10 0
y (mm)
-5
5
-10
0 -15
-30 -25 -20 -15 -10 -5 0 5 10
x (mm) x (mm)
(b) (c)
Figure 6.9. (a) gives an overview of the pre and post test data of a BCD sealing segment
and the position of (b) and (c), which are displayed in higher magnification.
(b) The deformation at the sealing segment’s end is of the same order of
magnitude as the wear. (c) The change of the inner diameter also causes a
permanent deformation of the outer diameter.
136
wear (mm)
z(
mm)
◦
ϕ( )
(a) Cap segment.

wear (mm)
z(
mm
)
◦
ϕ( )
Figure 6.10. The wear pattern of the ring segments results from subtracting the post
configuration from the pre configuration.
137
wear (mm)
ϕ (◦ )
(a) Cap segment.

wear (mm)
ϕ (◦ )
Figure 6.11. Wear pattern of the ring segments viewed in axial direction. The data is the
same as in 6.10. Each line represents an axial position given in the legend.
The dashed lines in (b) are the part of the sealing segment which is between
the high pressure side of the ring and the pressure balancing groove.
138
Table 6.1. Detailed wear and wear volume data. The pressure balancing groove splits a
sealing segment’s contact area in two sections. One section is on the head end
side (high pressure side) of the pressure balancing groove, the other on the
crank end side. The results for these areas are provided separately.
variable value unit
area data
theoretical area virgin state cap segment 138.3 mm2
theoretical area virgin state sealing segment head end 103.1 mm2
theoretical area virgin state sealing segment crank end 295.2 mm2
total theoretical BCD area virgin state 1073.3 mm2
measured area cap segment P 90.1 mm2
measured area sealing segment P head end 60.3 mm2
measured area sealing segment P crank end 221.0 mm2
measured area cap segment 89.3 mm2
measured area sealing segment head end 60.1 mm2
measured area sealing segment crank end 220.2 mm2
wear volume data

measured wear volume cap segment P 32.55 mm3
measured wear volume sealing segment P head end 36.58 mm3
measured wear volume sealing segment P crank end 163.95 mm3
measured wear volume cap segment 55.37 mm3
measured wear volume sealing segment head end 40.40 mm3
measured wear volume sealing segment crank end 181.39 mm3
up-scaled wear volume cap segment P 49.96 mm3
up-scaled wear volume sealing segment P head end 62.49 mm3
up-scaled wear volume sealing segment P crank end 219.08 mm3
up-scaled wear volume cap segment 85.77 mm3
up-scaled wear volume sealing segment head end 69.28 mm3
up-scaled wear volume sealing segment crank end 243.18 mm3
total measured wear volume 510.25 mm3
total up-scaled wear 729.76 mm3
radial wear data

mean radial wear cap segment P 0.361 mm
mean radial wear sealing segment P head end 0.606 mm
mean radial wear sealing segment P crank end 0.742 mm
mean radial wear cap segment 0.620 mm
mean radial wear sealing segment head end 0.672 mm
mean radial wear sealing segment crank end 0.824 mm
mean radial wear 0.689 mm
139
1.0
1
0.8 0.9
wear (mm)
0.6 0.8
0.4 0.7
0.2 0.6
0 0.5
8 0.4
6 0.3
z(
4 0.2
mm
2 0.1
)
0 0
0 90 180 270
270
ϕ (◦ )
Figure 6.12. Wear of all ring segments. The segments marked with “P” are the cap
segment at 90◦ and the sealing segment at 180◦ .
6.2.2 Measurement reproducibility
Repeated measurement of the same ring proves the reproducibility and capability of the
3D scanning method and the virtual aligning process:
A BCD ring is mounted, scanned and demounted twice. Figure 6.13 presents the difference
between the two scans for a whole BCD ring. A difference greater than 0.01 mm occurs
only at the circumferential ends of the scanned areas. These artefacts from mounting occur
since the measurement procedure tries to evaluate the inner diameter in circumferential
position as far as possible. The circumferential length of the inner diameter increases due
to wear, thus the measurement procedure for a virgin ring requires careful mounting at
the fixture.
Figure 6.14 shows the same data, while the viewing direction is the axial direction. For
each ring segment a separate graph presents the high repeatability and data quality of
this method. The differences of approximately ±0.01 mm is in the range of the sensitivity
of the CMM.
140
radial difference (mm)
0.03
0.02
0.02
0
0.01
-0.02 0
-0.04 -0.01
-0.06 -0.02
8 -0.03
6 -0.04
z(
4 -0.05
mm
2 -0.06
)
0 -0.07
0 90 180 270
270
ϕ (◦ )
Figure 6.13. Reproducibility test: The difference between two measurements of a single
BCD ring is depicted.
(a) Cap segment ”P” (b) Cap segment
ϕ (◦ ) ϕ (◦ )
(c) Sealing segment ”P” (d) Sealing segment
ϕ (◦ ) ϕ (◦ )
Figure 6.14. Reproducibility test: The difference between two measurements of a single
BCD ring are presented. The viewing direction is the axial direction. For
each segment (a) to (d) the data is drawn separately.
141
6.3 Test results
Out of a test series two tests are chosen for discussion. The first test, named A, shows two
things: Firstly what can happen if a ring operates outside it’s design limits i.e. becomes
too hot. Secondly a reason why several rings are used in a packing. In the second test
run, named B, rings experience common operating conditions. Test B is used later to
compare simulation and experiment.
6.3.1 Test A
Table 6.2 gives the parameters for test A. The rings in this test are named A1 (head end)
and A2 (crank end). Note that the used ring material (HY54) is not designed for bone-
dry operating conditions, thus the wear rates are significantly higher. The combination of
10 bar gas pressure and 2.15 m s−1 rod speed led to high rod temperatures. In comparison
to a real compressor the rod is not cooled by the fresh and cool gas, which enters the
compression chamber during suction. The cooling inside the rod is turned on, but does
not stop the rod temperature from rising above the material design temperature 150 ◦C.
Figure 6.15 shows the averaged rod and packing temperature, the leakage1 for both rings
and the gas pressure over time. The temperatures are averaged over the rod length or the
whole packing housing. The test duration is 192 h.
The temperature is very high within few minutes and stays above 150 ◦C. After only
17 h the rings fail the first time with leakage rates above 200 l min−1 . Since the pressure
is only regulated, but not controlled, the high leakage is accompanied by a pressure
drop. Nonetheless, the test was continued to check if the rings seal again. After failing,
the leakage rate dropped continuously and 6 h later the rings sealed fully for another
10 h. This pattern of sealing and failing repeated unexpectedly regularly with this set of
parameters. For closer consideration of such a cycle, view Figure 6.16, where the grey
highlighted area of Figure 6.15 is zoomed.
Photos of the worn rings can be seen in Figure 6.17 and Figure 6.18 for the rings A1
and A2, respectively. Both rings have segments sticking to each other after the test. The
segments seem welded together and snap apart when applying force.
A microscopy of the cap segments shows significant difference between a cap segment
with and without a sealing segment sticking to it, see Figure 6.19. The cap segment
which did not stick after test shows the usual homogeneity of HY54. Whereas the sticking
segment appears to be covered in fibrils, which are assumed to be PTFE fibrils, generated
during breaking the segments apart. The fibrils cover the cap segment from the wear gap
towards the cap segment’s ends. The density of fibrils seems highest near the wear gap
and decreases towards the segment’s end. This is in accordance to the calculated contact
pressure which is highest at the sealing segments ends.
1
The leakage rates given in this section are in standard litre per minute. The term “standard” refers
to the standard conditions of the gas, i.e. 101 325 Pa(=1 atm) and 273.15 K(=0 ◦C).
142
Figure 6.15. Protocol of the averaged rod and packing temperature, the leakage and the
gas pressure over time for test A. The area highlighted in grey between the
test time of 15 and 45 h is zoomed and shown in Figure 6.16.
143
Figure 6.16. Zoom from Figure 6.15. The rings fail and seal repeatedly. Position 1) marks
the failing of the rings. The high leakage is accompanied by pressure loss and
an instant drop of the rod temperature. The leakage rate drops between 1)
and 2) until the rings seal again. Meanwhile the rod temperature increases
until a maximum at position 3) is reached, the rings are still sealing. After
3) the rod temperature slowly decreases, while the rings seal until the rings
fails again at 4).
144
Table 6.2. Test parameters for test A.
parameter value unit

gas type nitrogen
gas dryness bone-dry
pressure 10 bar
average rod speed 2.15 m s−1
ring material HY54
backup material bronze
rod inside cooling yes
static leakage test at 6 bar
ring A1 8.6 l min−1
ring A2 13.2 l min−1
P
a) b) c)
2)
P P P
1)
3)
Figure 6.17. Worn ring A1: a) shows the worn ring form the pressure side, note that
two segments are sticking together at position 1). To distinguish between
equal segments, the “P”-segments are marked, according to the convention
presented above, see Figure 6.4. In b) and c) the dynamic contact surfaces
are depicted for the sealing segments and the cap segments, respectively.
The contact area of the sealing segments is very smooth and of homogeneous
appearance. The P-cap segment in c) shows inhomogeneous wear. A region
appearing brighter is marked with 2). On the caps pressure marks of the
sealing segments can be seen, 3). The area of the open wear gap appears
brighter and seems to be covered in either debris or dirt.
145
a) P b) 4)
1)
5)
P
2) 3) P
3) 2)
c)
b) c)
3) 2)
6)
Figure 6.18. Worn ring A2. To distinguish between equal segments, the “P”-segments
are marked, according to the convention presented above, see Figure 6.4. a)
shows the ring from the crank side, the garter spring is still mounted. A
gap between the P-cap segment and the sealing segments is visible, 1). The
sealing segments are touching only in one place, since they are sticking to
the cap segment 2) and 3). In b) the garter spring is removed and the P-cap
segment is taken from the other segments. It shows pressure marks of the
sealing segments. The wear gap 4) is visible between the pressure marks.
It is (naturally) bigger than the wear gap of the unmounted, unloaded wear
gap at 5). Picture c) was taken before dismounting the garter spring (similar
to a). The spring is visible at 6). The garter spring force is not sufficient to
separate the sticking segments in 2) and 3).
146
a1) b1)
b2)
a2)
5mm 5mm
a2) b2)
a3) b3)
1mm 1mm
a3) b3)
0.2mm 0.2mm
Figure 6.19. Microscopy of a cap segment after test A. The left pictures a1) to a3) show
a cap segment in different magnifications. No sticking was present between
cap and sealing segment at the test end. The right pictures b1) to b3) show a
cap surface after the corresponding sticking sealing segment has been broken
away. The higher magnifications b2) and b3) show that the surface is covered
with thin white fibrils. It is assumed that these are PTFE fibrils.
The wear measurement results in the Figures 6.20 and 6.12 for ring A1 and A2, respecti-
vely. The tables 7.1 and 6.1 give detailed data of the ring segment’s wear, the former is
in the appendix, the latter in section 6.2.1. Although the photographs of A1 (Fig. 6.17)
indicate that a certain section of the P-cap segment is worn differently, this cannot be
seen in the wear measurement. It is interesting that, although the rings failed and sealed
repeatedly the wear profile is very smooth and shows the expected pattern: The maxi-
147
mum appears at the sealing segment’s ends and in the circumferential middle of the cap
segments. Nevertheless this test is not fit to be compared to a steady-state simulation,
since the rod and thus the material temperature changed frequently and over a great
range.
0.9
0.8
0.8
wear (mm)
0.6
0.7
0.4
0.6
0.2 0.6
0 0.4
8
0.3
6
4 0.2
z(
2 0.1
mm
0 0
)
0 90 180 270
270
ϕ (◦ )
Figure 6.20. Wear of the ring A1.
The fact that a packing consists of several rings indicates that such a behaviour of sea-
ling and failing appears independently from the material, design and manufacturer. Fi-
gure 6.21 summarizes a theory, which explains this process of repeated sealing and failing.
After starting a test/compressor the rings seal and thus the temperature of rod and ring
increases. With increasing temperature the elastic modulus and the structural stiffness
decrease. This leads to higher contact forces since a more compliant structure carries
less load. The frictional force and thus the frictional heating rises along with the contact
forces, leading to a further increase in temperature. This cycle self-accelerates until a
quasi-stationary temperature level is reached. The material fails and the ring extrudes
between rod and backup ring/cup and is consumed by wear or until the ring segments
start to stick together.
It is assumed that a certain material dependent combination of temperature, contact
pressure and time allows the ring segments to stick together. If these conditions are met,
the ring segments stick together, the ring’s structural stiffness increases significantly and
allows to carry more of the gas pressure load than before. This reduces the contact force
and with it the frictional heating, resulting in a decrease in rod and ring temperature.
Figure 6.16 shows clearly: the rod temperature increases between 1) and 3) and then
slowly decreases until 4). The temperature decreases more slowly than it has previously
increased. This is explained as follows: As soon as the segments stick together the struc-
ture carries an ever increasing portion of the load as it slowly wears away. If the sticking
segments break loose again, they would geometrically compensate the wear, just like they
148
are designed to do. Here the segments stick firmly enough together and cannot compen-
sate wear. The contact pressure decreases and thus the temperature drops from 190 ◦C
to 170 ◦C.
In contrast to the previous (thermally upwards) cycle this cycle has the potential to
enormously self-accelerate. If the contact pressure is reduced enough due to ring wear a
leakage path may form. This gas flow cools ring and rod. The colder the ring becomes
the bigger the leakage path may become, which accelerates the cooling further. Thus at
a certain point, Figure 6.16 position 4), the rod temperature drops within minutes below
the lower measurement limit 75 ◦C. Here in Figure 6.16 at 4) it took less than 4 min for
the rod temperature to drop from 150 ◦C below 75 ◦C.
Finally with the low rod temperature, ring and rod are not conform any more and high
leakage occurs. The ring fails sealing, but over time conformity can again be achieved due
to wear, closing leakage paths until the ring seals again. Over time the sticking segments
may also break loose again. This is then a starting point for a next cycle.
149
start sealing temperature
te tact
co
tim erat essu

m
n
p pr
e ure re
time + wear
quasi steady-
contact state possible structural
force stiffness
(high) leakage
sticking
temperature
leakage
accelerates
contact structural
this cycle
force stiffness
Figure 6.21. Possible life scheme of a ring where the segments start sticking together.
The solid lines show cogent correlations. The dashed lines show possible
correlations. A proper ring seals at the beginning and enters a loop: it
heats up, loses structural stiffness, which increases the contact forces and
again the temperature. This cycle may stabilize into a quasi steady-state. A
combination of temperature, contact pressure and time can lead to sticking,
breaking the quasi steady-state. Sticking increases the structural stiffness,
thus decreases the contact forces leading to lower temperatures and again to
higher structural stiffness. This cycle accelerates if leakage occurs. If a ring
cools down rapidly the ring usually fails to seal due to thermally induced
strains, which open leakage paths. Nevertheless the ring can start sealing
again, if time and wear create conformity of ring and rod.
6.3.2 Test B
Test B has been performed with a gas pressure of 30 bar, a mean rod speed of 1.2 m s−1
and with bone-dry nitrogen. Table 6.3 shows the test parameters.
150
Table 6.3. Test B - test parameters
parameter value unit

gas type nitrogen
gas dryness bone-dry
pressure 30 bar
average rod speed 1.2 m s−1
ring material HY54
backup material bronze
rod inside cooling yes
static leakage test at 6 bar
ring B1 17 l min−1
ring B2 12 l min−1
Figure 6.22 presents the averaged rod and packing temperature, the leakage for both
rings and the gas pressure over time. The temperatures are averaged over the rod length
or the whole packing housing. The test duration is 359 h. The test shows an average
rod temperature of 84 ◦C. The temperature profile is very constant over the whole test
time. The maximum rod temperature is 94 ◦C. During the first 285 h the rings sealed
well, the leakage is always below 60 l min−1 for both rings. Then the rings start to lose
sealing performance. On average the leakage goes up once every 83 min in the last 75 h.
Figure 6.23 is a zoom of the grey highlighted area of Figure 6.22. Four losses in sealing
performance between the test time of 325 and 330 h have been monitored. A measurement
point has been taken every 3 min. Each leakage is represented by a single data point only.
These events are very short and no change in the rod temperature is measured, nonetheless
the pressure drops according to the increased leakage.
Figure 6.24 presents images of the disassembled rings after test B. The rings show no
sticking. The cap segments show pressure marks of the sealing segments. The dynamic
sealing surfaces appear homogeneous and smooth. The Figures 6.25 and 6.26 show the
results of wear measurement for the rings B1 and B2, respectively. The tables 7.2 and
7.3, in the appendix, give detailed data of the ring segment’s wear. In comparison to ring
B1, the ring B2 shows a very symmetrical wear pattern. The P-cap segment of ring B1
is nearly virgin, the maximum wear is 0.1 mm and the average radial wear is 0.08 mm,
whereas the other cap segment shows 0.35 mm average radial wear, with a maximum in
the segment’s middle of 0.62 mm. In comparison the cap segments of ring B2 show a
maximum wear of 0.74 mm (P-cap) and 0.79 mm (other cap segment). This indicates
that the sealing segments were sticking slightly to the P-cap segment of B1. Thus it is
assumed that the ring B1, especially the region of the P-cap segment is the reason for
repeated short-term leakage. A slight sticking is assumed since the ring starts to seal
again quickly after leaking.
Ring B2 is suitable for a comparison with the developed calculation. The rod and packing
151
temperatures stayed constant over the test duration and no signs of sticking or other
symmetry breaking effects are visible. The comparison follows in section 6.4.2.
Figure 6.22. Protocol of the averaged rod and packing temperature, the leakage of both
rings and the gas pressure over time for test B. Figure 6.23 presents a zoom
into the grey highlighted area.
152
Figure 6.23. Zoom from Figure 6.22. The rings fail four times for a very short period,
each time an according pressure drop is monitored. The rod temperature
stays constant since the rings start to seal again quickly after leaking.
153
a1) P b1) P
Ring B1 Ring B2
P P
pressure side pressure side
a2) P b2) b3)
Ring B1
crank side P P P
Figure 6.24. Worn rings B1 and B2, a) and b), respectively. A “P” marks the P-segments,
according to the convention in Figure 6.4. The dynamic contact surface
appears smooth and homogeneous. Their ring segments were not sticking
together when the test ended and the rings were unmounted. Nonetheless a
pressure mark of the sealing segments can be seen on the cap segments in
b3). [35]
154
1.2
1.2
1.0
wear (mm)
0.8 1.0
0.6
0.4 0.8
0.2
0 0.6
8
0.4
6
4 0.2
z(
2
mm
0 0
)
0 90 180 270
270
ϕ (◦ )
Figure 6.25. Wear of the ring B1. The ring shows an irregular pattern. The P-cap segment
is nearly in its virgin condition, with a maximum wear of 0.10 mm. This
indicates that the sealing segments were sticking to the P-cap segment.
1.2
1.2
1.0
wear (mm)
0.8 1.0
0.6
0.4 0.8
0.2
0 0.6
8
0.4
6
4 0.2
z(
2
mm
0 0
)
0 90 180 270
270
ϕ (◦ )
Figure 6.26. Wear of the ring B2. The segments are equally worn. The pattern looks as
expected and is used for comparison with a calculation.
155
6.4 Comparison of experiment and calculation
6.4.1 Frictional force
On the testing device presented in 4.2 a pressure inlet in the T-cup allows to fill the T-cup
with pressurized gas (i.e. air). So packing rings can be tested when sealing a pressure
differential. Such a test allows to measure the effect of the gas pressure distribution in
the dynamic sealing surface and to verify the numerically correct implementation.
6.4.1.1 Measurement
The BCD rings are mounted on the rod after cleaning with ethanol. The force is set to
zero while the rod with the mounted packing rings is hanging freely from the load cell.
Then the packing is fully assembled and pressurized. A certain initial pressure differential
eases the initial sealing, thus the high pressure levels are measured first.
Before the first test run starts, the rod is manually moved up and down for several
millimetres to ensure that the rings are sealing. The measurements are performed similar
to the situation in a real packing. To allow the pressurized rings to properly adjust to
their duty, they run several cycles with maximum speed (1000 mm min−1 ). Then the
series of measurements starts with a stroke of 200 mm and at least 4 cycles, a speed of 500
and 1000 mm min−1 and different pressure levels between 0 and 10 bar, Figure 6.27, 6.28
and 6.29. A manometer before the inlet measures the gas pressure behind the pressure
control valve.
200
150
friction force (N)
100 0 bar
2 bar
50 4 bar
5 bar
0 6 bar
-50 8 bar
9 bar
-100 10 bar
-150
-200
0 50 100 150 200
rod displacement (mm)
Figure 6.27. Frictional force measurement with a BCD ring (under given load) on the
steel rod at a relative velocity of 500 mm min−1 . Each pressure level was
tested with 4 cycles.
156
200
frictional force (N) 150
100
50 0 bar
2.5 bar
0 5 bar
7.5 bar
-50 10 bar
-100
-150
-200
0 50 100 150 200
Figure 6.28. Friction force measurement with a BCD ring on the steel rod at a relative
velocity of 1000 mm min−1 . For each pressure level 4 cycles were run.
200
150
100
friction force (N)
0 bar
50
2 bar
0 4 bar
6 bar
-50 8 bar
10 bar
-100
-150
-200
0 50 100 150 200
Figure 6.29. Friction force measurement with a BCD ring on the WC rod at a relative
velocity of 500 mm min−1 . For each pressure level a minimum of 5 cycles
were run.
157
Table 6.4 presents the test results and Figure 6.30 the resulting frictional forces. The
coefficient of friction is higher on the steel rod than on the WC rod and increases with
relative velocity.
The contact force FC for the ring is calculated from the measured frictional force FF
and the coefficient of friction µ. FC allows to compare the different measurement series,
Figure 6.31. Note that FC is independent of the tribological system, hence different ring
designs can be compared even if the tribological system is unknown. Through this repre-
sentation of the data the effect of the pressure load and the gas pressure distribution in
the dynamic sealing surface already contains all the microscopic effects of any tribological
system.
A linear fit represents all data points very well. This indicates that the gas pressure
distribution is equally present for each measurement series. Any deviation from the fit
results mainly from the manual pressure regulation.
Table 6.4. Contact force FC from a single BCD ring on a 50.8 mm rod, depending on the
gas pressure.
material relative velocity µ pressure FF FC

(mm/min) (bar) (N) (N)
500 0.205 10 92.6 452
500 0.205 9 85.6 418
500 0.205 8 77.7 379
500 0.205 6 60.7 296
500 0.205 5 51.4 251
500 0.205 4 41.3 202
HY54
steel
500 0.205 2 22.9 112

500 0.205 0 5.6 27
1000 0.208 10 97.5 466
1000 0.208 7.5 77.5 371
1000 0.208 5 53.3 255
1000 0.208 2.5 28.8 138
1000 0.208 0 5.9 28
500 0.19 10 87.7 469
WC/Co/Cr
500 0.19 8 71.9 378

HY54
500 0.19 6 55.3 291

500 0.19 4 40.1 211
500 0.19 2 23.7 125
500 0.19 0 5.0 26
158
100
steel rod 500 mm min−1
frictional force (N) 80 WC rod 500 mm min−1
60
40
20
0
0 1 2 3 4 5 6 7 8 9 10
Figure 6.30. Frictional force FF of a single BCD ring on a 50.8 mm rod, depending on the
sealed gas pressure differential.
500
400 WC rod 500 mm min−1
linear fit
contact force (N)
300
200
100
linear fit:
FC = 43.585po + 30.505
0
0 1 2 3 4 5 6 7 8 9 10
Figure 6.31. Contact force FC from a single BCD ring on a 50.8 mm rod, depending on
the gas pressure differential. The linear fit approximates all data points very
well.
159
6.4.1.2 Comparison with calculation
The numerical calculations presented in section 5.5.4 show the effect of the incompressible
or compressible gas pressure distribution in the dynamic sealing surface.
The simulation including the compressible gas pressure distribution in the dynamic sealing
surface is compared to the experimentally determined contact force. Since the effect of
the garter spring is small, it is disregarded in the calculation. For a comparison the
experimental contact force results are reduced by the contact force resulting from the
garter spring. The calculation is in excellent agreement with the experiment, measured
in the 0 bar run.
This allows to compare the results from the numerical simulation with the experiments
performed, see Figure 6.32. The excellent agreement between experiments and calculation
validates the formulation of the compressible gas pressure distribution in the dynamic
sealing surface by Lindner-Silwester [43] and its numerical implementation.
500
400 WC rod 500 mm min−1
numerical simulation
contact force (N)
300
200
100
0
0 1 2 3 4 5 6 7 8 9 10 11
Figure 6.32. Comparison of simulation and experiment: Contact force FC from a single
BCD ring on a 50.8 mm rod, depending on the sealed gas pressure differential.
The experimental data is corrected by the contact force resulting from the
garter spring load.
Now the frictional force of a single BCD on a WC rod with Ø=50.8 mm is calculated
dependent on the sealed pressure differential. The crank end pressure is ambient. The
determined coefficient of friction 0.19 leads to the forces in Figure 6.33.
160
800
700
frictional force (N) 600
500
400
300
200
100
0
0 20 40 60 80 100
Figure 6.33. Frictional force FF for a single BCD ring on a WC rod with Ø=50.8 mm, de-
pending on the sealed gas pressure difference. The gas pressure distribution
is calculated with compressible gas. The coefficient of friction is 0.19.
6.4.2 Wear
The results presented in this section have been published in [35].
A comparison of the wear of a packing ring between calculation and experiment is only
possible if the wear coefficient is known and can be inserted into the calculation. During
the experiment, section 6.3.2 (Test B), two BCD rings where loaded with 29.8 bar for
359 h (15 days). The mean rod speed was 1.19 m s−1 . The maximum temperature during
the test was 94 ◦C, the average temperature 84 ◦C.
In a tribological test (see section 4.1.4, Table 4.3) the coefficient of wear is determined at
80 ◦C giving k = 7.2 · 10−7 mm3 km−1 N−1 , which is used as an input parameter for the
calculation.
Figure 6.26 shows the wear of the ring segments. From these results the total wear
volume of each segment can be computed. The up-scaled wear data for the segments is
again given in Table 6.5. These measurements are in good agreement with the simulation
results corresponding to the same wear time of 15 days. The calculation yields a wear
volume of 466 mm3 for the sealing segment and 92 mm3 for the cap segment. These values
are between 13 and 21 % higher than the measured ones. This deviation shows that the
wear coefficient used in the calculation was slightly too high.
In experiments and in calculations the wear pattern does not significantly change in axial
direction, see e.g. Figure 6.26 and Figure 5.19, respectively. Figure 6.34 from [35] gives
161
a wear comparison at the nodal axial positions 2.08 and 6.44 mm from the cup face (as
explained in Figure 5.20). Calculation and measurement are in very good agreement,
which confirms the predictive power of the developed simulation model. It is important
to note that the coefficient of wear k has been determined in an independent experiment
and the calculation is free of adjustment or fitting parameters.
Table 6.5. Measured (and up-scaled) wear volumes for the segments of a BCD ring after
a 359 h test and the calculation result.
segment type position total wear volume

experiment calculation
(◦ ) (mm3 ) (mm3 )
seal 0 368 466
cap “P” 90 80 92
seal “P” 180 398 466
cap 270 76 92
all 921 1118
162
Figure 6.34. Calculated and measured wear for a BCD ring [35]. Top: Sketch of the ring’s
dynamic sealing surface and the paths A and B at which the comparison is
made. Middle: Path A, representing the cap segment of the BCD ring.
Bottom: Path B, representing the sealing section of the sealing segment of
the ring.
163
Chapter 7
Conclusion and Outlook
The presented calculation model for packing ring wear bridges for the first time the gap
between laboratory testing and real-world applications.
The calculation procedure is able to handle the wear of packing rings up to several mil-
limetres and gives insight into the wear process. It further allows to monitor quantities,
which are hardly or not at all measurable, e.g. stresses, strains, contact pressures, the
wear pattern and the evolution of these quantities.
The numerical implementation was validated against an analytical model, derived for a
simplified ring configuration. The developed theory, the assumptions and the model of
the ring wear were successfully validated against measurements performed on purpose-
developed test equipment. All material parameters were obtained from independent tests.
In none of the comparisons an adjustment parameter is used. The agreement between the
frictional-force measurements and the calculation is excellent. The agreement between
the wear tests and the calculation is good, qualitatively and quantitatively. This shows
the predictive power of the simulation and opens windows of opportunity to gain a sound
understanding of packing ring wear.
This calculation procedure can now be used i) to benchmark ring designs without the un-
predictability of any tribologic system, setting the wear coefficient and all other boundary
conditions equally. ii) to investigate how symmetry breaking effects like ring segments
sticking together or rod-cup misalignments influence the ring wear. iii) to investigate how
stopping and restarting a compressor effects the packing rings. To this end a material
model of higher complexity will be needed. iv) as part of the simulation of a whole packing
case with several interacting rings. Here further calculations have to be implemented.
This work lays the basis for future computer-aided ring designs improving the efficiency
of compressors.
164
Appendix
165
HY54
The properties of HY54 provide a very versatile material grade that has been successfully applied
in a multitude of applications, with gas conditions from wet through to dry, in both lubricated and
non-lubricated service. This material has shown itself to be a broad range problem solving material
where standard carbon filled PTFE materials do not give desired lifetime. Successful experience in
compressor applications covers air, methane, ethylene, propylene, helium, hydrogen, and nitrogen Air
and other mixed gas process streams.
Industrial Gases
Physical Properties
Property Method Value

Natural Gas
COTE - Radial x 10-6/C (20-200 °C) ASTM D696 73
COTE - Axial x 10-6/C (20-200 °C) ASTM D696 107

Refinery
Maximum operating temperature (°C) - 200
Density (g/cm3) ASTM D792-00 2.07 ± 0.05

Olefins
Shore D Hardness ASTM D2240-04 64 ± 2
Tensile strength at break (MPa) ASTM D638-03 15.2 ± 0.9 Alcohols
Elongation at break (%) ASTM D638-03 86.4 ± 0.5
Chemicals
Operating range
Max. Gas Temperature (°C) Max. Pressure (bar)

Refrigeration
Packing Discharge Cylinder Ring Diff.

Discharge Design
Non-Lube Lube Non-Lube Lube
200 150 100 175 50 100
Operating restriction for oxygen-service: Compression ratio up to 3 and max. temperature 225°C
© HOERBIGER 2015
Issue 2015 04
All values are approximate and subject to change without notification.

The maximum material design temperature is calculated by considering suction and discharge conditions,
machine speed, cooling and loading. Typically: Tdesign = Tsuction + 2/3(Tdischarge – Tsuction).
Additional operating conditions need to be considered when making material selections.
The data presented are guidelines only; consult HOERBIGER to ensure the correct material is specified.
Table 7.1. Detailed wear and wear volume data for ring A1.
variable value unit
area data
wear volume data

radial wear data

167
Table 7.2. Detailed wear and wear volume data for ring B1.
variable value unit
area data
wear volume data

radial wear data

168
Table 7.3. Detailed wear and wear volume data for ring B2.
variable value unit
area data
wear volume data

radial wear data

169
Developed code
Start calculation
This file contains all necessary input parameters to call the (Main BCD V04.py) backbone
file, which starts a wear calculation including all sub calculations and scripts.
1 #!/usr/bin/python
2 import os
3 os.system("python -u Main_BCD_V04.py \
4 --inp1 ID2Z_BCD_V11 \
5 --user1 UMM3D_BCD_V03_user1 \
6 --startcounter 0 \
7 --counter 51 \
8 --jobmain Main_2Z_BCD_V11 \
9 --jobsub Sub_2Z_BCD_V11_therm \
10 --inpmain Main_2Z_BCD_V11 \
11 --inpsub Sub_2Z_BCD_V11 \
12 --usermain UMM3D_BCD_V05.f \
13 --inpgenmain Main_Input_Generator_BCD_V04 \
14 --inpgensub Sub_Input_Generator_BCD_V05 \
15 --tempfilename Knoten_Temp \
16 --contactNodeSetName INNEN \
17 --wearElementSetName INNEN_ADA_MESH \
18 --neighborNumber 6 \
19 --inc 10 \
20 --highpressuretemp 6.1 \
21 --lowpressuretemp 0.1 \
22 --modelType 3D \
23 --abaquspath /opt/abaqus/Commands/abq6142 \
24 --cpus 1 \
25 | tee 0_LOG.log")
26 # The option -u in conjunction with a Python file makes Python work
,→ without buffering!
170
Backbone of the calculation
This file (Main BCD V04.py) starts the initial loading, then the neighbour nodes are
searched and the first gas pressure distribution in the dynamic sealing surface is calculated
before starting the wear calculation. The wear calculation and the pressure calculation
alternate until the ring is worn away or a maximum number of wear calculations (wear
time) is reached.
1 import sys
2 import os
3 import argparse # to parse arguments
4 from datetime import datetime
5 os.system('python -V')
6 start_time = datetime.now()
7 print('Starting time = '+str(start_time))
8 # Class containing all Parameters which are parsed
9 class Parameters:
10 pass
11 parameter = Parameters() # instance of the class
,→ Parameters
12 # Parsing arguments
13 parser = argparse.ArgumentParser(description='Coupled Abaqus Jobs
,→ including UserDefinedSubroutines.')
14 # Adding all necessary and possible arguments to the parser.
15 parser.add_argument('--inp1', type=str, required=True, help='Name of
,→ the first input file, providing the geometry (including sets) and
,→ a first (dummy) step.')
16 parser.add_argument('--user1', type=str, required=False, help='Name of
,→ the first Fortran file, containing the UserDefinedSubroutines for
,→ the first(dummy/initialization) calculation.')
17 parser.add_argument('--counter', type=int, required=True,
,→ help='Counter counts the number of fully run cycles.')
18 parser.add_argument('--startcounter',type=int,required=False,
,→ help='This is the first value of the counter, so an analysis can
,→ be restarted at any step.')
19 parser.add_argument('--jobmain', type=str, required=True, help='Name
,→ of the main job containing the global model. This model is used
,→ repeatedly.')
20 parser.add_argument('--jobsub', type=str, required=True, help='Name
,→ of the sub job containing the local model. This model is used
,→ repeatedly.')
21 parser.add_argument('--inpmain', type=str, required=True, help='Name
,→ of the input file for the main job.')
171
22 parser.add_argument('--inpsub', type=str, required=True, help='Name
,→ of the input file for the sub model job.')
23 parser.add_argument('--usermain', type=str, required=True, help='Name
,→ of the Fortran file, containing the UserDefinedSubroutines for the
,→ main model.')
24 parser.add_argument('--usersub', type=str, required=False, help='Name
,→ sub model.')
25 parser.add_argument('--inpgenmain', type=str, required=True,
,→ help='Name of the input file generator for the main model.')
26 parser.add_argument('--inpgensub', type=str, required=True, help='Name
,→ of the input file generator for the sub model.')
27 parser.add_argument('--tempfilename', type=str,required=True,
,→ help='Name of the *.txt file where all temperature values for all
,→ nodes are stored.')
28 parser.add_argument('--contactNodeSetName', type=str, required=True,
,→ help='Name of the node set containing all contact nodes (in
,→ CAPITAL LETTERS). Will be passed to find_node_neighbors.py.')
29 parser.add_argument('--wearElementSetName', type=str, required=True,
,→ help='Name of the element set which is affected by mesh smoothing
,→ via UMeshMotion later and used by find_node_neighbors.py.')
30 parser.add_argument('--neighborNumber', type=int, required=True,
,→ help='Number of the neighbors which will be searched behind each
,→ contact node and used for the mesh smoothing algorithm.')
31 parser.add_argument('--highpressuretemp', type=float, required=True,
,→ help='Pressure of the cylinder side. This is the higher pressure.
,→ The maximum is 10 MPa (100bar). This maximum was agreed upon on
,→ 13.08.2015. This value is also the higher temperature in the sub
,→ calculation/model. The ring is loaded with 0 pressure on the crank
,→ side. On the cylinder side the load is
,→ highpressuretemp-lowpressuretemp. This gives the advantage that
,→ the pressure on the crank side mus t not be specified and applied
,→ on changing element surfaces.')
32 parser.add_argument('--lowpressuretemp', type=float, required=True,
,→ help='Pressure of the crank side. This is the lower pressure. The
,→ minimum is 0.1 MPa (1bar). This value is the lower temperature in
,→ the sub calculation/model. The ring is loaded with 0 pressure on
,→ the crank side. On the cylinder side the load is
,→ highpressuretemp-lowpressuretemp. This gives the advantage that
,→ the pressure on the crank side mus t not be specified and applied
,→ on changing element surfaces.')
172
33 parser.add_argument('--modelType', type=str, required=True, help='Two
,→ model types are accepted: 3D and AX. 3D means a 3D Simulation with
,→ the z-axis being the length axis of the rod, AX means axisymmetric
,→ simulation, where x is the radial direction, y in rod direction.')
34 parser.add_argument('--abaquspath', type=str, required=True,
,→ help='Name of the abaqus executable.')
35 parser.add_argument('--cpus', type=int, required=True, help='Number of
,→ cpus which shall be used for the Abaqus calculation.')
36 parser.add_argument('--inc', type=float,required=True, help='This is
,→ the increment which is maximaly used when calculating the wear.')
37 args = parser.parse_args(namespace=parameter)
38 # Start of plausibility checks
39 if parameter.user1 == None:
40 print('No user subroutine was given for the initialization. If a
,→ subroutine for the main part of the analysis was given, this
,→ subroutine is used instead.')
41 if parameter.usermain != None:
42 print('The usual subroutine is used.')
43 parameter.user1 = parameter.usermain
44 print(parameter.user1)
45
46 if parameter.startcounter == None or parameter.startcounter == 0:

47 print('No startcounter or startcounter = 0 was given. Analysis
,→ starts at time zero with the initialization.')
48 if parameter.startcounter > parameter.counter:
49 sys.exit('\n\n\n ERROR - The startcounter is greater or equal than
,→ the counter.\nReconsider the input values. \n\n')
50
51 if parameter.highpressuretemp > 10:

52 sys.exit('\n\n\n ERROR - The given pressure value is greater than
,→ 10 MPa (100 bar). This is out of the application range. \n
,→ Reconsider the given input. \n\n')
53 elif parameter.highpressuretemp < 0:
54 sys.exit('\n\n\n ERROR - The given pressure value is smaller than
,→ 0 MPa (0 bar). This is out of the application range. \n
55
56 if parameter.lowpressuretemp < 0.1:

57 sys.exit('\n\n\n ERROR - The given pressure value is smaller than
,→ 0.1 MPa (1 bar). This is out of the application range. \n
58 elif parameter.lowpressuretemp > parameter.highpressuretemp:
173
59 sys.exit('\n\n\n ERROR - The given (low) pressure value is bigger
,→ than the (high) pressure value. \n Reconsider the given input.
,→ \n\n')
60
61 # End of plausibility checks

62
63 print('Start of the Coupled Analysis\n')

64 path_working_dir = os.path.dirname(os.path.abspath(__file__))
65 # The directory of the current __file__ is cut down to the directory
,→ without the filename.
66 print('The current working direcotry is: '+str(path_working_dir))
67 if parameter.startcounter == 0:
68 print('The Coupled Analysis begins with the initialization.')
69 parameter.startcounter = 1
70 # Start the first Abaqus job:
71 print('Start the 1st Abaqus job:\nName of the
,→ job:\t\t'+str(parameter.jobmain)+'\nName of the input
,→ file:\t\t'+str(parameter.inp1)+'.inp\n\n')
72 if 0 != os.system(parameter.abaquspath+"
,→ job="+str(parameter.jobmain)+"_0"+"
,→ inp="+str(parameter.inp1)+".inp"+"
,→ user="+str(parameter.user1)+" cpus="+str(parameter.cpus)+"
,→ interactive | tee "+str(parameter.jobmain)+"_0.log"):
73 sys.exit("Error at Calculation _0")
74 print("\n1st Abaqus job finished. \n")
75 if 0 != os.system(parameter.abaquspath+' python -u get_set_info.py
,→ --path '+str(path_working_dir)+' --jobname
,→ '+str(parameter.jobmain)+' --neighborNumber
,→ '+str(parameter.neighborNumber)):
76 sys.exit('Error during get_set_info.py.')
77 if False ==
,→ os.path.isfile('neighbor_nodes_'+parameter.jobmain+'.txt'):
78 # parameter.neighborNumber +1 because one more node is
,→ needed to give the wear direction for the edge nodes
79 if 0 != os.system(parameter.abaquspath+' python -u
,→ find_node_neighbors.py --path '+str(path_working_dir)+'
,→ --jobname '+str(parameter.jobmain)+' --contactNodeSetName
,→ '+str(parameter.contactNodeSetName)+' --wearElementSetName
,→ '+parameter.wearElementSetName+' --neighborNumber
,→ '+str(parameter.neighborNumber+1)+' --modelType
,→ '+str(parameter.modelType)):
80 sys.exit('Error during find_node_neighbors.py.')
81 else:
174
82 print('\nThe neighbor node file exists, the search is
,→ skipped.\n\n')
83 else:
84 print('This is a restart analysis. The calculation loop starts at
,→ loop number '+str(parameter.startcounter)+'.')
85 i = parameter.startcounter
86
87 print('Starting the calculation loop:\n')

88 # Loop to gain a solution for the coupled 'temperature' - wear
,→ problem:
89 for i in range(parameter.startcounter,parameter.counter,1):
90 print('\nLoop number '+str(i)+' :\nGeneration of the input file
,→ for the sub model.')
91 if 0 != os.system(parameter.abaquspath+' python
,→ '+str(parameter.inpgensub)+'.py'+' --jobmain
,→ '+str(parameter.jobmain)+' --jobsub '+str(parameter.jobsub)+'
,→ --inpsub '+str(parameter.inpsub)+' --counter '+str(i-1)+'
,→ --path '+str(path_working_dir)+' --inpgensub
,→ '+str(parameter.inpgensub)+' --highpressuretemp
,→ '+str(parameter.highpressuretemp)+' --lowpressuretemp
,→ '+str(parameter.lowpressuretemp)):
92 sys.exit('Error in generation of the Submodel in loop number
,→ '+str(i)+'.')
93 print('\nDone. Continuing with the sub model Abaqus job.\n')
94 if 0 != os.system(parameter.abaquspath+'
,→ job='+str(parameter.jobsub)+'_'+str(i)+'
,→ inp='+str(parameter.inpsub)+'_'+str(i-1)+' interactive | tee
,→ '+str(parameter.jobsub)+'_'+str(i)+'.log'):
95 sys.exit('Error in calculation of the Submodel in loop number
,→ '+str(i)+'.')
96 print('Done. Continuing with the generation of the input file for
,→ the main model.\n')
97 if 0 != os.system(parameter.abaquspath+' python
,→ '+str(parameter.inpgenmain)+'.py'+' --jobmain
,→ '+str(parameter.jobmain)+' --jobsub '+str(parameter.jobsub)+'
,→ --inpmain '+str(parameter.inpmain)+' --tempfilename
,→ '+str(parameter.tempfilename)+' --counter '+str(i)+' --path
,→ '+str(path_working_dir)+' --inpgenmain
,→ '+str(parameter.inpgenmain)+' --wearElementSetName
,→ '+str(parameter.wearElementSetName)+' --inc
,→ '+str(parameter.inc)+' --highpressuretemp
,→ '+str(parameter.highpressuretemp)+' --lowpressuretemp
,→ '+str(parameter.lowpressuretemp)):
175
98 sys.exit('Error in generation of the Mainmodel in loop number
,→ '+str(i)+'.')
99 if 0 != os.system(parameter.abaquspath+'
,→ job='+str(parameter.jobmain)+'_'+str(i)+'
,→ inp='+str(parameter.inpmain)+'_'+str(i)+'
,→ user='+str(parameter.usermain)+'
,→ oldjob='+str(parameter.jobmain)+'_'+str(i-1)+' interactive |
,→ tee '+str(parameter.jobmain)+'_'+str(i)+'.log'):
100 sys.exit('Error in calculation of the Mainmodel in loop number
,→ '+str(i)+'.')
101 print('Done. Loop number '+str(i)+' finished. Current time:
,→ '+str(datetime.now())+'\n\n')
102
103 end_time = datetime.now()

104 print('Analysis done. Calculation finished. Current date and time:
,→ '+str(end_time))
105 delta_time = end_time-start_time
106 print('The analysis finished in '+str(delta_time))
Initial FORTRAN subroutine
The definition of UFILD is necessary already in the initial loading of the structure, only
then UFIELD can be used in the following (restart-) wear calculations.
1 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
2 CCCCC
3 CCCCC UFIELD - Userdefined FIELD Variables
4 CCCCC
6
7 SUBROUTINE UFIELD(FIELD,KFIELD,NSECPT,KSTEP,KINC,TIME,NODE,
8 1 COORDS,TEMP,DTEMP,NFIELD)
9 C
10 INCLUDE 'ABA_PARAM.INC'
11 C
12 DIMENSION FIELD(NSECPT,NFIELD), TIME(2), COORDS(3),
13 1 TEMP(NSECPT), DTEMP(NSECPT)
14 C
15 FIELD = 0.0
16 RETURN
17 END
176
FORTRAN subroutines to calculate wear
Below are all necessary FORTRAN subroutines, which allow to calculate the wear of a
packing ring, based on the mechanisms governing this process.
2 CCCCC
3 CCCCC MODUL is used to share data between
4 CCCCC different user-defined subroutines
5 CCCCC
7
8 MODULE Information
9
10 CHARACTER*256 JOBNAME
11
12 CHARACTER*80, DIMENSION(:), allocatable ::

13 * node_set_names, element_set_names
14
15 INTEGER neighbor_node_nr, number_of_elements,

16 * number_of_node_sets, number_of_nodes,
17 * number_of_element_sets, num_contact_nodes,
18 * num_edge_nodes, LENJOBNAME
19
20 INTEGER, DIMENSION(:), allocatable ::

21 * node_set_count, i_element_set_count,
22 * nodes_temp_array
23
24 INTEGER, DIMENSION(:,:), allocatable ::

25 * N2EConn, nodes_in_contact, neighbors
26
27 DOUBLE PRECISION, DIMENSION(:), allocatable ::

28 * temp_array
29
30 DOUBLE PRECISION, DIMENSION(:,:), allocatable ::

31 * coords_nodes_in_contact, uvarm_neighbor_nodes,
32 * coords_allnodes, disp_neighbors
33
34 DOUBLE PRECISION, DIMENSION(:,:,:), allocatable ::

35 * applied_disp
36
37 C Information:
38 C N2EConn = Array that gives the all elements and
39 C their connected nodes
177
40 C neighbor_node_nr = number of neighbors of the contact node
41 C determined by the python script
42 C number_of_elements = number of all elements the model contains
43 C number_of_nodes = number of all nodes the model contains
44 C number_of_element_sets
45 C number_of_node_sets
46 C num_contact_nodes = number of nodes in contact
47 C between rod and ring (cup not considered)
48 C num_edge_nodes = number of nodes placed on an edge
49
50 C coords_allnodes = containing the coordinates of all nodes

51 C (:,1) - x-coords
52 C (:,2) - y-coords
53 C (:,3) - z-coords
54 C JOBNAME = name of the job, read from parameter file
55 C LENJOBNAME = length in characters of the JOBNAME
56
57 C Following 2 arrays have the same index for 1 data-pair

58
59 C nodes_temp_array = containing the node numbers of the nodes

60 C which have temperature values in temp_array
61 C temp_array = containing all the temperature values
62 C for the nodes in the nodes_temp_array
63
64 C node_set_names = Array containing all node set names

65 C node_set_count = number of nodes in a set
66
67 C element_set_names = Array containing all element set names

68 C i_element_set_count = number of elements in a set
69
70 C nodes_in_contact = array containing all nodes which are in

71 C contact with the rod (the nodes in contact
72 C with the cup are not included here ! )
73 C the node numbers are at index 1; index 2
74 C marks if the node is on the boarder of a
75 C surface (1=edge)or in the surface (=0)
76 C
77 C coords_nodes_in_contact (:,1) = X
78 C (:,2) = Y
79 C (:,3) = Z
80 C neighbors = array containing all neighbor nodes (:,1)
81 C are the contact nodes and then (:,2) too
82 C (:,num_neigh) neighbors which are aligned
83 C radially - calculated by a python script
178
84 C and read by fortran
85 C disp_neighbors = displacement of the neighbors, in radial
86 C direction (:,2 -> neighbor_node_nr)
87 C applied_disp = displacement of the neighbors
88 C (:,2 -> neighbor_node_nr; :); this variable
89 C is passed to UEXTERNALDB and then to UVARM
90 C for display of the displacment; the third
91 C index is for the x, y and z amount of the
92 C wear (z is 0)
93 C uvarm_neighbor_nodes = array containing the sum of all the
94 C displacement which a node (neighbor or
95 C contact) experiences over time
96 C
97 save
98 END MODULE Information
99
101 CCCCC
102 CCCCC Subroutine UEXTERNALDB allocate and initialize all
103 CCCCC global variables
104 CCCCC
106
107 SUBROUTINE UEXTERNALDB(LOP, LRESTART, TIME, DTIME, KSTEP, KINC)

108 C Initialization is called only once and at the first run
109 USE Information
111
112 C LOP = 0 the subroutine is called at the start of the analysis.

113 C LOP = 1 the subroutine is called at the start of the current
114 C analysis increment. The subroutine can be called
115 C multiple times at the beginning of an analysis
116 C increment if the increment fails to converge and a
117 C smaller time increment is required.
118 C LOP = 2 the subroutine is called at the end of the current
119 C analysis increment. When LOP=2, all information that you
120 C need to restart the analysis should be written to
121 C external files.
122 C LOP = 3 the subroutine is called at the end of the analysis.
123 C LOP = 4 the subroutine is called at the beginning of a restart
124 C analysis. When LOP=4, all necessary external files
125 C should be opened and properly positioned and all
126 C information required for the restart should be read
127 C from the external files.
128
179
129 DIMENSION TIME(2)
130 CHARACTER*256 zeile, OUTDIR, filename
131 CHARACTER*18 total_time
132 CHARACTER zahl*5
133 C local variables:
134 INTEGER node_number, merker, nr_of_cols, problem
135 C problem ... an integer error code returned by OPEN
136 INTEGER, DIMENSION (:,:), allocatable ::
137 * N2E_mixed
138 DOUBLE PRECISION temperature, coordinates(3)
139 C ex ... if file exists = .TRUE.
140 LOGICAL ex
141
142 WRITE(*,*) 'starte UEXTERNALDB mit LOP = ',LOP

143 WRITE(*,*) 'LRESTART = ',LRESTART,' TIME = ',TIME
144 WRITE(*,*) 'DTIME = ',DTIME,' KSTEP = ',KSTEP,' KINC =',KINC
145
146 IF ((LOP.EQ.4).AND.(KSTEP.EQ.0).AND.(KINC.EQ.0)) THEN

147 CALL GETOUTDIR(OUTDIR,LENOUTDIR)
148 WRITE(*,*) 'current working directory:', TRIM(OUTDIR)
149 WRITE(*,*) '======= INITIALISIERUNG PHASE ========'
150
151 node_number = 0
152 merker = 0
153 nr_of_cols = 0
154 temperature = 0.0
155 coordinates = 0.0
156
157 C Opening of the parameter file - this file includes all parameters
158 C which are needed for the FORTRAN code, the variables are saved to the
159 C module "Information".
160 filename(1:LENOUTDIR) = OUTDIR(1:LENOUTDIR)
161 filename((LENOUTDIR+1):) ='/parameter.txt'
162 OPEN(150,FILE=filename,STATUS='OLD')
163 DO i = 1,4,1
164 READ(150,*) zeile
165 WRITE(*,*) 'read has read the line: ', zeile
166 IF (INDEX(zeile,'JOBNAME').EQ.1) THEN
167 READ(150,*) JOBNAME
168 READ(150,*) LENJOBNAME
169 WRITE(*,*) 'jobname = ',TRIM(JOBNAME), ' CHAR: ', LENJOBNAME
170 ELSEIF (INDEX(zeile,'NODE_SETS').EQ.1) THEN
171 READ(150,*) number_of_node_sets
172 WRITE(*,*) 'number of node sets ', number_of_node_sets
173 ELSEIF (INDEX(zeile,'ELEMENT_SETS').EQ.1) THEN
180
174 READ(150,*) number_of_element_sets
175 WRITE(*,*) 'number of element sets ', number_of_element_sets
176 ELSEIF (INDEX(zeile,'NEIGHBOR_NODE_NUMBER').EQ.1) THEN
177 READ(150,*) neighbor_node_nr
178 WRITE(*,*) 'neighbor nodes = ', neighbor_node_nr
179 ENDIF
180 ENDDO
181 CLOSE(150)
182 filename = ''
183 ALLOCATE(node_set_names(number_of_node_sets))
184 ALLOCATE(node_set_count(number_of_node_sets))
185 ALLOCATE(element_set_names(number_of_element_sets))
186 ALLOCATE(i_element_set_count(number_of_element_sets))
187 node_set_names(1:number_of_node_sets) = ''
188 node_set_count(1:number_of_node_sets) = 0
189 element_set_names(1:number_of_element_sets) = ''
190 i_element_set_count(1:number_of_element_sets) = 0
191
192 C Reading the set names

193 C *** NODE SETS ***
195 filename((LENOUTDIR+1):) ='/set_names_'//TRIM(JOBNAME)//'.txt'
198 IF (INDEX(zeile,'NODE_SET_NAMES').EQ.1) THEN
199 DO i = 1,number_of_node_sets,1
200 READ(151,*) node_set_names(i)
201 READ(151,*) node_set_count(i)
202 IF ((INDEX(node_set_names(i),'INNEN').EQ.1) .AND.
203 * (INDEX(node_set_names(i),'INNEN_ADA_MESH').NE.1)) THEN
204 num_contact_nodes = node_set_count(i)
205 WRITE(*,*) 'Number of contact nodes',num_contact_nodes
206 ELSEIF (INDEX(node_set_names(i),'ALLNODES').EQ.1) THEN
207 number_of_nodes = node_set_count(i)
208 ALLOCATE(coords_allnodes(number_of_nodes,3))
209 coords_allnodes(1:number_of_nodes,1:3) = 0.0
210 WRITE(*,*) 'Total number of nodes:',number_of_nodes
211 ELSEIF (INDEX(node_set_names(i),'KANTE').EQ.1) THEN
212 num_edge_nodes = node_set_count(i)
213 WRITE(*,*) 'Number of edge nodes', num_edge_nodes
214 ENDIF
215 ENDDO
216 ELSE
217 WRITE(*,*) 'ERROR - LOP 4 USXTERNALDB - reading set names'
218 ENDIF
181
219 C *** ELEMENT SETS ***
221 IF (INDEX(zeile,'ELEMENT_SET_NAMES').EQ.1) THEN
222 DO i = 1,number_of_element_sets,1
223 READ(151,*) element_set_names(i)
224 READ(151,*) i_element_set_count(i)
225 IF (INDEX(element_set_names(i),'EL_ALL').EQ.1) THEN
226 number_of_elements = i_element_set_count(i)
227 ENDIF
228 ENDDO
229 ELSE
230 WRITE(*,*) 'ERROR - LOP 4 USXTERNALDB - reading element names'
231 ENDIF
232 CLOSE(151)
233 C Reading the neighbors nodes
234 C allocating neighbors: neighbor node nr +1 (contact nodes)
235 C +1 (neighbors to give direction for the nodes on an edge)
236 filename = ''
237 ALLOCATE(neighbors(num_contact_nodes,neighbor_node_nr+2))
238 ALLOCATE(disp_neighbors(num_contact_nodes,neighbor_node_nr+1))
239 ALLOCATE(applied_disp(num_contact_nodes,neighbor_node_nr+1,3))
240 neighbors(1:num_contact_nodes,1:neighbor_node_nr+1) = 0
241 disp_neighbors(1:num_contact_nodes,neighbor_node_nr+1) = 0.0
242 applied_disp(1:num_contact_nodes,1:neighbor_node_nr+1,3) = 0.0
243 C Reading contact nodes and their neighbors from *.txt file,
245 filename((LENOUTDIR+1):)='/neighbor_nodes_'//TRIM(JOBNAME)//'.txt'
247 DO i = 1,num_contact_nodes,1
248 READ(152,*) neighbors(i,1:neighbor_node_nr+2)
249 C WRITE(*,*) 'Reading neighbors: ', neighbors(i,:)
250 ENDDO
251 C START Sanity check: the neighbor array must be free of zeros
253 C WRITE(*,*) neighbors(i,:)
254 DO j = 1,neighbor_node_nr+2,1
255 IF (neighbors(i,j).EQ.0) THEN
256 WRITE(*,*) 'neighbors(',i,',',j,')=',neighbors(i,j)
257 ENDIF
258 ENDDO
259 ENDDO
260 C END Sanity check
261 WRITE(*,*) 'Neighbors read.'
262 filename = ''
263 CLOSE(152)
182
264 C Reading temperature of contact nodes
266 filename((LENOUTDIR+1):) ='/Knoten_Temp.txt'
267 ALLOCATE(temp_array(num_contact_nodes))
268 ALLOCATE(nodes_temp_array(num_contact_nodes))
269 temp_array(1:num_contact_nodes) = 0.0
270 nodes_temp_array(1:num_contact_nodes) = 0
273 READ(153,*) node_number, temperature
274 nodes_temp_array(i) = node_number
275 temp_array(i) = temperature
276 C WRITE(*,*) 'Reading temp-node file: ',
277 C * node_number, temperature
278 ENDDO
279 WRITE(*,*) 'Temp-file read.'
280 node_number = 0
281 i = 0
282 filename = ''
283 CLOSE(153)
284 C Reading contact nodes and their coordinates
285 ALLOCATE(nodes_in_contact(num_contact_nodes,2))
286 ALLOCATE(coords_nodes_in_contact(num_contact_nodes,3))
287 nodes_in_contact(1:num_contact_nodes,2) = 0
288 coords_nodes_in_contact(1:num_contact_nodes,3) = 0.0
290 filename((LENOUTDIR+1):) ='/node_INNEN.txt'
293 READ(154,*) node_number, coordinates(:)
294 IF (nodes_temp_array(i).EQ.node_number) THEN
295 nodes_in_contact(i,1) = node_number
296 nodes_in_contact(i,2) = 0
297 coords_nodes_in_contact(i,1:3) = coordinates(:)
298 C WRITE(*,*) 'nr: ',i,' node: ',
299 C * node_number,coords_nodes_in_contact(i,1:3)
300 ELSE
301 WRITE(*,*)'ERROR: nodes_in_contact and nodes_temp_array do',
302 *'not contain the same sequence of nodes! - Further calculations',
303 *' may be erroneous!'
304 ENDIF
305 ENDDO
306 WRITE(*,*) 'Contact nodes ring-rod read.'
307 filename = ''
308 CLOSE(154)
183
309
310 C Adding Information if tcontact node is an edge node or within a

311 C surface
313 filename((LENOUTDIR+1):) ='/node_KANTEN.txt'
316 i = 1
317 j = 1
318 DO WHILE (i.LE.num_edge_nodes)
319 C WRITE(*,*) nodes_in_contact(j,:)
320 IF (nodes_in_contact(j,1).EQ.node_number) THEN
321 nodes_in_contact(j,2) = 1
322 C WRITE(*,*) nodes_in_contact(j,:)
323 i = i + 1
324 j = j - 1
325 IF (i.LE.num_edge_nodes) THEN
327 ENDIF
328 ENDIF
329 j = j + 1
330 ENDDO
331 filename = ''
332 CLOSE(155)
333
334 C Reading all element numbers with their connectivities

336 filename((LENOUTDIR+1):) ='/element_EL_ALL.txt'
338 merker = 0
339 DO i = 1, number_of_element_sets, 1
340 WRITE(*,*) 'set_name = ', element_set_names(i)
341 IF (INDEX(element_set_names(i),'EL_ALL').EQ.1) THEN
342 merker = i
343 WRITE(*,*)'EL_ALL has',i_element_set_count(merker),'Element.'
344 ALLOCATE(N2EConn(i_element_set_count(merker),9))
345 ALLOCATE(N2E_mixed(i_element_set_count(merker),9))
346 N2EConn(1:i_element_set_count(merker),1:9) = 0
347 N2E_mixed(1:i_element_set_count(merker),1:9) = 0
348 ENDIF
349 ENDDO
350 IF (merker.EQ.0) THEN
351 WRITE(*,*) 'ERROR - EL_ALL NICHT GEFUNDEN'
352 ELSE
353 DO i = 1,i_element_set_count(merker), 1
184
354 READ(155,*) N2E_mixed(i,:)
355 ENDDO
356 DO i = 1,i_element_set_count(merker), 1
357 N2EConn(N2E_mixed(i,1),:)=N2E_mixed(i,:)
358 ENDDO
359 deallocate(N2E_mixed)
360 ENDIF
361 WRITE(*,*) 'All elements read.'
362 filename = ''
363 CLOSE(155)
364
365 C Reading the UVARM1 file

367 filename((LENOUTDIR+1):) ='/UVARM1.txt'
369 ALLOCATE(uvarm_neighbor_nodes(num_contact_nodes,
370 * neighbor_node_nr+1))
371 uvarm_neighbor_nodes(1:num_contact_nodes,
372 * 1:neighbor_node_nr+1) = 0.0
373 DO i = 1, num_contact_nodes, 1
374 READ(156,*) uvarm_neighbor_nodes(i,:)
375 ENDDO
376 WRITE(*,*) 'UVARM read.'
377 filename = ''
378 CLOSE(156)
379
380 WRITE(*,*) '====== INITIALISIERUNG ENDE =========='

381 ENDIF ! LOP = 4 END OF INITIALIZING
382
383 C => always update the uvarm!

384 IF ((LOP.EQ.1).OR.(LOP.EQ.2).OR.(LOP.EQ.3).OR.(LOP.EQ.4)) THEN
385 WRITE(*,*) 'Updating UVARM.'
387 DO j = 1, neighbor_node_nr+1, 1
388 uvarm_neighbor_nodes(i,j) =
389 1 uvarm_neighbor_nodes(i,j) +
390 2 sqrt(applied_disp(i,j,1)**2 +
391 3 applied_disp(i,j,2)**2 +
392 4 applied_disp(i,j,3)**2)
393 applied_disp(i,j,:) = 0.0
394 ENDDO
395 C WRITE(*,*) neighbors(i,1), uvarm_neighbor_nodes(i,1)
396 ENDDO
397 WRITE(*,*) 'UVARM updated.'
398 ENDIF
185
399
400 C LOP = 2: end of current analysis increment

401 IF (LOP.EQ.2) THEN
402 filename = ''
403 total_time = ''
404 WRITE(*,*) 'Writing UVARM to file.'
405 C Format of total_time and zahl dependent on their value
406 C WRITE(*,*) 'TIME(2)', TIME(2)
407
408 IF (TIME(2).LT.10) THEN

409 WRITE(total_time,'(f7.4)') TIME(2)
410 ELSEIF ((TIME(2).GE.10) .AND.(TIME(2).LT.100)) THEN
424 ENDIF
425
426 C WRITE(*,*) 'total_time >', total_time,'<'

427
428 nr_of_cols = neighbor_node_nr + 1

429
430 IF ((nr_of_cols.GT.1).AND.(nr_of_cols.LT.10)) THEN

431 WRITE(zahl,'(I1)') nr_of_cols
432 ELSEIF((nr_of_cols.GT.9).AND.(nr_of_cols.LT.100))THEN
434 ELSEIF (nr_of_cols.GT.99) THEN
436 ELSEIF((neighbor_node_nr.EQ.0).OR.(neighbor_node_nr.GT.999)) THEN
437 WRITE(*,*) 'ERROR - neighbor_node_nr either=0 or >999'
438 ENDIF
439
440 C WRITE(*,*) 'zahl', zahl

441 C WRITE(*,*) 'TIME(2)', TIME(2)
442 C WRITE(*,*) 'JOBNAME', TRIM(JOBNAME)
443 C WRITE(*,*) 'total_time',TRIM(total_time)
186
444

447 filename((LENOUTDIR+1):)='/UVARM1_'//TRIM(JOBNAME)//'_'//
448 1 TRIM(ADJUSTL(total_time))//'.txt'
449
450 C WRITE(*,*) 'filename: ',TRIM(filename)

451
452 C Check if the filename exists - added 27.2.2017:

453 C (if a 'NEW' file is opened and this file already exists an error
454 C occurs and Abaqus stops, thus the existing file is deleted and
455 C rewritten - this minor inaccuracy should hardly be noticable because
456 C this should only happen if the increment steps is small and the
457 C absolute calculation time is big)
458
459 OPEN(159,FILE=TRIM(filename),STATUS='REPLACE')
461 WRITE(159,"("//TRIM(ADJUSTL(zahl))//"(F14.10,:,', '))")
462 * uvarm_neighbor_nodes(i,:)
463 C WRITE(*,*) uvarm_neighbor_nodes(i,:)
464 ENDDO
465 zahl = ''
466 filename = ''
467 CLOSE(159)
468 WRITE(*,*) 'UVARM written to file.'
469
470 ENDIF ! LOP = 2 END OF CURRENT INCREMENT

471
472 IF (LOP.EQ.3) THEN

473 filename = ''
474 WRITE(*,*) 'LOP 3: Writing UVARM to UVARM1.txt'
477 filename((LENOUTDIR+1):) ='/UVARM1.txt'
478 OPEN(157,FILE=TRIM(filename),STATUS='REPLACE')
479 nr_of_cols = neighbor_node_nr + 1
480 IF ((nr_of_cols.GT.1).AND.(nr_of_cols.LT.10)) THEN
482 ELSEIF((nr_of_cols.GT.9).AND.(nr_of_cols.LT.100))THEN
484 ELSEIF (nr_of_cols.GT.99) THEN
486 ELSEIF((neighbor_node_nr.EQ.0).OR.(neighbor_node_nr.GT.999)) THEN
487 WRITE(*,*) 'ERROR - neighbor_node_nr either=0 or >1000'
488 ENDIF
187
490 WRITE(157,"("//TRIM(ADJUSTL(zahl))//"(F16.12,:,', '))")
491 * uvarm_neighbor_nodes(i,:)
492 ENDDO
493 filename = ''
494 zahl = ''
495 CLOSE(157)
496 WRITE(*,*) 'LOP 3: UVARM to UVARM1.txt written.'
497 ENDIF ! LOP = 3 END OF CURRENT ANALYSIS
498
499
500 RETURN
501 END
502
503 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
504 CCCCC
505 CCCCC Subroutine URDFIL to read from the result file
506 CCCCC or to write into it
507 CCCCC
509
510 SUBROUTINE URDFIL(LSTOP, LOVRWRT, KSTEP, KINC, DTIME, TIME)

511
512 USE Information

514
515 DIMENSION ARRAY(513),JRRAY(NPRECD,513),TIME(2)

516 EQUIVALENCE (ARRAY(1),JRRAY(1,1))
517
518 INTEGER index1

519
520 index1 = 0
521
522 C WRITE(*,*) 'URDFIL - START'

523 C WRITE(*,*)'STEP=',KSTEP,'INC=',KINC,'TIME INC=',DTIME
524 C WRITE(*,*)'STEP TIME=',TIME(1),'TOTAL TIME=',TIME(2)
525
526 LSTOP = 0
527 C No overwriting with lovrwrt = 0, overwriting with = 1
528 LOVRWRT = 0
529
530 CALL POSFIL(KSTEP,KINC,ARRAY,JRCD)

531 DO K1 = 1, 999999
532 JRRAY(1,:) = 0
533 CALL DBFILE(0,ARRAY,JRCD)
188
534 IF (JRCD.NE.0) GO TO 101
535 KEY = JRRAY(1,2)
536 C KEY = 107: CURRENT NODE COORDINATES
537 IF (KEY.EQ.107) THEN
538 C WRITE(*,*) 'Node Nr: JRRAY = ',JRRAY(1,3)
539 C WRITE(*,*) 'Coords : ARRAY = ',ARRAY(4:6)
540 coords_allnodes(JRRAY(1,3),1:3) = ARRAY(4:6)
541 CALL FINDITEMINLIST(nodes_in_contact, num_contact_nodes,
542 * index1, JRRAY(1,3))
543 IF (index1.NE.0) THEN
544 coords_nodes_in_contact(index1,1:3) = ARRAY(4:6)
545 C WRITE(*,*) 'Coords von Node', JRRAY(1,3),
546 C * coords_nodes_in_contact(index1,:)
547 ENDIF
548 index1 = 0
549 ENDIF
550 ENDDO
551
552 101 CONTINUE

553 C WRITE(*,*) 'ENDE URDFIL'
554 C WRITE(*,*) 'neighbors (1:2,1)', neighbors(1:2,1)
555 RETURN
556 END
557
558
560 CCCCC
561 CCCCC Subroutine UMESHMOTION user-defined node movement
562 CCCCC
563 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
564
565 SUBROUTINE UMESHMOTION(UREF,ULOCAL,NODE,NNDOF,

566 * LNODETYPE,ALOCAL,NDIM,TIME,DTIME,PNEWDT,
567 * KSTEP,KINC,KMESHSWEEP,JMATYP,JGVBLOCK,LSMOOTH)
568 USE Information
570 C - Variables UMESHMOTION
571 DIMENSION ULOCAL(NDIM)
572 INTEGER NODE
573 C JELEMLIST(*)
574 DIMENSION ALOCAL(NDIM,*),TIME(2)
575 DIMENSION JMATYP(*),JGVBLOCK(*)
576 C - Variables GETNODETOELEMCONN
577 PARAMETER ( MAXNELEMS = 100)
578 DIMENSION JELEMLIST(MAXNELEMS),JELEMTYPE(MAXNELEMS)
189
579 C - Variables GETVRMAVGATNODE
580 DIMENSION ARRAY(15)
581
582 C ... other Variables

583
584 DOUBLE PRECISION WGLOBAL(NDIM),WLOCAL(NDIM),

585 * cpress, norm_coord(3), radius, radial_wear,
586 * coords_n(3), coords_k(3), betrag, wear_vector(3),
587 * wear_in_norm
588 INTEGER ind1(9), ind2(9), isp(2), izei(2),
589 * i_edge_flag
590
591 C betrag ... Betrag des Vektors von coords_k nach coords_n
592 C coords_k ... Koordinaten des Kontaktknoten (bzw. des Knotens der
593 C naeher an der Kolbenstange liegt)
594 C coords_n ... Koordinaten des Nachbarknoten (bzw. des Knotens der
595 C weiter weg von der Kolbenstange liegt)
596 C cpress ... Contact pressure at the current NODE
597 C i_edge_flag ... Flag if a node is on the edge or within a surface
598 C ind1, ind2 --> Indices returned from FIND Subroutines
599 C isp ... uebergabe der Spalten Nr von (1) bis (2) dann
600 C INDEXFIND ausgefuehrt
601 C izei ... uebergabe der Zeilen Nr von (1) bis (2) an
602 C INDEXFIND
603 C norm_coord ... normierte Koordinaten der Verschleiszrichtung
604 C (1 = x), (2 = y), (3 = z ist
605 C nicht von belangen fuer Knoten in der
606 C Kontaktflaeche, wir jedoch beruecksichtigt bei
607 C Eckknoten)
608 C radius ... benoetigt zum Normieren der Koordinaten
609 C radial_wear ... absoluter Verschleiszbetrag in radialer Richtung
610 C wear_vector ... Verschleiszvektor (Gibt due Richtung des
611 C Verschleiszes an und ist NICHT GLEICH dem
612 C Verscheibungsvektors WGLOBAL. WGLOBAL kann in
613 C eine andere Richtung zeigen, wenn ein Knoten an
614 C einer Kante sitzt.)
615 C wear_in_norm ... Inproduct of wear_vector and norm_coord
616
617 C setzen des Koordinatensystem auf 0 = global, 1 = lokal

618 C Write(*,*) 'entering umeshmotion'
619 LRTN = 0
620 UREF = 1
621 JTYP = 0
622 JRCD = 0
623 NELEMS = MAXNELEMS
190
624 JELEMLIST = 0
625 JELEMTYPE = 0
626 WGLOBAL = 0.0
627
628 ind1(1:9) = 0
629 ind2(1:9) = 0
630 isp(1:2) = 0
631 izei(1:2) = 0
632 C if i_edge_flag is not reset it raises error
633 i_edge_flag = 2
634 betrag = 0.0
635 norm_coord = 0.0
636 cpress = 0.0
637 coords_n = 0.0
638 coords_k = 0.0
639 radius = 0.0
640 radial_wear = 0.0
641 wear_vector = 0.0
642
643 C ACHTUNG: in dieser Subroutine wird nur das im Inputfile definierte

644 C Rechengebiet betrachtet !!! (*ADAPTIVE MESH)
645 C WRITE(*,*) '+++++++++++++++++++++++++++++++++++++++++++++++++'
646 C WRITE(*,*) 'wir sind bei Knoten ', NODE
647 C WRITE(*,*) 'LNODETYPE = ', LNODETYPE
648
649 C WRITE(*,*) 'bevor GETNODE... NELEMS = ',NELEMS

650
651 CALL GETNODETOELEMCONN(NODE, NELEMS, JELEMLIST, JELEMTYPE,

652 * JRCD, JGVBLOCK)
653 C WRITE(*,*) 'get node to element connectivity gives:'
654 C WRITE(*,*) 'node' , NODE
655 C WRITE(*,*) 'NELEMS' , NELEMS
656 C WRITE(*,*) 'jelemlist', JELEMLIST
657 C WRITE(*,*) 'jelemtype', JELEMTYPE(1:NELEMS)
658 C WRITE(*,*) 'jrcd', JRCD
659 C WRITE(*,*) '-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-'
660
661 C if an error pops up in the .msg file which says:

662 C ***ERROR: SYSTEM ERROR IN shr_getvmave -- INCONSISTENT VARIABLE
663 C LENGTHS.
664 C
665 C it is highly likely that in the *Surface Interaction the option
666 C Tracking Thickness was not set at all or set to a too low number
667 C example of the line:
668 C *Surface Interaction, name=FRICTIONLESS, TRACKING THICKNESS = 5
191
669
670 CALL GETVRMAVGATNODE(NODE,JTYP,'CSTRESS',ARRAY,JRCD,

671 * JELEMLIST,NELEMS,JMATYP,JGVBLOCK)
672
673 cpress = ARRAY(1)

674 C WRITE(*,*) 'Kontaktdruck fuer Node ',NODE,' = ', cpress
675 C WRITE(*,*) 'Kontaktdruckarray fuer Node ',NODE,' = ',array(1:9)
676
677 C for umeshmotion node has to see contact pressure and be in the
678 C nodes_in_contact set (so nodes which are in contact with the cup are
679 C not considered)
680 izei(1) = 1
681 izei(2) = num_contact_nodes
682 isp(1) = 1
683 isp(2) = neighbor_node_nr+1
684 CALL FINDINDEX(neighbors(izei(1):izei(2),isp(1):isp(2)),
685 * izei, isp, ind1, ind2, NODE)
686 i_edge_flag = nodes_in_contact(ind1(1),2)
687
688 IF ((cpress.GT.0.0).AND.(ind1(1).NE.0).AND.(ind2(1).EQ.1)) THEN

689 C WRITE(*,*) 'contact node'
691 C Berechnungen der Verschiebungen fuer das manuelle Smoothing der
692 C hinteren Nachbarknoten, um weniger Mesh-Distortion zu haben.
693
694 C Abhaengig ob es sich um einen Knoten an der Kante oder in der Flaeche
695 C handelt, wird der Absolutbetrag des Verschleiszes aufgebracht:
696 C
697 C Abschaetzung des Verschleisfaktors von etwa 1 micromenter/Stunde
698 C was etwa 2.8 * 10 hoch -7 mm/sec
699 radial_wear = 2.8E-7*cpress*DTIME
700
701 C CHECK START if wear is to big => nodes can leave the surface
702 C reasonable limit 0.005 mm (for the increment)
703 IF (radial_wear.GT.0.007) THEN
704 WRITE(*,*) 'WEAR VERRINGERT von ', radial_wear,
705 * ' auf 0.007.'
707 ENDIF
708 C CHECK END
709
710 C calculation of the wear_vector (in radial direction):

711 C Durch die Koordinaten des Knoten kann die Verschleiszrichtung
712 C bestimmt werden. diese ist in radialer Richtung. (Koordinaten
192
713 C werden normiert (durch Betrag (x,y, nicht z) dividiert) und
714 C damit der Verschleiszbetrag multipliziert, dieser Betrag hat
715 C dem Name: 'radius')
716 radius = SQRT(coords_nodes_in_contact(ind1(1),1)**2
717 1 +coords_nodes_in_contact(ind1(1),2)**2)
718 DO i = 1, 2, 1
719 norm_coord(i)=coords_nodes_in_contact(ind1(1),i)/radius
720 wear_vector(i) = radial_wear*norm_coord(i)
721 ENDDO
722 wear_vector(3) = 0.0
723
724
725 C --> IN DER FLAECHE

726 C es muessen nur noch die Werte von oben uebernommen werden
727 IF (i_edge_flag.EQ.0) THEN
728 WGLOBAL = wear_vector
729 C --> AN DER KANTE
730 ELSEIF (i_edge_flag.EQ.1) THEN
731 C normierten Vektor zum naechsten Nachbarn bestimmen:
732 C Koordinaten Kontaktknoten
733 coords_k(1:3)=coords_allnodes(neighbors(ind1(1),1),1:3)
734 C Koordinaten Nachbarknoten
735 coords_n(1:3)=coords_allnodes(neighbors(ind1(1),2),1:3)
736 C normierter Vektor von Kontaktknoten zum Nachbarknoten
737 betrag = SQRT((coords_n(1)-coords_k(1))**2+
738 * (coords_n(2)-coords_k(2))**2+
739 * (coords_n(3)-coords_k(3))**2)
740 IF (betrag.LT.0.0001) THEN
741 WRITE(*,*) 'ERROR - variable betrag zu klein', betrag
742 WRITE(*,*) 'NODE', NODE, coords_k
743 WRITE(*,*) 'neighbor',neighbors(ind1(1),2),coords_n
744 ENDIF
745 DO i = 1, NDIM
746 norm_coord(i) = (coords_n(i)-coords_k(i))/betrag
747 ENDDO
748 C Berechnung Verschleiszvektor:
749 C wear_vector(3) := 0
750 C => kein Beitrag im Inprodukt aus der dritten Komponente
751 wear_in_norm = wear_vector(1)*norm_coord(1)+
752 * wear_vector(2)*norm_coord(2)
753 DO i = 1, NDIM
754 C WRITE(*,*) 'radial_wear = ', radial_wear
755 WGLOBAL(i)=radial_wear**2/(wear_in_norm)*norm_coord(i)
756 C WRITE(*,*) 'WGLOBAL = ', WGLOBAL
757 ENDDO
193
758 ELSE
759 WRITE(*,*) 'ERROR - i_edge_flag = ',i_edge_flag
760 ENDIF
761
762 C CHECK START IF NODES LEAVE THE SURFACE

763
764 C Following additional IF statement is necessary to prevent nodes

765 C from wearing if they do not have contact to the (rigid) rod any more:
766 C IF the radial node coordinate is bigger than 0.0005 the radius of the
767 C rod there is no more movement of the node performed.
768 IF (radius.GT.25.41) THEN
769 WGLOBAL(1:NDIM) = 0.0
770 C Also radial_wear has to be 0. To stop the neighbors from moving.
772 WRITE(*,*) NODE,'WEAR WAS SET TO 0; radius=',radius
773 ENDIF
774 IF (radius.GT.25.409) THEN
775 WRITE(*,*) 'Node ',NODE,' has left the rigid:'
776 WRITE(*,*) 'Radius = ', radius
777 ENDIF
778 C CHECK END ----------------------------------------------------------
779
780 C uebernehmen der aufgebrachten Verschiebung in das globale Array

781 C applied_disp
782 DO i = 1, NDIM, 1
783 applied_disp(ind1(1),1,i) = WGLOBAL(i)
784 ENDDO
785
786 C Calculate the radial wear for the neighbor nodes:

787 DO i = 1, neighbor_node_nr
788 disp_neighbors(ind1(1),i+1) = radial_wear *
789 * (neighbor_node_nr+1-i)/(neighbor_node_nr+1)
790 ENDDO
791 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
792 C Aufbringen der Verschiebung, wenn der Knoten ein hinterer Nachbar ist
793
794 ELSEIF ((ind1(1).NE.0).AND.

795 * ((ind2(1).GT.1).AND.(ind2(1).LE.neighbor_node_nr+1))) THEN
796 C Datenbereitstellung:
797 C WRITE(*,*) 'neighbor detected'
798 i_edge_flag = nodes_in_contact(ind1(1),2)
799 radial_wear = disp_neighbors(ind1(1),ind2(1))
800 disp_neighbors(ind1(1),ind2(1)) = 0.0
801
802 IF (radial_wear.GT.0) THEN
194
803 radius = SQRT(
804 1 coords_allnodes(neighbors(ind1(1),ind2(1)),1)**2
805 2 + coords_allnodes(neighbors(ind1(1),ind2(1)),2)**2)
806 DO i = 1, 2, 1
807 wear_vector(i) = radial_wear *
808 * coords_allnodes(neighbors(ind1(1),ind2(1)),i)/radius
809 ENDDO
810 wear_vector(3) = 0.0
811
812 C wieder Differentiation ob der Knoten ein Kantenknoten ist oder nicht:
813 C --> IN DER FLAECHE:
814 IF (i_edge_flag.EQ.0) THEN
815 DO i = 1, NDIM, 1
816 WGLOBAL(i) = wear_vector(i)
817 ENDDO
818
819 C --> AN DER KANTE:

820 ELSEIF (i_edge_flag.EQ.1) THEN
821 C normierten Vektor zum naechsten Nachbarn bestimmen:
822 C Koordinaten vom aktuellen Nachbarknoten
823 coords_k(1:3)=coords_allnodes(neighbors(ind1(1),ind2(1)),:)
824 C Koordinaten vom naechsten Nachbarknoten
825 coords_n(1:3)=
826 * coords_allnodes(neighbors(ind1(1),ind2(1)+1),:)
827 C normierter Vektor von Knoten k zum Knoten n
828 betrag = SQRT((coords_n(1)-coords_k(1))**2
829 1 + (coords_n(2)-coords_k(2))**2
830 2 + (coords_n(3)-coords_k(3))**2)
831 DO i = 1, NDIM
832 norm_coord(i) = (coords_n(i)-coords_k(i))/betrag
833 ENDDO
834 C Berechnung Verschleisz:
835 C wear_vector(3) := 0
836 C => kein Beitrag im Inprodukt aus der dritten Komponente
837 wear_in_norm = wear_vector(1)*norm_coord(1)+
838 * wear_vector(2)*norm_coord(2)
839 DO i = 1, NDIM
840 WGLOBAL(i)=radial_wear**2/(wear_in_norm)*norm_coord(i)
841 ENDDO
842 ELSE
843 WRITE(*,*) 'ERROR - i_edge_flag = ', i_edge_flag
844 ENDIF
845 C uebertrage den Verschleiszbetrag zur Darstellung im UVARM
846 DO i = 1, NDIM, 1
195
847 applied_disp(ind1(1),ind2(1),i) = WGLOBAL(i)
848 ENDDO
849 ELSE
850 WGLOBAL = 0.0
851 ENDIF
852 ENDIF
853
854 C WRITE(*,*) 'WGLOBAL @ NODE',NODE, ':', WGLOBAL(:)

855
856 CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
857 C Transformation globale Verschleiszrichtung in die
858 C => lokale Verschleiszrichtung
859 C WLOCAL - wird berechnet => Berechnung ULOCAL (braucht Abaqus)
860 C WGLOBAL - von mir vorgegeben
861 C ALOCAL - von UMESHMOTION vorgegeben:
862 DO i = 1, NDIM
863 WLOCAL(i) = 0
864 DO j = 1, NDIM
865 WLOCAL(i) = WLOCAL(i) + WGLOBAL(j) * ALOCAL(j,i)
866 END DO
867 END DO
868 DO i = 1, NDIM
869 ULOCAL(i) = ULOCAL(i) + WLOCAL(i)
870 END DO
871
872
873 C WRITE(*,*) NODE, 'WLOCAL', WLOCAL(:)

874 C WRITE(*,*) 'ULOCAL', ULOCAL(:)
875 C WRITE(*,*) 'Ende UMESHMOTION'
876
877 RETURN
878 END
879
881 CCCCC
882 CCCCC Subroutine DLOAD to define user-defined loads
883 CCCCC
885
886 SUBROUTINE DLOAD(F,KSTEP,KINC,TIME,NOEL,NPT,LAYER,KSPT,

887 * COORDS,JLTYP,SNAME)
888
889 USE Information

891
196
892 DIMENSION TIME(2), COORDS(3)
893 CHARACTER*80 SNAME
894 DOUBLE PRECISION pressure(4),coords_node(4,3), cipt(2), F
895
896 INTEGER ind1, ind2, izei, nodes(4), izaehl

897 C JLTYP = 0 ... Surface-based load
898 JLTYP = 0
899 nodes = 0
900 pressure = 0D0
901 coords_node = 0D0
902 cipt = 0D0 !coordinateds of the integration point
903 C in the unity element
904 F = 0D0
905 C F = 1 MPa = 10 bar
906 C WRITE(*,*) ' '
907 C WRITE(*,*) 'Was ist mit DLOAD?', NOEL, N2EConn(NOEL,1)
908
909 C Searching the Nodes which are connected to the Element, checking if
910 C they are in contact with the rod => if they are in the array
911 C nodes_in_contact
912 izaehl = 0
913 DO j = 2, 9
914 izei = num_contact_nodes
915 ind1 = 0
916 IF (N2EConn(NOEL,j).NE.0) THEN
917 CALL FINDITEMINLIST(nodes_in_contact,izei,ind1,N2EConn(NOEL,j))
918 IF (ind1.NE.0) THEN
919 izaehl = izaehl + 1
920 nodes(izaehl) = N2EConn(NOEL,j)
921 coords_node(izaehl,:) = coords_nodes_in_contact(ind1,:)
922 pressure(izaehl) = temp_array(ind1)
923 ENDIF
924 ENDIF
925 ENDDO
926
927
928 IF (izaehl.NE.4) THEN

929 DO j = 2, 9
930 izei = num_contact_nodes
931 ind1 = 0
932 IF (N2EConn(NOEL,j).NE.0) THEN
933 WRITE(*,*) 'NOEL = ', NOEL
934 WRITE(*,*) 'j = ', j
935 WRITE(*,*) 'N2EConn(NOEL,j)', N2EConn(NOEL,j)
197
936 CALL FINDITEMINLIST(nodes_in_contact,izei,ind1,N2EConn(NOEL,j))
937 IF (ind1.NE.0) THEN
938 izaehl = izaehl + 1
939 WRITE(*,*) 'izaehl = ', izaehl
940 nodes(izaehl) = N2EConn(NOEL,j)
941 WRITE(*,*) 'nodes(izaehl) = ', nodes(izaehl)
942 coords_node(izaehl,:) = coords_nodes_in_contact(ind1,:)
943 WRITE(*,*) 'coords_node (izaehl,:)',coords_node(izaehl,:)
944 pressure(izaehl) = temp_array(ind1)
945 WRITE(*,*) 'pressure(izaehl) = ', pressure(izaehl)
946 ENDIF
947 ENDIF
948 ENDDO
949 WRITE(*,*)'E R R O R - W R O N G NODE NUMBER: ', izaehl
950 WRITE(*,*) 'NOEL = ', NOEL
951 WRITE(*,*) 'N2EConn(NOEL,2:9)', N2EConn(NOEL,2:9)
952 WRITE(*,*) '-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-'
953 ENDIF
954
955
956 C Coordinates of the integration point depending on the NPT (number of

957 C the integration point).
958 C coordinates: s (=x) t (=y)
959 IF (NPT.EQ.1) THEN
960 cipt(1) = -0.57735026918963
961 cipt(2) = -0.57735026918963
962 ELSEIF (NPT.EQ.2) THEN
963 cipt(1) = 0.57735026918963
964 cipt(2) = -0.57735026918963
966 cipt(1) = 0.57735026918963
967 cipt(2) = 0.57735026918963
969 cipt(1) = -0.57735026918963
970 cipt(2) = 0.57735026918963
971 ELSE
972 WRITE(*,*) 'ERROR Integration point number not 1, 2, 3 or 4!'
973 ENDIF
974 C Solving the trial function (Ansatzfunktion) in the pictorial space:
975 F = (pressure(1)*(1 - cipt(1) - cipt(2) + cipt(1)*cipt(2))
976 1 +pressure(2)*(1 + cipt(1) - cipt(2) - cipt(1)*cipt(2))
977 2 +pressure(3)*(1 + cipt(1) + cipt(2) + cipt(1)*cipt(2))
978 3 +pressure(4)*(1 - cipt(1) + cipt(2) - cipt(1)*cipt(2)))/4
979 RETURN
980 END
198
981
982
983
985 CCCCC
986 CCCCC UFIELD - user-definded field variable
987 CCCCC
989
990 SUBROUTINE UFIELD(FIELD,KFIELD,NSECPT,KSTEP,KINC,TIME,NODE,

991 1 COORDS,TEMP,DTEMP,NFIELD)
992 C
993 USE Information
995 C
996 INTEGER izei(2),isp(2),ind1(9),ind2(9)
997
998 DIMENSION FIELD(NSECPT,NFIELD), TIME(2), COORDS(3),

999 1 TEMP(NSECPT), DTEMP(NSECPT)
1000 C
1001 C WRITE(*,*) 'ENTERING UFIELD'
1002 C WRITE(*,*) 'KFIELD=',KFIELD
1003 C WRITE(*,*) 'NSECPT=',NSECPT
1004 C WRITE(*,*) 'TEMP=', TEMP
1005 C WRITE(*,*) 'NFIELD=',NFIELD
1006 C WRITE(*,*) 'FIELD=',FIELD
1007 C WRITE(*,*) '---------------------------'
1008
1009 izei(1) = 1
1010 izei(2) = num_contact_nodes
1011 isp(1) = 1
1012 isp(2) = neighbor_node_nr+1
1013 ind1 = 0
1014 ind2 = 0
1015 CALL FINDINDEX(neighbors(izei(1):izei(2),isp(1):isp(2)),
1016 * izei, isp, ind1, ind2, NODE)
1017 IF ((ind1(1).NE.0).AND.(ind2(1).NE.0)) THEN
1018 FIELD(1,1) = uvarm_neighbor_nodes(ind1(1),ind2(1))
1019 C WRITE(*,*) 'FIELD=',FIELD
1020 ELSE
1021 FIELD(1,1) = 0.0
1022 ENDIF
1023
1024 RETURN
1025 END
199
1026
1027
1028
1030 CCCCC
1031 CCCCC FINDINDEX - finds and returns the index where a searched
1032 CCCCC value is in an array
1033 CCCCC
1035
1036 SUBROUTINE FINDINDEX(iarray, izeile, ispalte, index1, index2,

1037 * iwert)
1038
1039 INTEGER izeile(2),ispalte(2),

1040 * iarray(izeile(1):izeile(2),ispalte(1):ispalte(2)),
1041 * index1(9), index2(9), iwert, izaehler
1042
1043 index1=0
1044 index2=0
1045 izaehler=1
1046
1047 DO i = ispalte(1), ispalte(2), 1

1048 DO j = izeile(1), izeile(2), 1
1049 IF (iarray(j,i).EQ.iwert) THEN
1050 C WRITE(*,*) 'INDEX gefunden: Zeilennr, Spaltennr',i,j,iarray(j,i)
1051 index1(izaehler) = j
1052 index2(izaehler) = i
1053 izaehler = izaehler + 1
1054 ENDIF
1055 C WRITE(*,*) 'FINDINDEX,Zeil,Spalt,WERT',j,',',i,',',iarray(j,i)
1056 ENDDO
1057 ENDDO
1058 C wird der Wert nicht gefunden, so werden index1 und index2 = 0 gesetzt
1059 END
1060
1062 CCCCC
1063 CCCCC FINDITEMINLIST - finds and returns index of a searched
1064 CCCCC value on a 1D array
1065 CCCCC (only for integer values AND the value
1066 CCCCC must only occur once within the array)
1067 CCCCC
1069
200
1070 SUBROUTINE FINDITEMINLIST(iarray, length, index1, iwert)
1071
1072 INTEGER length,iarray(1:length),

1073 * index1, iwert
1074
1075 index1 = 0
1076 DO i = 1, length, 1
1077 IF (iarray(i).EQ.iwert) THEN
1078 index1 = i
1079 C WRITE(*,*) 'Wert ',iwert,' auf Pos ',index1
1080 ENDIF
1081 ENDDO
1082 END
Obtain information about sets
This code generates an information file, which has to be provided to the FORTRAN wear
routine.
1 import argparse # ability to use and parse arguments

2 import sys
3 from odbAccess import * # ability to work with an Abaqus - ODB file
4
5 class Parameters:
6 pass
7 parameter = Parameters() # instance of the class Parameters
9 parser = argparse.ArgumentParser(description="Writing the input file
,→ for the Abaqus analysis of a sub model.")
10 parser.add_argument('--path', type=str, required=True,
,→ help='Path of the working directory.')
11 parser.add_argument('--jobname', type=str, required=True,
,→ help='Name of the job, where the neighbor nodes will be
,→ searched.')
,→ help='Number of neighbor nodes.' )
14 #
15 odb = openOdb(path=parameter.path+"/"+parameter.jobname+'_0.odb')
16 print('get_set_info.py - START.')
17 parameter_file = open('parameter.txt', 'w')
18 parameter_file.write('JOBNAME\n'+str(parameter.jobname)+'\n'+
,→ str(len(parameter.jobname))+'\n')
201
19 parameter_file.write('NEIGHBOR_NODE_NUMBER\n'+
,→ str(parameter.neighborNumber)+'\n')
20 set_file = open('set_names_'+str(parameter.jobname)+'.txt', 'w')
21 # Writing all node set names and the number of nodes this set
,→ contains to a file (set_file.txt)
22 # All nodes of a node set are written to a separate file (including
,→ coordinates @ beginning)
23 set_file.write('NODE_SET_NAMES\n')
24 lastFrame = odb.steps['Verschleiss'].frames[-1]
25 set_counter = 0
26 max_node_nr = 0
27
28 for node_sets in odb.rootAssembly.instances['PART-1-1'].

,→ nodeSets.keys():
29 set_counter += 1
30 print('Working on node set: '+str(node_sets))
31 set_file.write(str(node_sets)+'\n')
32 node_set_file = open('node_'+str(node_sets)+'.txt','w')
33 node_set_temp_array = [] # array for cross checking the node
,→ number => no double nodes
34 node_counter = 0
35 for nodes in odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets[node_sets].nodes:
36 node_counter += 1
37 node_set_temp_array.append(nodes.label)
38 if nodes.label > max_node_nr:
39 max_node_nr = nodes.label
40 for allnodes in lastFrame.fieldOutputs['COORD'].values:
41 if (allnodes.nodeLabel in node_set_temp_array):
42 node_set_file.write(str(allnodes.nodeLabel)+', '+',
,→ '.join(map(str,allnodes.data[0:3]))+'\n')
43 node_set_temp_array.remove(allnodes.nodeLabel)
44 if str(node_sets) == 'INNEN':
45 contact_node_nr = node_counter
46 if str(node_sets) == 'ALLNODES':
47 if node_counter < max_node_nr:
48 sys.exit('\n\n\n ERROR \n\n The maximum node number of
,→ ALLNODES is bigger than the node_counter. \n
,→ max_node_nr = '+str(max_node_nr)+'; node_counter =
,→ '+str(node_counter)+' \n This can lead to a memory
,→ leakage in the FORTRAN subroutines! \n\n Correction of
,→ the error by changing the ALLNODES node set: \n In
,→ ALLNODES all the nodes of the model (also reference
,→ points) have to be included. \n\n')
202
49 set_file.write(str(node_counter)+'\n')
50 node_set_file.close()
51 parameter_file.write('NODE_SETS,\n'+str(set_counter)+'\n')
52 set_counter = 0
53 # Writing all element set names and the number of elements this set
,→ contains to a file (set_file.txt)
54 # All elements of a element set and its connectivities are written
,→ to a separate file (including connectivities)
55 set_file.write('ELEMENT_SET_NAMES\n')
56 for element_sets in
,→ odb.rootAssembly.instances['PART-1-1'].elementSets.keys():
57 set_counter += 1
58 print('Working on element set: '+str(element_sets))
59 set_file.write(str(element_sets)+'\n')
60 element_set_file = open('element_'+str(element_sets)+'.txt','w')
61 element_counter = 0
62 for elements in odb.rootAssembly.instances['PART-1-1'].
,→ elementSets[element_sets].elements:
63 if str(elements.type) == 'C3D8':
64 element_counter += 1
65 if element_counter == 1:
66 element_set_file.write(str(elements.label)+',\t')
67 else:
68 element_set_file.write('\n'+str(elements.label)+',\t')
69 for conn in elements.connectivity:
70 element_set_file.write(str(conn)+',\t')
71 elif str(elements.type) == 'C3D6':
72 element_counter += 1
73 if element_counter == 1:
74 element_set_file.write(str(elements.label)+',\t')
75 else:
76 element_set_file.write('\n'+str(elements.label)+',\t')
77 for conn in range(0,8):
78 element_set_file.write('0'+',\t')
,→ #C3D6 werden nicht die Nachbarn, sondern Nullen
,→ eingetragen, damit bemerkt wird, wenn ein
,→ Wedge-Element im 'adaptive-mesh sektor' ist
79 print('\n WARNING: \n Elements of type C3D6 detected! They
,→ must not be in the neighbor node section!')
80 else:
81 print('\n WARNING: \n Elements of type
,→ '+str(elements.type)+' detected!')
82 set_file.write(str(element_counter)+'\n')
83 element_set_file.close()
203
84 parameter_file.write('ELEMENT_SETS,\n'+str(set_counter)+'\n')
85 set_file.close()
86 # Prepare empty file for the UVARM Variable - containing the right
,→ amount of zeros:
87 uvarm_file = open('UVARM1.txt','w')
88 for i in range(0,contact_node_nr,1):
89 for j in range(0,parameter.neighborNumber+1,1):
90 uvarm_file.write('0.0,\t')
91 uvarm_file.write('\n')
92 uvarm_file.close()
93 parameter_file.close()
94 print('get_set_info.py - DONE.\n\n')
Find the neighbour nodes
This Python script finds the neighbour nodes of the contact nodes.
1 import sys # operating system

2 import argparse # ability to use and parse arguments
3 import collections # for counting arguments etc. in a list
4 import numpy # for calculation of median
5 from odbAccess import * # ability to work with an Abaqus - ODB
,→ file
6 from math import sqrt
7 from datetime import datetime
8 from collections import Counter
9
10 start_time = datetime.now()
11 print("Starting time = "+str(start_time))
12
13 class Parameters:
14 pass
18 parser.add_argument('--path', type=str, required=True,
,→ help='Path of the working directory.')
19 parser.add_argument('--jobname', type=str, required=True,
,→ help='Name of the job, where the neighbor nodes will be
,→ searched.')
204
20 parser.add_argument('--contactNodeSetName', type=str, required=True,
,→ help='Name of the contact node set. These nodes will get
,→ neighbors.')
,→ via UMeshMotion later.')
,→ help='Number of neighbor nodes, which will be calculated here.' )
23 parser.add_argument('--modelType', type=str, required=True,
,→ help='Two model types are accepted: 3D and AX. 3D means a 3D
,→ Simulation with the z-axis being the length axis of the rod, AX
,→ means axisymmetrical simulation, where x is the radial direction,
,→ y in rod direction.')
25 #
26 odb = openOdb(path=parameter.path+"/"+parameter.jobname+'_0.odb')
27 print('find_node_neighbors.py - START.')
28 if parameter.contactNodeSetName not in
,→ odb.rootAssembly.instances['PART-1-1'].nodeSets.keys():
29 print('\n\n\n * * * E R R O R in find_node_neighbors.py * * *
,→ \n\n\n'+'find_node_neighbors.py cannot find the given node set
,→ called: '+parameter.contactNodeSetName+'\n\n')
30 elif 'INNEN_ADA_MESH' not in
,→ odb.rootAssembly.instances['PART-1-1'].elementSets.keys():
31 print('\n\n\n * * * E R R O R in find_node_neighbors.py * * *
,→ \n\n\n'+'find_node_neighbors.py cannot find the given element
,→ set called: INNEN_ADA_MESH \n\n')
32
33 #print('neighbor number = '+str(parameter.neighborNumber))

34
35 if parameter.modelType == 'AX':
36 sys.exit('THIS NEIGHBORNODE SEARCH IS FOR 3D MODELS ONLY!')
37
38 # the following dictionaries work only for models containing 1

,→ instance named ['PART-1-1']
39 # creating a dictionary for all nodes of the model, the node.label
,→ is the key, the node object the value
40 node_dictionary = {}
41 for all_nodes in odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets['INNEN_ADA_MESH'].nodes:
42 node_dictionary.update({all_nodes.label:all_nodes})
43 # creating a dictionary for all elements, the node.label is the key,
,→ the node object the value
44 element_dictionary = {}
205
45 for all_elements in odb.rootAssembly.instances['PART-1-1'].
,→ elementSets['INNEN_ADA_MESH'].elements:
46 element_dictionary.update({all_elements.label:all_elements})
47 # creating a list which contains all contact nodes and their
,→ neighbors
48 # 1st entry = contact node
49 # the following entries per sub-list are the neighbors
50 neighbors = []
51 #list containing the increment of numbers between the node and the
,→ next neighbor
52 delta_node_number = []
53 #gap/distance between nodes (total gap), used for proof if a
,→ suggestion concerning the node neighbor is correct
54 delta_gap_between_nodes = []
55 # Finding neighbors in radial direction:
56 # This procedure works only if the node numbers of the whole model
,→ go from 1 to n ?!
57 # This procedure works for problems where the nodes are placed
,→ cylindrically in space. The z coordinate is not considered in the
,→ choosing phase, because
58 # the correct neighbor node has to have the same z-coordinate (+-2%)
,→ as the current_node.
59 contact_node_counter = -1 #
,→ contact_node_counter counts the contact nodes for positioning
,→ correctly in the array: neighbors
60 suggestion = False # suggestion is a
,→ boolean which determinates if a suggestion is going to be made or
,→ not
61 first_time = True # first_time is a
,→ boolean which determinates if an action is performed the 1st
,→ time
62 suggestion_1_counter = 0
63 suggestion_2_counter = 0
64 for contact_nodes in odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets[parameter.contactNodeSetName].nodes:
65 contact_node_counter += 1
66 #
67 # print('contact_node_counter = '+str(contact_node_counter))
68 # print('current contact node considered =
,→ '+str(contact_nodes.label))
69 current_node = contact_nodes
,→ # current node is one of the contact nodes
70 i = 0
71 neighbors.append([contact_nodes.label])
206
72
73 while i < parameter.neighborNumber:

,→ # repeat for the number of neighbor nodes per contact node
74 # print('Searching neighbor for Node
,→ '+str(current_node.label))
75 # after 2 runs of searching the whole database a suggestion for the
,→ next neighbor node can be made, to speed up the process of
,→ searching
76 # the most_common_delta_node_number as well as the 2nd most common
,→ node number are recalculated every 10th loop
77 if (contact_node_counter == 3 or contact_node_counter%10==0)
,→ and contact_node_counter != 0:
78 counted_delta_node_numbers = Counter(delta_node_number)
79 most_common_delta_node_number = [0,0]
80 most_common_delta_node_number[0] =
,→ counted_delta_node_numbers.most_common()[0][0]
,→ # calculation of the most common increment between
,→ the neighbor nodes
81 most_common_delta_node_number[1] =
,→ counted_delta_node_numbers.most_common()[1][0]
,→ # calculation of the second most common increment
,→ between the neighbor nodes
82 median_displacement =
,→ numpy.median(delta_gap_between_nodes)
,→ # calculation of the median of the gaps between the
,→ nodes (average node displacement)
83 # print('the most common increments are
,→ ='+str(most_common_delta_node_number[0])+' and
,→ '+str(most_common_delta_node_number[1])+' the displacement
,→ median = '+str(median_displacement))
84 if contact_node_counter > 2:
85 suggestion = True #
,→ <== # this is the switch which allows to use the
,→ suggestion modus for faster searching
86
87 # SUGGESTION OF A NEIGHBOR NODE FOR SEARCH ACCELERATION:

88 # this algorithm is designed for meshes which have their node
,→ neighbors radially.
89 # in case of a 3D model
90 # the following suggestion is only valid for a ring, where the
,→ rotation axis is the Z axis and the neighbors have the same z
,→ coordinate
91 if (suggestion == True) and (parameter.modelType == '3D'):
92 first_time = True
207
93 most_common_delta = most_common_delta_node_number[0]
94 while suggestion == True:
95 # print('Current node = '+str(current_node.label))
96 if first_time == False:
97 most_common_delta =
,→ most_common_delta_node_number[1]
,→ # Second guess - if the most common guess
,→ goes wrong - second common is tried
98
99 if (current_node.label + most_common_delta) in
,→ node_dictionary:
100 suggested_neighbor =
,→ node_dictionary[current_node.label +
,→ most_common_delta] # suggested
,→ neighbor = current node number + median of
,→ the increment of node numbers
101
102 # print('Suggested node =

,→ '+str(suggested_neighbor.label))
103 # print('a suggestion for 3D is
,→ made:'+str(suggested_neighbor.label))
104
105 # further the z-coordinate of a standard ring is

,→ as a maximum 8mm, so if this coordinate is
,→ bigger than 8.3 an input-error occured
106 if current_node.coordinates[2] > 8.3:
107 sys.exit('\n\nERROR - the z (axial)
,→ coordinate of the ring was bigger or
,→ equal to 8.3mm. \n Check if the model has
,→ the z-axis as rod-axis. This might be an
,→ input error. \n If a ring of thickness
,→ greater than 8mm shall be calculated,
,→ change the limit accordingly.\n\n')
108 # the suggestion of a node can only be correct if
,→ the z (axial) coordinate is equal +- 0.01
109 if abs(current_node.coordinates[2] -
,→ suggested_neighbor.coordinates[2]) < 0.01:
110 length_current_node =
,→ sqrt(current_node.coordinates[0]**2+
,→ current_node.coordinates[1]**2)
,→ # lenght of the coordinate vector of the
,→ current_node
208
111 length_sugg_node =
,→ sqrt(suggested_neighbor.coordinates[0]**2+
,→ suggested_neighbor.coordinates[1]**2)
,→ # lenght of the coordinate vector of the
,→ suggested_neighbor
112 # print('x coord current: '+
,→ str(current_node.coordinates[0])+'
,→ suggestion:
,→ '+str(suggested_neighbor.coordinates[0]))
113 # print('y coord current: '+
,→ suggestion:
114 # print('z coord current: '+
,→ suggestion:
115 # print('length current: '+
,→ str(length_current_node)+' suggestion:
,→ '+str(length_sugg_node))
116 # print('x coord normiert: current '+
,→ str(current_node.coordinates[0]/
,→ length_current_node)+ ' suggestion:
,→ '+str(suggested_neighbor.coordinates[0]/
,→ length_sugg_node))
117 # print('y coord normiert: current '+
,→ str(current_node.coordinates[1]/
,→ length_current_node)+ ' suggestion:
,→ '+str(suggested_neighbor.coordinates[1]/
,→ length_sugg_node))
118
119 # test if the node and the neighbor have the

,→ same unity vector radially (the
,→ z-coordinate is not considered,
120 # but was considered previously: the
,→ z-coordinate also must be equal),
,→ tolerance = 0.01
121 # further the radial distance between the
,→ suggested and the current node must not be
,→ greater than 1.2*median_displacement and
,→ not smaller than 0.8*median_displacement
122 # further the radial distance of the neighbor
,→ must be greater than the radial distance
,→ of the neighbor
209
123 distance_radial = sqrt(
,→ (current_node.coordinates[0]-
,→ suggested_neighbor.coordinates[0])**2
,→ +(current_node.coordinates[1]-
,→ suggested_neighbor.coordinates[1])**2)
124 distance_zaxis_to_current_node = sqrt(
,→ (current_node.coordinates[0])**2
,→ +(current_node.coordinates[1])**2)
125 distance_zaxis_to_suggested_neighbor = sqrt(
,→ (suggested_neighbor.coordinates[0])**2
,→ +(suggested_neighbor.coordinates[1])**2)
126 # print('radial distance between the current
,→ and suggsted node =
,→ '+str(distance_radial))
127 if (abs(current_node.coordinates[0]/
,→ length_current_node-
,→ suggested_neighbor.coordinates[0]/
,→ length_sugg_node) < 0.012) and \
128 (abs(current_node.coordinates[1]/
,→ length_sugg_node) < 0.012) and \
129 (distance_radial < 1.5*median_displacement)
,→ and \
130 (distance_radial > 0.8*median_displacement)
,→ and \
131 (distance_zaxis_to_current_node <
,→ distance_zaxis_to_suggested_neighbor)
,→ and \
132 (abs(distance_zaxis_to_suggested_neighbor-
,→ distance_zaxis_to_current_node) >
,→ 0.45*median_displacement):
133
134 # update the distance between nodes and

,→ the delta in node numbers
135 delta_coordinates =
,→ current_node.coordinates-
,→ suggested_neighbor.coordinates
136 delta_gap_between_nodes.append(
,→ sqrt(delta_coordinates[0]**2 +
,→ delta_coordinates[1]**2 +
,→ delta_coordinates[2]**2))
210
137 delta_node_number.append( -
,→ current_node.label +
,→ suggested_neighbor.label)
138
139 # taking the suggested node for the next

,→ neighbor
140 current_node = node_dictionary
,→ [suggested_neighbor.label]
141 neighbors[contact_node_counter].
,→ append(current_node.label)
142
143 # additional information

144 if first_time == True:
145 # print('suggestion accepted - 1st
,→ guess was right!')
146 suggestion_1_counter += 1
147 else:
148 # print('suggestion accepted - 2nd
,→ guess was right!')
149 suggestion_2_counter += 1
150 i = i + 1
151 suggestion = True
152 break
153
154 else:
155 print('Suggestion was wrong - (node
,→ '+str(suggested_neighbor.label)+')')
156
157 # ========DEBUGGING=========
158 # if (abs(current_node.coordinates[0]
,→ /length_current_node-
,→ length_sugg_node) >= 0.025):
159 # print('Possibility 1 x coord:
,→ '+str(abs(current_node.coordinates[0]/
,→ length_sugg_node))+' > 0.01 =
,→ tolerance.')
160 # if (abs(current_node.coordinates[1]/
,→ length_sugg_node) >= 0.025):
211
161 # print('Possibility 2 y coord:
,→ '+str(abs(current_node.coordinates[1]/
,→ length_sugg_node))+' > 0.01 =
,→ tolerance.')
162 # if (distance_radial >=
,→ 1.5*median_displacement) or
,→ (distance_radial <=
,→ 0.8*median_displacement):
163 # print('Possibility 3 distance
,→ between nodes: 0.8*median_displacement
,→ = '+ str(0.8*median_displacement)+ ' <
,→ '+str(distance_radial)+' <
,→ '+str(1.5*median_displacement)+' =
,→ 1.5*median_displacement.')
164 # if distance_zaxis_to_current_node >=
,→ distance_zaxis_to_suggested_neighbor:
165 # print('Possibility 4 radial
,→ position of inner node = '+
,→ str(distance_zaxis_to_current_node)+'
,→ > '+ str(distance_zaxis_to
,→ _suggested_neighbor)+' = outer node.
,→ Contradiction!')
166 # if abs(
,→ distance_zaxis_to_suggested_neighbor
,→ -distance_zaxis_to_current_node) <=
,→ 0.45*median_displacement:
167 # print('Possibility 5 the radial
,→ position ('+str(
,→ distance_zaxis_to_suggested_neighbor
,→ -distance_zaxis_to_current_node)+')
,→ was smaller than
,→ 0.3*median_displacement =
,→ '+str(0.45*median_displacement))
168 # ========DEBUGGING=========
170 first_time = False
171 else:
172 suggestion = False
173
174 else:
212
175 # print('Suggestion (node
,→ '+str(suggested_neighbor.label)+') was
,→ wrong - z coordinates not equal.')
176 #===========================
177 # print('delta z =
,→ '+str(abs(current_node.coordinates[2] -
,→ suggested_neighbor.coordinates[2]))+' >
,→ 0.01 = tolerance')
178 #===========================
179

182 else:
184
185 else:
186 # print('Suggested node not available.')
187

190 else:
192
193 if suggestion == False:

194 # print('looking for neighbor of node
,→ '+str(current_node.label))
195 # print('this is the '+str(i)+' neighbor of the contact
,→ node '+str(contact_nodes.label))
196 i = i + 1
,→ # counter for the neighbor nodes
197 current_element_list = {}
,→ # contains all elements which are in connected to the
,→ current_node
198 for elements in odb.rootAssembly.instances['PART-1-1'].
,→ elementSets[parameter.wearElementSetName].elements:
199 if current_node.label in elements.connectivity:
,→ # checking if the elements of the
,→ wearElementSetName are connected with the current
,→ node
200 # print('element found '+str(elements.label))
201 current_element_list.update(
,→ {elements.label:elements}) # in
,→ connected => updating the dictionary
202 ##care much output:
213
203 # print('current element list =
,→ '+str(current_element_list))
204 neighbor_node_list = {}
,→ # contains all nodes of the elements which are
,→ connected to the current_node, except the
,→ current_node
205 for element in current_element_list:
207 # print('element of current_element_list:
,→ '+str(current_element_list[element]))
208 for n in current_element_list[element].connectivity:
209 # print('current connectivity element
,→ investigated =
,→ '+str(current_element_list[element].label)+'
,→ -> Node: '+str(n))
210 if n != current_node.label:
211 neighbor_node_list.update(
,→ {n:node_dictionary[n]})
212 # print('Possible neighbors of Node:
,→ '+str(current_node.label)+' are the nodes
,→ '+str(neighbor_node_list.keys()))
213
214 # Finding the neighbor node through the coordinates

215 # the 3D MODEL
216
217 if parameter.modelType == '3D':

218 smallest_distance = 0 #
,→ calculation of the smallest distance between the
,→ current_node and the other neighbor nodes
219 for all_nodes in neighbor_node_list: #
,→ when iterating over a dictionary the keys are
,→ returned
220 distance = sqrt((current_node.coordinates[0]-
,→ neighbor_node_list[all_nodes].
,→ coordinates[0])**2+
,→ coordinates[2])**2)
221 if (smallest_distance != 0) and (smallest_distance
,→ > distance):
222 smallest_distance = distance
214
223 elif smallest_distance == 0:
225 # print('smallest_distance between current_node and
,→ the nearest neighbor:'+str(smallest_distance))
226 # if the radial distance to the rotation center
,→ (z-axis assumed) of the current_node +
,→ 0.1*smallest_distance is bigger than the radial
,→ distance form the rotation center (z-axis
,→ assumed) of the neighbor node,
227 # the neighbor node is not further away from the
,→ rotation axis than the current_node => will not
,→ be the next neighbor => pop item from dictionary
228 # further if the z-coordinate is not == z-coordinate
,→ of the current_node +- 0.1*smallest_distance is
,→ bigger than the radial distance form the rotation
,→ center (z-axis assumed) of the neighbor node =>
,→ will not be the next neighbor => pop item from
,→ dictionary
229 for all_nodes in neighbor_node_list.keys():
230 if (neighbor_node_list[all_nodes].
,→ coordinates[0]**2
,→ +neighbor_node_list[all_nodes].
,→ coordinates[1]**2)**0.5 <
,→ ((current_node.coordinates[0]**2
,→ +current_node.coordinates[1]**2)**0.5
,→ +0.3*smallest_distance):
231 del neighbor_node_list[all_nodes]
232 elif (neighbor_node_list[all_nodes].
,→ coordinates[2] - current_node.coordinates[2])
,→ > 0.1*smallest_distance:
234 # print('after poping out not wanted neighbors =>
,→ neighbor_node_list = '+str(neighbor_node_list))
235 # the neighbor with the smallest distance to the
,→ current_node should be the wanted neighbor node
236 smallest_distance = 0
237 j = 0
238 while j < 2:
239 j = j + 1
240 for all_nodes in neighbor_node_list.keys():
215
241 distance = sqrt((current_node.coordinates[0]-
,→ coordinates[2])**2)
242 if (smallest_distance != 0) and
,→ (smallest_distance > distance):
244 elif (smallest_distance != 0) and
,→ (smallest_distance < distance):
245 # print('eliminated node:
,→ '+str(neighbor_node_list[all_nodes]))
247 elif smallest_distance == 0:
249
250
251 # ERRORS:
252 # A) if there is more than 1 key in the neighbor_node_list left
253 # B) if there is no key in the neigbhor_node_list
254 # => the watchdog has to raise some ERROR
255 watchdog = 0
256 for key in neighbor_node_list.keys():
257 if type(key) == int:
258 watchdog += 1
259 # ========DEBUGGING=========
260 # print(' THE WINNER OF THE NEIGHBOR NODE CONTEST
,→ FOR NODE '+str(current_node.label)+' IS ....
,→ ' + str(neighbor_node_list[key].label))
261 # print('coords current
,→ '+str(current_node.coordinates)+' coords
,→ neighbor
,→ '+str(neighbor_node_list[key].coordinates))
262 # print('coords current quadrat
,→ '+str(current_node.coordinates**2)+' coords
,→ neighbor quadrat '+
,→ str(neighbor_node_list[key].coordinates**2))
263 # print('delta coords quadrat
,→ '+str((current_node.coordinates-
,→ neighbor_node_list[key].coordinates)**2))
216
264 # ========DEBUGGING=========
265 a = (current_node.coordinates-
,→ neighbor_node_list[key].coordinates)**2
266 # print('sum of coords quadrat
,→ '+str(a[0]+a[1]+a[2]))
267 b = sqrt(a[0]+a[1]+a[2])
268 # print('delta distance = sqrt(sum)'+str(b))
269 # gap/displacement of the current and the
,→ neighbor node
270 delta_coordinates = current_node.coordinates-
,→ neighbor_node_list[key].coordinates
271 delta_gap_between_nodes.append(
,→ sqrt(delta_coordinates[0]**2 +
,→ delta_coordinates[1]**2 +
,→ delta_coordinates[2]**2))
273 # print(delta_gap_between_nodes)
274
275 # difference in node numbers between the current

,→ and the neighbor node
276 # print('current node
,→ label:'+str(current_node.label))
277 # print('neighbor node
,→ list:'+str(neighbor_node_list[key].label))
278 # print('delta node number = '+str(-
,→ current_node.label +
,→ neighbor_node_list[key].label))
279 delta_node_number.append( - current_node.label +
,→ neighbor_node_list[key].label)
281 # print('delta node number
,→ ='+str(delta_node_number))
282
283 current_node = neighbor_node_list[key]

284 neighbors[contact_node_counter].
,→ append(current_node.label)
285 if watchdog > 1:
286 sys.exit('\n\n\nERROR - there was more than 1 neighbor
,→ found - ALARM: watchdog =
,→ '+str(watchdog)+'\n\n\n')
287 elif watchdog == 0:
217
288 sys.exit('\n\n\nERROR - there was no neighbor node
,→ found for node '+str(current_node.label)+'\n Hint:
,→ maybe the region for the neighbor nodes does not
,→ contain the next wanted neighbor.\n\n')
289 elif watchdog == 1:
290 if suggestion == False and contact_node_counter > 2:
291 if current_node.label ==
,→ suggested_neighbor.label:
292 # print('although the suggestion was
,→ initially considered false it turned
,→ out to be correct!\nSwitching
,→ suggestion-mode on again. Setting:
,→ suggestion = True')
293 suggestion = True
294
295
296 # After finding all neighbor nodes they are printed to a file:
,→ neighbor_nodes_JOBNAME.txt
297 node_file = open('neighbor_nodes_'+str(parameter.jobname)+'.txt', 'w')
298
299 # Check if there are no duplicates in the list, which contains the
,→ neighbor nodes:
300 # 1st putting all node numbers into one array:
301 all_node_numbers = []
302
303 for lines in neighbors:

304 for nodes in lines:
305 node_file.write(str(nodes)+',\t')
306 all_node_numbers.append(nodes)
307 node_file.write('\n')
308 node_file.close()
309
310 # Sort the list with all neigbors and check for duplicates:
311 # If any duplicates are found print a warning message. (Thank you
,→ stackoverflow - question 9835762)
312 def list_duplicates(seq):
313 seen = set()
314 seen_add = seen.add
315 # adds all elements it doesn't know yet to seen and all other to
,→ seen_twice
316 seen_twice = set( x for x in seq if x in seen or seen_add(x) )
317 # turn the set into a list (as requested)
318 return list( seen_twice )
319
218
320 def list_without_duplicates(seq):
321 seen = set()
322 seen_add = seen.add
323 return [ x for x in seq if not (x in seen or seen_add(x))]
324
325 my_duplicates = []
326 my_duplicates = list_duplicates(all_node_numbers)
327
328 contact_nodes_without_clear_neighbors = []
329
330 if my_duplicates:
331 print("\n\n\n + + + + + WARNING - NODE NR DUPLICATES FOUND + +
,→ + + + \n\n Following node numbers occure more than once:
,→ \n"+str(sorted(my_duplicates))+"\n\n HINT: Adjust the settings
,→ of the tolerances which are used to determine if a suggested
,→ node is the real neighbor node. (The tolerances of the advanced
,→ guess as well as the tolerances of the 3D neighbor node
,→ search.\n\n")
332 for lines in neighbors:
333 for nodes in lines:
334 for doubles in my_duplicates:
335 if doubles == nodes:
336 contact_nodes_without_clear_neighbors.append(lines[0])
337 contact_nodes_without_clear_neighbors =
,→ list_without_duplicates(contact_nodes_without_clear_neighbors)
338 contact_nodes_without_clear_neighbors.sort()
339 print('The contact nodes with unclear neighbors are listed below:
,→ \n'+str(contact_nodes_without_clear_neighbors))
340 print('The median displacement was '+str(median_displacement)+'.
,→ The minimum and maximum displacement are
,→ '+str(min(delta_gap_between_nodes))+';
,→ '+str(max(delta_gap_between_nodes))+'.')
341
342 # determination where the mistake is:

343 # the delta in node numbers is printed into the file
,→ 'delta_neighbor_node_nr_JOBNAME.txt'
344 # to get the delta the node nr of the previous node has to be
,→ subtracted from the current node number
345 # example: contact node, neighbor1, neighbor2, neighbor3 => delta
,→ (regular pattern, 1st row = 0, nothing to subtract)
346 # 123, 1030, 1040, 1050 0, 0, 0,
,→ 0
347 # 124, 1033, 1043, 1053 1, 3, 3,
,→ 3
219
348 # 125, 1036, 1046, 1056 1, 3, 3,
,→ 3
349 # through the deltas a pattern shall be visualized - the pattern
,→ will be broken by the false neighbors
350
351 # the file contains following information:

352 # delta between contact nodes, delta between contact and neighbor, n
,→ times delta between neighbors
353 delta = []
354 delta_line = []
355 for i in range(0,len(neighbors)):
356 for j in range(0,len(neighbors[0])):
357 if i == 0:
358 delta_line.append(0)
359 else:
360 delta_line.append(neighbors[i][j]-neighbors[i-1][j])
361 delta.append(delta_line)
362 delta_line = []
363
364 delta_node_nr_file =
,→ open('delta_neighbor_node_nr_'+str(parameter.jobname)+'.txt',
,→ 'w')
365
366 for line in delta:

367 for nodes in line:
368 delta_node_nr_file.write(str(nodes)+',\t')
369 delta_node_nr_file.write('\n')
370 delta_node_nr_file.close()
371 print("\n For easier checking the difference between a contact node
,→ and its neighbors and the previous contact node and its
,→ neighbors is calculated and printed to the file called:
,→ delta_neighbor_node_nr_"+str(parameter.jobname)+".txt \n")
372
373 # determination which nodes are in the adaptive mesh segement, but
,→ were not used in as neighbor nodes
374 # this only makes sense IF all nodes in the mesh segment should
,→ be used as neighbor nodes. (e.g.: If just 3 out of 5 possible
,→ neighbors are used ther will be a lot of output)
375
376 else:
377 print("No duplicates in the node numbers were found, every node
,→ appears only once. \n Detailed data below.\n")
378
379 end_time = datetime.now()
220
380 print("Neighbor nodes found. Current date and time: "+str(end_time))
381 delta_time = end_time-start_time
382 print("The search lasted "+str(delta_time))
383 print("\nAdditional information:")
384 contact_node_counter = contact_node_counter + 1 # +1 because counter
,→ started at 0
385 print("Number of contact nodes: \t"+str(contact_node_counter))
386 print("Number of neighbors:
,→ \t"+str(parameter.neighborNumber*contact_node_counter))
387 print("Number of total suggestions = (contact node number-3) * number
,→ of neighbors =
,→ "+str((contact_node_counter-3)*parameter.neighborNumber))
388 print("Number of true suggestions:
,→ \t"+str(suggestion_1_counter+suggestion_2_counter))
389 print('Resulting from: 1st guesses: \t'+str(suggestion_1_counter)+'\t
,→ and 2nd guesses: \t'+str(suggestion_2_counter))
390 print("Most common deltas in node numbers (delta, incidence):
,→ \t"+str(counted_delta_node_numbers.most_common()[0])+',
,→ \t'+str(counted_delta_node_numbers.most_common()[1])+',
,→ \t'+str(counted_delta_node_numbers.most_common()[2]))
391 print('The script find_node_neighbors.py finished successfully,
,→ writing all neighbor nodes of the contact nodes to
,→ neighbor_nodes_'+str(parameter.jobname)+'.txt')
Input generator for the wear calculation
This code generates an input file for the next Abaqus wear restart calculation.
1 # This Phyton file shall generate a new knotenTempTxt.txt and a new

,→ input/restart file:
2 #
3 print("enter Main_Input_Generator_BCD_V04.py")
4 import argparse
5 from odbAccess import *
6
7 class Parameters:
8 pass
221
13 parser.add_argument('--path', type=str, required=True, help='Path of
,→ the working directory.')
,→ of the sub job containing the local model. This model is used
,→ repeatedly.')
16 parser.add_argument('--jobsub', type=str, required=True, help='Name of
,→ the sub job containing the local model. This model is used
,→ repeatedly.')
17 parser.add_argument('--inpmain', type=str, required=True, help='Name
,→ of the input file for the sub model job.')
18 parser.add_argument('--inpgenmain', type=str, required=True,
,→ help='Name of the input file generator for the sub model.')
19 parser.add_argument('--tempfilename', type=str, required=True,
,→ help='Name of the *.txt file where all temperature values for all
,→ nodes are stored.')
,→ via UMeshMotion later.')
21 parser.add_argument('--inc', type=float, required=True, help='This is
,→ the increment which is maximaly used when calculating the wear.')
,→ 13.08.2015. This value is considered as the high temperature of
,→ the submodel calculation.')
,→ minimum is 0 MPa (0bar). This value is considered as the low
,→ temperature of the submodel calculation.')
24
26 odb = openOdb(path=parameter.path+"/"+parameter.jobsub+
,→ '_'+str(parameter.counter)+'.odb')
27 odbmain = openOdb(path=parameter.path+'/'+parameter.jobmain+
,→ '_'+str(parameter.counter-1)+'.odb')
28 # print(parameter.path+"/"+parameter.jobsub+'_'+
,→ str(parameter.counter)+'.odb')
29
30 # Write all Node-Temperatures to a file.

31 knotenTempTxt = open(parameter.path+'/'+parameter.tempfilename+'.txt',
,→ 'w')
32 lastFrame = odb.steps['thermal_analysis'].frames[-1]
222
33 for nt11 in lastFrame.fieldOutputs['NT11'].values:
34 # print('Node = %d, Temp. = %6.4f' %(nt11.nodeLabel,
,→ nt11.data))
35 for nodes in
,→ odbmain.rootAssembly.instances['PART-1-1'].nodeSets['INNEN'].
,→ nodes:
36 if nt11.nodeLabel == nodes.label:
37 # The temp or pressure value is offset so that the
,→ distribution falls from
,→ highpressuretemp-lowpressuretemp to 0.
38 temp_value = nt11.data-parameter.lowpressuretemp
39 if temp_value < 0: # correct rounding/truncation errors
40 temp_value = 0.0
41 print('Temperature / Pressure value < 0 - reset to 0
,→ for node '+str(nt11.nodeLabel)+' with the corrected
,→ value '+str(temp_value)+'\nCorrected by
,→ '+str(parameter.lowpressurevalue)+'.\n')
42 knotenTempTxt.write('%d, %8.6f\n' %(nt11.nodeLabel,
,→ nt11.data-parameter.lowpressuretemp))
43 print('\nINFORMATION:\nThe pressure (temperature) distribution has
,→ been offset (reduced) by the low pressure (temperature) value:
,→ '+str(parameter.lowpressuretemp)+'\n\n')
44 knotenTempTxt.close()
45 #
46 #
47 print('Generation of knotenTempTxt successfully finished!')
48 # ----- Node Temperatures written -----
49
50 print('Start Main-Restart-File generation.')

51 restartFile = open(str(parameter.path)+'/'+str(parameter.inpmain)+
,→ '_'+str(parameter.counter)+'.inp', 'w')
52 restartFile.write('**This file is generated through the Phyton file
,→ '+str(parameter.inpgenmain)+'.py.\n')
53 restartFile.write('*HEADING\n')
54 restartFile.write('**Restart the Analysis\n')
55 restartFile.write('*RESTART, WRITE, FREQUENCY=1, OVERLAY, READ, END
,→ STEP\n')
56 restartFile.write('** AMPLITUDE=STEP applies the load instantaneously
,→ at the start of the step\n')
57 restartFile.write('*STEP, NAME = Verschleiss, nlgeom = YES, AMPLITUDE
,→ = STEP, INC = 2000\n')
58 restartFile.write('*VISCO, STABILIZE, CETOL=0.01\n')
59 restartFile.write(str(parameter.inc)+' , 120000, 1e-3,
,→ '+str(100*parameter.inc)+'\n')
223
60 # only write the adaptive mesh controls once:
61 if (parameter.counter==1):
62 restartFile.write('*ADAPTIVE MESH CONTROLS, NAME=Ada-1\n0.,1.\n')
63 restartFile.write('*ADAPTIVE MESH,
,→ ELSET='+str(parameter.wearElementSetName)+', CONTROLS=Ada-1,
,→ OP=MOD, FREQUENCY=1, MESH SWEEPS=1\n')
64 restartFile.write('*ADAPTIVE MESH CONSTRAINT, CONSTRAINT TYPE=SPATIAL,
,→ TYPE=DISPLACEMENT, USER\n')
65 restartFile.write(str(parameter.wearElementSetName)+'\n')
66 restartFile.write('*DSLOAD\n')
67 restartFile.write('CAP_2Z_SURF_PRESSURE_CAP_HIGH, P,
,→ '+str(parameter.highpressuretemp-parameter.lowpressuretemp)+'\n')
68 restartFile.write('SS_2Z_SURF_PRESSURE_SS_HIGH, P,
,→ '+str(parameter.highpressuretemp-parameter.lowpressuretemp)+'\n')
69 restartFile.write('TEMP_ALL, PNU, 1\n')
70 restartFile.write('*FIELD, USER\n')
71 restartFile.write('INNEN_ADA_MESH\n')
72 restartFile.write('*OUTPUT, HISTORY, FREQUENCY = 1\n')
73 restartFile.write('*ENERGY OUTPUT, ELSET = EL_ALL, VARIABLE = ALL\n')
74 restartFile.write('*OUTPUT, FIELD, FREQUENCY = 1\n')
75 restartFile.write('*ELEMENT OUTPUT, ELSET = EL_ALL\n')
76 restartFile.write('S, P, FV1, E, LE\n')
77 restartFile.write('*NODE OUTPUT, NSET = ALLNODES\n')
78 restartFile.write('CF, RF, U, COORD\n')
79 restartFile.write('*CONTACT OUTPUT, NSET = ALLNODES\n')
80 restartFile.write('CDISP, CSTATUS, CSTRESS, CFORCE\n')
81 restartFile.write('*NODE FILE, NSET = ALLNODES\n')
82 restartFile.write('COORD\n')
83 restartFile.write('*CONTACT FILE, NSET = INNEN\n')
84 restartFile.write('CSTRESS\n')
85 restartFile.write('*END STEP\n')
86 restartFile.close()
87 print('Main-Restart-File successfully written.')
88 exit()
Input generator gas pressure sub-calculation
This file creates an input file for the Abaqus calculation of the gas pressure distribution
in the dynamic sealing surface.
1 #!/usr/bin/env python
2 # This Phyton file shall generate a new input file for the SUBMODEL:
3 #
224
4 print("enter Sub_Input_Generator_BCD_V05.py")
5 import os
6 import sys
7 import argparse
8 import math
9 from odbAccess import *
10 #
11 class Parameters:
12 pass
17 parser.add_argument('--path', type=str, required=True, help='Path of
,→ the working directory.')
,→ of the main job containing the global model. This model is used
,→ repeatedly.')
20 parser.add_argument('--jobsub', type=str, required=True, help='Name of
,→ the sub job containing the local model. This model is used
,→ repeatedly.')
21 parser.add_argument('--inpsub', type=str, required=True, help='Name of
,→ the input file for the sub model job.')
22 parser.add_argument('--usersub', type=str, required=False,help='Name
,→ sub model.')
23 parser.add_argument('--inpgensub', type=str, required=True, help='Name
,→ of the input file generator for the sub model.')
,→ 13.08.2015. This valau is considered as the high temperature of
,→ the submodel calculation.')
,→ minimum is 0 MPa (0bar). This value is also the lower temperature
,→ in the sub calculation/model. This value has to be smaller than
,→ the highpressuretemp.')
26
27
225
29 odb = openOdb(path=str(parameter.path)+"/"+str(parameter.jobmain)+
,→ "_"+str(parameter.counter)+'.odb')
30 # GENERATION OF THE SUBMODEL INPUT FILE
31 inp = open(parameter.inpsub+'_'+str(parameter.counter)+'.inp', 'w')
32 inp.write('**This file is generated through the Phyton file
,→ '+str(parameter.inpgensub)+'.py.\n')
33 inp.write('*HEADING\n')
34 inp.write('*NODE\n')
35
36 # writing the desired nodeSets form the odb (main model) to the
,→ input file:
37 # P R E P E R A T I O N O F T H E D A T A
38 # 1. for the CAP-Segement
39 nodeList_contact_nodes_CAP = []
40 nodeDict_contact_nodes_CAP = {}
41 for allnodes in
,→ odb.rootAssembly.instances['PART-1-1'].nodeSets['TEMP_CAP'].nodes:
42 nodeList_contact_nodes_CAP.append(allnodes.label)
43 nodeDict_contact_nodes_CAP.update({allnodes.label:allnodes})
44
45 # 2. for the Sealing-Segment (SS)

46 nodeList_contact_nodes_SS = []
47 nodeDict_contact_nodes_SS = {}
48 for allnodes in
,→ odb.rootAssembly.instances['PART-1-1'].nodeSets['TEMP_SS'].nodes:
49 nodeList_contact_nodes_SS.append(allnodes.label)
50 nodeDict_contact_nodes_SS.update({allnodes.label:allnodes})
51
52 lastFrame = odb.steps['Verschleiss'].frames[-1]
53
54 nodeDict_all_nodes = {}
55 # Dictionary which contains all nodes used in the sub-model
,→ calculation and their coordinates
56 # To get the coordinates use nodeDict_all_nodes[node nr][0/1/2] with
,→ (0=x,1=y,2=z)
57 # Searching for the nodes and their coordinates of the desired
,→ element-sets
58 for nodes in lastFrame.fieldOutputs['COORD'].values:
59 if nodes.nodeLabel in nodeList_contact_nodes_CAP:
60 inp.write(str(nodes.nodeLabel)+', '+',
,→ '.join(map(str,nodes.data[0:3]))+'\n')
61 nodeDict_all_nodes.update({nodes.nodeLabel:nodes.data})
62 elif nodes.nodeLabel in nodeList_contact_nodes_SS:
226
63 inp.write(str(nodes.nodeLabel)+', '+',
,→ '.join(map(str,nodes.data[0:3]))+'\n')
64 nodeDict_all_nodes.update({nodes.nodeLabel:nodes.data})
65
66 # Getting the current position of the reference nodes for the

,→ boundary condition:
67 # 1. CAP
68 if len(odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets['CAP_REF_POS_TEMP'].nodes) > 1:
69 sys.exit('\n\nERROR in Sub_Input_Generator -\n There is more than
,→ one node in the CAP_REF_POS_TEMP set.\n\n')
70 else:
71 cap_ref_pos_nr = odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets['CAP_REF_POS_TEMP'].nodes[0].label
72
73 # 2. SS
74 if len(odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets['SS_REF_POS_TEMP'].nodes) > 1:
75 sys.exit('\n\nERROR in Sub_Input_Generator -\n There is more than
,→ one node in the SS_REF_POS_TEMP set.\n\n')
76 else:
77 ss_ref_pos_nr = odb.rootAssembly.instances['PART-1-1'].
,→ nodeSets['SS_REF_POS_TEMP'].nodes[0].label
78
79 # sanity checks:
80 # no more than one node in the REF_POS_TEMP sets => test performed
,→ above
81 # the z-coordinates of the two REF_POS_TEMP nodes have to be
,→ (nearly) equal
82 if abs(nodeDict_all_nodes[cap_ref_pos_nr][2] -
,→ nodeDict_all_nodes[ss_ref_pos_nr][2]) > 0.15:
83 if parameter.counter == 1:
84 sys.exit('\n\nERROR in Sub_Input_Generator -\n The z-coordinates
,→ of the REF_POS_TEMP nodes are not equal. Difference =
,→ '+str(abs(nodeDict_all_nodes[cap_ref_pos_nr][2] -
,→ nodeDict_all_nodes[ss_ref_pos_nr][2]))+'\n\n')
85 else:
86 print('\n\nWARNING from Sub_Input_Generator -\n The z-coordinates
,→ of the REF_POS_TEMP nodes are not equal. Difference =
,→ '+str(abs(nodeDict_all_nodes[cap_ref_pos_nr][2] -
,→ nodeDict_all_nodes[ss_ref_pos_nr][2]))+'\n\n')
87
88 # Write the node sets to the (sub-) input file

89 inp.write('*NSET, NSET = NODES_CAP')
227
90 k=-1
91 for allnodes in
92 k+=1
93 if (k%16) == 0:
94 inp.write("\n"+str(allnodes.label)+", ")
95 else:
96 inp.write(str(allnodes.label)+", ")
97
98 inp.write('\n*NSET, NSET = NODES_SS')

99 k=-1
100 for allnodes in
101 k+=1
102 if (k%16) == 0:
104 else:
106
107 # here the algorithm checks if the nodes which were initially
,→ selected as nodes for the boundary conditions are still valid:
108
109 # a) for the CYLINDER SIDE (higher temperature):

110 # it is checked if the length of the vector from the SS_Ref_Pos to
,→ the CAP_Ref_Pos is smaller than the length of the vector from
,→ CAP_Ref_Pos to the current node
111 # vector from the SS to CAP
112 SS_to_CAP = nodeDict_all_nodes[cap_ref_pos_nr] -
,→ nodeDict_all_nodes[ss_ref_pos_nr]
113 # length of the vector
114 abs_SS_to_CAP = (SS_to_CAP[0]**2 + SS_to_CAP[1]**2 +
,→ SS_to_CAP[2]**2)**0.5
115
116 inp.write('\n*NSET, NSET = TEMP_CYLINDER_SIDE')

117 k=-1
118 for allnodes in
119 print_now = False
120 if abs(nodeDict_all_nodes[allnodes.label][2] -
,→ nodeDict_all_nodes[cap_ref_pos_nr][2]) < 0.1:
121 # if the node is in the contact plane of SS and CAP
122 SS_to_allnodes = nodeDict_all_nodes[allnodes.label] -
,→ nodeDict_all_nodes[ss_ref_pos_nr]
228
123 # create vector from the reference
,→ node to the current node
124 if abs_SS_to_CAP <= (SS_to_allnodes[0]**2 +
,→ SS_to_allnodes[1]**2 + SS_to_allnodes[2]**2)**0.5:
125 # and if the node is further away
,→ than CAP_Ref_Pos from SS_Ref_Pos
,→ the node is a node which has to
,→ be in the boundary condition set
126 print_now = True
127 else:
,→ nodeSets['TEMP_CYLINDER_SIDE'].nodes:
129 if allnodes.label == nodes.label: # the node is not in the
,→ contact plane of SS and CAP, but in the
,→ TEMP_CYLINDER_SIDE set, it shall carry the boundary
,→ condition as initially decided
131 break # for acceleration --> if node is found in
,→ set, there is no need to search further
132
133 if print_now == True:

134 k+=1
135 if (k%16) == 0:
137 else:
139
140 # Further add all nodes, which are in the TEMP_CYLINDER_SIDE set,
,→ but on the CAP Segement
141 for allnodes in
,→ nodeSets['TEMP_CYLINDER_SIDE'].nodes:
144 if allnodes.label == nodes.label:
146 break # for acceleration (as above)
147

149 k+=1
150 if (k%16) == 0:
152 else:
229
154
155
156 # b) for the CRANK SIDE (lower temperature):

157 move_to_isolation_set = []
158 inp.write('\n*NSET, NSET = TEMP_CRANK_SIDE')
159 k=-1
160 for allnodes in
162 if abs(nodeDict_all_nodes[allnodes.label][2] -
,→ nodeDict_all_nodes[ss_ref_pos_nr][2]) < 0.1:
163 CAP_to_allnodes = nodeDict_all_nodes[allnodes.label] -
,→ nodeDict_all_nodes[cap_ref_pos_nr]
164 if abs_SS_to_CAP <= (CAP_to_allnodes[0]**2 +
,→ CAP_to_allnodes[1]**2 + CAP_to_allnodes[2]**2)**0.5:
165 # abs_SS_to_CAP == abs_CAP_to_SS (PER DEFINITION!)
167 else:
,→ nodeSets['TEMP_CRANK_SIDE'].nodes:
171 break
172

174 k+=1
175 if (k%16) == 0:
177 else:
179
180 for allnodes in

,→ nodeSets['TEMP_CRANK_SIDE'].nodes:
185 break
186 # if the sealing segments touch each other the nodes touching
,→ loose their boundary condition and the temperature will be
,→ calculated there
187 # this happens when either x or y coord is equal to (nearly) 0
230
188 if nodeDict_all_nodes[allnodes.label][0] < 0.0001 or
,→ nodeDict_all_nodes[allnodes.label][1] < 0.0001:
190 #move_to_isolation_set.append(allnodes.label)
191

193 k+=1
194 if (k%16) == 0:
196 else:
198
199 # generation of elements from these nodes - improve for 3D

200 inp.write('\n** DS4 is a heat transfer 4 node quadrilateral shell
,→ element')
201 inp.write('\n*ELEMENT, TYPE = DS4, ELSET = EL_CAP')
202 # Below the element definition of all shell elements are defined.
,→ Therefore the elements from the Main-file are searched for nodes
,→ which are in contact with the rod (node-set-name 'TEMP_SS' and
,→ 'TEMP_CAP' see above). So 4 out of 8 nodes are derived and then
,→ ordered:
203 # 1st the smallest z-coordinates on the dictionary
,→ (nodeList_temporary) keys 1,2 and then the higher z values 3,4.
,→ Afterwards the angle between the node and the xz-plane is
,→ calculated and the dictionary is reordered:
204 # key 1 = smaller z, smaller angle
205 # key 2 = smaller z, higher angle
206 # key 3 = higher z, smaller angle
207 # key 4 = higher z, higher angle
208 # At last the nodes are written into the SUB-input file in following
,→ order: key: 1,3,4,2
209
210 # 1st this is done for the CAP-Segment

211 for elInnen in odb.rootAssembly.instances['PART-1-1'].
,→ elementSets['EL_TEMP_CAP'].elements:
212 nodeList_temporary = {}
213 # temporary dictionary containing the nodes of a
,→ shell element
214 counter = 0 # counter => has to be 4 to generate a valid element
215 for i in range(0,8,1):
216 if elInnen.connectivity[i] in nodeList_contact_nodes_CAP:
217 counter += 1
218 nodeList_temporary.update({counter:
,→ nodeDict_contact_nodes_CAP[elInnen.connectivity[i]]})
231
219 if counter != 4:
220 # check if counter == 4 else => ERROR-WARNING
221 print('ERROR - there were not 4 but '+str(counter)+' nodes
,→ found which are in contact. Something went wrong!
,→ (sub_inp_generator)')
222 else:
223 j=0
224 while j < 4:
225 j += 1
226 current_node = nodeList_temporary[j]
227 # current_node can be swapped if the z-coordinate of the
,→ i-node is smaller than that of the current one
228 # print('j='+str(j))
229 for i in range(j+1,5,1):
230 if (abs(current_node.coordinates[2] -
,→ nodeList_temporary[i].coordinates[2]) > 0.0001)
,→ and ((current_node.coordinates[2] -
,→ nodeList_temporary[i].coordinates[2]) > 0):
231 nodeList_temporary.update({j:nodeList_temporary[i]})
232 nodeList_temporary.update({i:current_node})
233 current_node = None
234 # print('swap bei i='+str(i)+' und j='+str(j))
235 j=0
236 # if a swap is performed reset the counter i and j
,→ to rerun the swapping process
237 break
239 # print(str(i)) # another reordering of the z-pair with
,→ smaller z values concerning the angle between the
,→ node and the xz plane
240 if math.atan(nodeList_temporary[i].coordinates[1]/
,→ nodeList_temporary[i].coordinates[0]) >
,→ math.atan(nodeList_temporary[i+1].coordinates[1]/
,→ nodeList_temporary[i+1].coordinates[0]):
241 # print('swap')
242 current_node = nodeList_temporary[i]
243 nodeList_temporary.update({i:nodeList_temporary[i+1]})
244 nodeList_temporary.update({i+1:current_node})
245 # print nodes in the right order to get a valid, not twisted
,→ element:
246 inp.write("\n"+str(elInnen.label))
247 inp.write(',\t'+str(nodeList_temporary[1].label))
232
251
252 inp.write('\n*ELEMENT, TYPE = DS4, ELSET = EL_SS')

253 # then this is done for the Sealing-Segement
254 for elInnen in odb.rootAssembly.instances['PART-1-1'].
,→ elementSets['EL_TEMP_SS'].elements:
255 nodeList_temporary = {}
256 # temporary dictionary containing the nodes of a
,→ shell element
257 counter = 0
258 # counter => has to be 4 to generate a valid element
260 if elInnen.connectivity[i] in nodeList_contact_nodes_SS:
261 counter += 1
262 nodeList_temporary.update({counter:
,→ nodeDict_contact_nodes_SS[elInnen.connectivity[i]]})
263 if counter != 4:
264 # check if counter == 4 else => ERROR-WARNING
265 print('ERROR - there were not 4 but '+str(counter)+' nodes
,→ found which are in contact. Something went wrong!
,→ (sub_inp_generator)')
266 else:
267 j=0
268 while j < 4:
269 j += 1
270 current_node = nodeList_temporary[j]
271 # current_node can be swapped if the z-coordinate of the
,→ i-node is smaller than that of the current one
272 # print('j='+str(j))
273 for i in range(j+1,5,1):
274 if (abs(current_node.coordinates[2] -
,→ nodeList_temporary[i].coordinates[2]) > 0.0001)
,→ and ((current_node.coordinates[2] -
,→ nodeList_temporary[i].coordinates[2]) > 0):
275 nodeList_temporary.update(
,→ {j:nodeList_temporary[i]})
276 nodeList_temporary.update({i:current_node})
277 current_node = None
278 # print('swap bei i='+str(i)+' und j='+str(j))
279 j=0
280 # if a swap is performed reset the counter i and
,→ j to rerun the swapping process
281 break
233
283 # print(str(i))
284 # another reordering of the z-pair with smaller z values
,→ concerning the angle between the node and the xz
,→ plane
285 if math.atan(nodeList_temporary[i].coordinates[1]/
,→ nodeList_temporary[i].coordinates[0]) >
,→ math.atan(nodeList_temporary[i+1].coordinates[1]/
,→ nodeList_temporary[i+1].coordinates[0]):
286 # print('swap')
287 current_node = nodeList_temporary[i]
288 nodeList_temporary.update({i:nodeList_temporary[i+1]})
289 nodeList_temporary.update({i+1:current_node})
290 # print nodes in the right order to get a valid, not twisted
,→ element:
291 inp.write("\n"+str(elInnen.label))
296
297 inp.write('\n*SHELL SECTION, ELSET = EL_CAP, MATERIAL =

,→ imaginaryMaterial\n1.,1')
298 inp.write('\n*SHELL SECTION, ELSET = EL_SS, MATERIAL =
,→ imaginaryMaterial\n1.,1')
299 inp.write('\n*SURFACE, TYPE=ELEMENT, NAME=SURF_TEMP_CAP')
300 inp.write('\nEL_CAP, SPOS')
301 inp.write('\n*SURFACE, TYPE=ELEMENT, NAME=SURF_TEMP_SS')
302 inp.write('\nEL_SS, SPOS')
303 inp.write('\n*TIE, NAME=TIE_TEMP_SURFS, ADJUST=YES, TYPE=SURFACE TO
,→ SURFACE')
304 inp.write('\nSURF_TEMP_SS, SURF_TEMP_CAP')
305 inp.write('\n**Material Definition\n')
306 inp.write('*MATERIAL, NAME=imaginaryMaterial\n')
307 inp.write('*CONDUCTIVITY\n')
308 inp.write('0.0001, 0.0001\n')
309 inp.write('1., 1.\n')
310 inp.write('1000., 1000.\n')
311 inp.write('** PHYSICAL CONSTANTS\n')
312 inp.write('*PHYSICAL CONSTANTS, ABSOLUTE ZERO=0.\n')
313 inp.write('*INITIAL CONDITIONS, TYPE = TEMPERATURE\n')
314 inp.write('NODES_CAP, '+str(parameter.highpressuretemp/2)+'\n')
315 inp.write('NODES_SS, '+str(parameter.highpressuretemp/2)+'\n')
316 inp.write('** STEP DEFINITION\n')
317 inp.write('*STEP, NAME = thermal_analysis, NLGEOM = NO\n')
234
318 inp.write('*HEAT TRANSFER, STEADY STATE\n')
319 inp.write('1., 1., 1.\n')
320 inp.write('*BOUNDARY\n')
321 inp.write('TEMP_CYLINDER_SIDE, 11, 11,
,→ '+str(parameter.highpressuretemp)+'\n')
322 inp.write('TEMP_CRANK_SIDE, 11, 11,
,→ '+str(parameter.lowpressuretemp)+'\n')
323 inp.write('*OUTPUT, FIELD, FREQUENCY = 1\n')
324 inp.write('*ELEMENT OUTPUT\n')
325 inp.write('TEMP\n')
326 inp.write('*NODE OUTPUT\n')
327 inp.write('NT\n')
328 inp.write('*END STEP\n')
329 inp.close()
235
Acronyms and Symbols
Acronyms
BCD balanced cap design
CF carbon fibre
CMM 3D coordinate measurement machine
FE finite element
FEP fluorinated ethylene propylene
GF glass fibre
HVOF high velocity oxygen fuel spraying
HW HOERBIGER Wien
MoS2 molybdenum disulfide
PA polyamide
PEEK poylether ether ketone
PPDT poly-p-phenyleneterephtalamide
PPS polyphenylene sulfide
PTFE polytetrafluoroethylene
UFD ultra fine diamond
WC tungsten carbide
236
List of Symbols
The quantities, which seem multiply defined, are all needed to distinguish between di-
mensionless and dimensional quantities or to differentiate between the integrand and the
integration limit.
α angle
Ac contact area
As contact area specimen
β angle
b axial ring width
C0 integration constant
d beam deflection
D rod diameter
D̃ rod diameter
D(i) inner diameter
D(o) outer diameter
εb outer fibre strain
εL non-dimensional load parameter
εrr radial strain
εt non-dimensional wear parameter
E elastic modulus
errthreshold error threshold
FC contact force
FC t total contact force over all ring segments
FF frictional force
FS spring force
G shear modulus
237
G̃ shear modulus
h radial ring height
h̃ (radial) ring height
h0 original radial ring height
hcrit critical radial ring height
hs specimen height
J second momentum of area
k wear coefficient
k̃ wear coefficient
k̃2 wear coefficient
K virtual wear coefficient
l length
L beam length
µ friction coefficient
Mb bending momentum
nS number of springs
nseg number of ring segments
ν Poisson’s ratio
ω̃ angular velocity
p pressure
p0 pressure asymtotic expansion, first part, fast time scale
p1 pressure asymtotic expansion, second part, fast time scale
pc contact pressure
p̃c contact pressure
pc,flat contact pressure assuming microscopically perfectly flat surfaces
238
pc,rough contact pressure reduced by the gas pressure in the dynamic sealing surface which
occurrs due to microscopical surface roughness
pcrk crank side pressure
pcycle pressure during a compression cycle
p̃cycle pressure during a compression cycle
pcyl cylinder side pressure
p̃d discharge pressure
pg absolute gas pressure
pL gas pressure load
ϕend angular position of the sealing segment’s tip
ϕmax angular position of the wear maximum
p̃L gas pressure load
pnon per non-periodic pressure
po pressure on the outer diameter
pp process pressure
p̃p process pressure
pper periodic pressure
p̃s suction pressure
P pressure, slow time scale
P0 pressure asymtotic expansion, first part, slow time scale
P1 pressure asymtotic expansion, second part, slow time scale
Pc contact pressure fast time scale
R(o)
ρ ratio of
R(i)
r radial coordinate
Ra roughness (arithmetical mean deviation of the assessed profil)
R(i) inner radius of a ring
R̃(i) inner radius of a ring
239
R(o) outer radius of a ring
R̃(o) outer radius of a ring
σb outer fibre strain
S0 fourier coefficent
S0,nl fourier coefficent
Sn,1 fourier coefficent
Sn,2 fourier coefficent
τ time
t time
∆t time increment
t̃ time
t0 time zero
t̃0 time zero
T time
Tg temperature
Tmax maximum time in calculation
u radial displacement
uB deflection at the beam end
v velocity
ṽ velocity
vM mean velocity
ṽM mean rod velocity
V0 fourier coefficent
Vn,1 fourier coefficent
Vn,2 fourier coefficent
V velocity, slow time scale
240
w wear in the deformed state
ŵ wear in the undeformed state
w̃ wear in the deformed state
w∞ ultimate wear limit (deformed state)
w0 wear asymtotic expansion, first part, fast time scale
w1 wear asymtotic expansion, second part, fast time scale
w2 wear asymtotic expansion, third part, fast time scale
W wear slow time scale
W0 wear asymtotic expansion, first part, slow time scale
W1 wear asymtotic expansion, second part, slow time scale
z axial coordinate
zmax maximum distance from neutral fibre to outer fibre

∂pA
∂Ac
contact pressure change due to area change
∂pf
∂Ac
contact pressure change due to fluid pressure change
List of Figures
Figure 2.1. Compressor types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Figure 2.2. Cut of a compressor including the piston rings, the rider rings, the
pressure packing, the intermediate packing and the wiper packing.
All packings consist of multiple rings. The photo taken is from [63]. . 4
Figure 2.3. Piston ring designs: a) uncut ring, b) one piece radially straight-cut
ring, c) angle-cut rings: c1) one piece, c2) two piece or c3) three piece
design, d) one piece overlap/step-cut ring, e) two piece gas-tight cut
ring, f) radially straight-cut ring with pressure-balancing groove, g)
radially straight-cut ring pair. . . . . . . . . . . . . . . . . . . . . . . 6
241
Figure 2.4. Rider ring designs for single piece rings: a) uncut ring, b) radially
straight-cut ring, c) angle-cut ring, d) angular-cut ring with radial
pressure-relieve grooves, e) angular-cut ring with axial pressure re-
lieve grooves, f) angular-cut ring with radial and axial pressure re-
lieve grooves. Similar versions of d) to f) can be found with the cut
and the pressure relieve grooves in straight axial direction. Two-piece
rings have another cut shifted by 180◦ . . . . . . . . . . . . . . . . . . 7
Figure 2.5. Most common pressure breakers: a) single piece, double acting b)
single piece, single acting, and c) three piece, double acting d) three
piece, single acting. The designs can have either i) bore clearance
and no gap between the segments or ii) no bore clearance and gap
between the segments for three piece rings. The bore clearance or gap
are small compared to the ring’s other dimensions. . . . . . . . . . . . 8
Figure 2.6. Overview of seal rings in a packing: a) 3 piece radial cut ring, b) 3
piece tangential cut ring with wear stop, c) 3 piece tangential to rod
cut ring without wear stop, d) 6 piece tangential cut ring (“bridge”),
e) 4 piece ring design f) balanced cap design (BCD) ring (f1 shows
the pressure side, f2 the sealing side). . . . . . . . . . . . . . . . . . . 9
Figure 2.7. The BCD consists of two cap and two sealing segments. The wear
gap, the pressure balancing groove and the garter spring groove are
marked. A “P” marks the pressure side to prevent reverse installation. 10
Figure 2.8. Overview of backup rings in a packing: a) solid (i.e. single piece,
uncut) backup ring, b) 3 piece radial cut backup ring. . . . . . . . . . 11
Figure 2.9. Oil wiper designs: a) oil wiper with bigger outer diameter to block
splashing oil, b) oil wiper with smaller outer diameter to increase scra-
ping performance, c) and d) profiled oil wiper ring with two scraping
edges each. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Figure 3.1. Mechanisms governing ring wear. a) Gas pressure load on ring pcyl
and pcrk . b) Contact pressure pc,flat assuming perfectly flat contact
surfaces. c) Gas pressure pg in the sealing surfaces is added, which
results from surface micro-roughness. d) The true solid-solid contact
pressure pc,rough (highlighted green) takes the gas pressure drop pg into
account. e) pc,rough to which wear is proportional, if relative motion
occurs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Figure 3.2. Zoom into the dynamic sealing surface. The softer ring contacts the
harder rod, the surface roughness results in a gap. Here all surface
roughness is depicted for the rod. The gap is assumed to be of constant
height and allows to consider the gas flow in a Hele-Shaw configuration. 27
Figure 3.3. Schematic pressure function over time, for some cycles a zoom is shown
on the right. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Figure 3.4. a) Pressure over crank angle for the first and second sealing element.
b) Rod speed over crank angle. For both diagrams 0◦ is the head end
centre point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Figure 3.5. Schematic sketch of the different load changing situations. . . . . . . . 35
242
Figure 3.6. Comparison of the cyclic pressure for the linear (red) and non-linear
(blue) wear equation including the resulting “mean” values for both
models, S0 and S0,nl respectively. . . . . . . . . . . . . . . . . . . . . 40
Figure 3.7. Validation of the model: The diagram shows the accumulated wear
(per ring height) over the number of revolutions. The calculation of
the fast time scale t is drawn as red line, every tenth calculated point
on the slow time scale T is marked with a blue circle. The smaller
plot shows the wear over the 107 th revolution, here each of the 36
calculated points is marked with a red plus. . . . . . . . . . . . . . . 43
Figure 3.8. The change in the given pressure load (dotted line) and the resulting
contact pressure over time, calculated on both time scales. The zoom
shows the contact pressure of the 107 th revolution on the fast time
scale. A plus marks every calculated point in time. . . . . . . . . . . . 44
Figure 3.9. Time scale model comparison for the accumulated wear (per ring
height) over the number of revolutions for the changing load presented
in Figure 3.8. The calculation of the fast time scale t is drawn as red
line, every tenth calculated point on the slow time scale T is marked
with a blue circle. The smaller plot shows the wear over the 107 th
revolution, here each of the 36 calculated points is marked with a red
plus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Figure 4.1. Friction force tester. Left: Sketch. Right: Photograph of the mounted
friction force tester on the standard testing machine. . . . . . . . . . . 48
Figure 4.2. Side and top view of the friction testing device, including a cut, where
all parts are named. (1) pin hole, fixture to the force measurement de-
vice, (2) rod, (3) top flange, (4) T-cup, (5) bottom flange, (6) columns,
(7) bottom fixture to the testing machine, (8) screw, (9) pressure inlet,
(10) O-ring groove, (11a) upper ring, (11b) lower ring. . . . . . . . . . 49
Figure 4.3. The blue dashed lines mark the roughness measurement positions.
For the steel rod on all 6 positions a reading is taken. The WC rod is
measured on position 1 to 5, since position 0 is not coated. . . . . . . 50
Figure 4.4. Ra measurements on the steel rod at the locations marked in Figure 4.3. 51
Figure 4.5. Ra measurements on the WC rod at the locations marked in Figure 4.3. 51
Figure 4.6. The 20 piece ring sketched in 3D and its drawing, including a cross
section of a ring segment. . . . . . . . . . . . . . . . . . . . . . . . . . 52
Figure 4.7. The mounting device used for the segmented ring. From left to right:
Sleeve, spacer and segmented ring. . . . . . . . . . . . . . . . . . . . . 53
Figure 4.8. The mounting device with the assembled ring on a rod. The garter
springs are not depicted. . . . . . . . . . . . . . . . . . . . . . . . . . 53
Figure 4.9. The segmented ring is mounted on a rod, held by two garter springs.
In the background the T-cup with the upper ring groove is visible. . . 54
Figure 4.10. Sketch of a single ring segment and all acting forces. The force triangle
for the determination of the contact force FS is shown on the bottom. 55
Figure 4.11. Sketch of a garter spring: total and effective spring length. . . . . . . 56
243
Figure 4.12. Force-displacement diagram of the two springs mounted on the seg-
mented ring. The marked point represents the spring load in the
mounted configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Figure 4.13. Frictional force measurement on the steel rod. . . . . . . . . . . . . . 58
Figure 4.14. Frictional force measurement on the WC rod, for every rod speed 10
cycles were run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Figure 4.15. The coefficient of friction and its standard deviation over rod speed. . 61
Figure 4.16. Test rig front view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Figure 4.17. Test chamber showing the counter face fixed on the reciprocating,
heatable block, the displacement sensor and the specimen fixture. . . 63
Figure 4.18. Specimen types: a) shows the bigger 18 mm specimen, b) shows the
smaller 9 mm specimen in a steel adapter. . . . . . . . . . . . . . . . 63
Figure 4.19. Post test: The specimens (black) are still in the steel adapter. On
the counter surfaces a transfer layer is visible. The cross cut on the
counter face is still visible around the transfer layer. The numbers
below the picture are the labels of the stations. . . . . . . . . . . . . . 64
Figure 4.20. The displacement (wear) over time of 4 specimens occurring during
the same run. Station 1 and 2 are on one side of the test rig, station
3 and 4 on the other side. The specimen holders on one side tend to
symmetrically shift their position due to vibrations. . . . . . . . . . . 65
Figure 4.21. The mean displacement (wear) over time is given for each test side on
the tribo rig. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 4.22. Beam model in the undeformed (top) and deformed (bottom) state.
The beam dimensions L, b and h are derived from the BCD’s geometry. 68
Figure 4.23. Geometric values needed to determine the size of the beam. L is
calculated from the mean diameter and the angles α and β. . . . . . . 69
Figure 4.24. Dependence of the geometric BCD ring features on the rod size (equals
D(i) ): the outer diameter D(o) , the radial ring height h and the wear
gap in (mm) and the angles α, β and their combination α + β/2 in
(◦ ). A representative beam length L is calculated with equation (4.6). 70
Figure 4.25. The outer fibre elongation over the deflection of the beam end for
different inner (rod) diameters. The deflection represents the ring
wear. Top: For ring diameters less than 40 mm the bending strains
are bigger than 10 %, if a deflection (wear) of 3 mm is applied. Bottom:
Zooms of the results between 0 and 6 % strain - the bigger the rod,
the smaller the strains. . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Figure 4.26. Elastic modulus dependent on strain rate and temperature. The lines
connect the mean values of multiple data points for each temperature
and strain rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Figure 4.27. Non-linear elastic, temperature dependent material data derived from
5 h tension or compression creep tests. Elastic moduli of E = 350, 700
and 1400 MPa are added to the graph. . . . . . . . . . . . . . . . . . 75
244
Figure 5.1. Working principle within a single wear increment: a) loaded and defor-
med configuration, b) displacement of nodes due to wear (proportional
to contact force), c) bottom: worn configuration, top: worn material
volume. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Figure 5.2. Smoothing algorithms: a) Starting from a loaded and deformed con-
figuration, b) the developed smoothing algorithm moves all the nodes
inside a wear box to prevent element distortion. c) Wear of a single
element without smoothing. d) Effect of the first run of the Abaqus
smoothing algorithm to reduce element distortion. . . . . . . . . . . . 78
Figure 5.3. BCD quarter model: sealing segment (red) and cap segment (blue)
viewed from the crank side. The used cylinder coordinate system
originates in the rod axis and the ring’s axial sealing surface. ϕend
gives the position of the sealing segment’s tip, which changes due to
wear. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Figure 5.4. Flow chart of a wear calculation including the calculation of the pres-
sure distribution in the dynamic sealing surface and setting up a wear-
box [35]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Figure 5.5. Flowchart of the algorithm to find all neighbour nodes in the wear-
box. The counter i iterates over all nodes in the dynamic sealing
surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 5.6. Flow chart for the calculation of the nodal displacement (wear) with
the user-defined subroutine UMESHMOTION. The subroutine is cal-
led for each node in the model in ascending order. . . . . . . . . . . . 85
Figure 5.7. Pressure distribution calculation: a) 3D model of the loaded ring geo-
metry viewed from cylinder side. The red elements are extracted and
the nodes on the inner diameter are used to build the dynamic sealing
surface model. b) Dynamic sealing surface model derived from a).
On the highlighted nodes boundary conditions are applied: red - high
value, blue - low value. c) Resulting (incompressible) gas pressure
distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Figure 5.8. Worn axisymmetric ring at t = t1 . a) Loaded worn ring (pressure po
acting on the outer diameter, contact pressure pc acting on the inner
diameter), assumed to be in state of plane strain. b) Dimensions of
the worn ring in the undeformed state. . . . . . . . . . . . . . . . . . 88
Figure 5.9. Numerical implementation of the analytical model to calculate the
wear on the inner surface of a plane strain uncut ring. . . . . . . . . . 91
Figure 5.10. Case 1: Change of the wear gap (R(i) − D/2), the accumulated wear
w and the contact pressure pc over time for a ring with h0 = 8 mm on
a rod of diameter D = 57.16 mm. The ring wear stops before the ring
is fully worn away. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Figure 5.11. Case 2: Change of the wear gap (R(i) − D/2), the accumulated wear
w and the contact pressure pc over time for a ring with h0 = 5 mm on
a rod of diameter D = 57.16 mm. The ring wear continues until the
ring is fully worn away. . . . . . . . . . . . . . . . . . . . . . . . . . . 92
245
Figure 5.12. Variation of contact pressure pc with inner radius over D/2 < R(i) <
R(o) for different pressure loadings po (case A: po = 30 MPa, case B:
po = 21.2 MPa and case C: po = 10 MPa) . . . . . . . . . . . . . . . . 94
Figure 5.13. Ultimate wear w∞ referred to initial ring height h0 for rings of different
aspect ratios h0 /D that are loaded by the non-dimensional pressure
po /G. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Figure 5.14. Cross section of the algorithm verification models: a) the 2D axial-
symmetric model meshed with 50x50 CAX4 elements and b) the cross
section of the 3D model meshed with C3D8 elements. The wear-box
is shaded red. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Figure 5.15. Verification of the wear algorithm of the numerical FE model against
the analytical (numerically integrated) model for the wear in the de-
formed configuration w and the contact pressure pc for an uncut ring. 96
Figure 5.16. BCD ring model set-up for a 50 mm rod. Analytical rigid bodies are
displayed gray, the sealing segment red and the cap segment blue. Ed-
ges of the cap segment are trimmed to avoid non-physical high contact
pressures, which can lead to numerical instability. The garter spring
groove is modelled as rectangular groove to allow meshing without
tetrahedrons or distortion. . . . . . . . . . . . . . . . . . . . . . . . . 97
Figure 5.17. BCD ring for the 50 mm rod, quarter-model: a) The wear-box is co-
loured yellow for the sealing (red) and the cap (blue) segment. b) The
developed view of the dynamic sealing surface with the view direction
being the negative radial direction. The pressure balancing groove
area is coloured green. On the lower right the clearance between the
sealing segment and the wear stop, half the wear gap, is marked. . . . 99
Figure 5.18. Wear pattern of a (linear elastic) wear calculation of BCD ring on a
50 mm rod after 23.6 d of continuous operation. . . . . . . . . . . . . . 101
Figure 5.19. Wear pattern of the BCD ring quarter model on a 50 mm rod after 6
points in time. Where a) to f) marks advancing time from 6 × 105 s
(166.67 h) to 3.6 × 106 s (1000 h). . . . . . . . . . . . . . . . . . . . . 102
Figure 5.20. a) Wear pattern of a linear elastic wear calculation of BCD ring quar-
ter model on a 50 mm rod after 23.6 d. The pink plane represents
the position of a node path in tangential direction 2 mm from the cup
face, schematically shown in b). c) presents the evolution of the wear
of the picked node path. Blue lines are drawn for time increments of
∆t = 1.2 × 106 s. The red lines indicate every 10th line. . . . . . . . . 104
Figure 5.21. Position of the node, which is used to measure the wear gap. a)
shows the dynamic sealing surface, the sealing segment’s contact area
is coloured red and the cap’s area blue. In b) the black framed detail
of a) is magnified. The wear stop is indicated by a dotted line. The
distance from the considered node to the wear stop is monitored over
time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
246
Figure 5.22. The evolution of the contact force (blue) and the wear gap (green)
are shown over time. When the wear gap closes (0 mm gap) after
17.1 days, the contact force shows a discontinuity. After gap closing
the contact force decreases faster because from then on the ring has
a higher structural stiffness in circumferential direction which allows
it to carry a greater part of the external pressure load. . . . . . . . . 105
Figure 5.23. Wear pattern evolution. The tangential node path at z = 2 mm. The
rod diameter = 28.58 mm. The case names are A1, A2, A3 and A4
given clockwise starting at the upper left plot. . . . . . . . . . . . . . 107
Figure 5.24. Evolution of the integral contact force (blue) and the closing of the
wear gap (green) for the four investigated configurations A1 to A4,
which are placed clockwise starting at the top left, respectively. Rod
diameter 28.58 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Figure 5.25. Wear pattern evolution. The tangential node path is at z = 2 mm.
The rod diameter = 50 mm. The case names are B1 to B4 given
clockwise starting at the top left plot. . . . . . . . . . . . . . . . . . . 109
Figure 5.26. Evolution of the integral contact force of the BCD quarter-model and
the rod (blue) and the closing of the wear gap (green) for the four
investigated BCD configurations B1 to B4. Rod diameter 50 mm. . . . 110
Figure 5.27. Wear pattern evolution. The tangential node path is at z = 2 mm.
The rod diameter = 100 mm. The case names are C1 to C4 given
clockwise starting at the top left plot. . . . . . . . . . . . . . . . . . . 111
Figure 5.28. Evolution of the integral contact force of the BCD quarter-model and
the rod (blue) and the closing of the wear gap (green) for the four
investigated BCD configurations C1 to C4. The rod diameter = 100 mm.112
Figure 5.29. Evolution of the wear gap for the four quarter-models and the three
rod diameters: a) 28.58 mm, b) 50 mm and c) 100 mm. . . . . . . . . . 115
Figure 5.30. Evolution of the integral direction independent contact force over time
for the four quarter-models on the three rod diameters a) 28.58 mm,
b) 50 mm and c) 100 mm. . . . . . . . . . . . . . . . . . . . . . . . . . 116
Figure 5.31. Axial gas pressure distribution in the dynamic sealing surface at two
circumferential positions calculated with and without taking into ac-
count compressibility effects, drawn as solid and dashed lines, respecti-
vely. The lines represent different pressure levels ranging from 6 to
101 bar. a) At ϕ = 10◦ the pressure balancing groove allows the high
gas pressure act between 4 and 8 mm from the cup face. b) At ϕ = 70◦
the gas pressure decreases over cap and sealing surface. . . . . . . . . 118
Figure 5.32. Evolution of the contact force over time for different differential pres-
sures (MPa). The values given are valid for the quarter model. The
crank end pressure is ambient, the pressure load varies from 0.5 to
10 MPa. a) shows the contact force assuming an incompressible gas
pressure distribution in the dynamic sealing surface. b) gives the re-
sults assuming compressible gas properties. . . . . . . . . . . . . . . . 119
247
Figure 5.33. Contact force FC of a single BCD ring on a 50.8 mm rod, depending
on the sealed gas pressure differential and the type of gas pressure
distribution calculation. In the dynamic sealing surface the gas is
assumed to be i) non existent, ii) incompressible and iii) compressible. 120
Figure 5.34. Wear pattern evolution. Black dashed lines represent calculations
with incompressible gas properties. Blue solid lines represent the cal-
culations with compressible gas properties. The plots show three dif-
ferent pressure differentials: a) 3 MPa, b) 6 MPa and c) 10 MPa. The
crank end pressure is at ambient pressure (0.1 MPa). . . . . . . . . . . 121
Figure 5.35. The evolution of the contact force over time and its dependency on
the elastic modulus is shown for two different load levels a) 3 MPa
and b) 6 MPa. The crank end pressure is ambient pressure (0.1 MPa).
The elastic modulus is given in the legend. . . . . . . . . . . . . . . . 123
Figure 5.36. The wear pattern evolution and its dependency on the elastic mo-
dulus is shown for two different load levels a) 3 MPa and b) 6 MPa.
The crank end pressure is ambient pressure (0.1 MPa). The elastic
modulus is given in the legend. . . . . . . . . . . . . . . . . . . . . . . 124
Figure 6.1. Shematic drawing of the test rig, the mounted packing and the 8
temperature measurement positions. [35] . . . . . . . . . . . . . . . . 127
Figure 6.2. Cut through the packing. The temperature measurement positions
are marked and labeled according to Figure 6.1. . . . . . . . . . . . . 128
Figure 6.3. a) a cap segment and b) a sealing segment ready for 3D scanning.
The segments are mounted on a steel fixture on the CMM. For both
segments the origin is set at the intersection of the three planes drawn
in red (yz plane), green (xz plane) and blue (xy plane). . . . . . . . . 131
Figure 6.4. Sketch of the BCD ring geometry. Left: One cap and one sealing
segment are marked with “P” on the high pressure (cylinder) side.
The location of the reference points is shown. Right: Low pressure
(crank) side of the BCD, the two marked segments are on the upper
right. Both sketches show the coordinate system used for displaying
results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Figure 6.5. Raw measurement data from BCD ring segments. The origin is from
the CMM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Figure 6.6. “Pre” configuration: Shifted and rotated data from 3D scans. The
origins are reset between the calculated reference points. The fitted
diameters are drawn as black line, their centre points as yellow filled
circles, the fitted planes are shaded blue. Each calculated reference
point is represented by a filled red circle and results from the inter-
section of a fitted outer diameter and a fitted plane. . . . . . . . . . . 133
Figure 6.7. “Post” configuration: Shifted and rotated data from 3D scans with the
origins are reset between the calculated reference points on the edges.
The fitted diameters are drawn as black line, their centre points as
yellow filled circles, the fitted planes are shaded blue. Each calculated
reference point is represented by a filled red circle and results from
the intersection of a fitted outer diameter and a fitted plane. . . . . . 134
248
Figure 6.8. (a) Overview of the pre and post test data of a BCD cap segment. (b)
The deformation of the cap segment facing the contact area contacting
the sealing segments. . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Figure 6.9. (a) gives an overview of the pre and post test data of a BCD sealing
segment and the position of (b) and (c), which are displayed in higher
magnification. (b) The deformation at the sealing segment’s end is of
the same order of magnitude as the wear. (c) The change of the inner
diameter also causes a permanent deformation of the outer diameter. 136
Figure 6.10. The wear pattern of the ring segments results from subtracting the
post configuration from the pre configuration. . . . . . . . . . . . . . 137
Figure 6.11. Wear pattern of the ring segments viewed in axial direction. The data
is the same as in 6.10. Each line represents an axial position given in
the legend. The dashed lines in (b) are the part of the sealing segment
which is between the high pressure side of the ring and the pressure
balancing groove. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Figure 6.12. Wear of all ring segments. The segments marked with “P” are the
cap segment at 90◦ and the sealing segment at 180◦ . . . . . . . . . . . 140
Figure 6.13. Reproducibility test: The difference between two measurements of a
single BCD ring is depicted. . . . . . . . . . . . . . . . . . . . . . . . 141
Figure 6.14. Reproducibility test: The difference between two measurements of a
single BCD ring are presented. The viewing direction is the axial
direction. For each segment (a) to (d) the data is drawn separately. . 141
Figure 6.15. Protocol of the averaged rod and packing temperature, the leakage
and the gas pressure over time for test A. The area highlighted in
grey between the test time of 15 and 45 h is zoomed and shown in
Figure 6.16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Figure 6.16. Zoom from Figure 6.15. The rings fail and seal repeatedly. Position
1) marks the failing of the rings. The high leakage is accompanied by
pressure loss and an instant drop of the rod temperature. The leakage
rate drops between 1) and 2) until the rings seal again. Meanwhile
the rod temperature increases until a maximum at position 3) is rea-
ched, the rings are still sealing. After 3) the rod temperature slowly
decreases, while the rings seal until the rings fails again at 4). . . . . . 144
Figure 6.17. Worn ring A1: a) shows the worn ring form the pressure side, note
that two segments are sticking together at position 1). To distinguish
between equal segments, the “P”-segments are marked, according to
the convention presented above, see Figure 6.4. In b) and c) the
dynamic contact surfaces are depicted for the sealing segments and the
cap segments, respectively. The contact area of the sealing segments is
very smooth and of homogeneous appearance. The P-cap segment in
c) shows inhomogeneous wear. A region appearing brighter is marked
with 2). On the caps pressure marks of the sealing segments can be
seen, 3). The area of the open wear gap appears brighter and seems
to be covered in either debris or dirt. . . . . . . . . . . . . . . . . . . 145
249
Figure 6.18. Worn ring A2. To distinguish between equal segments, the “P”-
segments are marked, according to the convention presented above,
see Figure 6.4. a) shows the ring from the crank side, the garter spring
is still mounted. A gap between the P-cap segment and the sealing
segments is visible, 1). The sealing segments are touching only in one
place, since they are sticking to the cap segment 2) and 3). In b)
the garter spring is removed and the P-cap segment is taken from the
other segments. It shows pressure marks of the sealing segments. The
wear gap 4) is visible between the pressure marks. It is (naturally)
bigger than the wear gap of the unmounted, unloaded wear gap at 5).
Picture c) was taken before dismounting the garter spring (similar to
a). The spring is visible at 6). The garter spring force is not sufficient
to separate the sticking segments in 2) and 3). . . . . . . . . . . . . . 146
Figure 6.19. Microscopy of a cap segment after test A. The left pictures a1) to
a3) show a cap segment in different magnifications. No sticking was
present between cap and sealing segment at the test end. The right
pictures b1) to b3) show a cap surface after the corresponding sticking
sealing segment has been broken away. The higher magnifications b2)
and b3) show that the surface is covered with thin white fibrils. It is
assumed that these are PTFE fibrils. . . . . . . . . . . . . . . . . . . 147
Figure 6.20. Wear of the ring A1. . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Figure 6.21. Possible life scheme of a ring where the segments start sticking toget-
her. The solid lines show cogent correlations. The dashed lines show
possible correlations. A proper ring seals at the beginning and enters
a loop: it heats up, loses structural stiffness, which increases the con-
tact forces and again the temperature. This cycle may stabilize into
a quasi steady-state. A combination of temperature, contact pres-
sure and time can lead to sticking, breaking the quasi steady-state.
Sticking increases the structural stiffness, thus decreases the contact
forces leading to lower temperatures and again to higher structural
stiffness. This cycle accelerates if leakage occurs. If a ring cools down
rapidly the ring usually fails to seal due to thermally induced strains,
which open leakage paths. Nevertheless the ring can start sealing
again, if time and wear create conformity of ring and rod. . . . . . . . 150
Figure 6.22. Protocol of the averaged rod and packing temperature, the leakage
of both rings and the gas pressure over time for test B. Figure 6.23
presents a zoom into the grey highlighted area. . . . . . . . . . . . . . 152
Figure 6.23. Zoom from Figure 6.22. The rings fail four times for a very short
period, each time an according pressure drop is monitored. The rod
temperature stays constant since the rings start to seal again quickly
after leaking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
250
Figure 6.24. Worn rings B1 and B2, a) and b), respectively. A “P” marks the
P-segments, according to the convention in Figure 6.4. The dynamic
contact surface appears smooth and homogeneous. Their ring seg-
ments were not sticking together when the test ended and the rings
were unmounted. Nonetheless a pressure mark of the sealing segments
can be seen on the cap segments in b3). [35] . . . . . . . . . . . . . . 154
Figure 6.25. Wear of the ring B1. The ring shows an irregular pattern. The P-cap
segment is nearly in its virgin condition, with a maximum wear of
0.10 mm. This indicates that the sealing segments were sticking to
the P-cap segment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Figure 6.26. Wear of the ring B2. The segments are equally worn. The pattern
looks as expected and is used for comparison with a calculation. . . . 155
Figure 6.27. Frictional force measurement with a BCD ring (under given load) on
the steel rod at a relative velocity of 500 mm min−1 . Each pressure
level was tested with 4 cycles. . . . . . . . . . . . . . . . . . . . . . . 156
Figure 6.28. Friction force measurement with a BCD ring on the steel rod at a
relative velocity of 1000 mm min−1 . For each pressure level 4 cycles
were run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Figure 6.29. Friction force measurement with a BCD ring on the WC rod at a
relative velocity of 500 mm min−1 . For each pressure level a minimum
of 5 cycles were run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Figure 6.30. Frictional force FF of a single BCD ring on a 50.8 mm rod, depending
on the sealed gas pressure differential. . . . . . . . . . . . . . . . . . . 159
Figure 6.31. Contact force FC from a single BCD ring on a 50.8 mm rod, depending
on the gas pressure differential. The linear fit approximates all data
points very well. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Figure 6.32. Comparison of simulation and experiment: Contact force FC from a
single BCD ring on a 50.8 mm rod, depending on the sealed gas pres-
sure differential. The experimental data is corrected by the contact
force resulting from the garter spring load. . . . . . . . . . . . . . . . 160
Figure 6.33. Frictional force FF for a single BCD ring on a WC rod with Ø=50.8 mm,
depending on the sealed gas pressure difference. The gas pressure
distribution is calculated with compressible gas. The coefficient of
friction is 0.19. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Figure 6.34. Calculated and measured wear for a BCD ring [35]. Top: Sketch of
the ring’s dynamic sealing surface and the paths A and B at which the
comparison is made. Middle: Path A, representing the cap segment
of the BCD ring. Bottom: Path B, representing the sealing section of
the sealing segment of the ring. . . . . . . . . . . . . . . . . . . . . . 163
251
List of Tables
Table 2.1. PTFE compounds with a single inorganic filler . . . . . . . . . . . . . 18

Table 2.2. PTFE compounds with multiple inorganic fillers . . . . . . . . . . . . 19
Table 2.3. PTFE compounds with another polymer (compound) . . . . . . . . . 19
Table 2.4. Gas dryness and typical examples [32] . . . . . . . . . . . . . . . . . . 22
Table 3.1. Estimation of εt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Table 3.2. Estimation of εL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Table 3.3. Parameters for the validation . . . . . . . . . . . . . . . . . . . . . . . 42
Table 4.1. Detailed velocity dependent measurement data from the determina-
tion of the coefficient of friction for HY54 running on a steel rod or
on a WC rod. The mean value and its standard deviation is given
for the rod movement upwards, downwards and independent from the
movement direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Table 4.2. Test parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Table 4.3. Resulting coefficient of wear. . . . . . . . . . . . . . . . . . . . . . . . 66
Table 5.1. Exemplary cases for ring wear. . . . . . . . . . . . . . . . . . . . . . . 90

Table 5.2. Cases considered while screening the wear behaviour of the BCD ring. 100
Table 5.3. Plot arrangement in the consecutive figures. The symbol ’*’ represents
a wildcard character to be substituted by the letters A, B or C. The
number in the case name can be cross-referenced to Table 5.2. Case
*4 is placed below case *1 which allows to better see the effects of the
pressure balancing groove. . . . . . . . . . . . . . . . . . . . . . . . . 106
Table 5.4. Circumferential position ϕmax of the maximum wear. For A4 two
values are given, since the maximum wear at the calculation’s end
occurs at 0◦ . For comparability the maximum position of the 3rd red
line is given. Case *4 has no pressure balancing groove. The symbol
’*’ represents a wildcard character to be substituted by the letters A,
B or C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Table 5.5. Time in days until the wear gap is closed, for all investigated variati-
ons. Case *4 has no pressure balancing groove. . . . . . . . . . . . . . 114
Table 5.6. Contact force of the ring variations. Case *4 has no pressure balancing
groove. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
252
Table 5.7. Contact force FC for a single (full) BCD ring on a 50.8 mm rod, depen-
ding on the head-end gas pressure. The contact pressure is listed for i)
perfectly flat surfaces assuming no gas pressure in the sealing surface
FC, flat , ii) an incompressible gas pressure distribution FC, incompressible
and iii) a compressible gas pressure distribution FC, compressible acting
in the dynamic sealing surface. . . . . . . . . . . . . . . . . . . . . . . 120
Table 6.1. Detailed wear and wear volume data. The pressure balancing groove
splits a sealing segment’s contact area in two sections. One section
is on the head end side (high pressure side) of the pressure balancing
groove, the other on the crank end side. The results for these areas
are provided separately. . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Table 6.2. Test parameters for test A. . . . . . . . . . . . . . . . . . . . . . . . . 145
Table 6.3. Test B - test parameters . . . . . . . . . . . . . . . . . . . . . . . . . 151
Table 6.4. Contact force FC from a single BCD ring on a 50.8 mm rod, depending
on the gas pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Table 6.5. Measured (and up-scaled) wear volumes for the segments of a BCD
ring after a 359 h test and the calculation result. . . . . . . . . . . . . 162
Table 7.1. Detailed wear and wear volume data for ring A1. . . . . . . . . . . . . 167
Table 7.2. Detailed wear and wear volume data for ring B1. . . . . . . . . . . . . 168
Table 7.3. Detailed wear and wear volume data for ring B2. . . . . . . . . . . . . 169
253
Bibliography
[1] J. F. Archard. Contact and rubbing of flat surfaces. Journal of Applied Physics,
24(8):981, 1953.
[2] M. A. Ashraf, R. Ahmed, O. Ali, N. H. Faisal, A. M. El-Sherik, and M. F. A.

Goosen. Finite element modeling of sliding wear in a composite alloy using a free-
mesh. Journal of Tribology, 137(3):031605–031605–15, 2015.
[3] T. F. Babuska, A. A. Pitenis, M. R. Jones, B. L. Nation, W. G. Sawyer, and N. Ar-

gibay. Temperature-dependent friction and wear behavior of PTFE and MoS2. Tri-
bology Letters, 63(2):15, 2016.
[4] P. R. Barry, P. Y. Chiu, S. S. Perry, W. G. Sawyer, S. B. Sinnott, and S. R. Phillpot.

Effect of temperature on the friction and wear of PTFE by atomic-level simulation.
Tribology Letters, 58(3):856, 2015.
[5] B. Bhushan. Introduction to tribology. John Wiley & Sons, Ltd, Hoboken and New
Jersey, second edition edition, 2013.
[6] T. A. Blanchet and F. E. Kennedy. Sliding wear mechanism of polytetrafluoroethy-

lene (PTFE) and PTFE composites. Wear, 153(1):229–243, 1992.
[7] T. A. Blanchet and Y.-L. Peng. Wear resistant irradiated FEP/unirradiated PTFE
composites. Wear, 214(2):186–191, 1998.
[8] E. M. Bortoleto, A. C. Rovani, V. Seriacopi, F. J. Profito, D. C. Zachariadis, I. F.

Machado, A. Sinatora, and R. M. Souza. Experimental and numerical analysis of
dry contact in the pin on disc test. Wear of Materials 2013, 301(1):19–26, 2013.
[9] B. J. Briscoe, L. H. Yao, and T. A. Stolarski. The friction and wear of

poly(tetrafluoroethylene)-poly (etheretherketone) composites: An initial appraisal
of the optimum composition. Wear, 108(4):357–374, 1986.
[10] D. L. Burris and Sawyer, W. G. A low friction and ultra low wear rate PEEK/PTFE
composite. Wear, 261(3):410–418, 2006.
[11] X. Chen and S. Hui. Ratcheting behavior of PTFE under cyclic compression. Polymer
Testing, 24(7):829–833, 2005.
254
[12] X.-H. Cheng, Y.-J. Xue, and C.-Y. Xie. Friction and wear of rare-earth modified
glass-fiber filled PTFE composites in dry reciprocating sliding motion with impact
loads. Wear, 253(7–8):869–877, 2002.
[13] J.-W. Cho, S. Jeon, H.-Y. Jeong, and S.-H. Chang. Evaluation of cutting efficiency
during TBM disc cutter excavation within a korean granitic rock using linear-cutting-
machine testing and photogrammetric measurement. Tunnelling and Underground
Space Technology, 35:37–54, 2013.
[14] J.-W. Cho, S. Jeon, S.-H. Yu, and S.-H. Chang. Optimum spacing of TBM disc cut-
ters: A numerical simulation using the three-dimensional dynamic fracturing method.
Tunnelling and Underground Space Technology, 25(3):230–244, 2010.
[15] A. W. Christiansen, E. Baer, and S. V. Radcliffe. The mechanical behaviour of

polymers under high pressure. The Philosophical Magazine: A Journal of Theoretical
Experimental and Applied Physics, 24(188):451–467, 1971.
[16] M. Conte and A. Igartua. Study of PTFE composites tribological behavior. Wear,
296(1):568–574, 2012.
[17] E. Cummings. Mechanical packing design and theory of operation, 2011.
[18] Dassault Systmes. Abaqus documentation, 2017.
[19] DEWESoft. version 7.0.6. DEWESoft, Trbovlje and Slovenia, 2012.
[20] H. Domininghaus, P. Elsner, P. Eyerer, and T. Hirth. Kunststoffe: Eigenschaften

und Anwendungen; mit 275 Tabellen. Springer, Heidelberg [u.a], 8., neu bearb. und
erw. aufl edition, 2012.
[21] W. Eifler, E. Schlücker, U. Spicher, and G. Will. Küttner Kolbenmaschinen - Kolben-

pumpen, Kolbenverdichter: Kolbenpumpen, Kolbenverdichter, Brennkraftmaschinen.
Vieweg+Teubner Verlag, 7 edition, 2009.
[22] N. Feistel. Beitrag zum Betriebsverhalten trocken laufender Dichtsysteme zur Ab-
dichtung der Arbeitsräume von Kreuzkopfkompressoren. PhD thesis, Universität Er-
langen, Erlangen and Nürnberg, 2001.
[23] N. Feistel. Performance improvement of dry-running sealing systems by optimiza-

tion of wear compensation. Conference of the European Forum for Reciprocating
Compressors, 9:239–247, 2014.
[24] R. P. Feynman, R. B. Leighton, and M. Sands, editors. The Feynman Lectures on

Physics. Addison Wesley, Boston, 2 edition, 2005.
[25] S. Fouvry, C. Paulin, and T. Liskiewicz. Application of an energy wear approach

to quantify fretting contact durability: Introduction of a wear energy capacity con-
cept. Tribology at the Interface: Proceedings of the 33rd Leeds-Lyon Symposium on
Tribology (Leeds, 2006), 40(10):1428–1440, 2007.
255
[26] K. Friedrich, Z. Lu, and A. M. Hager. Recent advances in polymer composites’
tribology. Wear, 190(2):139–144, 1995.
[27] P. C. Hanlon. Compressor Handbook. McGraw-Hill, USA, 2001.
[28] V. Hegadekatte, N. Huber, and O. Kraft. Finite element based simulation of dry sli-
ding wear. Modelling and Simulation in Materials Science and Engineering, 13(1):57–
75, 2005.
[29] V. Hegadekatte, N. Huber, and O. Kraft. Modeling and simulation of wear in a pin
on disc tribometer. Tribology Letters, 24(1):51–60, 2006.
[30] V. Hegadekatte, S. Kurzenhäuser, N. Huber, and O. Kraft. A predictive modeling

scheme for wear in tribometers. Tribological Contacts and Component Life: Procee-
dings of the 34th Leeds-Lyon Symposium on Tribology, 41(11):1020–1031, 2008.
[31] HOERBIGER Compression Technology. Hoerbiger specification, piston rod recondi-

tioning, 2008.
[32] HOERBIGER Compression Technology. Material selection matrix, 2016.
[33] HOERBIGER Compression Technology. Packing ring standard, 29.11.2017.
[34] C. J. Hooke, S. N. Kukureka, P. Liao, M. Rao, and Y. K. Chen. The friction and
wear of polymers in non-conformal contacts. Wear, 200(1–2):83–94, 1996.
[35] A. Kaufmann and T. Lindner-Silwester. New insight into the wear of packing rings:
model, calculation, experiment. Conference of the European Forum for Reciprocating
Compressors, 11:250–259, 2018.
[36] A. Kaufmann, T. Lindner-Silwester, and T. Antretter. Modelling dry wear of piston

rod sealing elements of reciprocating compressors considering gas pressure drop across
the dynamic sealing surface. Journal of Tribology, 140(4):042201–042201–7, 2018.
[37] J. Khedkar, I. Negulescu, and E. I. Meletis. Sliding wear behavior of PTFE compo-
sites. Wear, 252(5–6):361–369, 2002.
[38] N. V. Klaas, K. Marcus, and C. Kellock. The tribological behaviour of glass filled
polytetrafluoroethylene. Special Issue in Memory of Professor Tony Ball, 38(9):824–
833, 2005.
[39] S.-Q. Lai, L. Yue, T.-S. Li, and Z.-M. Hu. The friction and wear properties of
polytetrafluoroethylene filled with ultrafine diamond. Wear, 260(4):462–468, 2006.
[40] J. Lengiewicz and S. Stupkiewicz. Efficient model of evolution of wear in quasi-

steady-state sliding contacts. Wear, 303(1–2):611–621, 2013.
[41] C. Li and F. Yan. A comparative investigation of the wear behavior of PTFE and
PI under dry sliding and simulated sand-dust conditions. Wear, 266(7–8):632–638,
2009.
256
[42] H.L. Li, Z.W. Yin, D. Jiang, L.Y. Jin, and Y.Q. Cui. A study of the tribological
behavior of transfer films of PTFE composites formed under different loads, speeds
and morphologies of the counterface. Wear, 328–329:17–27, 2015.
[43] T. Lindner-Silwester. Advances in fundamental understanding of the dynamic sealing

action in packing systems. Conference of the European Forum for Reciprocating
Compressors, 5:40–50, 2007.
[44] T. Lindner-Silwester and C. Hold. The BCD packing ring - a new high performance
design. Conference of the European Forum for Reciprocating Compressors, 7:112–119,
2010.
[45] X. Liu, S. Liu, and H. Ji. Numerical research on rock breaking performance of water
jet based on sph. Powder Technology, 286:181–192, 2015.
[46] Z. P. Lu and K. Friedrich. On sliding friction and wear of PEEK and its composites.
10th International Conference on Wear of Materials, 181–183, Part 2(0):624–631,
1995.
[47] O. A. Mamaev, Y. K. Mashkov, R. I. Kosarenko, and V. P. Pivovarov. Durability of

a polymer composite material and service life of packing elements of contact movable
sealing devices. Journal of Friction and Wear, 29(2):127–132, 2008.
[48] F. J. Martı́nez, M. Canales, S. Izquierdo, M. A. Jiménez, and M. A. Martı́nez. Finite

element implementation and validation of wear modelling in sliding polymer–metal
contacts. Wear, 284–285(0):52–64, 2012.
[49] MATLAB. version 9.2.0 (R2017a). The MathWorks Inc., Natick and Massachusetts
and USA, 2017.
[50] H. C. Meng and K. C. Ludema. Wear models and predictive equations: their form
and content. 10th International Conference on Wear of Materials, 181–183, Part
2(0):443–457, 1995.
[51] S. Mukras, N. H. Kim, W. G. Sawyer, D. B. Jackson, and L. W. Bergquist. Numerical

integration schemes and parallel computation for wear prediction using finite element
method. Wear, 266(7–8):822–831, 2009.
[52] I. Páczelt and Z. Mróz. On the analysis of steady-state sliding wear processes. Tri-
bology International, 42(2):275–283, 2009.
[53] I. Páczelt and Z. Mróz. Solution of wear problems for monotonic and periodic sliding
with p-version of finite element method. Higher Order Finite Element and Isogeo-
metric Methods, 249–252:75–103, 2012.
[54] C. Paulin, S. Fouvry, and C. Meunier. Finite element modelling of fretting wear
surface evolution: Application to a Ti–6A1–4V contact. Wear, 264(1–2):26–36, 2008.
[55] M. Peigney. Simulating wear under cyclic loading by a minimization approach. In-
ternational Journal of Solids and Structures, 41(24–25):6783–6799, 2004.
257
[56] Y. M. Pleskachevsky and V. A. Smurugov. Thermal fluctuations at PTFE friction
and transfer. Wear, 209(1–2):123–127, 1997.
[57] C. D. Radcliff. Development of piston rings for reciprocating compressors. In Com-

pressors and their systems 2001, pages 201–212, London, 2001. Professional Engi-
neering Publishers.
[58] P. J. Rae and E. N. Brown. The properties of poly(tetrafluoroethylene) (PTFE) in

tension. Controlled Macromolecular Synthesis and Controlled Architectures - Supra-
molecular Polymer Assemblies, 46(19):8128–8140, 2005.
[59] P. J. Rae and D. M. Dattelbaum. The properties of poly(tetrafluoroethylene) (PTFE)

in compression. Polymer, 45(22):7615–7625, 2004.
[60] M. Rao, C. J. Hooke, S. N. Kukureka, P. Liao, and Y. K. Chen. The effect of PTFE
on the friction and wear behavior of polymers in rolling-sliding contact. Polymer
Engineering & Science, 38(12):1946–1958, 1998.
[61] A. Rezaei, W. van Paepegem, P. de Baets, W. Ost, and J. Degrieck. Adaptive finite
element simulation of wear evolution in radial sliding bearings. Wear, 296(1–2):660–
671, 2012.
[62] S. K. Rhee. Wear equation for polymers sliding against metal surfaces. Wear,
16(6):431–445, 1970.
[63] T. C. Robinson. Reciprocating compressor cutaway display, 2014. Photo is used with
the permission from HOERBIGER.
[64] J. A. Sauer, D. R. Mears, and K. D. Pae. Effects of hydrostatic pressure on the me-
chanical behaviour of polytetrafluoroethylene and polycarbonate. European Polymer
Journal, 6(7):1015–1032, 1970.
[65] A. Schelling and H. H. Kausch. The influence of long term reciprocating dry friction
on the wear behaviour of short fibre reinforced composite materials. Tribology of
composite materials, pages 227–238, 1990.
[66] T. L. Schmitz, J. E. Action, D. L. Burris, W. G. Sawyer, and J. C. Ziegert. Me-

asurement uncertainty in tribological wear rate testing. technical papers-society of
manufacturing engineers-all series-, 2004.
[67] X. Shen, Y. Liu, L. Cao, and X. Chen. Numerical simulation of sliding wear for self-
lubricating spherical plain bearings. Journal of Materials Research and Technology,
1(1):8–12, 2012.
[68] Y. J. Shi, X. Feng, H. Y. Wang, C. Liu, and X. H. Lu. Effects of filler crystal
structure and shape on the tribological properties of PTFE composites. Tribology
International, 40(7):1195–1203, 2007.
[69] W. Shibo, C. Niu, and B. Teng. Tribological behavior of polytetrafluoroethylene:

Effect of sliding motion. Journal of Tribology, 139(1):011301–011301–7, 2016.
258
[70] V. A. Smurugov, A. I. Senatrev, V. G. Savkin, V. V. Biran, and A. I. Sviridyonok.
On PTFE transfer and thermoactivation mechanism of wear. Wear, 158(1–2):61–69,
1992.
[71] H. Sui, H. Pohl, U. Schomburg, G. Upper, and S. Heine. Wear and friction of PTFE
seals. Wear, 224(2):175–182, 1999.
[72] K. Tanaka and S. Kawakami. Effect of various fillers on the friction and wear of
polytetrafluoroethylene-based composites. Wear, 79(2):221–234, 1982.
[73] T. Tevrüz. Tribological behaviours of carbon filled polytetrafluoroethylene (PTFE)

dry journal bearings. Wear, 221(1):61–68, 1998.
[74] C. Thomas. Innenkühlung der Kolbenstange von trockenlaufenden Kolbenverdichtern.

TUDpress, Dresden, 2014.
[75] I. Tzanakis, M. Conte, M. Hadfield, and T. A. Stolarski. Experimental and analytical

thermal study of PTFE composite sliding against high carbon steel as a function of
the surface roughness, sliding velocity and applied load. Wear, 303(1–2):154–168,
2013.
[76] Y. Uchiyama and K. Tanaka. Wear laws for polytetrafluoroethylene. Wear,

58(2):223–235, 1980.
[77] J. M. Urueña, A. A. Pitenis, K. L. Harris, and W. G. Sawyer. Evolution and wear

of fluoropolymer transfer films. Tribology Letters, 57(1):9, 2015.
[78] E. Vanswijgenhoven, S. Cutulic, K. Kenis, G. de Regter, O. Crols, and P. Pennings.

The effect of pneumatic gas pressure on the wear behaviour of PTFE. Wear, 264(7–
8):494–498, 2008.
[79] Z. Wang, K. Kou, Z. Liu, D. Zhang, H. Bi, M. Chao, and Q. Zhao. Structural
characterization and high-temperature compressive creep of PTFE-based composites
filled with inorganic nanoparticles. Polymers for Advanced Technologies, 23(3):545–
550, 2012.
[80] Wikipedia contributors. Hele-Shaw flow, 2018. [online; accessed 29-September-2018].
[81] Wikipedia contributors. Compressor, 5 July 2018 07:47 UTC. [online; accessed
20-July-2018].
[82] D. Xiang and C. Gu. A study on the friction and wear behavior of PTFE filled with
ultra-fine kaolin particulates. Materials Letters, 60(5):689–692, 2006.
[83] F. Yan, Q. Xue, and S. Yang. Debris formation process of PTFE and its composites.
Journal of Applied Polymer Science, 61(7):1223–1229, 1996.
[84] X.-B. Yang, X.-Q. Jin, Z.-M. Du, T.-S. Cui, and S.-K. Yang. Frictional behavior
investigation on three types of PTFE composites under oil-free sliding conditions.
Industrial Lubrication and Tribology, 61(5):254–260, 2009.
259
[85] B. Yu and A. W. Khair. Numerical modeling of rock ridge breakage in rotary cutting.
In 1st Canada - U.S. Rock Mechanics Symposium, page 8, ARMA, 2007. American
Rock Mechanics Association.
[86] Z.-Z. Zhang, Q.-J. Xue, W.-M. Liu, and W.-C. Shen. Friction and wear properties of
metal powder filled PTFE composites under oil lubricated conditions. Wear, 210(1–
2):151–156, 1997.
[87] F. Ziegler. Mechanics of solids and fluids. Springer-Verlag, New York, 2nd ed edition,
1995.
260

Kaufman - Piston Ring Dry Running

Uploaded by

Copyright:

Available Formats

Kaufman - Piston Ring Dry Running

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kaufman - Piston Ring Dry Running

Uploaded by

Copyright:

Available Formats

Chair of Mechanics

Wear of dry-running piston rod sealing rings:

Dipl.-Ing. Andreas Kaufmann, BSc

Matriculation Number: 00835288

Diese Arbeit präsentiert eine Berechnungsmethode zur Verschleißvorhersage für troc-

2 State of the art 3

3 Mechanisms governing wear 23

6 Experimental validation 125

7 Conclusion and Outlook 164

Developed code 170

Acronyms and Symbols 236

List of Figures 241

List of Tables 252

Sealing elements are performance determining components in reciprocating compressors

Common industry practise is to perform experiments on modified test compressors with

State of the art

positive displacement dynamic

reciprocating rotatory axial centrifugal

single double diaphragm scroll ring liquid screw vane lobe

Figure 2.1. Compressor types

discharge discharge intermediate

2.1 Piston rings

2.2 Pressure packing rings

axial groove position

CAP SEGMENT D-D D

garter spring groove D

2.3 Wiper rings

Polytetrafluoroethylene (PTFE) is an excellent tribological material. Its wear properties

The literature can be grouped into several categories:

• The type of motion:

◦ reciprocating (either circumferentially or linearly)

• The counter surface is either

◦ the same material as the specimen or

In comparison to other commodity and engineering polymers, PTFE is a very heavy

• −97 ◦C - glass transition temperature of amorphous PTFE

The crystal transformation temperatures around room temperature have to be kept in

filler type amount (wt%) source annotation

filler type and amount (wt%) source

Table 2.3. PTFE compounds with another polymer (compound)

second polymer amount (wt%) source annotation

In combination with other polymers PTFE is used as a filler to decrease a compound’s

2.5 Gas dryness

Table 2.4. Gas dryness and typical examples [32]

name dew point example compression process

wet > -20 air, chemical industry (CO2 )

Mechanisms governing wear

3.1 Contact pressure

pc,rough = pc,flat − pg . (3.2)

pc,flat cylinder side pc,flat

pcrk pcrk pc,rough

3.3 Consideration of time scales

p̃L (t̃) = p̃cycle (t̃) + p̃p (t̃)

0 90 180 270 360

0 90 180 270 360

Archard’s equation (3.9) can be written in the following form

dw(t) 2π k̃ p̃s ṽM

Table 3.1. Estimation of εt

name value range dimension

pcycle (t) = p̃cycle (t̃)/p̃s (t̃0 ) (3.15)

Table 3.2. Estimation of εL

• Case C) εL εt means the pressure on the ring is constant, Figure 3.5c.