Explosion Analysis of Piperack

Analysis of Explosion Load Effects in
Pipe-racks
Explosion simulation and its respective
structural response on pipe-racks on a
offshore topside module
Aiwei Su
Marine Technology
Submission date: May 2012
Supervisor: Bernt Johan Leira, IMT
Norwegian University of Science and Technology

Department of Marine Technology
Preface
This paper represents the work conducted in the spring semester of 2012, as a part of
TMR4900; Master Thesis at the Institute of Marine Technology, Department of Marine
Structures at NTNU (The Norwegian University of Science and Technology).
I would like to thank my supervisor Prof.Bernt J.Leira for modifying the project text on
the basis of the initial scope of project as proposed by Aker Solutions (Aker Engineering
and Technology) in Oslo, and also for his valuable support and guidance throughout
the work process. I would also like to extend my thanks to the contact persons at
Aker Solutions; Anne Bolstad and yvind Andersen for coming up with suggestions on
the topic descriptions and the overall scope of project. They have also been helpful in
providing me with necessary structural drawings, and answering questions Ive had along
the way.
Finally I would like to thank key personnel at the library at the Marine Technology
Center for their help with locating relevant literature.
The workload of this thesis turned out to be manageable within the time scope of the
spring semester. The completed thesis is roughly separated into three parts; The first
part being a literature study where previous work done by various sources are presented.
The second part is a modeling and analysis part where the pre-processing work is de-
scribed in detail. The third part is a result and discussion part where the post-processing
work is presented. Emphasis in the second and third part of this project, has been on
familiarization with the modeling and analysis capabilities in Abaqus/CAE Standard and
the explicit package (Abaqus/Explicit), together with post-processing and visualization
of results using Python and Matlab scripting.
Aiwei Su, Trondheim. May 28, 2012
i
Abstract
An explosion on a typical offshore installation is a critical, however rare accidental event

which may result in damage and loss of property, and in the worse case loss of lives.
Top-side sections on an offshore installation which are exposed to an explosion blast should
be designed to withstand such an event in a way that further escalation of the damage on
personnel or property is avoided. There is however limited information on the modeling
and analysis procedures for dealing with such events, and most classification rules on
this subject give limited information and guidelines on the practical implementation of
such effects for general analysis purposes. This thesis has been an investigation on these
matters.
The basis of this Master thesis has been on a pipe rack structure from one of Aker
Solutions top-side projects. It has been assumed that this structure is exposed to
an accidental hydrocarbon explosion, and the resulting structural response has been
investigated. A parameter study has been performed by varying blast load levels and
durations. The effects due to the weight of the piping on the structure have also been
studied. The main task of this thesis has been to compare simplified, static analysis
models against dynamic analysis models as to calculate dynamic amplification factors
(DAF), the purpose of this is to better understand the dynamic behavior of the structure
due to the blast loading. The intention is that the dynamic amplification factors are to
be used further by the engineers at Aker Solutions for dimensioning and designing of
similar structures.
Results showed that the blast loads as used in the analyses did not cause any structural
responses within the materials plastic range. Further on, it was found that the magnitude
of dynamic amplification is depending on both the blast duration and the weight of the
pipes on the pipe rack. Typical results indicated that this dynamic amplification was
reduced with increased blast duration.
ii
Sammendrag
En eksplosjon p a en offshore-installasjon er en kritisk, men sjelden ulykkelseshendelse som

kan fre til store deleggelser, skade p
a struktur og i det verste tilfellet tap av liv. Top-side
seksjonen p a en offshore-installasjon som kan vre eksponert for eksplosjonsulykker br
designes pa en slik m ate at eskalering av allerede pafrt skade pa struktur eller personell
unngas. Det er desverre begrenset informasjon a finne om konkrete modellering og
analyseprosedyrer, og de fleste klasseregler p a dette omr adet er svrt generelle og gir
begrenset informasjon og veiledning omkring den praktiske implementeringen av slike
eksplosjonslaster for generelle analyser. Gjennom arbeidet med denne oppgaven har den
praktiske implementeringen av eksplosjonslaster i forhold til dynamiske analyser blitt
underskt.
Utgangspunktet for denne oppgaven har vrt et rrstativ fra et av Aker Solutions
top-sideprosjekter. Det har blitt antatt at strukturen har blitt eksponert for en hy-
drokarboneksplosjon, og den resulterende strukturelle responsen har blitt underskt. Et
parameterstudie har blitt utfrt gjennom a variere eksplosjonslast og varighet. Effekten
av a ha forskjellig vekt pa rrene p
a stativet har ogsa blitt underskt. Hovedfokuset i
oppgaven har vrt a beregne dynamisk amplifikasjonsfaktorer (DAF), med den hensikt a
ke forstaelsen for hvordan strukturen responderer p a eksplosjoner. Disse faktorene vil
videre kunne brukes av ingenirer hos Aker Solutions for dimensjonering og design av
liknende strukturer. Resultater viste av eksplosjonslastene som brukt i analysene ikke
frte til strukturell respons innenfor materialets plastiske omr ade, videre ble det ogs
a
pavist at mengden dynamisk amplifikasjon er sterkt avhengig av b ade eksplosjonsvarighet
og rrvekten p a stativet. Typiske resultater indikerte av den dynamiske amplifikasjonen
ble redusert med kt eksplosjonsvarighet.
iii
Contents
1 Introduction 1
1.1 Explosion incident event tree . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 General introduction on explosion/blast loads . . . . . . . . . . . . . . . 2
1.3 Deflagrations and detonations . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Interaction between blast and structure 6

2.1 Reflection load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Dynamic pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Load on a structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Structural response 12
3.1 Dynamic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 SDOF-methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 MDOF-methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Geometric non-linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Solution methods for non-linear analyses . . . . . . . . . . . . . . . . . . 18
3.4.1 Implicit integration method . . . . . . . . . . . . . . . . . . . . . 18
3.4.2 Explicit integration method . . . . . . . . . . . . . . . . . . . . . 18
4 Blast loads on topside structural elements 19

4.1 Mitigation schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Pipe rack system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Pressure distribution and magnitude . . . . . . . . . . . . . . . . . . . . 26
4.4 Ductility Level Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5 Analysis by Abaqus 30
5.1 Modeling of geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Material settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.3 Structural elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.4 Model mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.4.1 Mesh sensitivity (Convergence study) . . . . . . . . . . . . . . . . 36
iv
CONTENTS v
5.4.2 CPU-time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.5 Mass model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.6 Load model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.7 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.8 Geometric non-linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.9 Damping model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.10 Step settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.10.1 Self-weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.10.2 Blast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.10.3 Post-blast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.10.4 Incrementation settings . . . . . . . . . . . . . . . . . . . . . . . . 53
5.11 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.11.1 Field-Output settings in Abaqus . . . . . . . . . . . . . . . . . . . 54
5.11.2 Integration and section points in Abaqus . . . . . . . . . . . . . . 56
5.11.3 Scripting in Abaqus and Matlab . . . . . . . . . . . . . . . . . . . 57
6 Analysis results 58
6.1 Eigenmodes and eigenfrequencies . . . . . . . . . . . . . . . . . . . . . . 58
6.2 Static analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.3 Dynamic analyses with simplified blast load model . . . . . . . . . . . . . 62
6.4 Dynamic Amplification Factors (DAF) . . . . . . . . . . . . . . . . . . . 66
6.4.1 Stress (Von-Mises) at critical locations . . . . . . . . . . . . . . . 66
6.4.2 Reaction Forces (Support Reactions) . . . . . . . . . . . . . . . . 78
7 Discussion of results 84
7.1 Static and dynamic analyses . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.2 DAF for critical locations on the pipe rack . . . . . . . . . . . . . . . . . 85
7.2.1 Using the upper bound (Method A) . . . . . . . . . . . . . . . . . 86
7.2.2 Averaging over all elements, use the upper bound (Method B) . . 87
7.2.3 Averaging over all elements, use curve fit techniques (Method C) . 89
7.2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.3 DAF for support reactions . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.3.1 Using the upper bound (Method A) . . . . . . . . . . . . . . . . . 94
7.3.2 Averaging over all nodes, use the upper bound (Method B) . . . . 94
7.3.3 Averaging over all nodes, use linear curve fit (Method C) . . . . . 95
7.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8 Conclusion and further work 102

8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.2 Recommendations for future work . . . . . . . . . . . . . . . . . . . . . . 103
Bibliography 105
CONTENTS vi
A Piping weight A. 1
B Response range corresponding to eigenmodes B. 1
C Pipe span chart C. 1
D Python Script for extracting RF (Magnitude) D. 1
E Python Script for extracting RF (X, Y and Z-direction) E. 1
F Variation of DAF against blast duration. F. 1
G Variation of DAF against piping weight (arrangement). G. 1
H Variation of V.M stresses along path/beam. H. 1
I Matlab script: Read in data and calculate DAF. I. 1

List of Figures
1.1 Explosion event tree. From the lecture notes by J.Amdahl. TMR 4195
Design of Offshore Structures, 2009. . . . . . . . . . . . . . . . . . . . . . 2
1.2 Characteristic shape of the pressure diagram for a) shock wave (detonation),
and b) pressure wave (deflagration)[1]. . . . . . . . . . . . . . . . . . . . 3
1.3 Simplified pressure diagram for a) shock wave (detonation), and b) pressure
wave (deflagration)[1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Important factors to be considered when idealising a blast load with respect
to the assumed structural response [2]. . . . . . . . . . . . . . . . . . . . 5
2.1 Reflection coefficients for different values of the overpressure, as functions

of the angle of incidence[1]. . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Schematic representation of the disturbance of the blast for a box shaped
structure [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 A schematic representation of the pressure variation through time is shown
for a finite reflective surface [1]. . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Drag coeffiecients given for various simple structural shapes [1]. . . . . . 10
2.5 A schematic representation of the pressure variation through time is shown
for a finite reflective surface [1]. . . . . . . . . . . . . . . . . . . . . . . . 11
3.1 Simple beam subjected to a static load [3]. . . . . . . . . . . . . . . . . . 13

3.2 The SDOF method implemented on a stiffened plate blast response problem
[4] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Variation of damping ratio with respect to response or natural frequency [5]. 17
4.1 Typical topside module arrangement [6]. . . . . . . . . . . . . . . . . . . 20

4.2 Figure showing the importance of proper ventilation as to reduce the effects
of an explosion[4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 Importance of correct positioning of equipment in case of an explosion
event [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.4 Blast scenario with pressure release. Dynamic pressure attacking perpen-
dicular on the pipe rack. . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 PDMS (Plant design manager system) CAD representation of the process-
ing area and the pipe rack position. Figure provided by Aker Solutions. . 23
vii
LIST OF FIGURES viii
4.6 Isolated view of the pipe rack and neighbouring structural elements, with
dynamic pressure attacking perpendicular to the span of the pipe rack. . 24
4.7 The pipe rack is wielded directly to the upper deck girders as shown. The
pipe rack is also connected to the primary structure (column) as shown
marked with green. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.8 www.reliner.com, showing a typical pipe bracket for the offshore industry. 25
4.9 PDMS (Plant Design Manager System) 3D representation of a part of
the pipe rack with pipes and U-brackets. Visualization provided by Aker
Solutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.10 The weight of a pipe is assuming to be represented as point loads on the
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.11 Simplified pressure diagram as provided from Aker Solutions for the dy-
namic pressure. According to Aker Solutions, the under-pressure will have
an absolute value of about 10% of the initial over-pressure. . . . . . . . . 27
5.1 Stress-strain capacity table, provided by Aker Solutions (from project scope) 32
5.2 Stress-strain capacity curve, provided by Aker Solutions (from project scope) 32
5.3 Euler-Bernoulli beam formulation [7]. . . . . . . . . . . . . . . . . . . . . 33
5.4 Timoshenko beam formulation accounts for the additional shear angle [7]. 33
5.5 Bending of Timoshenko beam [8]. . . . . . . . . . . . . . . . . . . . . . . 34
5.6 Force and moment equilibrium for an infinitesimal segment of the beam. 35
5.7 Figure showing the eigen modes for a particular beam with both ends
clamped. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.8 Showing the variation in global seeding for three different kind of seeding
densities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.9 Showing the two beams used in the stress convergence study. . . . . . . . 38
5.10 Mesh sensitivity for beam 1 and 2. Blast duration 50 ms, 0.2 bar drag
pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.11 Variation in CPU time for different mesh sizes. A trend line was created
based on these data, a power function can be derived from these. This
function gives the required CPU time (y) for a given input x (mesh size). 40
5.12 Inertia properties (point masses) assigned to represent piping weight . . . 41
5.13 Illustration of piping arrangement 1, 2 and 3. . . . . . . . . . . . . . . . 42
5.14 Line load applied on the beam elements in Abaqus . . . . . . . . . . . . 43
5.15 Amplitude toolset (tabular feature) in Abaqus . . . . . . . . . . . . . . . 44
5.16 Triangular pressure load profile corresponding to the input given in the
amplitude toolset (see also diagram provided by Aker Solutions in figure
4.11. Td is the duration of the blast, it is assumed to be between 50 and
200 ms) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.17 Fixed boundary conditions (in all 6 DOFs) are added to the connections
between pipe rack and the rest of the structure . . . . . . . . . . . . . . 45
LIST OF FIGURES ix
5.18 Figure illustrating the damping effect (stiffness proportional damping)

for the lowest eigenmode, position: element 899, Section point (SP) 13.
Von-Mises stress variation [Pa] over blast and post-blast domain. . . . . . 47
5.19 Figure illustrating the damping effect (stiffness proportional damping)
for the lowest eigenmode, position: element 899, Section point (SP) 13.
Von-Mises stress variation [Pa] over blast domain only. . . . . . . . . . . 48
5.20 Figure illustrating the damping effect (stiffness proportional damping) for
the lowest eigenmode, position: Node 152. Reaction force [N] in the blast
direction (Y-direction) over blast domain only. . . . . . . . . . . . . . . . 49
5.21 Stress variation for element 424, 50 ms blast duration, 1 second post-blast
step. Stresses given for 4 integration points. Linearly applied structural
self-weight and piping weight. . . . . . . . . . . . . . . . . . . . . . . . . 51
5.22 Stress variation for element 424, 50 ms blast duration, 1 second post-blast
step. Stresses given for 4 integration points. Instantaneously applied
structural self-weight and piping weight as for piping arrangement 1. . . . 52
5.23 Figure illustrating the effect of aliasing on a sinusoidal curve [9]. . . . . . 54
5.24 Figure illustrating the effect of aliasing for element 424. Aliasing effects
cause a misleading representation of the post-blast response. . . . . . . . 55
5.25 Numbering of integration points for output [10]. . . . . . . . . . . . . . . 56
5.26 Section points for rectangular hollow section (RHS profile), beam in space.
The four standard section points are 1, 5, 9 and 13 [10]. . . . . . . . . . . 56
6.1 High stress area for the cases with pipe rack including piping arrangements
1,2 and 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 High stress area for the case with pipe rack only, no piping weight applied. 61
6.3 Illustration of the structural response during blast and post-blast. . . . . 62
6.4 Stress (High stress areas marked red. Boundary conditions are marked green. 63
6.5 Dynamic response with respect to Von-Mises stresses on element 614 due
to change in piping arrangement (Green curve represents arrangement 3,
blue curve is the pipe rack without piping weight). The characteristic
response peak is seen during the blast for all three cases. . . . . . . . . . 64
6.6 Variation of reaction force in Y-direction (blast-direction) for 4 selected
nodes at end connections. . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.7 The difference between method 1 and method 2 for extracting stresses for
calculating the DAF. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.8 An example of maximum dynamic stress extraction from the blast step.
Section points 1 and 5 (red and blue curve represent the highest stresses
during the blast step). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.9 Variation of DAF with respect to blast duration for method 1 and 2. . . 69
6.10 The locations of the 12 selected elements to be used for the extended DAF
calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
LIST OF FIGURES x
6.11 An example of DAF values obtained from the 50 ms blast case, for ar-
rangement 1. The DAFs are calculated for four section-points separately,
in addition to the average of all section points. . . . . . . . . . . . . . . . 70
6.12 Plot of the variation of DAF with respect to blast duration, for all 4
section points for all 12 elements (0.4 bar, arrangement 3). The high DAFs
represented by the response from element 689 and 702 are clearly expressed
by the graph. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.13 By removing the contribution from element 689 and 702, a more general
picture of the response is seen. It should be emphasized that the response
expressed by these DAFs represents 91.7% of the total number of DAFs
for this particular piping arrangement. . . . . . . . . . . . . . . . . . . . 71
6.14 Plots of the variation of DAF with respect to blast duration for all section
points, for the different piping arrangements. Blast drag pressure 0.2 bar.
It is shown from these figures that the magnitude of the DAF decreases
with increasing blast duration. It should be noted that the data points for
element 689 and 702 have been removed in this figure. . . . . . . . . . . . 72
6.15 Plots of the variation of DAF with respect to blast duration for all section
points, for the different piping arrangements (neglecting the DAFs from
element 689 and 702). Blast drag pressure 0.4 bar. It should be noted that
most DAF values are below 2, with exceptions of some elements in the 50
ms blast case. It is shown from these figures that the DAF decreases with
increasing blast duration. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.16 The following figure illustrates the scattering of data points for DAFs as
line plots. The largest concentration of data points are located roughly
between 0.5 and 2. The extreme DAF points only represent a small number
of points. Please note that the range of scattering is largest for the 50 ms
blast case. When looking at the 50 ms 0.2 bar blast case, 90.6% of the
DAFs are below 2. For the 50 ms 0.4 bar blast case, 85.4% of the DAFs
are below 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.17 The following figure illustrates the scattering of data points for DAFs as
line plots. The once four DAFs for each element is now reduced to one
averaged DAF, this reduced the size of the scattering range of data points.
Once again, it should be emphasized that the majority of DAFs are below
the value of 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.18 The following scatter plots illustrate the scattering of data points for DAFs
(averaged section points). These plots do not contain the data points from
element 689 and 702. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.19 The variation of DAF plotted against piping weight (arrangement) as a
scatter plot. The figure illustrates the scattering of data points. These
plots do not contain the data points from element 689 and 702. . . . . . 78
6.20 Location and node numbering for the extraction of reaction forces. . . . . 79
LIST OF FIGURES xi
6.21 Coordinate system with respect to the direction of the blast, and the
direction of the gravity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.22 Variation of DAF (based on reaction force in X-direction) with respect
to blast duration. Please note that most DAFs are below 10, with the
exception of the DAF for nodes 65, 95, 71 and 158. . . . . . . . . . . . . 81
6.23 Variation of DAF (based on reaction force in Y-direction) with respect to
blast duration. The Y-direction is the direction of the blast. The majority
of the DAFs are found to be below 1.5. . . . . . . . . . . . . . . . . . . . 82
6.24 Variation of DAF (based on reaction force in Z-direction) with respect to
blast duration. The Z-direction is the direction of the gravity. Please note
that the majority of the DAFs are below 1.5. . . . . . . . . . . . . . . . . 83
7.1 Showing the response stress peak observed under blast. The figure also
show that the differences between the lowest and highest stresses (Delta
peak stress) remain almost constant for the different piping weights during
the blast. This indicates that the stress response during the blast may be
independent of the piping weight. . . . . . . . . . . . . . . . . . . . . . . 85
7.2 Upper bound values for DAF, method A. 0.2 bar blast pressure (not
including data points from element 689 and 702). . . . . . . . . . . . . . 86
7.3 Upper bound values for DAF, method A. 0.4 bar blast pressure (not
including data points from element 689 and 702). . . . . . . . . . . . . . 87
7.4 Upper bound values for DAF, method B. 0.2 bar blast pressure. . . . . . 88
7.5 Upper bound values for DAF, method B. 0.4 bar blast pressure. . . . . . 88
7.6 Variation of DAF with respect to blast duration, method C. Please note
that this graph does not include the averaged data points from element
689 and 702. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.7 Variation of DAF with respect to blast duration, method C. The following
graph includes the averaged data points from element 689 and 702. . . . 90
7.8 Variation of DAF with respect to arrangement (piping weight), method
C. The following graph does not include the averaged data points from
element 689 and 702. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.9 Variation of DAF with respect to arrangement (piping weight), method C.
The following graph includes the averaged data points from element 689
and 702. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.10 According to method A, the DAF value can be taken directly from the
upper bound. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.11 According to method B, the DAF value can be taken directly from the
upper bound from the averaged DAF over all nodes. . . . . . . . . . . . . 95
7.12 Linear least squared curve fit for the data points of RF1 (Reaction force
in X-direction), DAF vs. blast duration. . . . . . . . . . . . . . . . . . . 96
in Y-direction; direction of blast), DAF vs. blast duration. . . . . . . . . 96
LIST OF FIGURES xii
in Z-direction; direction of gravity), DAF vs. blast duration. . . . . . . . 97
in X-direction), DAF vs. piping arrangement. . . . . . . . . . . . . . . . 98
in Y-direction; direction of gravity), DAF vs. piping arrangement. . . . . 99
7.17 Nonlinear least squared curve fit for the data points of RF3 (Reaction
force in Z-direction; direction of gravity), DAF vs. piping arrangement. . 99
A.1 Piping weight recalculated as concentrated masses in Excel. . . . . . . . A. 1
C.1 Span chart provided by Aker Solutions . . . . . . . . . . . . . . . . . . . C. 1
F.1 The following graph shows the variation of DAF for the first section points
with respect to blast duration, using method 1 to extract the stresses.
Stress is taken for one element per case only . . . . . . . . . . . . . . . . F. 1
F.2 The following graph shows the variation of DAF for the first section points
with respect to blast duration, using method 2 to extract the stresses.
Stress is taken for one element per case only . . . . . . . . . . . . . . . . F. 2
G.1 Averaged DAF plotted against piping weight (arrangement) for all elements.
0.2 bar blast drag pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . G. 1
G.2 Averaged DAF plotted against piping weight (arrangement) neglecting the
data points for element 689 and 702. 0.2 bar blast drag pressure. . . . . . G. 2
G.3 Averaged DAF plotted against piping weight (arrangement) for all elements.
0.4 bar blast drag pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . G. 2
G.4 Averaged DAF plotted against piping weight (arrangement) neglecting the
data points for element 689 and 702. 0.4 bar blast drag pressure. . . . . . G. 3
H.1 Integrated Von-Mises stress along path, horisontal beam 1. . . . . . . . . H. 1

H.2 Integrated Von-Mises stress along path, horisontal beam 2. . . . . . . . . H. 2
H.3 Von-Mises stress variation along path, horisontal beam 1. . . . . . . . . . H. 3
H.4 Von-Mises stress variation along path, horisontal beam 2. . . . . . . . . . H. 4
List of Tables
1.1 Definition of response domains by various sources . . . . . . . . . . . . . 4
3.1 Damping ratios for different structural systems . . . . . . . . . . . . . . . 17
4.1 Occurence events on the North Sea Platforms in the UK, Norwegian and
Dutch sector for 25 years (1973-97) [11] . . . . . . . . . . . . . . . . . . . 28
4.2 Ductility values for design of topside offshore installations . . . . . . . . . 29
4.3 Damage levels - Ductility ratios . . . . . . . . . . . . . . . . . . . . . . . 29
5.1 Approximate global sizes with its corresponding mesh size for the pipe
rack beam model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2 Damping constants used for the investigation of the damping effect on the
structural response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.1 Results from the static simplified model with 0.2 bar drag pressure for two
different section points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 Results from the static simplified model with 0.4 bar drag pressure for two
different section points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.1 DAF for 0.2 bar drag pressure (DAF vs. blast duration). The values in
parentheses are the DAF accounted for element 689 and 702. . . . . . . . 93
7.2 DAF for 0.4 bar drag pressure (DAF vs. blast duration). The values in
parentheses are the DAF accounted for element 689 and 702. . . . . . . . 93
7.3 DAF for different piping arrangements found by method C. Arrangement
1 being the heaviest loaded condition. The values in parentheses are the
DAFs accounted for element 689 and 702. . . . . . . . . . . . . . . . . . 93
7.4 DAF values for support reactions (reaction forces) in X-direction. The
values in parentheses are the DAF for 0.4 bar drag pressure. . . . . . . . 100
7.5 DAF values for support reactions (reaction forces) in Y-direction (direction
of the blast). The values in parentheses are the DAF for 0.4 bar drag
pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
xiii
LIST OF TABLES xiv
7.6 DAF values for support reactions (reaction forces) in Z-direction (direction
of the gravity). The values in parentheses are the DAF for 0.4 bar drag
pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.7 DAF values for support reactions (reaction forces) in X-direction. The
values in parentheses are the DAFs for 0.4 bar drag pressure. . . . . . . . 101
7.8 DAF values for support reactions (reaction forces) in Y-direction. The
7.9 DAF values for support reactions (reaction forces) in Z-direction. The
Nomenclature
Ps incident peak overpressure

tp positive phase duration
is impulse
Pr reflection pressure
PD dynamic (drag) pressure
ts time in which Pr decreases to Ps and PD
S characteristic dimension of a surface exposed to a blast
U velocity
c0 velocity of sound in air
p0 atmospheric pressure
CD drag coefficient
s air density within a blast
us (t) velocity of combustion products
klm load mass transformation factor
x maximum plastic displacement
xel maximum elastic displacement
ductility ratio, coefficient of kinetic friction
nominal (engineering) stress
true true (cauchy) stress
nominal (engineering) strain
true true (cauchy) strain
M ises Von Mises stress
y , f yield stress, stress in y-direction
ij stresses in the principle directions i=j=1,2,3
E Youngs modulus
I area moment of inertia
FD structural damping force
FE internal elastic force
g structural damping factor
ci characteristic constant for the i th damping mechanism
u velocity vector
xv
LIST OF TABLES xvi
N normal pressure between body and surface

C damping matrix
M mass matrix
K stiffness matrix
mass proportional damping constant
stiffness proportional damping constant
damping ratio
i natural frequency for the i th eigenmode
d damping stress
Del the materials current elastic stiffness
strain rate
t time
r(t) response given time t
shear angle
rotation angle of beam cross-section
M bending moment
Fs shear force
u axial displacement
w deflection of beam axis
x axial (normal) strain
xz shear strain in xz-direction
G shear modulus
xz shear stress in xz-direction
material density
k shear correction factor
curvature
q transverse distributed load
m distributed external moment
fs sampling frequency
fmax maximum frequency of signal
Rdyni dynamic response for element or node i
Rstati static response for element or node i
Abbreviations
AP I American Petroleum Institute

DN V Det Norske Veritas
F EA Finite element analysis
F EM Finite element method
CAD Computer aided design
RHS Rectangular hollow section
GU I Graphical user interface
DOF Degrees of freedom
GBS Gravity based structure
T LP Tension leg platform
C 3P O See-Threepio
R2 D2 Artoo Detoo
M DOD Multi-degree of freedom
SDOF Single-degree of freedom
CF D Computational fluid dynamics
ERP Explosion relief panel
DAL Design accidental load
F LACS Flame accelerator simulator (CFD-Software)
ODB Output database
SP Section point
RF Reaction force
DAF Dynamic amplification factor
xvii
Chapter 1
Introduction
1.1 Explosion incident event tree

An explosion on a typical offshore installation is the result of several incidents which
together create a chain reaction leading up to the final explosion. An accidental release
of a gas or liquid, for instance due to a leakage might ignite immediately, creating a fire.
However if the gas or liquid release is within a confined space, it may lead to formation
of a combustible fuel-air cloud, which after ignition will create a hydrocarbon explosion.
The primary objectives for assessing blast scenarios and to provide blast resistant designs
are [12]:
(a) Personnel safety
(b) Controlled shutdown
(c) Financial considerations
Blast resistant design should provide a level of safety for persons in a building that is no
less than that for persons outside the building in the event of an explosion. Preventing
cascading events due to the loss of control of process units not involved in an accident is
another important objective in blast resistant design. A blast incident in one processing
area should not be allowed to affect the safe operation or shutdown of other units or
areas.
Preventing or minimizing financial losses is another objective of blast resistant design.
Good blast resistant design should, depending on the size and location of the blast utilize
the ductility of the material and cause a limited degree of damage to the structure, such
that it can be either repaired or have a section replaced.
The total explosion event tree is illustrated below:
1
CHAPTER 1. INTRODUCTION 2
Figure 1.1: Explosion event tree. From the lecture notes by J.Amdahl. TMR 4195 Design
of Offshore Structures, 2009.
1.2 General introduction on explosion/blast loads

An explosion is a rapid increase in volume and release of energy in an extreme manner,
usually with the generation of high temperatures and pressure gases. A shock or pressure
wave is created practically instantaneously after an explosion, which is referred to as a
blast. This blast will move at high speed through surrounding air, often in supersonic
speed. If the explosion takes place within a confined space, the effects and damages can
often be extensive. Debris (fragments) from an explosion will fly at high velocity and
impose further damage to the surroundings and in the worst case cause injury or death.
An explosion load is characterized by the following parameters[1]:
Shape of shock or pressure wave
Rise time
Maximum pressure (incident peak overpressure), Ps
Positive phase duration, tp
1
Impulse, is = Ps tp
2
The shape of the shock or pressure wave is dependent on several factors[4]:
Type of fuel and oxidizer
Size and fuel concentration of the combustible cloud

Location of ignition point
Strength of ignition source
Size, location and type of explosion vent areas
Location and size of structural elements and equipment
Mitigation schemes
1.3 Deflagrations and detonations

Explosions are separated into two types; (a) detonations and (b) deflagrations. Defla-
grations are characterized by pressure waves moving in subsonic speeds (typically <
1000 m/s), while detonations are more extreme and represent shock waves moving in the
supersonic range (1000-2000 m/s)[4]. A typical gas explosion is of the deflagration type,
while an explosion of condensed explosives (or a very powerful gas explosion) is in the
detonation range.
Figure 1.2: Characteristic shape of the pressure diagram for a) shock wave (detonation),
and b) pressure wave (deflagration)[1].
A detonation is characterized by an instantaneous pressure rise (no rise time), and often a
negative pressure after tp (positive phase duration). The maximum value of this negative
pressure does not normally play any important role, since this pressure is of a magnitude
much smaller than the positive peak overpressure. A deflagration is characterized by
a rise time, and a slower decrease to zero within the positive phase duration time. A
common simplification of the shape of the pressure diagrams is to linearize the variation,
as shown in figure 1.3.
Figure 1.3: Simplified pressure diagram for a) shock wave (detonation), and b) pressure
wave (deflagration)[1].
When idealizing a blast load, it is suggested by many sources to study the ratio between
the blast duration and the structural eigenperiod. Three different domains, classified as
respectively impulsive, dynamic and quasi-static can be defined with respect to the value
of this ratio. The definitions of the different domains are summarized in table 1.1.
Source Impulsive domain Dynamic domain Quasi-static domain

td td td
Amdahl J. [4] 0.3 0.3 3.0 3.0
T T T
td td td
ERS [2] 0.4 0.4 2.0 2.0
T T T
Table 1.1: Definition of response domains by various sources
It is seen that a great amount of simplifications on the representative pressure history

can be done when dealing with the quasi-static domain, however for an impulsive or
dynamic domain, the requirements on the accuracy of the blast load representation are
much higher. Reference is made to the Blast and Fire Engineering Project for Topside
Structures (Work Package No. BR1/BR2 [2]), and the table as presented in figure 1.4.
Figure 1.4: Important factors to be considered when idealising a blast load with respect
to the assumed structural response [2].
Chapter 2
Interaction between blast and

structure
When a blast meets a structure or an obstacle, it will be disturbed. The disturbance

causes the loading on the obstacle to differentiate from the time-pressure path of the
undisturbed blast, taking a much more complex form which in turn is heavily dependent
on the shape of the structure and direction of the blast wave relative to the structure.
Two effects from blast and structure interaction are important to mention, these are
reflections and dynamic pressures.
2.1 Reflection load

When a blast wave hits a surface, it will be reflected, and a wave moving in the direction
opposite to the incident wave is created. This causes the reflection surface to be loaded
by an overpressure, a reflection load Pr of the reflected wave. This overpressure is larger
than the peak overpressure Ps of the incident wave, and a reflection coefficient is used to
account for this effect:
Pr
rk = (2.1)
Ps
The value of this coefficient is dependent on several factors, including the angle of incidence
between the wave-front and the reflected surface (0-90 deg) and the type of wave (shock
wave or pressure wave). Figure 2.1 shows the reflection coefficients for different values of
the overpressure, as functions of the angle of incidence.
A simplified schematic representation of the disturbance of the blast caused by a box
shaped structure is shown in fig 2.2
6
CHAPTER 2. INTERACTION BETWEEN BLAST AND STRUCTURE 7
Figure 2.1: Reflection coefficients for different values of the overpressure, as functions of
the angle of incidence[1].
Figure 2.2: Schematic representation of the disturbance of the blast for a box shaped
structure [1].
The reflection pressure is often added together with the incident peak overpressure Ps
and the dynamic pressure PD . The time in which the reflected pressure starts to decrease
to Ps and PD is denoted ts .
3S
ts = (2.2)
U
Where S is the characteristic dimension of the surface, while U is the velocity determined
by:
s
6 Ps
U = c0 1+ (2.3)
7 p0
in which c0 is the velocity of sound in air (340m/s), and p0 is the atmospheric pressure.
Figure 2.3: A schematic representation of the pressure variation through time is shown
for a finite reflective surface [1].
2.2 Dynamic pressure

A blast will, when meeting a structure or an obstacle create a dynamic load on the
structure. This is caused by air displacement in the direction of the blast-wave. The
air displacement from a blast is referred to as an explosion wind, and causes a dynamic
pressure given by the following formula [1]:
PD = CD 0.5 s us (t)2 (2.4)

whereby: CD is the so-called drag coefficient, which is dependent on the shape of the
structure (projected area) and its orientation relative to the blast front. Figure 2.4 shows
the different values of CD given for various simple structural shapes.
s is the air density within the blast (kg m3 ), and us (t) is the velocity of the air particles
(combustion products) (m s1 )
Figure 2.4: Drag coeffiecients given for various simple structural shapes [1].
It should be noted that the dynamic pressure exerts the dominant blast effect on open
frame structures, framed structures with frangible cladding, and on small structures or
components such as poles, stacks etc.
2.3 Load on a structure

It is often common to differentiate between three types of extreme conditions when
dealing with blast wave interactions on a structure. Situation (A) in fig 2.5 represents a
case where the blast wave runs over a large surface without hindrance. The load on the
surface will for this case be taken equal to the overpressure PS of the incident wave.
In case (B), the blast wave is on a path perpendicular to a surface of very large dimensions.
There will be minimal rarefaction effects from the edges, and the load on the surface can
be taken as the overpressure from the reflected blast wave Pr .
For case (C) we are dealing with an object of small dimensions, the rarefaction progresses
so quickly that any reflection can be neglected. Furthermore, one can assume that the
pressure difference between the front and back part of the part is so small that only
the dynamic pressure PD is considered. It should be noted that in most structures, a
combination of these three cases must be considered.
Figure 2.5: A schematic representation of the pressure variation through time is shown
for a finite reflective surface [1].
Chapter 3
Structural response
When a structure is exposed to a dynamic (transient) load, such as a blast load, it will
react to this load by deforming. Structures can also vibrate; this is a phenomenon which
is governed by its own natural periods of vibration. Because of the extensive magnitude
of a blast load compared to the loads which the structure is designed to withstand under
normal operating conditions (dead-, live- and wind loads etc.), together with the small
probability that a structure will be exposed to such a load during its life-time, it is
therefore not economical to design a structure for this type of load. Consequently, it is
possible to accept, in practice, a certain degree of damage in the form of ductility. The
definition of ductility will be addressed later in this paper (see section 4.4). For severe
blast cases, the study of plastic deformations will be necessary, hence linear elastic theory
will not be sufficient.
Studying the change in deformations, strains and the maximum stresses over the explosion
time history will be the main task during a dynamic blast assessment.
3.1 Dynamic analysis

Structural dynamic analyses are required to assess certain blast problems. Because both
load and response vary with time, it is evident that a dynamic problem does not have a
single solution as with a static problem. Instead, one must establish a response history
covering all the relevant phenomena which are of interest in the study.
A characteristic feature of a dynamic analysis is illustrated in fig 3.1. If a simple beam is
subjected to a static load p, its internal forces are in equilibrium with the present load.
On the other hand, if the load p(t) is added dynamically, the mass and stiffness of the
structure play a role in the distribution of the internal forces in it. The internal moments
and shears in the beam must equilibrate not only with the externally applied force p(t),
but also the inertia forces resulting from the accelerations of the beam. It should be noted
12
CHAPTER 3. STRUCTURAL RESPONSE 13
that if motions are so slow that the inertia forces are negligibly small, one can treat the
problem as a static problem; however this is usually not the case with a blast problem.
Figure 3.1: Simple beam subjected to a static load [3].
A structural dynamic analysis will be dependent on whether the system is being treated
as a single-degree of freedom (SDOF) system, or a multi-degree of freedom (MDOF)
system. The response of a SDOF system can be evaluated directly from the solution of a
single differential equation of motion, while the MDOF system is made complicated due
to the necessity of discretization. A numerical solution of a MDOF system equation can
be found using non-linear finite element method.
3.1.1 SDOF-methods
SDOF-methods, such as Biggs method is commonly used to assess the blast response
in an early design stage, and allows for easy hand calculations. This method is based
on simplifying the structural problem into a mass-spring system without damping, and
includes a load-mass transformation factor to account for various boundary and load
conditions [4]. These transformation factors are based on approximations to classical
beam and plate theory for deflections in the elastic range and plastic hinge or yield-line
theory in the plastic range. Biggs method can be easily adopted for dynamic analysis
of structural topside components such as floor plates and blast walls subjected to blast
loads[2].
The SDOF system represented in fig 3.2 can be expressed with the following formula [4]:
klm M y + Ky = F (t) (3.1)
Whereas klm is the load-mass transformation factor.

A numerical approach based on the Rayleigh-Ritz method is also applicable. This method
represents a versatile way to formulate a load-deflection curve using deflection functions.
The assumed mode of deflection can be described by one or more component deflections or
modes with unknown magnitudes which, when combined, approximate the exact displaced
shape. Structural models with composite members like stiffened plates can be easily
analysed using this method [2].
Figure 3.2: The SDOF method implemented on a stiffened plate blast response problem
[4]
It is shown that SDOF-methods as mentioned above provide an accurate estimate of the

blast response for stiffened plates. The accuracy of the method is dependent on how good
the resistance function represents the real situation.
3.1.2 MDOF-methods
For scenarios such as a topside frame structure being exposed to a blast load, the
problem can no longer be simplified as a SDOF-system, and more advanced methods
have to be implemented. The most common method is to use the Finite Element
Method (FEM)/Finite Element Analysis (FEA), utilizing its ability to handle complicated
geometries and boundaries.
3.2 Geometric non-linearity

Geometric non-linearity effects are due to the presence of large strain, small strains but
finite displacements/and or rotations, and loss of structural stability. Slender structures
such as bars may experience large displacements and rotations with small strains [13], in
which geometric non-linear effects have to be accounted for if one is to obtain accurate
results. Due to the nature of a large-displacement problem, an updated reference state is
needed for the structure (updated geometry), and stress-equilibrium is thus calculated on
the basis of the latest reference state.
3.3 Damping
When exposing a non-conservative dynamic system to an external force, the presence of
damping forces causes a dissipation of energy in the system which eventually causes a
reduction in vibration amplitude, and ultimately stops the motion. If the energy in a
non-conservative system is to be kept at a constant level, an external source must supply
energy to the system at a rate equal to the rate of energy dissipation.
The most common types of damping are:
Structural damping (hysteretic damping)
Viscous damping
Coulomb damping (dry friction damping)
Structural damping (Hysteretic damping) is due to internal friction, or friction at connec-
tions between elements of a structural system. For an elastic system (a system vibrating
so that the induced stresses are within the elastic range), the ith structural damping
force FD,i is proportional in magnitude to the internal elastic force FE,i (i.e the the forces
caused by stressing of the structure), and opposite in direction to the velocity vector ui .
This expression is expressed by:
FD,i = i g FE,i (3.2)
whereas g is a constant (structural damping factor) and i is the unit imaginary number
[14] [15].
Viscous damping is found in many engineering systems such as shock absorbers, hydraulic
dashpots and sliding of a body on a lubricated surface. This kind of damping is generally
used for systems which are vibrating or moving in a fluid, being air, oil or similar. The
ith viscous damping force is expressed as:
FD,i = ci ui (3.3)
whereas ci is a characteristic constant for the ith damping mechanism, and ui is the
velocity vector [14]. For free vibrations of a one DOF system with a viscous model, the
amplitudes will decay exponentially.
Coulomb damping or dry friction is due to the motion of a body on a dry surface. This
results in a damping force which is constant, only dependent on the normal pressure N
between the moving body and the surface, and , the coefficient of kinetic friction. The
Coulomb damping force is expressed by:
FD = N (3.4)
In the ABAQUS code, a classic viscous damping model called Rayleigh damping is used
to represent the effect of damping. It should be emphasized that Rayleigh damping is
only a generalization of the damping properties of a real structure, and it is used in order
to avoid the need to form a damping matrix based on the physical properties of the real
structure. The formulation of this damping matrix is based on assuming proportionality
to the mass and stiffness matrices through equation 3.5:
C =M +K (3.5)
here is the mass proportional damping constant, and is the stiffness proportional
damping constant.
Relationships between modal equations and orthogonality conditions allow this equation
to be rewritten as equation 3.6 [16]
i
i = + (3.6)
2i 2
Here i is the damping ratio and i is the natural frequency of the ith mode.
The Rayleigh damping model assumes that the damping ratio depends upon the frequency
through the two constants that are usually chosen based upon matching the system
damping to the experimental values at two of the systems natural frequencies [17].
The first term in equation 3.6 is a mass damping term. The factor introduces damping
forces caused by the absolute velocities of the model and so simulates the idea of the
model moving through a viscous ether (a permeating, still fluid, so that any motion of
any point in the model causes damping), in analogy to the viscous damping equation
3.3. This damping factor defines mass proportional damping, in the sense that it gives
a damping contribution proportional to the mass matrix for an element. The second
term in equation 3.6 is a stiffness term. The stiffness proportional damping constant in
ABAQUS introduces damping proportional to the strain rate. is interpreted as defining
viscous material damping in ABAQUS, which creates an additional damping stress, d
proportional to the total strain rate. This is expressed in equation [15].
d = Del (3.7)
whereas is the strain rate. Further wise, it should be noted that Del is defined as the
materials current elastic stiffness in both ABAQUS Standard and the Explicit module.
Figure 3.3 illustrates the contribution from the mass and stiffness damping terms to the
total damping ratio. Generally speaking, the mass proportional Rayleigh damping ,
damps the lower frequencies, while the stiffness proportional Rayleigh damping , damps
the higher frequencies.
Figure 3.3: Variation of damping ratio with respect to response or natural frequency [5].
Simple variations of Rayleigh damping are the mass-proportional damping (=0) and
stiffness proportional damping (=0). During an early screening phase, for any MDOF
systems of a certain complexity, it is difficult to determine meaningful values of the
two damping constants and . In practice, it is therefore more common to tune in
the damping properties to provide a certain level of damping (often given as percent of
critical damping) based on known values for damping ratios from similar structures. The
constants can then be calculated on the basis of this. Table 3.1, provides an overview
over approximate values for , the viscous damping ratio for common materials.
Type of system , damping ratio (% of critical)

Reinforced steel [18] 0.3-1
Unclad welded steel structures (e.g. steel stacks) [19] 0.3
Unclad bolted steel structures [19] 0.5
Steel material (bare structure) [20] 0.1-0.2
Table 3.1: Damping ratios for different structural systems
Further discussion upon damping in Abaqus are found in section 5.9.

3.4 Solution methods for non-linear analyses

In FEA (Finite Element Analysis), direct numerical integration methods are used to
find approximate solutions to the non-linear response problems. These are considered
standard solution methods in e.g Abaqus CAE. The direct numerical integration methods
are separated into two methods; implicit and explicit methods.
3.4.1 Implicit integration method
Implicit methods imply that the displacements at the new time step, t + t, are expressed
by the velocities and accelerations at the new time step, in addition to the historical
information at previous time steps. Many of the implicit methods are unconditionally
stable for linear systems; there is no mathematical limit on the size of the time increment
that can be used to integrate a linear system. The method can be expressed with equation
3.8.
r(t + t) = f {r(t), r(t

+ t), r(t + t), r(t),
r(t), ...} (3.8)
The implicit integration method is the basis solution method for Abaqus CAE, i.e the
Abaqus (Standard) Implicit Package.
3.4.2 Explicit integration method
The explicit scheme is based on obtaining the displacements at a new time step, t + t by
the displacements, velocities and accelerations of previous time steps, as seen in equation
3.9 [21].
r(t + t) = f {r(t), r(t),

r(t), r(t t), r(t
t), r(t t), ...} (3.9)
It should be noted that the central difference operator, which is the most commonly
used explicit operator for stress analysis applications in the Abaqus Explicit Package
(Abaqus/Explicit), is only conditionally stable; the stability limit being approximately
equal to the time for an elastic wave to cross the smallest element dimension in the model
[22]. The applications for explicit methods are thus limited to problems of very short
durations, i.e impulse load problems or similar.
Chapter 4
Blast loads on topside structural

elements
A topside module is a large, pre-built unit that includes accommodation, production and
drilling zones which is typically built separate from the rest of the substructure, which
can be either a TLP, jacket, FPSO or Semi-submersible. The topside module will be
installed on top of the rest of the substructure during the late phase of assembly.
A typical topside module is designed as a bearing system consisting of a system of
trusses, beams and stiffened plates. The purpose of the structure is to support as much
equipment as possible while keeping the overall self-weight to a minimum, the design
should also withstand unique load combinations which include both the pre-service
conditions (fabrication and installation) and the in-service conditions (environmental,
functional and live-loads). Accidental loads such as blast loads are treated separately, as
these often require extensive non-linear structural and CFD analyses.
The typical arrangement of a topside module is shown on fig 4.1. A topside module can
be looked on as a compact process plant, and it should be noted that the module is
separated into zones, spread over several deck levels. A structural steel framing supports
the entities of the process areas which consist of: the wellhead area at which pipes carrying
hydrocarbon fluids from the reservoir terminate; the production separators in which these
fluids are separated; and the compression facilities at which the gases from the separators
are compressed, dried, purified and prepared for export.
19
CHAPTER 4. BLAST LOADS ON TOPSIDE STRUCTURAL ELEMENTS 20
Figure 4.1: Typical topside module arrangement [6].
One of the worst scenarios that can occur on an offshore installation is a hydrocarbon
explosion. In an unconfined area, the pressure waves from the blast are emitted in all
possible directions in a matter of milliseconds, capable of travelling over long distances
and creating widespread damage. In a confined area, there is a large risk of increased
over pressures which are capable of causing extensive localized damage to both equipment
and structure. Entities having large surfaces such as walls will attract high loads from
an explosion. For small entities such as piping lines or cable trays, the effect of blast
pressure is at minimum, but the blast waves may drag and pull these entities off the
supports, leading to possible rupture which may escalate the initial event [23].
Normal safety conventions require the installation of blast walls (corrugated or stiffened
plates) around these respective areas [24], in which the purpose is to minimize the damage
of a possible explosion event.
One should bear in mind that the topside module is an extremely complex structural
system, consisting of structural members, piping, equipment, cables and other appurte-
nances that can obstruct the free movement of the pressure waves. To be able to calculate
the response for such a complex case, it is necessary to do simplifications. It should be
emphasized that these simplifications should be done in such a way that a certain level of
conservatism is ensured.
4.1 Mitigation schemes

The magnitude of the damage caused by a blast can be reduced by simple mitigation
schemes easily implemented in an early design phase of a topside module. These miti-
gation schemes are especially important for enclosed spaces, such as in a process area.
Ventilation openings and special designed release ducts can easily help reducing the level
of overpressure from an explosion as seen in fig 4.2, effectively leading the pressure wave
towards a safe direction where it can be left to escape.
Figure 4.2: Figure showing the importance of proper ventilation as to reduce the effects
of an explosion[4].
The placement of any structural component or equipment relative to the blast may also
reduce the damage imposed to these objects. If obstacles are placed in such a way that
they leave an escape route for the pressure wave toward the vent areas, the localized
damage within the confined space may be reduced dramatically.
Figure 4.3: Importance of correct positioning of equipment in case of an explosion event

[4].
Other mitigation systems include ERP (Explosion Relief Panels) and water deluge systems
[25].
4.2 Pipe rack system

A pipe rack is a structure designed to support metallic pipes and cables going from one
area of the topside module to another. The major pipe racks on a typical offshore topside
are concentrated around the processing areas, going from one processing unit to another,
supporting the pipes which transport either hydrocarbons, gases or other chemicals used
for processing and production.
A conventional offshore pipe rack is a steel-based truss structure wielded together by
RHS (Rectangular hollow section) profiles. Due to the ease of production, all wielded
RHS connections have eccentricities. Investigations by [11] on topside structures exposed
to explosions, showed that the ductility of joints played a prominent role as a key of
safety element that can be translated to performances of a structural system.The effects
of connection eccentricities will not be assessed in this paper, but may be an area of
investigation as a part of a further work. In this paper, a structural model without node
eccentricities at the connections is analysed.
A pipe rack is wielded directly on the primary topside framing structure. For the blast
scenario as proposed by Aker Solutions, the main span of the pipe rack is perpendicular to
the dynamic pressure from the blast (See figure 4.4). Strictly speaking, this is considered
an undesirable position, as the pressure is attacking perpendicular to the span of the pipe
rack structure.
Figure 4.4: Blast scenario with pressure release. Dynamic pressure attacking perpendicular
on the pipe rack.
Figure 4.5: PDMS (Plant design manager system) CAD representation of the processing
area and the pipe rack position. Figure provided by Aker Solutions.
All pipes are connected to the pipe rack by special U-brackets, this is shown in figure 4.9
and 4.8. For larger pipes, special brackets are used.
Figure 4.6: Isolated view of the pipe rack and neighbouring structural elements, with
dynamic pressure attacking perpendicular to the span of the pipe rack.
Figure 4.7: The pipe rack is wielded directly to the upper deck girders as shown. The
pipe rack is also connected to the primary structure (column) as shown marked with
green.
Figure 4.8: www.reliner.com, showing a typical pipe bracket for the offshore industry.
Figure 4.9: PDMS (Plant Design Manager System) 3D representation of a part of the
pipe rack with pipes and U-brackets. Visualization provided by Aker Solutions.
Due to the fastening method of the pipes using simple brackets, it should be reasonable
to assume that these pipes do not contribute significantly to the overall stiffness and
structural integrity of the pipe rack system. The weight of the pipes on the pipe rack
structure on the other hand, will be of an importance. Because the number of pipes,
including their dimensions and weights will be subjects to changes through the project,
simplifications on the piping weight and arrangement has been done. For all analyses,
the arrangements as shown in figure 5.13 are used.
Approximate weights of the main pipes have been calculated using a weight table given
by Aker Solutions (see appendices A and C). The pipe weight will be distributed over the
span of the pipe rack, assuming the weight to be represented by a point mass (inertia) on
the corresponding horisontal brace.
Figure 4.10: The weight of a pipe is assuming to be represented as point loads on the
structure.
4.3 Pressure distribution and magnitude

The pipe rack consists of a truss system of RHS (rectangular hollow section) profiles
which is of small dimensions compared to the actual blast. As previously mentioned in
section 2.2 and 2.3, the reflection pressure can therefore be neglected, and the only
relevant pressure to be taken account for in the analysis will be the dynamic pressure
PD . This is a simplification also suggested by [26] for the analysis of blast responses on
primary steelwork such as isolated columns and beams.
As suggested by [27], the explosion pressure distribution can be considered uniformly

distributed for components and sub-structures. This applies for a pipe rack.
A simplified pressure diagram shown in figure 4.11 has been provided by Aker Solutions.
For explosion scenarios, Aker assumes the duration to be in the range of 50-200 ms.
Figure 4.11: Simplified pressure diagram as provided from Aker Solutions for the dynamic
pressure. According to Aker Solutions, the under-pressure will have an absolute value of
about 10% of the initial over-pressure.
As seen from section section 2.2, the dynamic pressure is given as the following expression:
PD = CD 0.5 s us (t)2 (4.1)
If we assume a hydrocarbon gas explosion being in the deflagration range, the velocity of
the combustion particles are to be assumed to be travelling in subsonic speed (less than
343 m/s). According to [28], typical combustion gases found on an offshore processing
facility are ethane, propane, butane and methane. These have vapor densities in the
range of roughly 0.5-2 kg/m3.
Predicting the magnitude of the different variables in the calculation of the dynamic
pressure for a blast scenario is a complicated task. Tabulated values of the drag coefficient,
CD is only given for simple geometries such as a box or a cylinder. The velocity and the
density of the combustion particles during a deflagration blast is very hard to determine
as these depend on many factors. However, simplifications can be made as to give a first
estimate. According to [29], the major portion of the loading on an open-frame structure
is the drag pressure contribution. The drag coefficient CD for an individual member such
as a I-beam or rectangle is about 1.5, however when the whole frame is considered it has
been suggested to reduce CD to 1.0. This is because various members shield one another
to a certain extent from the effects of the full blast loading.
If we assume the explosion gas to be represented by s = 0.5 kg/m3 , and with an air
velocity just under subsonic (340 m/s), CD equal to 1.0 we get a dynamic pressure of
roughly 0.3 bar. The proposed design value (DAL) for the dynamic pressure as given
by Aker Solutions is between 0.2 and 0.4 bar (These pressure values are obtained from
blast analyses by an external risk consultant company using Gexcons FLACS software).
We see that simple hand calculations of the dynamic pressure, based on the assumed
parameters give a reasonable value which is in the same magnitude as the one proposed
by Aker.
Table 4.1 shows data obtained over a time period of 25 years (1973-97) from various blast
events and the corresponding recorded overpressure level. These blast values are in the
same range as provided by Aker.
Overpressure (bar) 0.2 0.2-1 1-2 2

All events (total 34) 24 8 2 0
Annual exceedance freq. 2 104 4 10 -2 104
5
2 10 -4 105
5
2 105
Table 4.1: Occurence events on the North Sea Platforms in the UK, Norwegian and Dutch
sector for 25 years (1973-97) [11]
4.4 Ductility Level Analysis

A ductile material is defined as a material which allows plastic deformation without losing
its strength before fracture occurs, while a brittle material allows little or no plastic
deformation. In general, structures must be designed to behave elastically under normal
loading conditions (operational conditions), while plastic displacements are allowed during
a blast scenario. A measure of the maximum plastic displacement is given by the ductility
ratio defined by:
x
= (4.2)
xel
Where x is the maximum plastic displacement and xel is the maximum elastic displacement.
Two important levels of assessment are defined following the API (American Petroleum
Institute) guidelines concerning blast load design, these are the so-called Design Level
analysis, and the Ultimate Strength analysis (Ductility Level analysis) [26]. The
Design Level analyses are conventional linear elastic load cases similar to those used
routinely for design with the normal code based elastic failure criteria applied. For a
Ductility Level analysis, critical ductility values are to be calculated and studied.
Ductility values that have been used for designing topside offshore installations are as
follows according to [30]:
Structure type Ductility value or deg

Load bearing panels 1 (BP and Exxon)
Beams and non-load bearing 5 (BP and Exxon)
Rotation beam-column 2 or 12 deg (BP and Exxon)
Secondary structures 1.5 (Brown and Root)
Blast wall 2.5 (Brown and Root)
Steel beam rotation (compact section) 20 or 12 deg(Mobil)
Plates rotation 20 or 12 deg (Mobil)
Table 4.2: Ductility values for design of topside offshore installations
Yasseri [31] proposed the following ductility criteria based on damage levels, shown in
table 4.3.
Ranking/Damage level Acceptable limits

High 6
Substantial 46
Moderate 24
Light 12
Negligible 1
Table 4.3: Damage levels - Ductility ratios
The damage levels from table 4.3 correspond to the following descriptions:
(a) High: Beyond repair and total loss
(b) Substantial: Repairable damage for main load bearing system but major loss for
equipment with shutdown
(c) Moderate: Repairable damage and interruption to production; resume production
after inspection, repair and replacement
(d) Light: Minimum repairs and no interruption to production
(e) Negligible: No repairs required
Chapter 5
Analysis by Abaqus
Non-linear dynamic analyses are required to assess the blast response problems because of
the dynamic nature of the blast load, together with the expected plastic response from the
structure. The finite element (FEA) software Abaqus CAE (version 6.10-2) has been used
for pre- and post processing. The Abaqus Implicit module has been used to analyse the
static cases, while the Abaqus Explicit module has been used for the non-linear dynamic
analyses. Abaqus is a versatile FEA software used throughout the world for stress,
heat transfer, and other types of analysis in mechanical, structural, civil, biomedical,
and related engineering applications. Abaqus CAE is capable of pre-processing, post-
processing, and monitoring the processing stage of solving a FEM problem. It should be
noted that the first stage can be done by other compatible CAD software as well, or even
a text editor. However caution should be made with respect to exporting and importing
models from other CAD software, as errors and compatibility-issues can arise if one deals
with complex models or problems.
While Abaqus CAE supports automatic time incrementation within an implicit time
integration algorithm (Hilber-Hughes-Taylor) for nonlinear dynamic analysis, the Abaqus
Explicit Package uses the explicit time integration algorithm. Abaqus Explicit Package
is thus more suitable for the analysis of brief transient dynamic events, such as the one
being assessed in this paper. For a description on the incrementation settings used for
the pipe rack blast model, reference is made to section 5.10.4.
It should be noted that two different models were made; one static model and one dynamic
model. While these two models have the same geometry, boundary conditions, masses and
mesh densities, the static model features a linear elastic material. The dynamic model
has an elasto-plastic material assigned to the structure and a time-varying load model.
Parameter studies based on varying the magnitude of the blast load, blast duration and
the weight of the piping were performed, and a direct comparison between the static
model and the more realistic dynamic model was done. The results were studied and a
connection between these two models have been investigated through the calculation of
30
CHAPTER 5. ANALYSIS BY ABAQUS 31
dynamic amplification factors.
5.1 Modeling of geometry

The pipe rack structure was modeled as wire elements with given sectional profiles
corresponding to the correct dimensions as from the drawings provided by Aker Solutions.
Five different RHS (Rectangular hollow section) profiles were used:
RHS 200x200x8
RHS 200x200x10
RHS 200x100x8
RHS 200x200x12.5
RHS 350x350x12.5
Both beams and truss elements are used in two and three dimensions to model slender,
rod-like structures in Abaqus CAE. Beam and truss elements represent structures in which
the cross-section is assumed to be small compared to the length. The main difference
between these two is that the beam element provides axial strength and bending stiffness,
while the bending stiffness is neglected for the truss element [32]. For the analysis, all
wire elements have been assigned beam section properties. The non-linear geometry
(Nlgeom) setting has been applied to all elements.
5.2 Material settings

Elastic material properties have been assigned to the static analysis model, while material
non-linearity has been accounted for in the dynamic model, by establishing an elasto-
plastic material in Abaqus. The classical metal plasticity model in Abaqus is intended
for applications such as crash analyses, metal forming, and general collapse studies [32].
It should also be noted that the plasticity model in Abaqus must be used in conjunction
with the linear elastic material model, and that true (Cauchy) stress and log strain values
are to be used [33].
Nominal stress (engineering stress) is calculated assuming the cross-sectional area not
to change during deformation; this is a valid approach if the expected deformations are
small. However for a case with large deformations, the change in cross-sectional area
is significant and true (Cauchy) stress has to be implemented. We see from figure 5.2
that both nominal and true stresses are almost identical and linear elastic up to the yield
strain. After this point, the true stress is larger than the nominal stress when the strain
increases. This is due to the reduction of the cross sectional area due to deformation.
The equations that convert nominal stress to true stress are as follows [34]:
true = (1 + ) (5.1)
true = ln(1 + ) (5.2)
In the elasto-plastic non-linear material model, it has been suggested by Aker Solutions
to use the table as from figure 5.1 and 5.2 for Carbon Steel grade S355.
Figure 5.1: Stress-strain capacity table, provided by Aker Solutions (from project scope)
Figure 5.2: Stress-strain capacity curve, provided by Aker Solutions (from project scope)
5.3 Structural elements

It should be noted that while Abaqus/Implicit supports both Euler-Bernulli (B23, B23H,
B33, and B33H) and Timoshenko (B21, B22, B31, B31OS, B32, B32OS) beam elements,
Abaqus/Explicit supports the latter type only.
The Explicit linear beam element B31 has been assigned to the model. The B31 element is
a 2-node linear beam in space based on the Timoshenko beam formulation. Unlike Euler-
Bernoulli beam elements, Timoshenko beams also incorporate a first order correction for
transverse shear deformation effects. The kinematic assumption from the Timoshenko
beam theory says that a plane section originally normal to the centroid remains plane,
but in addition also shear deformations occur. This is in contrast to Euler-Beam theory
where the plane section normal to the midline is assumed remaining unchanged during
deformation. An additional angle will be introduced in the Timoshenko beam theory
due to the additional shear effect, this is defined as the shear angle. is the rotation
angle of the normal. The slope of the midline for will be w0 (x) = + . Reference is
made to figure 5.3 and 5.4 for a comparison between these two basic beam theories.
Figure 5.3: Euler-Bernoulli beam formulation [7].
Figure 5.4: Timoshenko beam formulation accounts for the additional shear angle [7].
The effects of the shear contribution can be shown by deriving the bending moment M
and Fs , the shear force in the cross section. This will be done following an approach as
presented by Yao. W and Zhong. W [8].
Figure 5.5: Bending of Timoshenko beam [8].
If the x-axis is defined along the undeformed beam axis, and the xz-plane is taken as the
deflection plane as shown in figure 5.5, the displacements u(x, z) and w(x,
z) at any point
(x,z) on the beam can be described by two generalized displacements, i.e the deflection of
the beam axis w(x)

and the rotational angle of the beam cross-section (x). These are
expressed by the equations in 5.3.

u(x, z) = z (x), w(x,
z) = w(x)
(5.3)
By using the strain-displacement relation [35] the following strain components are ob-
tained:
d dw
x = z , xz = (5.4)
dx dx
If the stress-strain relation is used, and the effect of the stress quantities from y is
neglected, the two expressions are further rewritten:
d 1 dw 1
z = x , = xz (5.5)
dx E dx G
By multiplying the first equation in 5.5 by z on both sides and integrating over the
cross-section area on both sides, and repeating this integration for the second equation,
equation 5.6 and 5.7 are obtained.
d
Z
1 1
I = zx dA = M (5.6)
dx E E
A
Z
dw 1 1
A( ) = xz dA = Fs (5.7)
dx G G
A
Here I is the moment of inertia about the y-axis of the cross-section of the beam, A
is the cross-sectional area of the beam and E is the Youngs modulus. G is the shear
modulus. The final expressions for the bending moment M and the shear force Fs in the
cross section are:
= EI
M , Fs = kGA (5.8)
w
Whereas
(x) = , and = .
x x
When deriving these formulas, an assumption was made that the shear stresses over the
cross-section remained constant. This is not correct in real-life, hence a shear correction
factor k is introduced. This factor is dependent on the assumed shear stress distribution
over a cross-section. It should be noted that k implies approaching the classic
Euler-Bernoulli beam solution, where the shear stress contribution is neglected.
Figure 5.6: Force and moment equilibrium for an infinitesimal segment of the beam.
By further using moment and force equilibrium on an infinitesimal segment of the beam,
the strong form of the equation of motion in terms of displacements can be derived. The
final expression is given as equation 5.9. The transverse distributed load is represented by
q, m
is the distributed external moment. Reference is made to [8] for the full derivation.
2 w

w
[kGA( )] + q = A 2

x x t (5.9)
w 2
(EI ) + kGA( ) + m = I 2

x x x t
5.4 Model mesh
5.4.1 Mesh sensitivity (Convergence study)
When meshing the model, one need to consider the mode shapes that may be excited in
the response, and choose a refinement of the mesh based on these observations. For a
particular beam with both ends clamped as in figure 5.7, several eigenmodes are to be
expected. The mesh resolution for this case, should be such that it can fully capture both
the 1st and the higher 4th eigenmode.
Figure 5.7: Figure showing the eigen modes for a particular beam with both ends clamped.
One should also bear in mind that designing the mesh for the static case may be different
than designing the mesh for the dynamic case, where the response might be excited by
high-frequency loads. In finite element modeling, a finer mesh will typically result in
a more accurate solution. However, an increase in mesh density will also increase the
computation time. A mesh sensitivity study has been performed on the beam model.
The purpose is to study the stress sensitivity with respect to the stress distribution of a
part of the model, and to find an optimal mesh density which gives adequate results with
the shortest CPU time as possible.
It should be noted that automatic global seeding has been used in Abaqus, which means
that the mesh size is created automatically based on a maximum size criteria given as
an approximate global size-input. Abaqus will mesh the model based on this input, and
ensure that this criterion is met. For a beam model such as the pipe rack, automatic
global seeding will give seeds of constant density, and thus an even mesh distribution over
the global model.
Table 5.1 shows approximate global sizes with its corresponding mesh size (percent of
whole beam length) for the pipe rack model:
Approximate global size Approximate mesh size

(Automatic global seeding) (% of whole beam)
0.4 13-14
0.3 8-11
0.2 7-8
0.1 3-4
0.07 2-3
0.04 1
0.01 less than 1
Table 5.1: Approximate global sizes with its corresponding mesh size for the pipe rack
beam model
Figure 5.8: Showing the variation in global seeding for three different kind of seeding
densities.
The mesh sensitivity study was done based on looking at two horizontal beam elements
(see figure 5.9), and integrating the stress fields over the length of the elements. Due to
the dynamic analysis varying in time, a specific time step (time = 0.9763 s) was chosen.
The stress field integration has been performed for all mesh density cases at the same
time step. Reference is made to appendix H.
Figure 5.9: Showing the two beams used in the stress convergence study.
By studying figure 5.10, it is seen that the integrated stress will converge towards a fixed
value for finer meshes. It is shown evidently, that mesh sizes 0.3 and 0.4 are unsuitable for
such analyses as they will lead to overestimation of stresses compared to the converged
value. For beam 1, mesh 0.3 and 0.4 give an overestimation of the integrated stress of
roughly 20 to 30 percent. For beam 2, this overestimation is between roughly 50 and 100
percent. Such overestimated stress levels are not acceptable. The same figure also shows
that a mesh size 0.1 will cause a slight underestimation of the stress level for both beam
1 and beam 2. Convergence is achieved for finer mesh less than 0.1. For detailed stress
field plots for beam 1 and beam 2, please see appendix H.
Figure 5.10: Mesh sensitivity for beam 1 and 2. Blast duration 50 ms, 0.2 bar drag
pressure.
5.4.2 CPU-time
The CPU Time will increase for finer mesh in the model. Figure 5.11 shows the CPU time
for a given mesh size for the undamped pipe rack beam model. Because a parameter study
is to be performed, a large amount of analyses are to be run. It is therefore essential that
the analysis model has a CPU time of such duration that effectiveness in each analysis
run can be achieved, and still result in adequate results. It is previously shown in section
5.4.1 that mesh sizes below 0.2 are unsuitable to use for the model. From figure 5.11 it is
seen that mesh sizes above 0.1 will cause a rapid increase in CPU time.
Figure 5.11: Variation in CPU time for different mesh sizes. A trend line was created
based on these data, a power function can be derived from these. This function gives the
required CPU time (y) for a given input x (mesh size).
Based on a consideration of a trade-off between computational time and relative accuracy

of results, an approximate global size of 0.2 has been chosen to be used for the automatic
global seeding in Abaqus. The yielded results are expected to be slightly conservative, as
discussed in the previous section 5.4.1.
5.5 Mass model

The structural mass has been represented in Abaqus by applying gravity loads on all
elements. The weight of the pipes attached to the pipe rack has been modelled by applying
inertia (point masses) properties at prescribed points on the beam model. It should be
noted that inertia (point masses) in Abaqus are specified as mass, and not weight [36].
Reference is given to appendix A for an overview of the inertia (point masses) used. It
has been assumed that each pipe spans across the whole length of the pipe rack, thus
distributing its weight evenly on each horisontal beam, this is visualized in figure 5.12.
Figure 5.12: Inertia properties (point masses) assigned to represent piping weight
The piping arrangement on a pipe rack will vary, both with respect to the number of
pipes, sizes and weights, and are prone to changes throughout a project. From a general
perspective, the size of the pipes (including some requirements on spacing between them)
limits the number of pipes which can be added to a pipe rack. Electrical wirings are
present as well, but are negligible in an analysis due to small weight and size. To be able
to solve the problem in the most general way, it is needed to generalize the problem; the
effect on the change of piping will therefore be represented by several assumed piping
configurations, hence named arrangement 1, 2 and 3. These are shown in figure 5.13.
Arrangement 1 represents the heaviest loaded condition, with overall weight decreasing
for 2 and 3 respectively.
Figure 5.13: Illustration of piping arrangement 1, 2 and 3.
5.6 Load model

When using a wire element assigned with a profile, a uniform distributed load is applied
in Abaqus using a line load. A line load is applied as a uniform load over the wire element
(unit=N/m). Because the dynamic pressure is expressed in Pascals, it is multiplied with
a characteristic element RHS profile height to obtain the correct line load. For this
pipe rack, two type of RHS profiles are used; RHS200 and RHS350. Meaning that the
cross-section height of the profiles are 200 and 350mm respectively.
N
Dynamic pressure = 0.2 bar = 2 104 P a ( )
m2
N
Dynamic pressure = 0.4 bar = 4 105 P a ( 2 )
m
The line loads corresponding to these pressures (on RHS350) are thus:
N N
Line load (0.2 bar) = 20 000 2
0.35 m = 7000
m m
N N
Line load (0.4 bar) = 40 000 2 0.35 m = 14 000
m m
The line loads corresponding to these pressures (on RHS200) are:

N N
Line load (0.2 bar) = 20 000 2
0.2 m = 4000
m m
N N
Line load (0.4 bar) = 40 000 2 0.2 m = 8000
m m
The line load has been applied to the whole structure as shown in figure 5.14. Simplification
on the effects of shielding has been taken account for by using a drag coefficient CD equal
to 1.0, as suggested in section 4.3.
Figure 5.14: Line load applied on the beam elements in Abaqus
Because a dynamic analysis is time-dependent, it is necessary to create a time-varying

load profile in Abaqus to correctly represent the effects of the blast load. The amplitude
toolset in Abaqus has been used to define the pressure load profile for the dynamic
analyses. By using the tabular feature, the load profile was defined by directly giving the
input values for a profile amplitude corresponding to a specific time. This is illustrated
in figure 5.15 for the pressure load profile for the 50 ms blast.
Figure 5.15: Amplitude toolset (tabular feature) in Abaqus
Figure 5.16: Triangular pressure load profile corresponding to the input given in the
amplitude toolset (see also diagram provided by Aker Solutions in figure 4.11. Td is the
duration of the blast, it is assumed to be between 50 and 200 ms)
5.7 Boundary conditions

All structural elements wielded directly to the girders on the deck above are assume
fixed in all 6 DOFs. For simplification, the three connections between pipe rack and
the support beams are assumed fixed as well. An overview of the boundary conditions
imposed to the model in Abaqus is found in figure 5.17.
Figure 5.17: Fixed boundary conditions (in all 6 DOFs) are added to the connections
between pipe rack and the rest of the structure
5.8 Geometric non-linearity

Non-linear geometric effects have been taken accounted for both in the static and dynamic
analyses by turning on the Nlgeom feature in Abaqus CAE.
5.9 Damping model

General damping in Abaqus is represented by Rayleigh damping. To investigate the
damping contribution from the stiffness and mass proportional damping term, the pipe
rack model, 50 ms blast duration and 0.2 bar overpressure is analysed in ABAQUS/Explicit.
Abaqus requires input for the mass and stiffness proportional damping coefficients and
. By assuming stiffness proportional damping, the mass proportional damping constant
() is set to zero in the rewritten Rayleigh damping expression (equation 3.6), the result
is equation 5.10.

= (5.10)
2
Solving for yields:

2
= (5.11)

Similarly the mass proportional damping term can be found by setting the stiffness
proportional constant () to zero.
Solving for yields:
= 2 (5.12)
Table 5.2 shows damping constants calculated to study the damping effect (The nat-
ural frequency used in these calculations is taken as the eigenfrequency for the lowest
eigenmode). By using these constants, analyses are run with both mass and a stiffness
proportional damping model respectively. The results from the stiffness proportional
damping analyses for an element are shown in figure 5.18. It should be noted that the
damping effect from the mass proportional damping is of negligible magnitude.
Damping ratio, (percent of critical)

0.1 0.029 1.39e-4
0.2 0.058 2.78e-4
0.3 0.086 4.16e-4
0.4 0.115 5.55e-4
Table 5.2: Damping constants used for the investigation of the damping effect on the
structural response
When comparing the different stiffness proportional damping models over the blast and
post-blast domain (see figure 5.18), it is seen that a damping ratio of 0.1 percent of
critical damping gives the most significant damping effect among all stiffness proportional
damping models. The global Von-Mises stress level from the maximum peak due to the
blast loading in the post-blast domain will be reduced with roughly 70 percent compared
with the no damping case, assuming 0.1 percent damping ratio. When 0.2 percent
damping ratio is used, the reduction is around 50 percent compared to the case with no
damping. Figure 5.19 shows that 0.4 percent of critical damping gives the largest damping
effect when looking at the blast domain only. For this particular case, a reduction of
roughly 7 percent on the Von-Mises stress is observed for the damped model compared
to the undamped one.
Further wise, figure 5.20 illustrates the damping effect on the reaction force (support
reaction) in the direction of the blast (Y-direction). The largest reduction in the reaction
force between the non-damped model and the damped model is roughly 13.
It is shown through these investigations, that if no damping model is used, structural
elements will oscillate forever due to the lack of energy dissipation in the computational
model. Because of the many uncertainties of choosing a proper damping constant,
including the risk of producing non-conservative results, a choice was made to neglect
damping in all dynamic analysis models presented in this paper. This argument is
also backed up by investigations on explosion analyses by Hong and J.Y Richard [37].
According to their research, damping effects can be ignored in explosion analyses, if
conservative stress levels are of interest. In addition, if the structural response happens to
be plastic, Yandzio and Gough showed that in an inelastic analysis, the energy dissipated
through plastic deformation is significantly greater than that dissipated by structural
damping [38].
Figure 5.18: Figure illustrating the damping effect (stiffness proportional damping) for
the lowest eigenmode, position: element 899, Section point (SP) 13. Von-Mises stress
variation [Pa] over blast and post-blast domain.
the lowest eigenmode, position: element 899, Section point (SP) 13. Von-Mises stress
variation [Pa] over blast domain only.
the lowest eigenmode, position: Node 152. Reaction force [N] in the blast direction
(Y-direction) over blast domain only.
5.10 Step settings

Analysis steps in Abaqus are used to separate an analysis into manageable parts. Within
each step, the user is free to control the different loading- and boundary condition
parameters. The initial step is a standard step in Abaqus, which is created automatically
at the beginning of the models step sequence. This initial step cannot be renamed,
edited, replaced, copied or deleted.
The initial step allows the user to define boundary conditions, predefined fields and
other interactions that are applicable at the beginning of the analysis. For the pipe rack
analyses, the boundary conditions are applied in the initial step, and will thus propagate
through the rest of the steps in the analysis. In addition the initial step, 3 other steps
are present in the analysis. These are:
1. Self-weight (gravity load, mass from pipes)
2. Blast (positive and neg. load amplitude). Different step duration depending on
duration of blast (i.e 50, 100, 150 or 200 ms)
3. Post-blast (only gravity load is set active, no propagation of blast load from previous
step)
5.10.1 Self-weight
The Self-weight step is where the gravity (structural self-weight of the structure) and
piping weight is applied linearly on the structure using Abaqus in-built RAMP function.
This function uses 0.5 seconds to linearly increase the structural self weight up to a
quasi-static stress level. This is illustrated in figure 5.21. If the structural weight is loaded
instantaneously, the element will experience oscillatory stress variation before the initial
blast is initiated. This will give a wrong representation of the stress variation during the
actual blast, which is illustrated in figure 5.22. It is shown from the same figure that
the oscillations will reach the maximum global amplitude earlier for the instantaneously
applied self-weight and piping weight, than for the case with linear increase. Both methods
however show a similar stress level for the maximum amplitude during post-blast.
Figure 5.21: Stress variation for element 424, 50 ms blast duration, 1 second post-blast
step. Stresses given for 4 integration points. Linearly applied structural self-weight and
piping weight.
Figure 5.22: Stress variation for element 424, 50 ms blast duration, 1 second post-blast
step. Stresses given for 4 integration points. Instantaneously applied structural self-weight
and piping weight as for piping arrangement 1.
5.10.2 Blast
The Blast step is where the blast drag pressure is introduced. Reference is made to
section 4.3 and 5.6 for a more detailed description of the load itself. The amplitude tool
set (tabular feature) in Abaqus was used to define this amplitude profile. This amplitude
takes account for both the positive and the negative part of the drag pressure profile.
This step has the same duration as the blast itself, i.e 50, 100, 150 or 200 ms depending
on which case that was being analysed.
5.10.3 Post-blast
The Post-blast step is a step without any loads imposed on the structure, but with the
structural self-weight (gravity load) propagating from the previous steps. The purpose of
this step is to study the post-blast response of the structure. An important aspect of this
step is to ensure correct settings for the ODB-field output to avoid aliasing during signal
processing. Reference is made to section 5.11.1.
Determining the length of the post-blast step is a topic of discussion. It is observed
from results that the maximum response amplitude, e.g Von-Mises stresses are higher
during the post-blast step than under the blast step. However, the structures post-blast
response is not important for calculating the DAF, which is the main scope of this thesis.
For calculating stresses to determine DAF-values, the duration of the post-blast step has
been decreased to keep the analysis time to a minimum. For the study of the damping
effect on the response during post-blast, this step has been purposely increased to be able
to fully capture this.
5.10.4 Incrementation settings
The incrementation settings in Abaqus are being set to the automatic configuration
(Automatic time incrementation). This means that Abaqus/Explicit will automatically
adjust the increment size when the analysis is proceeding depending on the stability limit
[39].
Fixed time incrementation is supported by Abaqus/Explicit, but when this is used,
Abaqus/Explicit will not check if the computed response is stable during the step. Validity
of the response has to be checked for each analyses by carefully checking the energy
history and other response variables, this is time consuming and rarely recommended
[39].
5.11 Post-processing
5.11.1 Field-Output settings in Abaqus
The recording of the data during an analysis is done by sampling. In Abaqus CAE, this
is controlled via the Field-Output Editor. To obtain a realistic representation of the
response, and to avoid aliasing (i.e distortion of sampled result data), it is important to
choose a sampling frequency which is high enough to cover all variations, but on the same
time, be of such level that the size of the output file does not become unnecessary large.
The effect of distortion of data due to aliasing is illustrated in figure 5.23. The original
sine wave has a maximum frequency of 1 kHz. In order to define the amplitude and phase
of the original sine wave, it is necessary to use a sampling frequency of at least 2fmax (2
kHz), in which 2fmax is defined as the Nyquist sampling rate according to the Sampling
Theorem [40]. The requirement is given as expression 5.13.
fs > 2 fmax (5.13)
here fs is the sampling frequency, and fmax is the maximum frequency of the signal.
If a sampling rate of 3 kHz is used, which is above the required sampling rate, the data
will be recorded without aliasing. The 1.1 kHz sampling rate however, is below the
required rate, and the resulting curve becomes a misleading representation of the original
sine wave.
Figure 5.23: Figure illustrating the effect of aliasing on a sinusoidal curve [9].
The safest way to ensure that aliasing does not occur in Abaqus is to request output at
each increment. However, this is often impractical as it may create very large output files.
Figure 5.24 shows the effect of aliasing (distortion of sampled data) for element 424. A
low sampling rate may cause the maximum stress peak not to be accurately recorded; it
may also give a very misleading representation of the high-frequency oscillations which
characterizes the post-blast response.
The sampling rate (output rate) has been adjusted for each step in the analysis using the
Field Output Editor in Abaqus CAE. The sampling rate for the self-weight step is low, as
it is expected a linear increase of both stresses and displacements until it reaches the final
quasi-static value. Further wise, the sampling rate has been increased for the blast and
post-blast steps. This is to be able to accurately record the high-frequency variations of
both stresses and displacements during both blast and post-blast.
Figure 5.24: Figure illustrating the effect of aliasing for element 424. Aliasing effects
cause a misleading representation of the post-blast response.
5.11.2 Integration and section points in Abaqus
Abaqus operates with two different definitions for output formats such as the stress; one
being integration point, and the other one being section point. For the 2-noded linear B31
beam element used in this thesis, the integration point is defined as shown in figure 5.25.
Figure 5.25: Numbering of integration points for output [10].
Figure 5.26: Section points for rectangular hollow section (RHS profile), beam in space.
The four standard section points are 1, 5, 9 and 13 [10].
5.11.3 Scripting in Abaqus and Matlab
When the Abaqus/CAE GUI is used to create a model and to visualize the results, com-
mands are issued internally by Abaqus/CAE after each user operation. The programming
language called Python is used by the GUI to create the commands which are sent to
the Abaqus/CAE kernel. Abaqus/CAE has the functionality to open and run custom
scripts written in the Python language, and scripting thus represents a powerful tool
which can automate many tasks which may be repetitive and time-consuming if done
manually through the Abaqus/CAE GUI.
For this thesis, simple Python scripts have been used for extracting result data from the
Output Database (ODB) file and writing these into a text file for further post-processing
in Matlab. An Output Database file contains all the analysis results as requested by the
user when setting up the analysis in Abaqus/CAE. The extracted result data from the
ODB were the reaction forces (RF:Magnitude, X-direction, Y-direction, Z-direction) for
all 23 supports (end connections) on the model, together with the Von-Mises stresses at
critical locations on the model. This was done for all the static and the dynamic analyses.
A Matlab-script was written to calculate the DAF-values, using read-in data from the
.dat files (text files) produced by the Abaqus Python script.
Reference is made to appendix D, E and I.
Chapter 6
Analysis results
This chapter presents the results from the parameter study performed on the pipe rack
using both the static and the dynamic model. A major part of this chapter has been to
investigate the behaviour of the structure during blast by comparing static and dynamic
analysis models. Calculation of dynamic amplification factors (DAF) for both critical
locations and end connections (support reactions) represents a way of understanding how
the pipe rack responds to blast loads. The structural response will be described in terms
of variation in stresses (Von-Mises, mises [Pa]) and reaction forces (Newton, [N]). The
DAF is calculated as the ratio between these parameters.
A parameter study has been performed on the pipe rack, where the following parameters
have been accounted for:
Blast load duration, Td : 50, 100, 150 and 200 ms
Blast load drag pressure level, Pd : 0.2 bar and 0.4 bar
Structure and piping weight: Structure only, structure including piping arrangement
1, 2 and 3
6.1 Eigenmodes and eigenfrequencies

By using the linear perturbation feature in Abaqus, the eigenfrequencies are found
corresponding to the first 10 eigenmodes. Four cases were studied; the first case took
account for the bare structure (self-weight only), the rest involved the structure including
piping arrangement 1, 2 and 3 as described in section 5.5. By studying the ratio between
the eigenperiod and the duration of the blast, one can determine whether the responses
should be classified as being in the impulsive, dynamic or quasi-static domain. According
to Aker Solutions, the design blast load will be of duration between 50-200 ms. By
58
CHAPTER 6. ANALYSIS RESULTS 59
comparing these values with the obtained eigenperiods from Abaqus, the response range
for the structure has been studied.
For a structure where both the self-weight and the piping are included in the analysis, it
is seen that responses due to blast loads of duration between 50 and 100 ms are within
the definition of a dynamic domain, according to the Blast and Fire Engineering Project
[2]. According to J.Amdahl, blast durations between 50 and 150 ms are within the
dynamic domain, with 200 ms blast duration in the transition zone between dynamic
and quasi-static. For the bare structure (piping neglected), the responses are generally
within the quasi-static domain for all blast durations. When the weight of the pipes are
decreased (as described under section 5.5), the response range is expected to approach
the quasi-static domain. Reference is made to appendix B.
As previously stated in section 1.3, a blast load problem within the dynamic domain, will
require the pressure history to be idealised as triangular shaped loading functions. The
peak value should be preserved as to capture the response of the structure in an accurate
way. For the quasi-static domain, the load duration is not too important if the response
is purely elastic. However, this becomes important when response is plastic (reference
is made to figure 1.4). For the analyses done in this thesis, a triangular shaped loading
function has been used for all cases.
6.2 Static analyses

Static analyses have been performed with the purpose of studying the static structural
response for the simplified analysis model. The static model features a static drag pressure
load. Because this static drag pressure is time independent, the effects on the structural
response due to the duration of the blast are neglected.
It has been observed that the highest stress concentrations are found at element 614
for the cases with piping arrangement 1, 2 and 3. For the case with the bare structure
only, element 702 is shown to experience the maximum stress. Element 614 is located at
the connection between a larger (RHS350) and a smaller RHS profile (RHS200) beam.
Element 702 is located at the connection between an angled brace and chord. Reference
is made to figure 6.1 and 6.2. The maximum Von-Mises stress for each respective element
is presented in table 6.1 for 0.2 bar drag load, and table 6.2 for the 0.4 bar drag load. It
should be noted that all stresses are within the elastic range of the material.
Arrangement mises,max (MPa), SP1 mises,max (MPa), SP2 Element number

1 55.6 55.3 614
2 49.0 48.7 614
3 33.4 33.0 614
No pipes 29.3 32.4 702
Table 6.1: Results from the static simplified model with 0.2 bar drag pressure for two
different section points.
Arrangement mises,max (MPa), SP1 mises,max (MPa), SP2 Element number

1 71.3 70.6 614
2 64.8 64.1 614
3 61.8 61.5 702
No pipes 59.5 63.0 702
Table 6.2: Results from the static simplified model with 0.4 bar drag pressure for two
different section points.
Figure 6.1: High stress area for the cases with pipe rack including piping arrangements
1,2 and 3.
Figure 6.2: High stress area for the case with pipe rack only, no piping weight applied.
6.3 Dynamic analyses with simplified blast load model

Dynamic analyses have been performed with the purpose of studying the transient
structural response. Because a blast case is of a dynamic nature, the structure is expected
to react dynamically in a likewise manner.
The general response of the structure is observed to follow a common trend for the
different blast cases. During the actual blast, the structure will see a slight lateral
displacement in both positive and negative Y-direction (where the Y-direction represents
the direction of the blast). This is due to the positive and negative contribution from the
blast drag pressure. Figure 6.3 is provided for illustrative purposes. During post-blast,
the horizontal beams which support the pipes will start to oscillate vertically. These
oscillatory displacements are initiated by the response during the actual blast, and are
driven by the piping weight. Hence, the heavier the pipes on the pipe racks are the higher
stresses do they induce on the pipe rack. The highest stresses are found at the connection
points where the vertical beams meet the horizontal beams supporting the pipes (see
figure 6.4).
Figure 6.3: Illustration of the structural response during blast and post-blast.
Figure 6.4: Stress (High stress areas marked red. Boundary conditions are marked green.
When observing the variation of Von-Mises stresses at selected elements on the structure,
it is possible to capture the response during the actual blast as a characteristic peak.
This is illustrated in figure 6.5. The differences in the response due to the different piping
arrangements are clearly seen.
When observing the variation of reaction forces (RF) at selected nodes at the end
connections, a similar response peak can be observed. The reactions forces will work
against the direction of the blast load, and because the initial blast load direction is in
the negative Y-direction, the positive drag pressure amplitude will give positive reaction
forces, and vice versa for the negative drag pressure amplitude. This is illustrated in
figure 6.6.
Figure 6.5: Dynamic response with respect to Von-Mises stresses on element 614 due to
change in piping arrangement (Green curve represents arrangement 3, blue curve is the
pipe rack without piping weight). The characteristic response peak is seen during the
blast for all three cases.
Figure 6.6: Variation of reaction force in Y-direction (blast-direction) for 4 selected nodes
at end connections.
6.4 Dynamic Amplification Factors (DAF)

A dynamic load may have a significantly larger effect on the structure than a static load
of the same magnitude due to the structures inability to respond quickly to the loading.
This is especially true for structural responses due to blast loads. The DAF reflects
the increase of response due to a dynamic load, and is defined as the ratio between the
dynamic response parameter (e.g Von Mises stress, reaction force, z-displacement etc.)
and its respective static response parameter. This is seen from equation 6.1, for position
or element i.
Rdyni
DAFi = (6.1)
Rstati
The purpose of determining dynamic amplification factors is to try to find whether a
relationship exists between a dynamic analysis model and its corresponding simplified
static analysis model. In practice, the DAF will serve as a simple input for a structural
engineer, as he or she will be able to quickly estimate stress values or reaction forces for
the dynamic blast case by simply using a static analysis software.
Calculating DAF values involves comparing time independent analyses, i.e static analyses
with time depending dynamic analyses. For dynamic analyses, the response of the
structure in terms of stresses and reaction forces will change in time. This has been
proven to be a challenge when trying to find the correct result values for calculation of
DAFs for critical locations on the pipe rack. By creating a set of standard procedures
for extracting stresses has thus proven practical, these methods are presented in section
6.4.1.
The most logic way of extracting stresses and reaction forces for determining DAFs would
be to look at the response during the blast step. This is because the DAF is to be
calculated by comparing the static blast model with its corresponding dynamic model.
The static blast model with its static load represents a simplified blast model, in which
the purpose is to model the structural response due to the blast only. Moving into the
post-blast range will question the validity of comparing the static model with the dynamic
model, the post-blast response is thus irrelevant for the calculation of DAF values.
6.4.1 Stress (Von-Mises) at critical locations

Dynamic amplification factors are to be calculated based on elements in the model which
show a high degree of utilization, i.e high stress elements. Two approaches for calculating
the DAF have been implemented. The first method starts with locating the element from
the static analysis which experiences the highest stress, the DAF is calculated on the
basis of this static stress together with the largest stress for this particular element over
the blast-duration of its corresponding dynamic analysis.
The second approach starts with locating the element with the highest global stress
over all blast-time steps in the dynamic analysis. Since the element with the highest
stress changes its location through time, the particular element which has the highest
stress during the blast may not be the same element which has the highest global stress
initially from the static analysis. The DAF is calculated on the basis of the location of
this element, also taking account for the same element in the static analysis. Figure 6.7
summarizes the different concepts behind method 1 and 2.
Figure 6.7: The difference between method 1 and method 2 for extracting stresses for
calculating the DAF.
Because the element which experiences the largest stress throughout a dynamic analysis,
may not be the same element which experiences the largest stress in the corresponding
static analysis, it is expected that the DAF calculated using the second method will be
somewhat higher than the first method. As the stress for one element is given for four
different section points, it is seen that two section points always give the highest stresses.
Reference is made to figure 6.8.
Figure 6.8: An example of maximum dynamic stress extraction from the blast step.
Section points 1 and 5 (red and blue curve represent the highest stresses during the blast
step).
By using method 1 and method 2, the DAF is plotted against blast duration. Not
surprisingly, method 2 gives higher DAFs than method 1. The calculated DAF vary
between 0.9 and 1.6 depending on blast duration. It can be observed that the DAF is
significantly higher for the 50 ms blast duration than for the other cases, and with a
reduction of DAF when the blast duration in increased. Again, it should be noted that
only one stress value is extracted out from each blast case. Reference is made to figure
6.9 and appendix F.
Figure 6.9: Variation of DAF with respect to blast duration for method 1 and 2.
Because the procedures for calculating the DAFs as previously presented involved extrac-
tion of only one stress value (from one element) per case, there may be a risk of ignoring
other high stress elements which may be in the same magnitude as the maximum stress.
To cope with this issue, a set of 12 elements from critical locations on the pipe rack
structure has been selected for each blast case, and method 2 has been used for stress
extraction for the calculation of DAFs. Reference is made to figure 6.10 for an overview
over the location of these 12 elements.
Figure 6.10: The locations of the 12 selected elements to be used for the extended DAF
calculations.
Figure 6.11: An example of DAF values obtained from the 50 ms blast case, for arrange-
ment 1. The DAFs are calculated for four section-points separately, in addition to the
average of all section points.
The results from these calculations show the same trend as seen from previous graphs,
that is the 50 ms blast being the most critical case. Element 689 and 702 consequently
show high DAFs for all blast cases and arrangements, with the highest DAF being close
to 100 for some analyses. This is highly unexpected. It should be noted that the DAFs
for element 689 and 702 represent 8.3% of all calculated DAF values.
By plotting the variation of DAF with respect to blast duration, the contributions from
element 689 and 702 are clearly seen. By removing these two contributions from the
graphs, a more general picture is created of the response, these results will represent
91.7% of the calculated DAFs. Reference is made to figure 6.12 and 6.13.
Figure 6.12: Plot of the variation of DAF with respect to blast duration, for all 4 section
points for all 12 elements (0.4 bar, arrangement 3). The high DAFs represented by the
response from element 689 and 702 are clearly expressed by the graph.
Figure 6.13: By removing the contribution from element 689 and 702, a more general
picture of the response is seen. It should be emphasized that the response expressed
by these DAFs represents 91.7% of the total number of DAFs for this particular piping
arrangement.
Figure 6.14: Plots of the variation of DAF with respect to blast duration for all section
points, for the different piping arrangements. Blast drag pressure 0.2 bar. It is shown
from these figures that the magnitude of the DAF decreases with increasing blast duration.
It should be noted that the data points for element 689 and 702 have been removed in
this figure.
Figure 6.15: Plots of the variation of DAF with respect to blast duration for all section
points, for the different piping arrangements (neglecting the DAFs from element 689 and
702). Blast drag pressure 0.4 bar. It should be noted that most DAF values are below 2,
with exceptions of some elements in the 50 ms blast case. It is shown from these figures
that the DAF decreases with increasing blast duration.
Averaging the DAF for all section points
As each data point on figure 6.14 and 6.15 represents a DAF for a certain element and
section point, there is a certain scattering of data points for each blast duration. By
combining the data from each arrangement into one figure, it is seen that the majority
of DAF values are concentrated in the range between 0.5 and 2. Reference is made to
figure 6.16. It proved to be a challenge when trying to determine a characteristic DAF
for each blast duration; if the highest DAF is taken for each blast duration, this value
may not characterize the majority of the DAFs calculated for the same duration, further
wise there is a risk of obtaining too conservative values.
A solution to this issue is to assume that the DAF for each element can be taken as the
average of the four DAFs for each section point. This reduced the number of data points
for each element, thus reducing the scattering range for each blast duration case. The
result of this are curves which describe the general change of DAF with respect to the
blast duration in a much organized manner than previously. Reference is made to figure
6.17.
Figure 6.16: The following figure illustrates the scattering of data points for DAFs as
line plots. The largest concentration of data points are located roughly between 0.5 and
2. The extreme DAF points only represent a small number of points. Please note that
the range of scattering is largest for the 50 ms blast case. When looking at the 50 ms 0.2
bar blast case, 90.6% of the DAFs are below 2. For the 50 ms 0.4 bar blast case, 85.4%
of the DAFs are below 2.
Figure 6.17: The following figure illustrates the scattering of data points for DAFs as
line plots. The once four DAFs for each element is now reduced to one averaged DAF,
this reduced the size of the scattering range of data points. Once again, it should be
emphasized that the majority of DAFs are below the value of 2.
To investigate the possible relationship between the DAF and the blast duration, the
variation of DAF is plotted against the blast duration for all elements except element 689
and 702. Reference is made to figure 6.18. It is shown that there is a clear relationship
between the DAF and the blast duration, in which the DAF decreases with increased
blast duration. The scattering range is consistently larger for the 0.4 bar than the 0.2
bar case.
Figure 6.18: The following scatter plots illustrate the scattering of data points for DAFs
(averaged section points). These plots do not contain the data points from element 689
and 702.
By averaging the DAF for all section-points on each element, and plotting the variation of
DAF with respect to the piping arrangement using method 2 for the same critical points
on the structure, it is harder to see a relationship between the DAF and piping weight.
This might indicate that the DAF will remain constant regardless of the change in piping
weight (arrangement). The scattering range of DAF is between 2.1 and 1.2. Reference is
made to figure 6.19 and appendix G. These results are further treated under section 7.2.
Figure 6.19: The variation of DAF plotted against piping weight (arrangement) as a
scatter plot. The figure illustrates the scattering of data points. These plots do not
contain the data points from element 689 and 702.
6.4.2 Reaction Forces (Support Reactions)

Dynamic amplification factors (DAFs) are to be calculated for all support reactions (end
connections in the model). In Abaqus, these support reactions can be found by extracting
the nodal reaction forces (RF-forces) in the X, Y and Z-direction. There are in total 23
end connections in the model, figure 6.20 shows the location and node numbering for
these.
Figure 6.20: Location and node numbering for the extraction of reaction forces.
The DAF is calculated using expression 6.1, which involves a direct comparison between
the dynamic and the static model. The reaction force from the dynamic case, is taken as
the maximum reaction force during the blast load step. This is the same procedure as
previously presented in section 6.4. If one assume that the DAF can be averaged over all
arrangements, it is shown that the DAF will decrease with increased blast duration in
the same manner as with the DAF for Von Mises stresses at critical locations on the pipe
rack.
Investigation on the reaction forces in the Y- and Z-directions (RF2 and RF3) is of most
importance, as these two forces represent the support reactions in the direction of the
blast and gravity respectively. For reaction forces in the blast direction (RF2), the DAF
is shown to be between 1.6 and 0.95. For the Z-direction (RF3) the DAF is found to be
between 2 and 1. Reference is made to figure 6.23 and 6.24.
For the reaction forces in the X-direction (RF1), there are four nodes which show high
DAF values up to 70. Reference is made to figure 6.22. The reason for this can be
explained by the differences in the response between the static and dynamic analyses. In
the static analyses, the external load working on the structure is the static blast load in
the Y-direction. To fulfill static equilibrium, there will mainly be reaction forces in the
Y-direction opposing this external load, thus, making the reaction force contribution from
the X-direction minimal. For the corresponding dynamic analyses however, the structure
will respond to the dynamic blast load in such a manner that it will see significant
displacements in the X-direction. The reaction forces in X-direction will reflect these
directional changes, and thus give a significant contribution in the X-direction for the
dynamic analyses. Since the DAF is calculated on the basis of the ratio between the
contribution from the dynamic and the static analysis, the DAF value becomes high due
to the significant differences between these two reaction force parameters. Figure 6.21
shows the coordinate system for the analysis models.
Figure 6.21: Coordinate system with respect to the direction of the blast, and the direction
of the gravity.
Figure 6.22: Variation of DAF (based on reaction force in X-direction) with respect to
blast duration. Please note that most DAFs are below 10, with the exception of the DAF
for nodes 65, 95, 71 and 158.
Figure 6.23: Variation of DAF (based on reaction force in Y-direction) with respect to
blast duration. The Y-direction is the direction of the blast. The majority of the DAFs
are found to be below 1.5.
Figure 6.24: Variation of DAF (based on reaction force in Z-direction) with respect to
blast duration. The Z-direction is the direction of the gravity. Please note that the
majority of the DAFs are below 1.5.
Chapter 7
Discussion of results
In the following chapter, the results from the analyses will be discussed in able to
understand the general response of the pipe rack, when being exposed to blast loads. The
results will be further treated in able to obtain the final values as requested through the
initial project scope, which was to find dimensioning DAF-values for critical locations on
the pipe rack and support reactions.
7.1 Static and dynamic analyses

Both the static and dynamic analyses show that the stress variations on the structure are
within the materials elastic range. One might think that blast loads will cause plastic
deformations on the structure, which is indeed the case with blast walls and similar
structures as seen from for instance the research paper by L.A Louca and R.M.Mohamed
Ali [11]. However, for a truss work structure as the pipe rack being modelled, each
structural member is of a small area presented to the explosion. Due to this fact, the
direct overpressure loading on the pipe rack is of relatively small magnitude. In these
circumstances, the dynamic pressure, or drag load is to be considered. Because the
dynamic pressure is calculated based on the projected area of a member, the resulting
pressure load was not of such a magnitude that it resulted in any plastic response of the
structure.
Because the main project scope was to investigate the dynamic amplification of the
structure by comparing dynamic analyses against static analyses, it is merely the ratio
between stresses which are of interest, and not the stress value itself.
The dynamic analyses showed the effect of the structural blast- and post-blast response
due to the positive and negative drag pressure contribution from the blast load. For most
elements, the blast-response was captured as a stress peak during the blast. Reference is
made to figure 7.1. Due to the lack of a damping model, the stress levels in the post-blast
84
CHAPTER 7. DISCUSSION OF RESULTS 85
response range represented a great uncertainty. For the project scope however, this proved
to be irrelevant as the blast-response was the main focus.
Figure 7.1: Showing the response stress peak observed under blast. The figure also show
that the differences between the lowest and highest stresses (Delta peak stress) remain
almost constant for the different piping weights during the blast. This indicates that the
stress response during the blast may be independent of the piping weight.
7.2 DAF for critical locations on the pipe rack

Dynamic amplification factors have been found based on the change of Von-Mises stresses
on the structure for the static and dynamic analyses. Results show that the DAF varies
with the blast duration, in addition to this, there is also a significant amount of scattering
of data points for each blast duration, this is due to each critical element on the pipe rack
having different DAFs. Two elements (689 and 702) stand out because they have DAFs
up to 100. Seen from a general perspective, the DAFs represented by these two elements
stand for only a small percentage of the total DAFs, this makes up for an argument for
not taking account for these two elements in further calculations, if one is to study the
general response.
As the main task is to find general DAF values which are to characterize critical locations
on the pipe rack, certain assumptions and simplifications must be made to obtain any
reasonable results. A few suggestions on the approach of determining general, dimensioning
DAF-values for critical locations on the pipe rack are given below in section 7.2.1, 7.3.2
and 7.2.3.
7.2.1 Using the upper bound (Method A)
If the high DAFs from element 689 and 702 can be ignored, and if it is assumed that the
once four DAFs for each section point in an element can be replaced with one averaged
DAF, the upper bound values can be taken as to represent the DAFs for the critical
locations on the pipe rack. This is hereby referred to as method A. Reference is made to
figure 7.2 and 7.3.
Figure 7.2: Upper bound values for DAF, method A. 0.2 bar blast pressure (not including
data points from element 689 and 702).
Figure 7.3: Upper bound values for DAF, method A. 0.4 bar blast pressure (not including
data points from element 689 and 702).
7.2.2 Averaging over all elements, use the upper bound (Method
B)
Method B involves taking the average of the DAFs over all section points and elements
for each blast duration. This reduced the number of DAFs to only one value per blast
duration. The upper bound is then taken for the remaining data points. This is illustrated
in figure 7.4 and 7.5.
Figure 7.4: Upper bound values for DAF, method B. 0.2 bar blast pressure.
Figure 7.5: Upper bound values for DAF, method B. 0.4 bar blast pressure.
7.2.3 Averaging over all elements, use curve fit techniques (Method
C)
Method C involved averaging the data points for all section points and elements for each
blast duration, in the same manner as for method B. Linear or nonlinear least squares
regression is then used to a curve to the remaining data points. The mathematical
expression for this curve will express the variation of DAF over the blast duration. The
benefit of using method C is that any DAF can be calculated using the resulting equation
given a particular input parameter (e.g blast duration).
DAF vs. blast duration
It is seen that the data points and the variation of DAF with respect to the blast duration
are fitted best to a logarithmic curve.
Figure 7.6: Variation of DAF with respect to blast duration, method C. Please note that
this graph does not include the averaged data points from element 689 and 702.
Figure 7.7: Variation of DAF with respect to blast duration, method C. The following
graph includes the averaged data points from element 689 and 702.
DAF vs. piping arrangement (piping weight)
By plotting the DAF over the piping arrangement using method C, very small variations
of the DAF was seen for varying piping arrangements. An exception is for arrangement 3
which showed deviation from the general trend, however by using linear regression to fit
a curve to the remaining data points, it shown that the variation of DAF is very small, if
not constant. For the case with element 689 and 702 accounted for, the DAF was found
to be between 1.7-1.8 and 2.2-2.3 for the 0.2 and 0.4 bar cases respectively. If element 689
and 702 were not accounted for, the DAF seemed to have a near constant value of 1.2 for
the 0.2 bar case, and a varying DAF between 1.2-1.3 for the 0.4 bar case. Reference is
made to figure 7.9 and 7.8.
Figure 7.8: Variation of DAF with respect to arrangement (piping weight), method C.
The following graph does not include the averaged data points from element 689 and 702.
Figure 7.9: Variation of DAF with respect to arrangement (piping weight), method C.
The following graph includes the averaged data points from element 689 and 702.
7.2.4 Summary
It is seen that different methods for obtaining the DAF for critical locations on the
pipe rack, give different results. Results which include the data points from element 689
and 702 give noticeable higher DAF-values than for results without these two elements
accounted for. The DAF values representing the 0.4 bar drag pressure are also higher
than those for 0.2 bar drag pressure. While the DAF decreases with increased blast
duration, it seems to remain constant with respect to piping weight (arrangement). Table
7.1, 7.2 and 7.3 summarize the DAF values found using the different methods.
Blast duration DAF Method A DAF Method B DAF Method C

50 ms 2.3 (-) 1.4 (2.5) 1.4 (2.4)
100 ms 1.7 (-) 1.2 (1.7) 1.2 (1.8)
150 ms 1.4 (-) 1.1 (1.3) 1.1 (1.5)
200 ms 1.3 (-) 1.0 (1.4) 1.0 (1.3)
Table 7.1: DAF for 0.2 bar drag pressure (DAF vs. blast duration). The values in
parentheses are the DAF accounted for element 689 and 702.

50 ms 2.8 (-) 1.5 (3.4) 1.5 (3.3)
100 ms 1.9 (-) 1.3 (2.1) 1.3 (2.4)
150 ms 1.6 (-) 1.1 (1.6) 1.2 (1.8)
200 ms 1.6 (-) 1.1 (1.7) 1.1 (1.4)
Table 7.2: DAF for 0.4 bar drag pressure (DAF vs. blast duration). The values in
parentheses are the DAF accounted for element 689 and 702.
Arrangement 0.2 bar 0.4 bar

1 1.2 (1.7) 1.3 (2.2)
2 1.2 (1.7) 1.2 (2.2)
3 1.2 (1.7) 1.2 (2.2)
No pipes 1.2 (1.8) 1.2 (2.3)
Table 7.3: DAF for different piping arrangements found by method C. Arrangement 1
being the heaviest loaded condition. The values in parentheses are the DAFs accounted
for element 689 and 702.
7.3 DAF for support reactions

The different methods for determining DAF for support reactions are the same as
previously presented in section 7.2 for critical locations on the pipe rack.
7.3.1 Using the upper bound (Method A)
By using the upper bound from the results (see figure 6.22, 6.23 and 6.24) the DAF values
could be taken directly from the graphs for each case. An example of this is shown in
figure 7.10.
Figure 7.10: According to method A, the DAF value can be taken directly from the upper
bound.
7.3.2 Averaging over all nodes, use the upper bound (Method
B)
Method B involved taking the average of the DAFs over all nodes for each blast duration.
This reduced the number of DAFs to only one value per blast duration. The upper bound
was then taken for the remaining data points. This is illustrated in figure 7.11.
Figure 7.11: According to method B, the DAF value can be taken directly from the upper
bound from the averaged DAF over all nodes.
7.3.3 Averaging over all nodes, use linear curve fit (Method C)
DAF vs. blast duration
From the results of the variation of DAF for reaction forces represented by figure 6.22, 6.23
and 6.24, scattering of data points are observed for each blast duration. A generalization
has been done in the same manner as with the DAFs for critical locations on the pipe
rack; by assuming that all DAFs for all nodes can be reduced to one generalized, averaged
DAF. The data points are thus reduced to one DAF value per blast duration.
A study of the results showed that the DAF decreases with increased blast duration in
such a way that a linear curve can be fitted to the data points. Linear least squares
regression has been used to generate curves which show the relationship between the
DAF and the blast duration. The best fit (i.e the curve which has the R2 -value closest
to 1) is obtained from the results from the reaction forces in Z-direction, which is the
direction of the gravity. Reference is made to figure 7.14.
Another observation made is that the two curves for the reaction forces in Y-direction
(each representing the 0.2 and 0.4 bar drag pressure) almost coincide with each other.
This indicates that the reaction forces in Y-direction are independent of the magnitude
of the blast. Reference is made to figure 7.13.
Figure 7.12: Linear least squared curve fit for the data points of RF1 (Reaction force in
X-direction), DAF vs. blast duration.
Y-direction; direction of blast), DAF vs. blast duration.
Z-direction; direction of gravity), DAF vs. blast duration.
DAF vs. piping arrangement (piping weight)
Method C was utilized, and the data points for each DAF were reduced to one value per
piping arrangement. The change of DAF with respect to piping arrangement was then
plotted. For the reaction forces in the X-direction, linear curves were fitted to the data
points. These showed a decrease in DAF for reduced piping weight for both the 0.2 and
the 0.4 bar cases. Reference is made to figure 7.15.
For the reaction forces in the Y-direction, the change of DAF is between 1.1 and 1.2. The
DAF is decreasing with reduced piping weight for both 0.2 and 0.4 bar cases in the same
manner as with the X-direction. Reference is made to figure 7.16.
The reaction forces in the Z-direction were fitted to second order polynomial curves. For
this direction, the DAF increases with reduced piping weight for both the 0.2 and 0.4
cases. Reference is made to figure 7.17.
X-direction), DAF vs. piping arrangement.
Y-direction; direction of gravity), DAF vs. piping arrangement.
Figure 7.17: Nonlinear least squared curve fit for the data points of RF3 (Reaction force
in Z-direction; direction of gravity), DAF vs. piping arrangement.
7.3.4 Summary
As with section 7.2.4, it is seen that different methods for obtaining DAF for support
reactions (reaction forces) on the pipe rack, give different results. It should be noted
that the DAF values for the support reactions (reaction forces) in the Y-direction (i.e the
direction of the blast) seem to be constant, this indicates that the DAF for this direction
is independent of the magnitude of the drag pressure.
Table 7.4, 7.5 and 7.6 summarize the obtained results of the DAF with respect to the
blast duration for the different methods. Table 7.7, 7.8 and 7.9 summarize the obtained
results of the DAF with respect to the piping arrangement for the different methods.

50 ms 45.9 (63.7) 4.9 (7.0) 5.8 (7.5)
100 ms 72.8 (58.5) 6.2 (7.0) 4.8 (6.2)
150 ms 34.4 (33.8) 3.7 (4.5) 3.8 (4.9)
200 ms 16.1 (26.3) 2.5 (3.5) 2.9 (3.6)
Table 7.4: DAF values for support reactions (reaction forces) in X-direction. The values
in parentheses are the DAF for 0.4 bar drag pressure.

50 ms 1.6 (1.6) 1.2 (1.2) 1.2 (1.2)
100 ms 1.6 (1.5) 1.2 (1.2) 1.2 (1.2)
150 ms 1.5 (1.5) 1.2 (1.1) 1.2 (1.2)
200 ms 1.5 (1.5) 1.1 (1.1) 1.1 (1.1)
Table 7.5: DAF values for support reactions (reaction forces) in Y-direction (direction of
the blast). The values in parentheses are the DAF for 0.4 bar drag pressure.

50 ms 2.0 (5.7) 1.2 (1.4) 1.2 (1.4)
100 ms 1.8 (4.1) 1.2 (1.3) 1.2 (1.3)
150 ms 1.5 (2.7) 1.1 (1.2) 1.1 (1.2)
200 ms 1.5 (2.1) 1.1 (1.1) 1.1 (1.1)
Table 7.6: DAF values for support reactions (reaction forces) in Z-direction (direction of
the gravity). The values in parentheses are the DAF for 0.4 bar drag pressure.
Arrangement DAF Method A DAF Method B DAF Method C

1 188.3 (51.7) 7.4 (5.2) 6.7 (6.6)
2 87.3 (180.1) 4.4 (8.1) 5.1 (5.9)
3 18.6 (45.4) 2.9 (4.8) 3.5 (5.2)
No pipes 16.7 (55.7) 2.6 (3.9) 2.0 (4.4)
Table 7.7: DAF values for support reactions (reaction forces) in X-direction. The values
in parentheses are the DAFs for 0.4 bar drag pressure.

1 1.8 (1.6) 1.2 (1.2) 1.2 (1.2)
2 1.6 (1.6) 1.2 (1.2) 1.2 (1.2)
3 1.6 (1.6) 1.2 (1.2) 1.2 (1.2)
No pipes 1.6 (1.6) 1.1 (1.1) 1.1 (1.1)
Table 7.8: DAF values for support reactions (reaction forces) in Y-direction. The values

1 1.5 (1.8) 1.1 (1.1) 1.1 (1.1)
2 1.5 (2.5) 1.1 (1.2) 1.1 (1.2)
3 1.9 (4.2) 1.1 (1.3) 1.1 (1.3)
No pipes 3.4 (17.8) 1.1 (1.5) 1.2 (1.5)
Table 7.9: DAF values for support reactions (reaction forces) in Z-direction. The values
Chapter 8
Conclusion and further work
8.1 Conclusion
This paper has been an investigation on the phenomenon of hydrocarbon explosions,
the resulting blast load, and the structural response following such an event. Although
a hydrocarbon explosion is considered a rare accidental event, effort should be made
to reduce the amount of structural damage if an explosion was to occur. If blast load
scenarios are not considered for critical structural elements close to a processing area,
an initial blast accident can easily cause further escalation of damage, especially if pipes
containing hydrocarbon liquid or gases rupture. A literature study has been done to
gather as much relevant information about the topic as possible. Although similar work
has been done by various sources such as The British Gas Research and Technology (ERS)
[2], Hamdan [23], Christian Michelsen Research (CMR) Programme [41] and Scheler EA.
et al.[30], none of these sources focus on the practical implementation of the dynamic
drag/overpressure on the actual structure. The different sources deal mostly with the
structural response of blast walls (confined spaces) and thick support beams. There
are also a lot of uncertainties in the different sources concerning the drag effects of the
dynamic pressure, this is an effect of more importance for truss works and open frame
structures than for instance blast walls and other structures with large surfaces.
A pipe rack (truss) structure which is located outside a topside hydrocarbon processing
area has been the object of the study. The blast scenario is assumed to be causing a
drag pressure acting perpendicular to the longitudinal direction of the pipe rack and
its pipes. In order to fully understand the dynamic response of the pipe rack structure
during the blast, a parameter study has been performed by varying the blast load and
duration. Further on, dynamic amplification factors (DAFs) have been calculated for
critical locations on the pipe rack and at end connections by comparing dynamic and
static analysis models. These dimensioning DAFs are to reflect the dynamic nature of
the structural response during blast. Because the main project scope is to investigate
102
CHAPTER 8. CONCLUSION AND FURTHER WORK 103
the dynamic amplification of the structure by comparing dynamic analyses against static
analyses, it is merely the ratio between stresses which are of interest. With this said,
the uncertainties regarding the determination of the magnitude of the drag load on the
structure is greatly reduced.
Familiarization with the dynamic features in Abaqus/Explicit (especially concerning
performing analyses of an impulsive character) and the post-processing of time dependent
data have been emphasized. A simplified beam model of the pipe rack structure was
created, in this model, connection (joint) eccentricities have been neglected. A mesh
sensitivity (stress convergence) study was performed, and the mesh density was determined
based on a consideration of a trade-off between computational time and relative accuracy
of results. The piping weights on the pipe rack have been taken account for by using
inertia masses, but the physical effects of the drag pressure on the pipes have been
ignored for the sake of simplification. This simplification was based on assuming that
the pipes are connected to the pipe rack with simple U-braces, which makes the piping
stiffness-contribution to the overall structure negligible. Damping has been neglected as
it is shown not to affect the responses to any large degree for blast analyses.
It should be noted by the reader that the structural responses due to the blast loading
were within the materials elastic range, and no permanent deformations were observed.
If plastic deformations have been observed, a ductility level analysis as presented under
section 4.4 would have been performed. Three different methods were used to determine
the general, dimensioning DAF for each blast case, and the results from all methods were
presented.
Final results showed that the dimensioning DAF is dependent on both the blast duration
and the weight of the pipes on the structure. For critical locations on the pipe rack,
the DAF decreased with increased blast duration. The DAF remained almost constant
regardless of change in piping weight.
For the end connections (support reactions), the DAF decreased with increased blast
duration in the X- and Z-direction, while remaining almost constant for the Y-direction
(blast direction) regardless of both changes in blast duration and piping weight. The
DAF decreased with reduced piping weight in the X-direction, while it increased with
reduced piping weight in the Z-direction.
8.2 Recommendations for future work

Further work should focus on analysing pipe rack structures of different sizes, RHS-
profiles and designs as to obtain more DAF-values, such that a more general picture of
the structural response of pipe racks can be obtained. Further work should also focus on
investigating the effects of connection eccentricity on the structure with respect to the
DAF, this will require a new model to be created. A parameter study with additional
CHAPTER 8. CONCLUSION AND FURTHER WORK 104
drag pressure levels and blast durations may also be considered. In addition to this,
variable mesh densities should be introduced, as to better capture the stresses near joints
and critical areas.
CFD analyses, simulating the dispersion of the pressure and its effect on the pipes
and structure, including possible shielding or blocking effects may be studied. A more
accurate drag coefficient CD for the drag pressure equation (as from equation 4.1) should
be calculated. However, this may require familiarization with specialized CFD-software
used to simulate explosions, i.e FLACS or Kameleon FireEx (KFX).
Bibliography
[1] Ir. W.P.M Merx. Methods for the determination of possible damage to people and
objects resulting from releases of hazardous materials. CPR 16E, TNO, first edition,
1992.
[2] British Gas Research and Technology (ERS). Blast and fire engineering project for
topside structures, Work package No.BR2. 1991.
[3] W.Clough. R and Penzien. J. Dynamics of Structures. McGraw-Hill International
Editions, second edition, 1993.
[4] Amdahl J. Lecture notes in subject TMR 4195 Design of Offshore Structures, Dept.
Marine Structures, Norwegian Institute of Technology, Trondheim. 2011.
[5] Orcaflex Manual 9.4a. Orcina Ltd., 2010.
[6] http://www.offshoretechnology.com/projects/britannia/.
[7] Cirak. F. Lectures in Plates and Shells: Theory and Computations. Structural
Mechanics Lab, Department of Engineering, University of Cambridge, 1999.
[8] Zhong. W et al. Yao. W. Sympletic Elasticity. World Scientific Publishing Co. Pte.
Ltd., 2009.
[9] Abaqus/CAE 6.10-2 Documentation. Getting Started with Abaqus: Interactive
Edition, publisher =.
[10] Abaqus/CAE 6.10-2 Documentation. User Manual: 26.3.8 Beam element library.
Dassault Systemes Simulia Corp., 2010.
[11] L.A. Louca and R.M Mohamed Ali. Improving the ductile behaviour of offshore topside
structures under extreme loads. Elsevier/Deparment of Civil and Environmental
Engineering, Imperial College London UK, 1994.
[12] Task Committee on Blast Resistant Design Society of Civil Engineers. Design of
Blast Resistant buildings in petrochemical facilities. ASCE Publications, 1997.
[13] Madenci E. and Guven I. The Finite Element Method and Applications in Engineering
using ANSYS. Springer Science+Business Media, LLC, 2006.
105
BIBLIOGRAPHY 106
[14] Walter C. Hurty and Moshe F.Rubinstein. Dynamics of Structures. Prentice-Hall

Series in Engineering of the Physical Sciences, 1964.
[15] Abaqus/CAE 6.10-2 Documentation. User Manual: Section 23.1.1 Material damping.
[16] Edward L.Wilson. Three-Dimensional Static and Dynamic Analysis of Structures:
A Physical Approach with Emphasis on Earthquake Engineering. Computers and
Structures, Inc., 1999.
[17] Senthil S. Vel Brian P. Baillargeon and Jeffery S. Koplik. Utilizing ABAQUS to
Analyze the Active Vibration Suppression of Structural Systems. ABAQUS Users
Conference, 2004.
[18] ABS Consulting Ltd. Design, materials and connections for blast-loaded structures,
Research Report 205. Health and Safety Executive (UK), 2006.
[19] Buick Davison and Graham W.Owens. Steel Designers Manual 6th Edition. Blackwell
Publishing, 2006.
[20] H. Bachmann et al. Vibration Problems in Structures. Birkhauser Verlag, Berlin,
1995.
[21] Moan T. TMR4305, Advanced Structural Analysis, Chapter 12 - Nonlinear Analysis.
2011.
[22] Abaqus/CAE 6.10-2 Documentation. Theory Manual. Dassault Systemes Simulia
Corp., 2010.
[23] Hamdan F. Structural strengthening of offshore topsides structures as part of explosion
risk reduction methods. The Steel Construction Institute, Research Report 489, 2006.
[24] NORSOK N-003. Action and Action effects; 8.2 Fires and explosions. Standards
Norway (Norsk Standard), second edition, 2007.
[25] FABIG. FABIG Technical Note 2: Explosion Mitigation Systems. The Steel Con-
struction Institute, 1994.
[26] Tam V. et. al Walker S., Corr B. New Guidance on Fire and Explosion Engineering.
21st International Conference on Offshore Mechanics and Artic Engineering (OMAE
2002), 2002.
[27] Det Norske Veritas. DNV-RP-C204 Design Against Accidential Loads. DNV, Det
Norske Veritas, 2010.
[28] Technical Staff of General Monitors. A Guide to the Characteristics of Combustible
Gases and Applicable Detection Technologies. General Monitors.
[29] Demeter G.Fertis. Dynamics and Vibration of Structures. John Wiley & Sons Inc,
1973.
BIBLIOGRAPHY 107
[30] Valdivieso JB Wood BJ. Scheler EA, Salter HS. The modeling of gas explosions and
the response of topside structures SPE,. First International Conference on Health,
Safety and Environment (Hague), 1991.
[31] Yasseri S. Iso-damage diagrams for blast resistant design. FABIG Newsletter Issue
42, 2005.
[32] Abaqus/CAE 6.10-2 Documentation. Section 12, The Property Module. Dassault
Systemes Simulia Corp., 2010.
[33] Abaqus/CAE 6.10-2 Documentation. Section 20.2.3, Rate-dependent yield. Dassault
Systemes Simulia Corp., 2010.
[34] Michael L. Berins. SPI plastics engineering handbook of the Society of the Plastics
Industry, Inc. Kluwer Academic Publishers, fifth edition, 2002.
[35] Irgens. F. Formelsamling i Mekanikk. Tapir Akademisk Forlag, third edition, 1999.
[36] Abaqus/CAE 6.10-2 Documentation. User Manual: Section 32.1 Defining inertia.
[37] Hong Chen and J. Y. Richard Liew. Explosion and Fire Analysis of Steel Frames
Using Mixed Element Approach. JOURNAL OF ENGINEERING MECHANICS
ASCE/JUNE 2005, 2005.
[38] M. Gough and E. Yandzio. Protection of Buildings Against Explosions. The Steel
Construction Institute, 1999.
[39] Abaqus/CAE 6.10-2 Documentation. User Manual: Section 6.3.3 Explicit dynamic
analysis. Dassault Systemes Simulia Corp., 2010.
[40] Sverre Steen and Jan V.Aarsnes. TMR7 Experimental Methods in Marine Hydrody-
namics Compendium. Institutt for Marin Teknikk, 2010.
[41] Wingerden K.V. Bjerketvedt, Bakke J.R. GEXCON Gas Explosion Handbook. Gas
Safety Programme 1990-1992.
Appendix A
Piping weight
Figure A.1: Piping weight recalculated as concentrated masses in Excel.
A. 1
Appendix B
Response range corresponding to

eigenmodes
B. 1
Appendix C
Pipe span chart
Figure C.1: Span chart provided by Aker Solutions
C. 1
Appendix D
Python Script for extracting RF

(Magnitude)
#Script for extracting Reaction Forces from 23 Nodes (at end connections) on the pipe rack:
#(The numerical symbol firkant is the comment symbol in Python)
from abaqus import *

from abaqusConstants import *
o1 = session.openOdb(name=C:/Temp/RHS200_02bar_50ms_arr1.odb)
session.viewports[Viewport: 1].setValues(displayedObject=o1)
odb = session.odbs[C:/Temp/RHS200_02bar_50ms_arr1.odb]
session.xyDataListFromField(odb=odb, outputPosition=NODAL, variable=((RF,
NODAL, ((INVARIANT, Magnitude), )), ), nodePick=((PIPE RACK NEWEST-1,
23, ([#82004020 #0 #402150e1 #20000120 #3084c000 #2 ], )), ), )
x0 = session.xyDataObjects[RF:Magnitude PI: PIPE RACK NEWEST-1 N: 6]

D. 1
APPENDIX D. PYTHON SCRIPT FOR EXTRACTING RF (MAGNITUDE) D. 2

session.writeXYReport(fileName=C:/Temp/RHS200_02bar_50ms_arr1.rpt, xyData=(x0, x1, x2,
x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15, x16, x17, x18,
x19, x20, x21, x22))
Appendix E
Python Script for extracting RF (X,

Y and Z-direction)
#Script for extracting Reaction Forces from 23 Nodes (at end connections) on the pipe rack:
from abaqus import *

from abaqusConstants import *
o1 = session.openOdb(name=C:/Temp/RHS200_02bar_50ms_arr1.odb)
session.viewports[Viewport: 1].setValues(displayedObject=o1)
odb = session.odbs[C:/Temp/RHS200_02bar_50ms_arr1.odb]
session.xyDataListFromField(odb=odb, outputPosition=NODAL, variable=((RF,
NODAL, ((COMPONENT, RF1), (COMPONENT, RF2), (COMPONENT, RF3), )), ),
nodePick=((PIPE RACK NEWEST-1, 23, (
[#82004020 #0 #402150e1 #20000120 #3084c000 #2 ], )), ), )
x0 = session.xyDataObjects[RF:RF1 PI: PIPE RACK NEWEST-1 N: 6]

E. 1
APPENDIX E. PYTHON SCRIPT FOR EXTRACTING RF (X, Y AND Z-DIRECTION)E. 2

APPENDIX E. PYTHON SCRIPT FOR EXTRACTING RF (X, Y AND Z-DIRECTION)E. 3

session.writeXYReport(fileName=RHS200_02bar_50ms_arr1.rpt, xyData=(x0,
x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15, x16, x17,
x18, x19, x20, x21, x22, x23, x24, x25, x26, x27, x28, x29, x30, x31, x32,
x63, x64, x65, x66, x67, x68))
Appendix F
Variation of DAF against blast

duration.
Figure F.1: The following graph shows the variation of DAF for the first section points
with respect to blast duration, using method 1 to extract the stresses. Stress is taken for
one element per case only
F. 1
APPENDIX F. VARIATION OF DAF AGAINST BLAST DURATION. F. 2
Figure F.2: The following graph shows the variation of DAF for the first section points
with respect to blast duration, using method 2 to extract the stresses. Stress is taken for
one element per case only
Appendix G
Variation of DAF against piping

weight (arrangement).
Figure G.1: Averaged DAF plotted against piping weight (arrangement) for all elements.
0.2 bar blast drag pressure.
G. 1
APPENDIX G. VARIATION OF DAF AGAINST PIPING WEIGHT (ARRANGEMENT). G. 2
Figure G.2: Averaged DAF plotted against piping weight (arrangement) neglecting the
data points for element 689 and 702. 0.2 bar blast drag pressure.
Figure G.3: Averaged DAF plotted against piping weight (arrangement) for all elements.
0.4 bar blast drag pressure.
APPENDIX G. VARIATION OF DAF AGAINST PIPING WEIGHT (ARRANGEMENT). G. 3
Figure G.4: Averaged DAF plotted against piping weight (arrangement) neglecting the
data points for element 689 and 702. 0.4 bar blast drag pressure.
Appendix H
Variation of V.M stresses along

path/beam.
Figure H.1: Integrated Von-Mises stress along path, horisontal beam 1.
H. 1
APPENDIX H. VARIATION OF V.M STRESSES ALONG PATH/BEAM. H. 2
Figure H.2: Integrated Von-Mises stress along path, horisontal beam 2.

Figure H.3: Von-Mises stress variation along path, horisontal beam 1.

Figure H.4: Von-Mises stress variation along path, horisontal beam 2.

Appendix I
Matlab script: Read in data and

calculate DAF.
1 %Master Thesis spring 2012 for Stud.Techn. Aiwei Su.

2 %The following file is to be used to calculate DAF (Dynamic ...
amplification
3 %factors) based on RF (Reaction forces) on end connections on pipe ...
racks:
4 %DAF=dynamic RF max/static RF max
5
6 infile1='RHS200 02bar 50ms arr1.dat'; %Name of ODB result file, ...
dynamic case
7 infile2='RHS200 02bar static arr1.dat'; %Name of ODB result file, ...
static case
8 rows1=51;
9 rows2=2;
10 DAF=zeros(5,12);
11 DAF SP1=zeros(2,12);
15
16 %Writing the element numbers into the first row in the DAF matrix ...
for the
17 %different section points.
18 Elem=[106,115,301,614,689,702,809,899,1111,1150,1175,1264];
19 DAF(1,:)=Elem;
20 DAF SP1(1,:)=Elem;
24
I. 1
APPENDIX I. MATLAB SCRIPT: READ IN DATA AND CALCULATE DAF. I. 2
25 %Using textreadfunction to read in the reaction forces (RF) over ...

the blast
26 %duration. Each node (end connection) is denoted ...
RFdyn N'nodenumber'. Dynamic
27 %and static case:
28 [time1,dynE106 SP1,dynE115 SP1,dynE301 SP1,dynE614 SP1,dynE689 SP1,
29 dynE702 SP1,dynE809 SP1,dynE899 SP1,dynE1111 SP1,dynE1150 SP1,
37 dynE1150 SP4,dynE1175 SP4,dynE1264 SP4] = textread(infile1,'%f\t
38 %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t
41 %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f',rows1,'headerlines',5);
42 [time2,statE106 SP1,statE115 SP1,statE301 SP1,statE614 SP1,
43 statE689 SP1,statE702 SP1,statE809 SP1,statE899 SP1,statE1111 SP1,
51 statE1111 SP4,statE1150 SP4,statE1175 SP4,statE1264 SP4] = textread(
52 infile2,'%f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t
53 %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t
54 %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t
55 %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f\t %f',rows2,'headerlines',5);
56
57 %Calculating the DAF for SP1:
58 DAF SP1(2,1)= max(dynE106 SP1)/max(statE106 SP1);
70
APPENDIX I. MATLAB SCRIPT: READ IN DATA AND CALCULATE DAF. I. 3

84
98

112
113 %Collecting all DAFs for for an element (i.e SP1,2,3 and 4) in one ...
place:
114 DAF(2,:)=DAF SP1(2,:);
115 DAF(3,:)=DAF SP2(2,:);
116 DAF(4,:)=DAF SP3(2,:);
117 DAF(5,:)=DAF SP4(2,:);
118
119 %xlswrite('DAF temp.xlsx',DAF) %Writing the result matrix into Excel ...
file.

Explosion Analysis of Piperack

Uploaded by

Copyright:

Available Formats

Explosion Analysis of Piperack

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

Explosion Analysis of Piperack

Uploaded by

Copyright:

Available Formats

Analysis of Explosion Load Effects in

Norwegian University of Science and Technology

Aiwei Su, Trondheim. May 28, 2012

An explosion on a typical offshore installation is a critical, however rare accidental event

En eksplosjon p a en offshore-installasjon er en kritisk, men sjelden ulykkelseshendelse som

2 Interaction between blast and structure 6

4 Blast loads on topside structural elements 19

8 Conclusion and further work 102

B Response range corresponding to eigenmodes B. 1

C Pipe span chart C. 1

D Python Script for extracting RF (Magnitude) D. 1

E Python Script for extracting RF (X, Y and Z-direction) E. 1

F Variation of DAF against blast duration. F. 1

G Variation of DAF against piping weight (arrangement). G. 1

H Variation of V.M stresses along path/beam. H. 1

I Matlab script: Read in data and calculate DAF. I. 1

2.1 Reflection coefficients for different values of the overpressure, as functions

3.1 Simple beam subjected to a static load [3]. . . . . . . . . . . . . . . . . . 13

4.1 Typical topside module arrangement [6]. . . . . . . . . . . . . . . . . . . 20

5.18 Figure illustrating the damping effect (stiffness proportional damping)

A.1 Piping weight recalculated as concentrated masses in Excel. . . . . . . . A. 1

C.1 Span chart provided by Aker Solutions . . . . . . . . . . . . . . . . . . . C. 1

H.1 Integrated Von-Mises stress along path, horisontal beam 1. . . . . . . . . H. 1

1.1 Definition of response domains by various sources . . . . . . . . . . . . . 4

3.1 Damping ratios for different structural systems . . . . . . . . . . . . . . . 17

Ps incident peak overpressure

N normal pressure between body and surface

AP I American Petroleum Institute

1.1 Explosion incident event tree

1.2 General introduction on explosion/blast loads

Size and fuel concentration of the combustible cloud

1.3 Deflagrations and detonations

Source Impulsive domain Dynamic domain Quasi-static domain

It is seen that a great amount of simplifications on the representative pressure history

Interaction between blast and

When a blast meets a structure or an obstacle, it will be disturbed. The disturbance

2.1 Reflection load

2.2 Dynamic pressure

PD = CD 0.5 s us (t)2 (2.4)

2.3 Load on a structure

3.1 Dynamic analysis

Figure 3.1: Simple beam subjected to a static load [3].

Whereas klm is the load-mass transformation factor.

It is shown that SDOF-methods as mentioned above provide an accurate estimate of the

3.2 Geometric non-linearity

FD,i = i g FE,i (3.2)

Type of system , damping ratio (% of critical)

Table 3.1: Damping ratios for different structural systems

Further discussion upon damping in Abaqus are found in section 5.9.

3.4 Solution methods for non-linear analyses

3.4.1 Implicit integration method

r(t + t) = f {r(t), r(t

3.4.2 Explicit integration method

r(t + t) = f {r(t), r(t),

Blast loads on topside structural

Figure 4.1: Typical topside module arrangement [6].

4.1 Mitigation schemes

Figure 4.3: Importance of correct positioning of equipment in case of an explosion event

4.2 Pipe rack system

4.3 Pressure distribution and magnitude

true = ln(1 + ) (5.2)