Mechanical Vibrations in Hydraulic Machines

MECHANICAL VIBRATIONS IN HYDRAULIC MACHINES
REPORT 2015:124
HYDROPOWER
Mechanical vibrations in hydraulic machines

Suggested vibration limits based on analysis of the IEC
TK4 vibration database
JONAS CARLSSON, SWECO ENERGUIDE AB
ISBN 978-91-7673-124-6 | 2015 ENERGIFORSK

Energiforsk AB | Phone: 08-677 25 30 | E-mail: [email protected] | www.energiforsk.se
Foreword
This report is an updated version of chapter 6 in the previously published Elforsk
report 12:70.
For the revision and integration of the current mechanical vibration standards for
hydraulic power generating and pumping plants ISO/IEC 7919-5 and 10816-5 IEC and
ISO are supporting an international workgroup (ISO/TC 108/SC 2 & IEC/TC 4 - JWG1
Vibration of Hydraulic Machines). To support the working group, analysis of the IEC
TK4 vibration database has been performed by the Swedish national workgroup. The
purpose of the analysis has been to form a statistical foundation for a recommendation
on vibration limits in a new integrated vibration standard.
The fourth step in the total scope of work represent the additional analysis that has
been performed and added to the report in this version. That corresponds to text in the
report marked with greyed background. The other parts of chapter 6 in Elforsk report
12:70 are unchanged but slightly restructured in this version.
Swedish delegates in JWG1 are ke Grahn, Vattenfall and Anders Bard, SWECO
Energuide. The author of this report is Jonas Carlsson, E.ON Vattenkraft (previously
SWECO Energuide). The project has been a part of Energiforsks R&D programme
Anlggningsteknik Vattenkraft 2013-2014. Vattenfall Vattenkraft, Fortum Generation,
E.ON Vattenkraft Sverige, Statkraft Sverige, Skellefte Kraft, Jmtkraft, Ume Energi,
Sollefteforsens, Holmen Energi, Karlstads Energi and Jnkping Energi are
participating companies in Anlggningsteknik Vattenkraft 2013-2014.
Stockholm April 2015
Cristian Andersson
Energiforsk
Summary
This report is an updated version of chapter 6 in the previously published Elforsk
report 12:70. For the revision and integration of the current mechanical vibration
standards for hydraulic power generating and pumping plants ISO/IEC 7919-5 and
10816-5 IEC and ISO are supporting an international workgroup (ISO/TC 108/SC 2 &
IEC/TC 4 - JWG1 Vibration of Hydraulic Machines). To support the working group,
analysis of the IEC TK4 vibration database has been performed by the Swedish national
workgroup. The purpose of the analysis has been to form a statistical foundation for a
recommendation on vibration limits in a new integrated vibration standard.
The fourth step in the total scope of work represent the additional analysis that has
been performed and added to the report in this version.
As a statistical foundation for the revised standard an international vibration database
have been developed. The analysis performed in this report are based on the database
versions Vib_DB_ Revision_E 1, Vib_DB_Revision_F 2 and Vib_DB_Revision_J 3. Revision
E contains 2392 rows, revision F contains 2472, whereas revision J contains 7355 rows.
Every row corresponds to one measurement. The database contains measurements on
all types of hydraulic power generating and pumping machines and commonly more
than one measurement on each machine.
During step 4 (Analyze the most recent database with methods established in earlier
steps in order to propose action limits for the different machine types) of the work
scope in the total project, database revision J of the database was considered. Database
revision J has been improved with a vast number of measurements. The filtered
database revision J was analyzed with the same method as for step 3. The median
values for vibration level and shaft oscillations are presented in a table where also the
suggested action limit values are presented.
The main conclusions from the previous work remain:
No clear correlation between vibration values and the unit specific parameters
such as head, rotational speed, runner diameter and radial bearing clearance
can be observed which implies that both shaft oscillations and vibration
velocities are relevant parameters.
Several shortcomings can be identified for the parameter utilized dynamic

bearing clearance (UDBC). High magnetic unbalance in the generator and high
hydraulic unbalance in the turbine gives small shaft oscillations and small
UDBC, bearing load can however be very high. A poorly aligned shaft
arrangement can also give low values on shaft oscillations and UDBC,
although high bearing load.
The future standard has to distinguish at least between turbine type (Francis,
Kaplan, Bulb, Pelton and Pump) and between bearing location (turbine
bearing and generator bearings). Preferable is also a separation into shaft
orientation (horizontal and vertical).
Corresponds to Vib_DB_Revision_E-(2010-09-20)
Corresponds to Vib_DB_Revision_F-(2011-11-24)
3 Corresponds to Vib_DB_Revision_J-(2013-03-29)
1
2
Reference values are suggested to be based on the median values, at least for
turbine and generator bearings.
Actions are suggested to be undertaken if the actual vibration value exceeds

1.6 and 2.5 times the reference value:
1.6 times the reference value corresponds roughly to the 75 percentile.
2.5 correspond approximately to the 90 percentiles.
Problem units are evenly distributed over the complete range of measured
vibration values. Surprisingly, machines labeled as problem in the database
do not have exceptionally high vibration levels if compared to other measured
values which are considered to be normal. However, the proportion of
machines marked as problem increases with higher vibration values. At
approximately 2.5 times the median value the proportion of problem units
increases more radically. This could prove that the proposed boundary levels
2.5 and 1.6 times the median value makes sense.
The suggested boundaries are significantly lower than the current boundary zones in
the existing standard. The suggestion was supported by analysis that shows that the
median value for the Burr-distributions is very close to the mean values for the
datasets. And since the median value method objectively excludes extreme values, this
method is promoted for finding the adequate reference values. The method is also
verified through comparisons of the median value for a dataset with only
measurements in best operating range and an unfiltered dataset. The resulting median
values is nearly equal for the two compared datasets.
During 2013-2014, analysis of database revision J was also conducted. Since the number
of measurements was increased with this version the aim of the analysis was to find
reference values which could be used for producing action limits. These limits were
produced by using the boundary levels recommended from earlier analysis and the
median values calculated from the database revision J. The recommended action limits
from this analysis is generally at the same level as calculated from previous revisions of
the database. Exceptions are shaft oscillations for Bulb units which results in increased
values compared to previous database revisions.
Suggested future work:
Verify the suggested action limits with more or improved data if the database
is revised.
Refine the problem definition in the database.
Identify relevant bearing groups from the database for parameter correlation.
Identify the relationship between the vibrations of the generator guide

bearings and the turbine guide bearing.
Explain the measured shaft movements that is larger than the specified
available bearing clearance.
It is also suggested to identify the impact of the current requirements of

IEC/ISO of 30 m p-p for stationary parts. Current ISO 10816-5 action limits of
30 m p-p is more severe for the majority of large-scale turbines than the
future recommended limits that will be expressed in mm/s.
List of contents
1
Introduction
1.1
International workgroup ISO/TC 108/SC 2 & IEC/TC 4 - JWG1 Vibration of

Hydraulic Machines
1.2
Scope of work
Analyzing method
2.1
Database
2.2
Filterning of data and assessment of the validity of the current database
2.3
Analyzing method
11
2.3.1 Step 1 and 2
11
2.3.2 Step 3 12
2.3.3 Step 4 12
3
Analysis
3.1
13
Analyze the database versus a number of different parameters in order

to find correlations and physical explanations on the findings
13
3.1.1 Head 13
3.2
3.3
3.1.2 Runner nominal speed
17
3.1.3 Runner diameter
20
3.1.4 Radial bearing clearance
23
3.1.5 General observations
26
3.1.6 Physical approach
26
3.1.7 Statistical approach
27
Analyze an unfiltered and more recent version of the database
31
3.2.1 Medium vibration levels for turbine types and shaft orientation
31
3.2.2 Comparison of median values between database revision E and F
34
3.2.3 Distribution of measured values and units marked as problem
35
Analyze the most recent database with methods established in earlier

steps in order to propose action limits for the different machine types
38
3.3.1 Median values from the database revision J
38
3.3.2 Analysis for determination of median values per machine group
39
Conclusions
43
Future work
45
References
46
Appendix
47
7.1
Appendix 1
47
7.2
Appendix 2
55
Introduction
1.1
INTERNATIONAL WORKGROUP ISO/TC 108/SC 2 & IEC/TC 4 - JWG1 VIBRATION OF

HYDRAULIC MACHINES
For the revision and integration of the current mechanical vibration standards for
hydraulic power generating and pumping plants ISO/IEC 7919-5 and 10816-5 IEC and
ISO are supporting an international workgroup (ISO/TC 108/SC 2 & IEC/TC 4 - JWG1
Vibration of Hydraulic Machines).
Swedish delegates in the international workgroup are ke Grahn and Anders Bard and
a Swedish working group has been formed for dealing with this topic.
1.2
SCOPE OF WORK
To support the international working group ISO/TC 108/SC 2 & IEC/TC 4 - JWG1
analysis of the updated TK4 vibration database has been performed by the Swedish
working group.
The purpose of the analysis has been to form a statistical foundation for a
recommendation on vibration limits in a new integrated vibration standard.
Earlier revisions of the database have been evaluated through a master thesis by
Junnosuke Oguma performed at Lule University of Technology in 2009, and pervious
project work at SWECO. A number of shortages were highlighted and no correlation
between measured values and dangerous vibration levels could be found. For
stationary parts, no analyses were made due to lack of measurement data in the
database. Present database has been expanded which proposes further study of the
database.
This report describes the analysis work carried out by the Swedish national workgroup
and the preliminary results this has led to. The work has been conducted in four steps:
1.
2.
3.
4.
Establish an assessment for the validity of the current database to see if data was
improved.
Analyze the database versus a number of different parameters in order to find
correlations and physical explanations on the findings.
Analyze an unfiltered and more recent version of the database.
Analyze the most recent database with methods established in earlier steps in
order to propose action limits for the different machine types.
Analyzing method
2.1
DATABASE
As a statistical foundation for the revised standard an international vibration database

have been developed. The analysis performed in this report are based on the database
versions Vib_DB_ Revision_E 4, Vib_DB_Revision_F 5 and Vib_DB_Revision_J 6. Revision
E contains 2392 rows, revision F contains 2472, whereas revision J contains 7355 rows.
Every row corresponds to one measurement. The database contains measurements on
all types of hydraulic power generating and pumping machines and commonly more
than one measurement on each machine.
2.2
FILTERNING OF DATA AND ASSESSMENT OF THE VALIDITY OF THE CURRENT

DATABASE
During step 1 and 2 of the work scope in this project, revision E of the database was
considered. The database contains numerous machines with more than one
measurement. In order for the statistics to be comparable each machine must be given
equal weight. Because of this the database has been filtered so that only one
measurement per machine remains. The most excluding filtering condition was that
rows not containing both bearing vibrations and shaft oscillations were removed. The
remaining measurements are chosen in order that it only contains vertical machines of
the type Kaplan and Francis, for the type Bulb also horizontal machines are included.
For the parameter relative output at measurement all measurements outside the
interval below were excluded.
Francis: 70-100%
Kaplan: 50-100%
Bulb: 0-100%
It was desirable to use measurements done close to the operation point at maximum
efficiency. Due to lack of relative flow data, this operation point was not possible to
define for all measurements. The parameter ISO machine group was considered and
for the types Kaplan and Francis, measurements with value 1 and 2 were filtered out.
For Bulb no filtration of this parameter was done. Table 1 below shows the number of
remaining measurements after filtration.
Type
Generator guide bearing

non drive end

drive end
Turbine guide bearing
Shaft
Shaft
Shaft
Bearing
Bearing
Bearing
Smax
Sp-p Sp-p VRMS
Smax
Sp-p Sp-p VRMS
Smax
Sp-p Sp-p VRMS
Francis 146
37
114
142
26
106
146
65
12
138
Kaplan 41
20
36
43
15
33
46
32
46
Bulb
Table 1: Number of existing measurements after filtration of database revision E
Corresponds to Vib_DB_Revision_E-(2010-09-20)
Corresponds to Vib_DB_Revision_F-(2011-11-24)
6 Corresponds to Vib_DB_Revision_J-(2013-03-29)
4
5
As can be seen from Table 1 only a fraction of the measured data remains after the
filtration. Earlier reports concluded that few measurements included vibration data on
both rotating and stationary parts. With revision E this has been improved, see table 2
below.
non drive end
Generator guide bearing drive

end
Spp or Smax +
bearing
housing
vibration
displacement
Spp or
Smax +
bearing
housing
vibration
velocity
Spp or Smax +
bearing
housing
vibration
displacement
Spp or Smax +
bearing
housing
vibration
velocity
Spp or Smax
+ bearing
housing
vibration
displacement
Spp or
Smax +
bearing
housing
vibration
velocity
Bulb
Francis
100
92
10
114
Kaplan
25
25
34
Type
Table 2: Number of measurements on both rotating and stationary parts in database revision E
Although the database has been improved, still it suffers from inconsistency. For this
reason the data set varies depending on the chosen parameter.
In the database revision F all data was considered during the analysis. For consistency
a separation between machine types and shaft orientation was done. Also, units
marked as problem was analyzed separately. Francis-, Kaplan- and Pump-units
which had undefined shaft orientation was assumed to be vertical if the runner
diameter exceeded 2 m. Bulb-units which had undefined shaft orientations was
assumed to be horizontal if the runner diameter exceeded 2 m.
10
Table 3 below summarizes the separation of data.

Type
Pelton
Francis
Kaplan
Pump
Bulb
Other
Shaft orientation
Total number
Problem units
horizontal
73
14
vertical
67
undefined
total
145
18
horizontal
161
vertical
1219
129
undefined
45
31
total
1425
167
horizontal
10
vertical
361
23
undefined
total
373
27
Horizontal
39
vertical
361
41
undefined
18
total
409
59
horizontal
50
vertical
undefined
total
51
total
69
2472
274
Total
Table 3: Number of measurements separated in turbine types, shaft orientation and problem units
During step 4 of the work scope in this project, database revision J of the database was
considered. Database revision J has been improved with a vast number of
measurements. The majority of data originates from Chinese measurements. Since
numerous of the Chinese measurements are conducted on only a few units the data
have been filtered according to the following method. Analysis of the database
included all measurements but with Chinese data reduced so that only maximum three
measurements per unit are included. Selection of measurements in the normal
operating range is prioritized and measurements marked as problem are excluded.
2.3
ANALYZING METHOD
2.3.1
Step 1 and 2
For step 1 (Establish an assessment for the validity of the current database to see if data
was improved) and 2 (Analyze the database versus a number of different parameters in
order to find correlations and physical explanations on the findings) of the work scope
database revision E was analyzed. Each data set was plotted in diagrams with different
11
parameters and vibration values. Due to the scattered data no clear trends were
observed. A curve fitting method based on the median values was then adopted. The
data was separated into smaller groups with the same numbers of data. For each group
the median values was calculated for abscissa and ordinate and then projected onto the
diagrams with the scattered source data. With this method extreme values are
objectively excluded and do not influence any assumed trend.
Also an attempt in finding the statistical distribution function of the vibration data has
been done. This was mainly done by software routines producing the best fit for each
data set. Each parameter analysis is made on each machine type, Kaplan and Francis.
For Bulb, no analysis was done since the available data set was considered too small.
Bearing location have been considered and also the arrangement of generator bearings.
2.3.2
Step 3
For step 3 (Analyze an unfiltered and more resent version of the database) of the work
scope in this project database revision F was analyzed. For the analysis the median
value was calculated for the groups defined in table 3. The different bearing locations
was considered and also if the measurements was marked as problem in the
database. The median values for the individual groups were then plotted in bar charts
for visualization of vibration levels. In addition, the median values from the filtered
database revision E for vertical Francis and Kaplan units was projected on the
corresponding charts for revision F. This for comparison with revision F. The
distribution of measured values has also been plotted and projected on the proportion
of units marked as problem.
2.3.3
Step 4
For step 4 (Analyze the most recent database with methods established in earlier steps
in order to propose action limits for the different machine types) the filtered database
revision J was analyzed with the same method as for step 3. The median values are
presented in a table where also the action limit values are presented.
12
Analysis
3.1
ANALYZE THE DATABASE VERSUS A NUMBER OF DIFFERENT PARAMETERS IN ORDER

TO FIND CORRELATIONS AND PHYSICAL EXPLANATIONS ON THE FINDINGS
A number of parameters have been evaluated for database revision E in order to find
correlation of vibration levels. The parameters considered are:
Head
Nominal speed
Runner diameter
Radial bearing clearance
All plots are presented below and also in appendix1.
3.1.1
Head
The utilized dynamic bearing clearance (UDBC) shows no or small tendency of relation
to increasing head. The turbine guide bearing (TGB) in Francis machines shows an
increasing tendency with increasing head. All other bearings show no correlation
between UDBC and head. See figure 1 and 2.
Figure 1: UDBC vs. Rated head for Francis
13
Figure 2: UDBC vs. Rated head for Kaplan
14
The shaft vibration displacement peak-to-peak value (Spp) shows no correlation to

head. For Kaplan units the data is very scattered and the fitted median curve fluctuates
heavily. See figure 3 and 4.
Figure 3: Spp vs. Rated head for Francis
Figure 4: Spp vs. Rated head for Kaplan
15
The bearing housing vibration velocity shows a small increase with higher heads. This
trend is most apparent for The turbine guide bearing (TGB) in Kaplan units. See figure
5 and 6.
Figure 5: Vibration velocity vs. Rated head for Francis
Figure 6: Vibration velocity vs. Rated head for Kaplan
No clear correlation between vibration values and head can be identified. However, for
The turbine guide bearing (TGB), an increase in vibration velocities can be observed for
higher heads, no or weak influence can be observed for generator guide bearing (GGB).
This is valid for both turbine types. For Francis turbines a small increase in vibration
level are found for heads above 200 meters.
16
3.1.2
Runner nominal speed
The utilized dynamic bearing clearance (UDBC) has a somewhat increasing trend for
turbine guide bearing (TGB) in Francis units at increasing speed. A similar but weaker
trend is present for generator guide bearing (GGB) in Kaplan units. Beyond these two,
no relation between UDBC and nominal speed can be found, see figure 7 and 8.
Figure 7: UDBC vs. Sync. speed for Francis
Figure 8: UDBC vs. Sync. speed for Kaplan
17
No relation between shaft vibration displacement peak-to-peak (Spp) and nominal

speed could be found. Measurement data for both Francis and Kaplan is very scattered
and the fitted median curve fluctuates between 50 and 150m for the whole speed
range, see figure 9 and 10.
Figure 9: Spp vs. Sync. Speed for Francis
Figure 10: Spp vs. Sync. speed for Kaplan
18
For turbine guide bearing (TGB) the bearing housing vibration velocity is clearly
increasing with higher nominal speed in both turbine types. For GGB no clear trend is
shown, see figure 11 and 12.
Figure 11: Vibration velocity vs. Sync. Speed for Francis
Figure 12: Vibration velocity vs. Sync. speed
No clear correlation between vibration values and nominal speed could be observed.
For turbine guide bearing correlation between vibration velocity and speed can be
identified.
19
3.1.3
Runner diameter
The utilized dynamic bearing clearance (UDBC) shows no clear trend for increasing
diameter. The turbine guide bearing in Francis units shows decreasing UDBC with
increasing runner diameter, this cannot be observed for the Kaplan turbine or for
generator guide bearings. See figure 13 and 14.
Figure 13: UDBC vs. Nominal diameter for Francis
Figure 14: UDBC vs. Nominal diameter Kaplan
20
No relation between shaft vibration displacement peak-to-peak (Spp) and runner

diameter could be identified. For Kaplan units the fitted 50%-probability curve
fluctuates heavily for both turbine guide bearing (TGB) and generator guide bearing
(GGB). For Francis turbines the vibration values are rather constant for increasing
diameter, see figure 15 and 16.
Figure 15: Spp vs. Nominal diameter for Francis
Figure 16: Spp vs. Nominal diameter for Kaplan
21
The bearing housing vibration velocity shows decreasing values with increasing
diameter for turbine guide bearing (TGB). This trend is most clear at smaller diameters
up to 2m. Above this diameter the vibration trend is relatively constant. No clear trend
can be recognized for generator guide bearing (GGB), see figure 17 and 18.
Figure 17: Vibration velocity vs. Nominal diameter for Francis
Figure 18: Vibration velocity vs. Nominal diameter for Kaplan
No clear correlation between vibration values and runner diameter could be observed.
22
3.1.4
Radial bearing clearance
The utilized dynamic bearing clearance (UDBC) shows a decreasing trend for all
turbine types when radial bearing clearance is increasing, see figure 19 and 20.
Figure 19: UDBC vs. Radial bearing clearance for Francis
Figure 20: UDBC vs. Radial bearing clearance for Kaplan
23
The shaft vibration displacement peak-to-peak (Spp) measurements are again very
scattered and the fitted median curve fluctuates over a wide range for both Kaplan and
Francis. No trend can be observed. In figure 21 and 22 the data is presented and also
the limit for 100% utilized dynamic bearing clearance. Note that some measurements
have larger shaft displacements than available bearing clearance, hence UDBC > 100%.
Figure 21: Spp vs. Radial bearing clearance for Francis
Figure 22: Spp vs. Radial bearing clearance for Kaplan
24
The bearing housing vibration velocity shows no clear trend for increasing radial
bearing clearance. For GGB the trend is increasing for Kaplan whiles for the same
bearing locations for Francis the median curve is rather constant. The turbine guide
bearing (TGB) in Kaplan shows a decreasing trend while the opposite is found in
Francis TGB, see figure 23 and 24.
Figure 23: Vibration velocity vs. Radial bearing clearance
Figure 24: Vibration velocity vs. Radial bearing clearance
Most of the radial bearing clearance measurements are gathered around 200m which
makes possible trends difficult to detect.
25
3.1.5
General observations
For Kaplan turbines utilized dynamic bearing clearance for the turbine guide
bearing (TGB) is generally lower than for the generator guide bearings (GGB),
19% and 26% respectively. For Francis, the opposite are found, 25% and 22%
respectively.
For Kaplan shaft oscillations are lower for TGB than for GGB, 85m and
115m. For Francis units the vibration values are about the same, 100m, for
all bearing locations.
The vibration velocities for Kaplan units are 0,7 mm/s for TGB and 0,3 mm/s
for GGB. Corresponding values for Francis turbines are 0,5 mm/s and 0,3
mm/s, respectively.
The turbine guide bearing shows generally higher vibration values than
generator bearings. This is valid for both Francis and Kaplan units.
Suspended type generators shows considerably lower vibration levels on both

upper and lower generator bearing than semi-umbrella type units.
3.1.6
Physical approach
The analysis shows that both the shaft oscillation and the vibration velocity
measurements in the database do not correlate to the studied parameters. Regarding
vibration velocity it is possible to keep a physical argumentation that shows a
relationship between vibration velocity and mechanical stress:
It can be shown that the size of the supporting structure for the turbine guide
bearing is proportional to the runner diameter. It can also be shown that the
size of the bearing brackets for the generator guide bearings is proportional to
the rotor diameter. If all bearing brackets have similar design criteria and
material properties, Hookes law implies that the permissible strain will be
constant. However, since strain and displacement is related through size,
increasing size will result in larger displacements. Consequently, the allowed
displacement is proportional to the turbine or generator diameter.
Previous studies have shown that circumferential velocity for the turbine is
rather constant for all types of reaction turbines. This implies that the
rotational speed is inversely proportional to the diameter of the runner.
The combined conclusion for this argumentation is that large machines

experience high deflections with low rotational frequency whereas small
machines experience small deflections with high rotational frequency. The
stress levels in the supporting structure is however equal for all machine sizes
and thus the vibration velocity is constant.
Regarding shaft oscillation and the lack of correlation with bearing journal diameter,
the physical explanation is that all bearing clearances are designed for supplying a
carrying oil film even at small shaft eccentricities [1]. In conjunction to this the IEEE Std
810-1987 (R2001) [2] specifies the total allowable run out for a shaft system to 76 m.
This is a maximum value independent of the shaft diameter. Clearly, the total allowable
run out must be accommodated within a normal bearing clearance.
26
3.1.7
Statistical approach
Since no clear correlation between parameters and measured vibration data were
observed the aim was to find a probability function that could fit the measured data set.
A good probability distribution could repair a database with too few data. Each
measured vibration value is assumed to be a stochastic variable and hence form a
continuous distributed sample. The sample is then discretized by assigning frequency
values to intervals of equal distance as to form a histogram. It is clear that all samples
have a positive skewness which indicates that the tail on the right side is longer than
the left side and the majority of the values lie to the left of the mean. The curve fitting
process was done by software routines and the best fit was produced by the Burr
distribution, see figure 25 and 26.
Figure 25: Discretized sample of shaft oscillation and projected the fitted Burr-distribution
Figure 26: Discretized sample of vibration velocity and projected the fitted Burr-distribution
27
The Burr distribution is applied in a variety of areas such as reliability studies and
failure time modeling. Unlike other failure time distributions such as Weibull and
Rayleigh, the Burr distribution contains two shape parameters. This makes the
distribution more versatile when fitted onto a sample. Below, in figure 27, is the fitted
Burr distribution for vibration velocities with definitions on the mode, mean and
median values.
Figure 27: Definitions of mode, mean and median
In table 4, the values mean, median and mode is summarized for each data set. The
median value can be derived from the measurement data or from the distribution. It is
separating the greater and lesser halves in the data set. The mean is also calculated
from the measured data set. The mode and the median values are taken from the
probability density function i.e. the Burr distribution. Since the distribution have a
positive skew, the mode (peak of the distribution) will lie to the left of the median
value.
28
Each data set and distribution is presented in appendix 2.

Shaft vibration Spp [m]
Data set
Type
Measured values
From distribution func.
Mean
Median
Mode
Median
Francis
138
100
62
100
Kaplan
163
117
71
122
Francis
115
93
62
91
Kaplan
139
85
45
93
Vibration velocity [mm/s]

Data set
Type
Measured values
From distribution func.
Mean
Median
Mode
Median
Francis
0,46
0,31
0,23
0,31
Kaplan
0,46
0,32
0,22
0,32
Francis
0,67
0,52
0,40
0,53
Kaplan
0,70
0,66
0,49
0,62
Table 4: Mean-, median- and mode- values for each data set and bearing location
An approach in controlling the accuracy for the distribution is made in table 5. The
method used is to compare the measured values (median and mean) with the
corresponding calculated values from the probability density function (median and
mean).
Shaft vibration Spp, accuracy of distribution
Data set
Type
Median
Mean
Francis
1,00
0,98
Kaplan
0,96
0,98
Francis
1,01
1,00
Kaplan
0,92
0,97
Type
Median
Mean
Francis
1,03
0,90
Kaplan
1,00
0,90
Francis
0,98
0,99
Kaplan
1,06
1,00
Vibration velocity, accuracy of distribution

Data set
Table 5: Accuracy of distribution. Calculated as median (measured) / median (from distribution) and mean
(measured) / mean (from distribution)
The accuracy for the Burr distribution is by this method in the range of 100-90%. The
best fit is produced for Francis turbines when shaft vibration is considered.
For safe and reliable running of the machine under normal operation conditions
requires that the vibration values should remain below certain limits. According to ISO
7919-5 and ISO 10816-5, the limits are defined by zone boundary values. The ratios
between the zone boundaries were, according to the standards, found through
discussions within the workgroup and with experts in the field. The ratios are 1.6x and
29
2.5x a specific reference value. 2.5 times the reference value corresponds to increase of
turbine vibration level that leads to essential change of its vibration state. Also, the ratio
of the product of the two suggested values to 2.5 is equal to the ratio of 2.5 to 1.6, hence
the ratio is within the Golden ratio which has a huge number of applications in the
nature. The reference value may be a subject of discussion, in this report the suggestion
is to use either the median value or the mode value. Table 6 below presents the
percentiles and the actual values for 1.6x and 2.5x the reference value.
Shaft oscillation Spp [m]
1.6x
2.5x
Data set
Type
Mode
Median
Mode
Median
Generator guide
bearing
Francis
50% (99)
73% (160)
72% (155)
88% (250)
Kaplan
47% (114)
73% (195)
69% (178)
88% (304)
Francis
55% (99)
75% (146)
71% (155)
91% (228)
Kaplan
39% (72)
71% (148)
59% (113)
85% (231)
Vibration velocity [mm/s]

1.6x
2.5x
Data set
Type
Mode
Median
Mode
Median
Generator guide
bearing
Francis
61% (0,37)
76% (0,50)
82% (0,58)
89% (0,78)
Kaplan
56% (0,35)
75% (0,51)
78% (0,55)
88% (0,80)
Francis
63% (0,64)
78% (0,85)
85% (1,00)
92% (1,33)
Kaplan
66% (0,78)
80% (0,99)
89% (1,23)
96% (1,55)
Table 6: Percentiles and actual values for 1.6x and 2.5x the reference values
Using the mode as reference value leads to considerably low boundary values,
sometimes lower than the median. Using the median value produces more realistic
values for reference. 1.6 corresponds to around 75% probability and 2.5 corresponds to
around 90% probability.
Figure 28 and 29 shows the boundaries for 1.6x and 2.5x reference value, here the 50%probability value, for both shaft vibration and vibration velocity, is used.
30
Figure 28: Boundary for 1.6 and 2.5 times the reference value (160 resp. 250m)
Figure 29: Boundary for 1.6 and 2.5 times the reference value (0.6 resp. 1.0mm/s)
3.2
ANALYZE AN UNFILTERED AND MORE RECENT VERSION OF THE DATABASE
3.2.1
Medium vibration levels for turbine types and shaft orientation
For step 3 (Analyse an unfiltered and more resent version of the database) of the work
scope in this project database revision F was analyzed. For the analysis the median
value was calculated for the groups defined in table 3. Figure 30-33 shows the vibration
levels for the specific turbine types and shaft orientations. Note that units marked as
problem is not included in the dataset.
31
160,00
140,00
120,00
Spp [m]
100,00
Pelton
Francis
80,00
Kaplan
60,00
Pump
40,00
20,00
0,00
GGB NDE
GGB DE
TGB
Figure 30: Level of Spp for vertical units
1,60
Vib. veloc.[mm/s]
1,40
1,20
1,00
Pelton
0,80
Francis
0,60
Kaplan
Pump
0,40
0,20
0,00
GGB NDE
GGB DE
Figure 31: Vibration velocity level for vertical units
32
TGB
160,00
140,00
120,00
Spp [m]
100,00
Pelton
80,00
Francis
Bulb
60,00
40,00
20,00
0,00
GGB NDE
GGB DE
TGB
Figure 32: Level of Spp for horizontal units
1,60
1,40
Vib. veloc. [mm/s]
1,20
1,00
Pelton
0,80
Francis
0,60
Bulb
0,40
0,20
0,00
GGB NDE
GGB DE
TGB
Figure 33: Vibration velocity level for horizontal units
The median values are summarized in table 7 below.

Vertical units
GGB NDE
GGB DE
TGB
Spp
Vrms
Spp
Vrms
Spp
Vrms
Pelton
109
0,60
126
0,56
88
0,48
Francis
100
0,35
99
0,35
108
0,56
Kaplan
116
0,44
104
0,37
67
0,69
33
Pump
139
0,57
102
0,46
110
1,37
Horizontal units
GGB NDE
Spp
GGB DE
Vrms
Spp
TGB
Vrms
Spp
Vrms
Pelton
60
0,47
84
1,30
74
1,45
Francis
63
0,55
80
0,55
148
0,94
Bulb
156
0,30
50
0,48
46
1,02
Table 7: Median values for the relevant groups
3.2.2
Comparison of median values between database revision E and F
The filtered dataset from database revision E which only contains measurements of
units in the best operating range is compared with the complete database revision F,
see figure 34-35. The result should be used to validate the median method.
120
Spp [m]
100
80
60
Best OP
40
All
20
0
GGB NDE
GGB DE
TGB
0,70
Vib. veloc. [mm/s]
0,60
0,50
0,40
Best OP
0,30
All
0,20
0,10
0,00
GGB NDE
GGB DE
TGB
Figure 34: Comparison of median values for vertical Francis units for dataset only containing measurements at
best operation point and dataset containing all measured values
34
120
Spp [m]
100
80
60
Best OP
40
All
20
0
GGB NDE
GGB DE
TGB
0,70
Vib. veloc. [mm/s]
0,60
0,50
0,40
Best OP
0,30
All
0,20
0,10
0,00
GGB NDE
GGB DE
TGB
Figure 35: Comparison of median values for vertical Kaplan units for dataset only containing measurements at
best operation point and dataset containing all measured values
3.2.3
Distribution of measured values and units marked as problem
Figure 36 and 37 below shows the distribution of measurements on generator guide

bearings (GGB) and turbine guide bearings (TGB) for vertical Francis and Kaplan units.
Both normal and units marked as problem is included in the distribution. Projected is
the proportion of problem units expressed as percentage of the total number of
measurements. Projected is also the median value and the proposed boundary limits.
35
100%
300
250
80%
Quantity
200
60%
150
40%
100
Normal
Problem
Median
1,6x median
2,5x median
20%
50
Problem percentage
0%
0
0
100 200 300 400 500 600 700

Group of Spp [m]
100%
300
250
80%
Quantity
200
60%
150
40%
100
Normal
Problem
Median
1,6x median
2,5x median
20%
50
Problem percentage
0%
0
0
100 200 300 400 500 600 700

Group of Spp [m]
Figure 36: Distribution of measurements for "normal" units (blue) and units marked as "problem" (red). The
solid black curve is a representation of the proportion of problem units in the distribution. Distribution at top
is for generator bearings and bottom is for turbine guide bearings.
36
100%
300
250
80%
Quantity
200
60%
150
40%
100
Normal
Problem
Median
1,6x median
2,5x median
20%
50
Problem percentage
0%
0
0
Group of vib.veloc. [mm/s]

100%
300
250
80%
Quantity
200
60%
150
40%
100
Normal
Problem
Median
1,6x median
2,5x median
20%
50
Problem percentage
0%
0
0
Group of vib.veloc. [mm/s]

Figure 37: Distribution of measurements for "normal" units (blue) and units marked as "problem" (red). The
solid black curve is a representation of the proportion of problem units in the distribution. Distribution at top
is for generator bearings and bottom is for turbine guide bearings.
37
3.3
ANALYZE THE MOST RECENT DATABASE WITH METHODS ESTABLISHED IN EARLIER

STEPS IN ORDER TO PROPOSE ACTION LIMITS FOR THE DIFFERENT MACHINE TYPES
3.3.1
Median values from the database revision J
The median values are calculated from the complete filtered revision J of the database,
see chapter 2.3.3. Table 8 and 9 shows the vibration levels and action limits for 1.6x and
2.5x the median value for the specific turbine types and shaft orientations. Note that
units marked as problem is not included in the dataset. Median values based on less
than 10 measurements are highlighted in red and values based on 10-30 measurements
are highlighted in yellow.
Machine type:
Shaft oscillation S_pp (m)
Bearing vibration V_rms (mm/s)
Francis vertical
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
Number of measurements
926
537
759
328
258
287
Median
108
111
99
0,6
0,3
0,3
Median x 1.6 (A/B)
174
178
158
0,9
0,5
0,5
Median x 2.5 (C/D)
271
278
247
1,4
0,8
0,9
Machine type:
Pump-Turbine vertical
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
228
173
185
155
102
145
Median
104
97
136
1,2
0,4
0,6
Median x 1.6 (A/B)
166
155
218
1,9
0,7
0,9
Median x 2.5 (C/D)
259
243
341
3,0
1,1
1,5
Machine type:
Kaplan vertical
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
253
161
113
129
82
102
Median
68
105
102
0,7
0,4
0,4
Median x 1.6 (A/B)
109
168
163
1,1
0,6
0,7
Median x 2.5 (C/D)
170
263
255
1,8
1,0
1,1
Machine type:
Pelton vertical
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
29
31
32
35
37
44
Median
83
88
107
0,5
0,5
0,6
Median x 1.6 (A/B)
133
141
171
0,8
0,8
1,0
Median x 2.5 (C/D)
208
221
267
1,2
1,3
1,5
Machine type:
Pump vertical
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
30
18
36
14
Median
121
106
111
0,1
0,4
0,4
Median x 1.6 (A/B)
194
170
178
0,2
0,6
0,6
Median x 2.5 (C/D)
303
266
278
0,3
0,9
1,0
Table 8: Median values and action limits for 1.6x and 2.5x the median value for vertical units
38
Machine type:
Francis horizontal
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
69
112
139
37
35
27
Median
173
80
65
0,9
0,5
0,6
Median x 1.6 (A/B)
277
127
104
1,4
0,8
1,0
Median x 2.5 (C/D)
432
199
162
2,3
1,3
1,5
Machine type:
Bulb horizontal (*)
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
42
29
24
34
24
33
Median
52
140
133
1,6
0,4
0,3
Median x 1.6 (A/B)
83
224
213
2,6
0,6
0,5
Median x 2.5 (C/D)
130
350
333
4,1
1,0
0,8
(*) Double regulated

Machine type:
Pelton horizontal
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
34
39
36
39
47
47
Median
74
83
61
1,0
1,3
0,7
Median x 1.6 (A/B)
118
133
97
1,6
2,1
1,1
Median x 2.5 (C/D)
185
208
151
2,4
3,3
1,8
Machine type:
Pump horizontal
T1
GE - DE
GE - NDE
T1
GE - DE
GE - NDE
26
30
39
Median
194
73
59
0,3
0,2
Median x 1.6 (A/B)
311
117
95
0,5
0,4
Median x 2.5 (C/D)
486
183
148
0,7
0,6
Table 9: Median values and action limits for 1.6x and 2.5x the median value for horizontal units
The calculated median values for the different types of machines and bearing locations
are generally at the same level as calculated from previous revisions. Exceptions are
shaft oscillations for Bulb units which have increased values versus previous revisions.
Note that although the vast number of measurements in database revision J, there are
still machine types that lack measurements for determination of action limits. Although
the database has been substantially increased the number of measurements of bearing
housings is the same as for previous revision.
3.3.2
Analysis for determination of median values per machine group
During discussions with the international work group, JWG1, there were requests for
presenting separate action values for housing vibrations for the different machine
groups. During this analysis an examination whether there existed a common factor for
the relationship between the groups was done. Since many of the measurements in the
database lacks information of machine groups, several filtrations was performed, see
table 10.
39
7R
8a R
8b R
9R
Information on
Filtration: machine
vibration
problems known
at measurement
IEC Machine type

classification
ISO Machine Group
Shaft orientation
Empty
F+K
Empty
F+K
V + empty
Empty
F+K
3 + empty
4 + empty
Empty
F+K
3 + empty
4 + empty
V + empty
Table 10: Explination of the different types of filtrations
None of the chosen filtrations could present a dataset from which there could be a clear
change in vibration magnitude between the groups, see table 11-13. Note that cells in
red are based on 10 actual measurements and cells in yellow are based on 11-30 actual
measurements.
Factor
(group
4
/group
3)
GGB NDE
Filtration: 7 R
8a
R 8b R
9R
Vib.velocity
RMS
number median
number
empty K 3
76
0,45
14
0,00
empty K 3
V+
empty 76
0,45
14
0,00
3+
empty K empty V
673
0,44
100
0,78
3+
V+
empty K empty empty 703
0,45
102
1,00
GGB NDE pos.1
Filtration: 7 R
8a
R 8b R
GGB DE pos.2
9R
Vib.velocity
Vib.velocity
RMS
RMS
number median
number median
number
Factor
(pos.1
/
pos.2)
empty K 4
83
empty K 4
V+
empty 83
4+
empty K empty V
680
0,34
86
0,34
69
1,00
4+
V+
0,45
88
0,37
73
1,20
Table 11: Type of filtration, calculated median value and common factor between the groups and for bearing
positions.
40
Factor
(group
4
/group
3)
GGB NDE
Filtration: 7 R
8a
R 8b R
9R
Vib.velocity
RMS
number median
number
empty F
381
0,45
14
1,78
empty F
V+
empty 381
0,45
14
1,78
empty F
3+
empty V
3493
0,33
279
1,00
empty F
3+
V+
empty empty 3666
0,33
280
1,00
GGB NDE pos.1
Filtration: 7 R
8a
R 8b R
GGB DE pos.2
9R
Vib.velocity
Vib.velocity
RMS
RMS
number median
number median
number
Factor
(pos.1
/
pos.2)
empty F 4
91
0,80
0,54
18
1,48
empty F 4
V+
empty 91
0,80
0,54
18
1,48
4+
empty F empty V
3203
0,33
268
0,33
251
1,00
4+
V+
empty F empty empty 3376
0,33
269
0,33
251
1,00
positions.
41
Factor
(group
4
/group
3)
GGB NDE
9R
Vib.velocity
RMS
number median
number
F
+
empty K 3
457
0,45
28
1,80
F
+
empty K 3
V+
empty 457
0,45
28
1,80
F
+ 3+
empty K empty V
4166
0,37
379
0,99
F
+ 3+
V+
0,37
382
1,00
Filtration: 7 R
8a
R 8b R
GGB NDE pos.1
GGB DE pos.2
Factor
(pos.1
/
pos.2)
9R
Vib.velocity
Vib.velocity
RMS
RMS
number median
number median
number
F
+
empty K 4
174
0,80
0,65
1,23
F
+
empty K 4
V+
empty 174
0,80
0,65
1,23
F
+ 4+
empty K tomma V
3883
0,37
354
0,33
310
1,11
F
+ 4+
V+
empty K tomma empty 4086
0,37
357
0,33
314
1,12
Filtrering: 7 R
8a
R 8b R
positions.
Since no clear change in vibration magnitude between the groups could be detected the
proposed action limits in table 8 and 9 should be valid for all machine groups.
42
Conclusions
No clear correlation between vibration values and the unit specific parameters such as
head, rotational speed, runner diameter and radial bearing clearance could be
observed. The lack of correlation implies paradoxically that both shaft oscillations and
vibration velocities are relevant parameters:
It is common to use functions that cover the trends and coefficients for the
calibration. An indication that the function is appropriate is that the
coefficients are independent of all parameters. This can be applied for the
physical parameters vibration level and shaft oscillations.
In section 3.1.6, a physical argumentation is presented based on stress levels in the

supporting structure for the guide bearings and that the circumferential velocity for the
turbine is rather constant for all types of reaction turbines. The conclusion from this
explanation is summarized in figure 38 below.
Vibration velocity
Speed
Diameter (size)
Figure 38: The relation between size, displacement and speed
The argumentation is further described in the presentation JWG1 Vibrations on

hydraulic Machines, The physical relevance for the vibration velocity parameter.
Several shortcomings can be identified for the parameter utilized dynamic bearing
clearance. High magnetic unbalance in the generator and high hydraulic unbalance in
the turbine gives small shaft oscillations and small UDBC, bearing load can however be
very high. A poorly aligned shaft arrangement can also give low values on shaft
oscillations and UDBC, although high bearing load.
Also a physical explanation of the shaft oscillations not correlating to the bearing
journal diameter is presented in section 4.6. The conclusion from this is that all bearing
clearances are chosen so that even small eccentricities will give rise to a supporting oil
film. The clearance is thus not a function of size. The minimum clearance is limited by
the different expansion rates of shaft and bearing due to the temperature rise at start of
the unit. It should also be mentioned that the majority of radial clearance values in the
database is specified as design value. The actual bearing clearance at the specific
measurement is therefore not known. This might explain some of the situations when
the shaft oscillations are larger than the specified bearing clearance, UDBC > 100%.
43
The future standard has to distinguish at least between turbine type (Francis, Kaplan,
Bulb, Pelton and Pump) and between bearing location (turbine bearing and generator
bearings). Preferable is also a separation into shaft orientation (horizontal and vertical).
A statistical study of database revision E found the measured data to be Burrdistributed. Vibration reference values are here suggested to be based on the median
value for at least turbine and generator bearings. Actions should be undertaken if the
actual vibration value exceeds 1.6 and 2.5 times the reference value:
1.6 times the reference value corresponds roughly to the 75 percentile
2.5 correspond approximately to the 90 percentiles.
The suggested boundaries are here significantly lower than the current boundary zones
in the existing standard. A more in depth and adequate statistical analysis could be
made in order for establishing valid vibration reference values. However, the analysis
shows that the median value for the shown Burr-distributions is very close to the mean
values for the datasets. And as the median value method objectively excludes extreme
values this method is promoted for finding the adequate reference values. The method
is also verified by figure 34 and 35 where the median value for a dataset with only
measurements in best operating range is compared with an unfiltered dataset. The
resulting median value is nearly equal for the two compared datasets.
An analysis of the distribution of units marked as problem in the database revision F
was also conducted. It was shown that problem units were evenly distributed over the
complete range of measured values. Surprisingly, machines labeled as problem in the
database do not have exceptionally high vibration levels if compared to other measured
values which are considered to be normal. However, the proportion of machines
marked as problem increases with higher vibration values. At approximately 2.5
times the median value the proportion of problem units increases more radically. This
could prove that the proposed boundary levels 2.5 and 1.6 times the median value
makes sense.
During 2013-2014, analysis of database revision J was conducted. Since the number of
measurements was increased with this version the aim of the analysis was to find
reference values which could be used for producing action limits. These limits were
produced by using the boundary levels recommended from earlier analysis and the
median values calculated from the increased database revision J. The recommended
action limits from this analysis is generally at the same level as calculated from
previous revisions. Exceptions are shaft oscillations for Bulb units which have
increased values versus previous revisions. Although the database has been
substantially increased the number of measurements of bearing housings is the same as
for previous revision.
44
Future work
During analyze work the database has been revised a number of times.
Current database was released 2013-11-20 and contains 7355 measurements. If
the database is revised with even more measurements, the median method
could be applied on the latest revision for verification of action limits.
The data suppliers should refine their definition for machines marked as
problem in the database.
Identification of relevant bearing groups (turbine/generator,

suspended/umbrella, closed shell/tilting pad). It is possible from the database
to filter out these groups which then can be internally analysed for parameter
correlation.
Identification of the relationship between the vibrations of the generator guide

bearings and the turbine guide bearing. This should then render a physical
explanation of the relationship. The rotordynamic behaviour of a shaft system
could be taken into consideration during this analysis.
Explanation of measured shaft movements that is larger than the specified

available bearing clearance
An identification of the impact of the current requirements of IEC/ISO of 30

m
p-p for stationary parts. Current ISO 10816-5 action limits of 30 m p-p is more
severe for the majority of large-scale turbines than the future recommended
limits that will be expressed in mm/s.
More in depth analysis of the physical relationships identified in this report
45
References
[1] M. Nsselqvist, Simulation and characterization of rotordynamic properties
for vertical machines, Doctoral thesis, Lule University of Technology, ISSN:
1402-1544;2012:1402-1544; 2012
[2] IEEE Standard for Hydraulic Turbine and Generator Integrally Forged Shaft
Couplings and Shaft Runout Tolerances, IEEE Power Engineering Society,
Reaffirmed 14 May 2001
46
Appendix
7.1
APPENDIX 1
47
48
49
50
51
52
53
54
7.2
APPENDIX 2
55
56
57
58
Mechanical vibrations in hydraulic

machines
This report is an updated version of chapter 6 in the previously published
Elforsk report 12:70.
For the revision and integration of the current mechanical vibration
standards for hydraulic power generating and pumping plants ISO/IEC
7919-5 and 10816-5 IEC and ISO are supporting an international
workgroup (ISO/TC 108/SC 2 & IEC/TC 4 - JWG1 Vibration of Hydraulic
Machines). To support the working group, analysis of the IEC TK4
vibration database has been performed by a Swedish national workgroup.
The purpose of the analysis has been to form a statistical foundation for a
recommendation on vibration limits in a new integrated vibration
standard.
Reference values are suggested to be based on median values from the
database. Actions are suggested to be undertaken if the actual vibration
value exceeds 1.6 and 2.5 times the reference value.
Another step forward in Swedish energy research

Energiforsk Swedish Energy Research Centre is a research and knowledge based organization
that brings together large parts of Swedish research and development on energy. The goal is to
increase the efficiency and implementation of scientific results to meet future challenges in the
energy sector. We work in a number of research areas such as hydropower, energy gases and
liquid automotive fuels, fuel based combined heat and power generation, and energy
management in the forest industry. Our mission also includes the generation of knowledge about
resource-efficient sourcing of energy in an overall perspective, via its transformation and
transmission to its end-use. Read more: www.energiforsk.se

Mechanical Vibrations in Hydraulic Machines

Uploaded by

Copyright:

Available Formats

Mechanical Vibrations in Hydraulic Machines

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

Mechanical Vibrations in Hydraulic Machines

Uploaded by

Copyright:

Available Formats

What is the purpose of analyzing the IEC TK4 vibration database?

What is the purpose of analyzing the IEC TK4 vibration database?

What statistical methods were used to analyze the database?

What statistical methods were used to analyze the database?

MECHANICAL VIBRATIONS IN HYDRAULIC MACHINES