Sie sind auf Seite 1von 20

GE Measurement & Control

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting


used for Condition Monitoring of a Hydro Turbine
Roengchai Chumai | Technical Leader | Bently Nevada Machinery Diagnostics Services | GE Oil & Gas |
roengchai.chumai@ge.com

Abstract
The paper discusses a classic resonance problem of a vibration probe mounting installed on a hydro
turbine during retrofit project commissioning. Fluctuating vibration was detected within the online machine
condition monitoring system. Experimental Modal Analysis (EMA) was performed to confirm the presence of
a resonance issue at each probe bracket mounting at the Turbine Guide Bearing (TGB). Actual natural
frequency was identified with associated mode shape and damping. Finite Element Analysis (FEA) was then
used to perform modal/frequency analysis of the probe bracket with calibrating to match the actual modal
impact test result. Consequently, the model was used for structure modification and redesign of a new
probe bracket with higher natural frequency to prevent natural frequency excitation. Multiple disciplines
and tools are required to analyze and solve this problem.

Introduction
Modal analysis has been used in many industries and applications to find natural frequencies along with
associated mode shapes within relevant frequency ranges in both engineering design and field
troubleshooting work [1, 2]. The main objective is to avoid excessive vibration amplitude due to resonance
excitation which can shorten the operating life of the test object or machine part due to fatigue, or even
prevent the machine from operating reliably in some cases. Experimental Modal Analysis (EMA) results are
also used for correlation with Analytical Modal Analysis (AMA) models which are constructed using Finite
Element Analysis (FEA) based on precise dimensions, geometry, material properties, and boundary
condition (e.g. restrains/constrains, pre-stretch, etc.). Once the FEA model is calibrated to match actual
modal test data from EMA with acceptable tolerance, the model can be used to find the optimum design
for future fabricating or manufacturing without a trial and error approach (saving cost and time). The
machine and associated parts are free from natural frequency excitation throughout operating speed
range. Consequently, vibration amplitude is acceptable for long-term continuous operation.

Application Information
This project was performed to provide an online machine condition monitoring and diagnostics system for
a hydro power generation plant covering four Francis hydro turbine-generator sets (4 x 250 MW) and two
Pelton hydro turbine-generator sets (2 x 37.5 MW). Examples of hydro turbine types installed in this plant
are shown in Figure 1 below. All units have casing vibration transducers installed at all bearings, and a
Keyphasor transducer on each rotor. Vibration signals are then connected to the existing vibration
monitor rack, serving as a simple machinery protection system. However, the existing machinery protection
system is not sufficient for machinery diagnostics and long-term machine condition monitoring. Hence, the
new project work scope includes designing and installing a shaft vibration transducer at all bearings;
proper monitors; software for an online Condition Monitoring System (CMS); and system integration
between the machine control and online machine condition monitoring system. Selected process variables
are integrated into the CMS via OPC (OLE for Process Control) protocol. The Francis hydro turbine-generator
set has three radial journal bearings installed: Generator upper and lower guide bearings and a turbine
guide bearing. The Pelton hydro turbine-generator set has two bearings: Generator upper and lower guide
bearings. The Francis turbine discussed in this paper incorporates a 15 blade runner, 26 wicket gates, and
stay vanes. The tilting pad thrust bearing has six pads.

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Project delivery depended on machine overhaul schedule so it was started with unit #3. The first set of
vibration probe brackets was designed based on engineering judgment and experience from past projects
without calculations. As commonly known, a shaft vibration probe measures relative vibration between the
rotor and the bearing housing where the probe is mounted. The probe bracket and its mounting must be
rigid with natural frequencies well above machine operating speed and possible exciting frequency range
depending on machine configuration and operation such as blade passing frequency, fluid induced
vibration, etc.

Figure 1: Cutaway of Francis (left) and Pelton (right) hydro turbines

A machine train diagram after all transducers were installed is shown in Figure 2 below. There are two
casing vibration transducers installed and two X-Y shaft vibration probes installed at each bearing. A
Keyphasor transducer is also installed to observe a notch on the shaft surface so as to provide the
vibration phase angle and supplementary machine speed. The online machine condition monitoring and
diagnostic software platform retrieves all vibration data from the vibration monitor rack and processes the
variables data from the machine control system. The data is processed, stored in a database, and
presented as various kinds of plots Such as trend, time base, orbit, spectrum, bode, polar, average shaft
centerline plots, and others. Plant personnel and machinery specialists can then utilize those plots to view
vibration data in correlation with applicable process or environmental variables to monitor and diagnose
machine problems.

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Figure 2: Machine train and transducer layout of unit #1 to #4 of Francis hydro turbine-generator set

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Vibration Analysis after Commissioning


As part of a Site Acceptance Test (SAT), a system functions check was carried out including a vibration data
review at all measurement points. Vibration readings were found to be normal except at the turbine guide
bearing shown in Figure 3. Vibration amplitude at one probe appeared to be fluctuating over a period of
steady state operation. The trend plot in Figure 4 shows fluctuating overall vibration amplitude at the Y
probe while the X probe appeared to be stable at a normal level at low load operating condition. The
vibration waveform data presented in an orbit plot (Figure 5) shows high frequency vibration induced at the
Y probe but normal frequency at the X probe. XY probes are installed at the same bearing. Consequently,
measured vibration frequency is normally similar, but naturally with some difference in amplitude. The
observed vibration characteristic was concluded to have been most likely not caused by real mechanical
vibration produced from the machine.

X Probe

Y Probe

Figure 3: Shaft vibration measurement at turbine guide bearing with original probe bracket design

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Recognized best practices suggest normally having seismic transducers installed adjacent to rotor
vibration proximity transducers. However, this could not be done in this case due to limitations of probe
mounting and fixture The reason being that the seismic transducers were installed by the Original
Equipment Manufacturer (OEM). When XY measurements were added during the provided retrofit, the
existing bracket accommodation did not provide adequate angular shift between the X-probe and the
visible seismic transducer. The situation was the same between the Y-probe and the next visible seismic
transducer.

Less than 10% load condition

Y Probe

X Probe

Figure 4: Trend plot of overall shaft vibration amplitudes (X and Y probes) measured at turbine guide bearing after commissioning

Y Probe

X Probe

Figure 5: Orbit plot of shaft vibration measured at turbine guide bearing after commissioning

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Vibration frequency at the guide bearing measurement point was reviewed and found predominantly at 95
Hz at the Y probe (Figure 6). There was some low amplitude vibration presented at 120 Hz at the X probe
(Figure 7). There was no high frequency vibration excited at the generator upper and lower bearings (Figure
8).
The waterfall plots show clear evidence of excitation at the turbine guide bearing, only with higher
amplitude at the Y probe. Nevertheless, those high vibration frequency components are not coincident with
any vibration frequency generated from the machine and associated components. Consequently, it was
suspected to be caused by a resonance problem, which natural frequency of the probe bracket and
mounting that could be excited, resulting in fluctuating vibration amplitude predominantly at the Y probe.
The difference of vibration amplitude and frequency between the X and Y probes could be due to varying
stiffness of the probe mounting, as the physical installation of these two probes is different. We are most
likely seeing guide bearing angular stiffness anisotropy (at the bearing shown in Figure 3). For example, the
X probe mounting may be influenced by the adjacent oil cooler structure, which can increase stiffness.
Somehow, higher frequency excitation in this case has lower energy as lower excited vibration amplitude.
The source of excitation was unknown at this stage. A modal impact test was required to identify the
natural frequency of the fixtures mechanical structure and associated mode shapes.

1X vibration component and its harmonics

Vibration component at 95 Hz

Figure 6: Waterfall plots of spectra versus time of the Y probe measured at the turbine guide bearing after commissioning

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

1X vibration component and its harmonics

Vibration component at 120 Hz

Figure 7: Waterfall plots of spectra versus time of X probe measured at turbine guide bearing after commissioning

1X vibration

Figure 8: Waterfall plots of spectra versus time of X probe measured at generator lower guide bearing after commissioning

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Experimental Modal Analysis


In order to confirm the suspected resonance problem, a modal impact test was carried out on each probe
bracket installed at the turbine guide bearing. A Bently Nevada SCOUT140-Ex portable data collector
(equivalent to Commtest vb8) was used together with an accelerometer and impact force hammer. The test
setup is shown in Figure 9. To get the best response the accelerometer was installed close to the end of the
probe bracket where the shaft vibration probe is installed.
The impact force hammer was then used to hit the bracket in the vertical direction, which is the selected
measurement plane in this case. Based on bracket confirmation and geometry, it is likely to deflect and
vibrate in the vertical direction. Impulse force input from the impact force hammer generates broadband
exciting frequencies. Frequencies that coincide with the probe bracket natural frequency show up as peak
vibration amplitude with a 90 degree phase shifted which can be picked up by the installed accelerometer.
There are generally an infinite number of natural frequencies of the test object but the frequency range of
interest that was selected for this particular test was 0 to 1000 Hz. Coherence values are used to check and
confirm data quality based on the relationship of input force and output vibration. It is expected to be more
than 90% for good and acceptable results.
It was identified that the first natural frequency of the Y probe mounting structure was 91.25 Hz with
99.83% coherence (Figure 10). For the X probe bracket it was 116.30 Hz with 100% coherence (Figure 11).
This confirms the high vibration frequency presented in vibration readings of the X and Y probes was due to
probe mounting structure resonance excitation. However, there is some difference in the observed
vibration frequency during turbine operation and during the test. This could be due to two different
reasons: First, a mass loading effect, since a miniature accelerometer was not available at the site. Hence,
we used a normal, large one with a heavy weight for the test instead. Accelerometer mass with a magnetic
base can alter the test object (probe bracket) mass, as happened in this case. This can affect the test result
accuracy, as the identified natural frequency might be lower than the actual value. However, it is
acceptable for a troubleshooting job to produce an approximate result. Second, there is different behavior
of the TGB shell (different modal mass, different modal stiffness) during turbine operation and during the
test. According to ISO 7919-5 standard[3], the lowest natural frequency of the transducer support structure
should be more than seven times that of the machine operating speed (333 rpm or 5.55 Hz) which is about
38.85 Hz in this case. The observed resonance frequency is much higher than this guideline. However, field
test data showed that it is not safe from excitation. Consequently, probe bracket and mounting
modification is required.

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

Heavy accelerometer
with magnetic base

Seismic vibration
transducer
Figure 9: Photo of modal impact test performed at site with heavy accelerometer
NT2_Francis_U3 - 2 - +Y - Modal Accelerance 1000 Hz - Coherence
Cursor A:

100

91.25 Hz

99.83 %

80
%

60
40
20
0
0

100

19/2/2013 11:01:29 AM

200

300

400

500
Hz

600

700

800

900

1,000

<add note>

g/N

NT2_Francis_U3 - 2 - +Y - Modal Accelerance 1000 Hz - Magnitude


8
7
6
5
4
3
2
1
0

Cursor A:

100

19/2/2013 11:01:29 AM

200

300

400

500
Hz

600

91.25 Hz

700

8.018 g/N

800

183.9 deg

900

1,000

<add note>

NT2_Francis_U3 - 2 - +Y - Modal Accelerance 1000 Hz - Phase


Cursor A:

91.25 Hz

183.9 ?

183.9 deg

300

200

100
0
0

100

19/2/2013 11:01:29 AM

200

300

400

500
Hz

600

700

800

900

1,000

<add note>

Figure 10: Modal impact test result of Y probe with original fixing

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

GE Measurement & Control

NT2_Francis_U3 - 1 - +X - Modal Accelerance 1000 Hz - Coherence


Cursor A:

100

116.3 Hz

100 %

80
%

60
40
20
0
0

100

19/2/2013 11:07:29 AM

200

300

400

500
Hz

600

700

800

900

1,000

<add note>

NT2_Francis_U3 - 1 - +X - Modal Accelerance 1000 Hz - Magnitude

Cursor A:

2.5

116.3 Hz

2.826 g/N

256 deg

g/N

2
1.5
1
0.5
0
0

100

19/2/2013 11:07:29 AM

200

300

400

500
Hz

600

700

800

900

1,000

<add note>

NT2_Francis_U3 - 1 - +X - Modal Accelerance 1000 Hz - Phase


Cursor A:

116.3 Hz

256 ?

256 deg

300
200
100
0
0

100

19/2/2013 11:07:29 AM

200

300

400

500
Hz

600

700

800

900

1,000

<add note>

Figure 11: Modal impact test result of X probe with original fixing

Mode#3
Mode#1
Mode#2

Figure 12: Frequency Response Function (FRF) of original fixing

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

10

GE Measurement & Control

After natural frequency was obtained as a preliminary check, it was confirmed to be a resonance problem
of the vibration probe installation. The test was continued to identify mode shape at each particular
frequency. An impact force hammer was used as a roving point and the accelerometer was fixed where it
was expected to sense high vibration amplitude near the end of the bracket. A grid of the measurement
points was drawn on the bracket, so that data collection could be carried out accordingly. Frequency
Response Function (FRF) was then calculated for each point as a ratio of acceleration/force. Curve fitting
might be implemented to find the damping value at each mode [4] and select the number of interest
modes E.g. the first three modes are identified in Figure 12. The first mode shape is shown in Figure 13
with maximum deflection at the bracket end where the vibration probe is installed. This can partially
explain how a resonance problem influenced the excessive vibration amplitude observed at the Y probe.
Similarly, the second mode shape is identified in Figure 14 with only minimal deflection. Therefore, we did
not consider it at later stages of analysis. The first mode is the main contributor to the observed problem.

Figure 13: Experimental modal analysis result of first mode shape with original bracket design

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

11

GE Measurement & Control

Figure 14: Experimental modal analysis result of second mode shape with original bracket design

The example in Figure 15 shows proper instrument setup to avoid a mass loading problem which can
affect the accuracy of the test result. It also presents an alternative transducer attachment for a more
complete modal test The transducer axis has the same orientation as the axis of the non-contacting
probe. However, we are still not able to improve our test due to change of modal characteristics (modal
mass and stiffness) for running versus not running unit conditions of the TGB shell that supports the
bracket, creating the mechanical system that resonates.
A two channel portable data collector with a modal test feature (such as SCOUT140-Ex or vb8) is required. A
miniature accelerometer is used together with a small magnetic base or wax for mounting to suit the
frequency range of interest. An appropriate impact force hammer should be used with consideration of tip
material and size to ensure sufficient frequency response range as well as enough force to strike the test
object in order to get a good response. The harder the tip is, the higher the frequency response range will
be. The product datasheet should be referred to for more detail and precise data.

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

12

GE Measurement & Control

Miniature accelerometer

Impact Force Hammer

Minimum 2 channels analyzer

Figure 15: Example of typical setup for modal impact test with SCOUT140-Ex and miniature accelerometer

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

13

GE Measurement & Control

Analytical Modal Analysis


Because we dont have the option of changing the modal characteristics of the TGB shell, we can only try to
improve the real situation by changing the construction of the bracket to improve the resonance
characteristics of the mechanical system formed by the TGB shell and transducer bracket. Before bracket
modification/redesign we used Finite Element Analysis (FEA) to model the original bracket and then
performed modal/frequency analysis to find its natural frequency and associate mode shape. The result
was calibrated with EMA result which was considered actual data. This approach can eliminate trial and
error, saving cost and fabrication/manufacturing time. The model was built with inputs of precise
dimensions and geometry, material properties, and boundary condition[5]. Bracket material was carbon
steel code JIS G3101 SS400 with elasticity 210000 N/mm2, Poissons ratio 0.26, and mass density 7860
kg/m3. The bracket was fixed at two spots where it was bolted to the bearing cover. This prevented the
bracket from moving in translation and rotationally. The contact surface between the bracket and bearing
housing was modeled as an elastic support which can control bracket movement in the vertical direction.
Support stiffness was interpolated to match modal analysis results with actual data from the EMA test
Hence the model was calibrated. Model assumption includes homogenous material properties, uniform
support stiffness throughout the contact face, properly tied mounting bolts, and the actual material used
being the same as the specification. The first mode of the original bracket was 92.7 Hz (Figure 13) which
closely matches the actual data in both frequency and mode shape.

Figure 16: First mode shape of original bracket design at 92.65 Hz

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

14

GE Measurement & Control

A new bracket was redesigned by changing the dimensions and geometry to increase thickness, add more
ribs, and then re-run modal analysis. Such bracket construction will result (through the bracket fixing) in
increasing of TGB shell stiffness in the horizontal direction as well. First mode of the new bracket was
increased to 179 Hz (Figure 17) which is about 32 times the machine running speed at 333.33 RPM (5.56 Hz).
This is considered safe from excitation. The new bracket was then fabricated and installed in the field.

Figure 17: First mode shape of new bracket design at 179.05 Hz

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

15

GE Measurement & Control

Vibration Results of New Probe Bracket


New probe brackets were redesigned and installed onsite as shown in Figure 18. FEA simulation showed
that the first mode was increased from 92 to 179 Hz which is approximately 32 times the machine running
speed. Vibration data appeared to show no sign of a probe bracket resonance problem as shown in orbit
plot (Figure 19) and spectrum plot (Figure 20). However, this data was collected at a machine operating
condition of 235 MW with 75% inlet guide vane (IGV) opened. The waterfall plot in Figure 21 was then used
to view the vibration spectrum of the Y probe at various machine operating conditions. It was found that
low amplitudes of resonance vibration at around 183 Hz were revealed at some operating conditions. This
implies that the resonance vibration problem relates to the machine operating condition such as inlet flow,
IGV position, generator load, etc.

X Probe

Y Probe

Figure 18: New bracket design installed at turbine guide bearing

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

16

GE Measurement & Control

Figure 19: Orbit shape of vibration data measured at turbine guide bearing with new probe bracket design

1X vibration component and its harmonics

Figure 20: Vibration spectrum of Y probe installed at turbine guide bearing with new bracket design when the unit was operated at
235 MW with 75% IGV opened

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

17

GE Measurement & Control

Resonance frequency at 183 Hz

Broad frequency seismic

No broad frequency vibration

Broad frequency seismic vibration

Figure 21: Waterfall plot of vibration spectra versus time measured by Y probe at turbine guide bearing with new bracket design

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

18

GE Measurement & Control

Possible Sources of Excitation


Operating process variables were reviewed as shown in Figure 22. We noticed a significant correlation
between the IGV position and the mechanical vibration of the turbine structure. It is typical of Francis
turbines, that below a percentage of nominal power, high fluid flow instabilities may be generated. This can
result in structural resonances of various stationary components of the hydro turbine Generator (HTG)
structure. In this unit, we discovered that the excitations resulting in resonance vibrations of turbine
structure are present when IGV position is less than 10% when you increase the inlet flow. This means
every time the unit is experiencing loading or unloading we can expect transient operation that generates
higher vibration levels. However, because the unit is generally operated at base load condition, the
structural vibrations do not significantly influence its aging Due to fatigue.
Flow-induced vibration was revealed by casing vibration data measured at the bearing housing You can
recognize two seismic transducers connected to the TGB shell (Figure 18) as broadband frequency
components (Figure 23). The shaft vibration probe bracket is mounted on the bearing housing. Some
vibration component frequencies could be coincident with the brackets own natural frequency. There was
no mechanical vibration resonance excitation revealed in shaft vibration data when no broadband
vibration frequency components were present in the casing vibration data. However, it appeared to be low
severity mechanical excitation at higher vibration frequency components as lower vibration levels. The new
probe bracket design is considered acceptable since the unit is not generally operated at low load
condition below 10% flow where there are still some low amplitudes of resonance vibrations. However,
now the noise-to-signal ratio is very small, and will not destroy diagnostic analysis of the physical condition
of the HTG. Although unnecessary, if required, the mechanical structure of probe fixing can be modified
further and better simulated using FEA for the TGB shell to get even higher first mode natural frequency.
This can minimize structural mechanical vibration excitations at low inlet flow conditions as well as reduce
vibration levels if the mechanical vibrations are excited at a higher frequency.

Inlet Flow, ml/s

Gen Load, MW

IGV Position, %

Casing vibration, YV
Casing vibration, XV
Turbine speed

Figure 22: Trend plots of seismic vibrations measured at TGB (top) correlating with process variables (bottom)

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

19

GE Measurement & Control

Broad frequency seismic vibration

No broad frequency vibration

Broad frequency seismic vibration

Figure 23: Waterfall spectrum of seismic vibrations measured at shell of TGB

Conclusions
Modal analysis is an effective tool for solving vibration problems as it can be used in the field for
troubleshooting, design phase, simulation study, as well as structure modification and improvement. This
paper discussed the approach of applying modal analysis to successfully solve vibration problems
observed from an online machine condition monitoring system installed on a hydro turbine-generator set
during project commissioning. EMA was used to identify a resonance problem of vibration probe mounting
at TGB, allowing the natural frequency and associated mode shape to be identified. FEA was then used to
model probe bracket structure based on precise dimensions and material properties before
modal/frequency analysis was carried out to find the calculated natural frequency and associated mode
shape. The result was then calibrated with EMA by adjusting the boundary condition before the structure
modification was studied and simulated for probe mounting improvement to increase the natural
frequency. FEA can save manufacturing costs and time without doing trial and error for new product
design. The new bracket design resulted in stiffness increase of the probe mounting, which was acceptable
and resulted in improving the noise-to-signal ratio.

References
[1] Avitabile, P., 2001, "Experimental Modal Analysis: A Simple Non-Mathematical Presentation," Sound and
Vibration.
[2] Vzquez, J. A., Cloud, C. H., and Eizember, R. J., 2012, "Simplified Modal Analysis for the Plant Machinery
Engineer," 41st Turbomachinery Symposium, Texas A&M University, Houston, Texas.
[3] International Standard Organization, 2005, "ISO 7919-5 Mechanical Vibration - Evaluation of Machine
Vibration by Measurements on Roating Shafts - Part 5 Machine Sets in Hydraulic Power Genrating and
Pumping Plants," Internation Standard Organization.
[4] Lee, M., and Richardson, M., 1992, "Determining the Accuracy of Modal Parameter Estimation Methods,"
IMAC X.
[5] Richardson, M. H., 1978, "Measurement and Analysis of the Dynamics of Mechanical Structures,"
Hewlett-Packard Conference for Automotive and Related Industries, Hewlett-Packard, Detroit.

Modal Analysis with SCOUT/vb8 Vibration Probe Mounting used for Condition Monitoring of a Hydro Turbine

20