SHOP ROTORDYNAMIC TESTING – OPTIONS, OBJECTIVES, BENEFITS AND PRACTICES

Dr. John A. Kocur, Jr. Dr. C. Hunter Cloud

ExxonMobil Research & Engineering BRG Machinery Consulting LLC
3225 Gallows Road, Rm 2A-0601 3241 Plank Road
Fairfax, VA 22037-0001 North Garden, VA 22959
(703) 846-2275 (434) 293-9926
[email protected] [email protected]

John A. Kocur, Jr. is a Machinery ASME, the Vibration Institute, the API 684 rotordynamics task
Engineer in the Equipment Engineering force, and the advisory committee for the Texas A&M
Division at ExxonMobil Research & Turbomachinery Symposium.
Engineering in Fairfax, Virginia. He
has worked in the turbomachinery field ABSTRACT
for 30 years. In his current capacity,
he provides support to the downstream, Understanding the lateral rotordynamic behavior is critical
upstream and chemical business lines in determining the reliability/operability of rotating equipment.
within ExxonMobil with expertise on Whether examining a centrifugal pump or compressor, steam or
vibrations, rotor/thermo dynamics, failure analysis and health gas turbine, motor or generator, rotating machinery share the
monitoring of rotating equipment. Prior to joining EMRE, he same need to accurately predict and measure dynamic behavior.
held the position of Manager of Product Engineering and Industrial specifications determining fit for purpose rely on the
Testing at Siemens Demag Delaval Turbomachinery. There accuracy of rotordynamic predictions where direct
Dr. Kocur directed the development, research, design, measurement is impractical or otherwise impossible in an
engineering and testing of the compressor and steam turbine industrial setting. Testing to confirm rotordynamic prediction
product lines. He has also held positions with Pratt & Whitney and behavior provides both the purchaser and vendor the
and Amoco Corporation. confidence that the design will meet project expectations.
Dr. Kocur received his BSME (1978), MSME (1982) and Rotordynamic shop testing has several options available to
Ph.D. (1991) from the University of Virginia and an MBA the project during acceptance tests at the vendor’s shop. These
(1981) from Tulane University. He has authored papers on options include mechanical run, string and full load/Type 1
rotor instability and bearing dynamics, lectured on hydrostatic testing as well as verification testing to validate unbalance
bearings, has sat as a committee chairman for NASA Lewis and response and stability predictions. Such testing has numerous
is a member of ASME. Currently, he holds positions within API advantages; the most important being the avoidance of
as 617 chair, 684 chair and member of Subcommittee on production disruptions involved with testing at the job site.
Mechanical Equipment (SOME). Each test option has associated costs as well as limitations as to
what lateral vibration characteristics are revealed.
C. Hunter Cloud is President of BRG Understanding these factors is vital to efficiently mitigate the
Machinery Consulting, LLC, in risks associated with the purchased equipment.
Charlottesville, Virginia, a company Applying best practices and an understanding of the
providing a diverse range of rotating industrial (API) test requirements are needed to derive the
machinery technical services. He began maximum benefit of each test option. The best practices not
his career with Mobil Research and only involve the test procedure but also the associated
Development Corporation in Princeton, analytical methods used to post process the measurement
NJ, as a turbomachinery specialist information. Whether performing a simple mechanical run test
responsible for application engineering, or more complex stability verification during ASME Type I
commissioning, and troubleshooting for testing, ensuring that a logical, repeatable and proven
production, refining and chemical facilities. During his 11 methodology is followed produces reliable evidence to confirm
years at Mobil, he worked on numerous projects, including the rotordynamic model and lateral vibration performance. The
several offshore gas injection platforms in Nigeria as well as rationale behind the API test requirements provides an
serving as reliability manager at a large US refinery. understanding of why that test is being performed and its
Dr. Cloud received his B.S. (Mechanical Engineering, correct application to the dynamic behavior.
1991) and Ph.D. (Mechanical and Aerospace Engineering,
2007) from the University of Virginia. He is a member of
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 Test options can be separated into two categories; tests that Rotordynamic behavior testing involves both direct
reveal portions of the dynamic behavior of the equipment to measurement and inference. Direct measurement of the
specific operating conditions and those used to verify the vibration is typically made at the journal locations. Internal
analytical predictions of that behavior. API mechanical, string vibration levels at other critical locations can only be inferred
and Type I (or full load) tests reveal the rotordynamic behavior from these measurements using the rotordynamic predictions.
of the equipment to a given set of conditions. These are used Thus, industrial specifications determining fit for purpose rely
specifically to determine acceptability of the design. on the accuracy of rotordynamic predictions where direct
Unbalance and stability verification testing is used to confirm measurement is impractical or otherwise impossible in an
(or provide confidence in) the rotordynamic model. industrial setting. Testing to confirm rotordynamic prediction
Confidence in the model permits extrapolation of the design and behavior provides both the purchaser and vendor the
(vendor) and operation (purchaser) beyond the machine’s as- confidence that the design will meet project expectations.
built and specific shop test conditions.
Rotordynamic shop testing has several options available to
the project during acceptance tests at the vendor’s shop. These
INTRODUCTION
options include tests to demonstrate operating behavior at a
Demands on turbomachinery continue to push designs specific condition (mechanical run, string and full load/Type 1
beyond experience limits in terms of speed, power, size, testing) and verification testing to validate unbalance response
pressure development and flow rate all the while demanding and stability predictions. Such testing has numerous
higher reliability and operability. To meet the stringent advantages; the most important being the avoidance of
objectives of the application, almost absolute knowledge of the production disruptions at the job site. Each test option has
behavior of the machine is necessary. To aide in this associated costs as well as limitations as to what lateral
understanding, advanced analytical methods have been vibration characteristics are revealed. Understanding these
developed in parallel with rigorous testing practices. factors is vital to efficiently mitigate the risks associated with
the purchased equipment.
Shop acceptance testing has long been used as the “final”
check of the turbomachinery design and is required for all Application of best practices and an understanding of the
special purpose equipment. With the advancement of industrial (API) test requirements are needed to derive the
computational methods, failure rates during testing have been maximum benefit of each test option. The best practice not
greatly reduced. However, experience has taught us analytical only involves the test procedure but also the associated
methods alone are insufficient to guarantee the “right the first analytical methods used to post process the measurement
time” philosophy that many reliability systems employ. information. Whether performing a simple mechanical run test
Whether used to provide data to baseline prototype equipment, or more complex stability verification during ASME Type I
to benchmark extensions of current experience limits, or for testing, ensuring that a logical, repeatable and proven
verification of proven practices, testing remains an integral part methodology is followed produces reliable evidence to confirm
of all reliability systems. the rotordynamic model and lateral vibration performance.
Understanding the dynamic behavior is critical in Test options can be separated into two categories; tests that
determining the reliability/operability of rotating equipment. reveal portions of the dynamic behavior of the equipment to
Whether designing a pump, compressor, turbine, motor or specific operating conditions (Vibration Demonstration Tests)
generator, all rotating machinery share the same need to and those used to verify the analytical predictions of that
accurately predict and measure dynamic behavior. Literature is behavior (Design Verification Tests). API mechanical run,
swamped with failures that resulted from both inadequate string and Type I (or full load/full pressure) tests reveal the
design and testing methodologies. While the potential for rotordynamic behavior of the equipment for a given set of
failure originates at the design stage, testing represents the final conditions. These are used specifically to determine
step to identify that potential. The importance of efficiently acceptability of the design in a pass/fail mode. Unbalance
employing both cannot be understated [1]. response and stability verification testing is used to confirm (or
provide confidence in) the rotordynamic model and analysis.
Turbomachinery is dominated by two classes of dynamic
Confidence in the model permits extrapolation of the design
behavior; rotordynamics and fluid dynamics. As their names
(vendor) and operation (purchaser) beyond the machine’s as-
imply, each focuses on a specific dynamic behavior;
built and specific shop test conditions.
rotordynamics on the rotating system’s vibration and fluid
dynamics on the mechanical interaction with the working fluid. The tutorial will cover the following aspects of
As our depth of understanding increases, the more interrelated rotordynamic testing:
these behaviors become. Decisions made in the design of one
• Decision to test: Why and on What Basis?
can impact the other with sometimes disastrous effects.
Nowhere is this more evident than centrifugal equipment, • Rotordynamic Testing
especially compression. For this reason, this tutorial will focus
on testing the rotordynamic behavior of centrifugal o Options
compressors. However, most of the principles and practices are o Objectives
applicable across all types of turbomachinery.
o Preparation
o Information / Knowledge Gain
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 o Benefits • Instability (high subsynchronous vibration)
• Recommended Practices o Risk: Instability drivers (gas density, speed, ∆P,
critical speed location), operating characteristics,
o Vibration Demonstration Tests
deposition plugging seal cavities 1
o Design Verification Tests
o Scenario: Internal rubbing, bearing damage,
component failure
DECISION TO TEST: WHY AND ON WHAT BASIS
o Consequence: Downtime impact for repair or
The need for rotordynamic testing stems from several
reconfiguration (extensive), impact of operation
objectives; to prove the behavior of the machine, to test the
restrictions
accuracy of the vendor’s predictions and to identify problems
before the machine is put into operation. The decision to test • High synchronous vibration levels
and which objectives to pursue result from several factors
related to the application. These factors are: o Risk: Unbalance sensitivity, high amplification
factors, critical speed encroachment, mode
• Risk-consequence analysis considering the following: involved, thermal instability
o Impact on operations given failure or performance o Scenario: Failure of casing attachments, bearing
deficiencies failure, internal rubbing
o Outage length due to location o Consequence: Downtime for repairs,
reconfiguration costs
o Experience with similar services
• Internal rubbing
• Safety, Health and Environment (SHE) impact of the
project o Risk: Sensitivity of rotor to unbalance, high
amplification factor, component involved (blade,
• Technology application experience (vendor or user)
laby seal, etc.), unbalance due to
o Prototype equipment erosion/corrosion aspects of working fluid
o New application of the technology o Scenario: Efficiency loss due to expanded seal/tip
clearances, blade failure
o New arrangement
o Consequence: Impact of performance loss on
• Variability of process conditions operation, downtime impact for repair
• Rotordynamic analysis results
These factors involve the project risks and consequences
and should be known at the beginning of the project when the
testing decisions are made. Detailed machine design and
analysis (which may act to mitigate some of these risks) are not
performed until latter stages of the project. Since one goal of
the testing may involve verification of the analysis, basing the
test decision on the project’s risks and consequences makes
sense.
Approaching the project from a risk-consequence analysis
establishes a logical framework upon which to make decisions
[2]. The framework provides a basis for efficiency and enables
the correct mitigation activities to be performed. A typical risk
matrix is shown in Figure 1.
The risk matrix should be used to address specific concerns
of the machinery. For rotordynamics, a possible outcome
should be weighed with the factors driving risk and those
driving the consequence. (Consequence here is defined as
being entirely economic.) Examples of problematic outcomes
to consider are:

Figure 1) Typical Risk Matrix

1
K.E. Atkins and R. X. Perez [4] discuss quantification of instability risk using
failure rate data.
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 The user must also consider the safety, health and every possible operating condition or combination of assembly
environmental impact of possible failures that can be attributed /machining tolerances be tested. To fully understand the
to the machinery in question. The impacts can result in the acceptability of the design and the rotordynamic behavior, we
release of gas from component failure due to high vibrations, must rely on the accuracy of the rotordynamic analysis.
failure to meet regulatory requirements due to unplanned Consequently, tests to determine the accuracy of those
outage or injury due to parts release during failure. The risks of predictions were developed. These additional tests were
each can be determined through a failure mode analysis [3] that developed to provide more than a pass/fail test, which is
incorporates the rotordynamic contributions to the identified essentially what the vibration demonstration tests are. They
failure modes. were implemented to determine the robustness of the design.
This is particularly useful when operating conditions are widely
Experience plays an important role in determining the
variable or when design experience limits are exceeded. The
extent of testing to perform. Experience in this case relates to
second type of testing option is referred to as design
both the vendor and user. Obviously, prototype machinery or
verification tests with the two most prevalent being:
equipment that extends the experience limits of the vendor
should be tested thoroughly. What may be overlooked is the • Unbalance response – This test would include the
experience of the user with that equipment. First application of more invasive testing required by API [8]
technology within a user company may benefit from additional
testing at the vendor’s shop. The testing can be used to better • Stability – Not currently specified by API standards.
understand the dynamics and what conditions or operating Verification testing also has the options of where these tests
nuances may affect the rotordynamic behavior. Finally, can be performed. In either case, the tests may be performed in
prototype components within the machinery may require a high speed balance bunker, during the mechanical test or
additional component testing to determine their impact on the during the full load testing. Each option adds additional
rotordynamic behavior of the machinery. considerations in what can be measured, dynamic effects
Finally, the results of the rotordynamics analysis should be included in the test and what portion of the dynamic behavior is
used to weigh the decision to test and which test to select. analyzed.
Machines shown sensitive to destabilizing forces or those Test Preparation
predicted to have low stability levels may benefit from stability
verification tests. Similarly, a rotor with high amplification Following the decision to perform a rotordynamic test, the
factors may influence the decision to perform unbalance user should decide which specifications to apply to the test.
response verification testing. Verification testing, as noted The specification should describe the objectives and
earlier, is intended to prove the accuracy of the predictive tools requirements of the test. For several of the test options
used to model the rotor behavior and thus assess the described above, API standards have described a specific test
acceptability of the design. The use of a proven rotordynamic procedure to be followed. The API standard details the
analysis is an effective and efficient mitigation strategy towards procedure, objective and requirements for the test. For the
reducing risk. more specialized testing, the user will need to develop their
own test specification. This can be done with the assistance of
ROTORDYNAMIC TESTING the vendor or by consulting industry specialists.

Options Test Objectives

Fundamentally, there are two types of testing options The objectives of the tests performed should be discussed at
available for lateral rotordynamic assessment. The first type, the initial stages of the project. Agreement on the test
referred to as vibration demonstration tests, demonstrates the objectives will assist in the determination of what equipment is
behavior of the as-built machine and/or train for a specific set needed, measurements to be taken, and conditions to be run.
of operating conditions. Generally, only the site specific Generally speaking, the objectives of each test are listed below.
instrumentation is used to measure lateral vibration behavior. Other specialty objectives can be added, but the ones listed
The tests do not address the accuracy of the model or tools used below would form the basis for each test.
to predict the rotordynamic behavior, nor do they attempt to Vibration Demonstration Tests
estimate or determine the robustness of the design. Those that
fall into this category are: Mechanical Run Test

• Mechanical Run – an example is the API required The mechanical run test as required by API is primarily a
mechanical test [5] vibration level check. Measured at the probes located at the
journal location, vibration levels are checked against the
• String – API 617 refers to this as a complete unit test specified limit for both overall and non-synchronous
[6] components. General mechanical performance is also
examined including bearing temperatures, close clearance
• Full load/Full pressure – Referred to as a Type I test
rubbing and seal performance typically up to maximum
by ASME [7]
continuous speed (MCS). Supercritical behavior is examined
Many important aspects of the rotordynamic behavior by determination of the amplification factor and separation
cannot be practically measured (i.e. internal vibration levels, margin of typically the 1st critical speed. (Obviously, the
separation margin to modes above operating speed) nor can behavior of modes above MCS remains undetermined.)
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
String Test “is the compressor stable” but “how stable is it.” The
measurements are then compared against the analytical
The string test is not much different than the mechanical
predictions to determine accuracy. Pettinato et al. [9] presented
test mentioned previously. As the name implies, the string test
a methodology employing this test during mechanical and
is performed with all or a major portion of the machinery train
performance testing (as required by API 617).
connected (typically everything but the driver.) The objectives
of this test are also similar to the mechanical test of a single Test Information / Knowledge Gain
body in that vibration levels are checked against limits,
To determine the extent of testing to perform, one needs to
mechanical performance is examined, and supercritical
understand the information or knowledge gained of the
behavior is analyzed. However, in this case, these are
rotordynamic behavior of the body or train. While similar
determined for the coupled train configuration. The string test
information can be obtained from several of the tests, the costs
is run to measure the coupled body dynamic behavior (when
associated with each determine the overall efficiency of
rigid couplings are used in the train) or the coupling spacer
obtaining the necessary information to mitigate risks identified
dynamics (for couplings with unusually long or heavy spacer
in the risk matrix. As before, the tests are separated into
tubes.)
vibration demonstration and verification testing. Vibration
Full Load / Full Pressure Test demonstration testing confirms the machinery can meet project
specifications for a given operating and as-built condition. No
Full load / full pressure tests are rarely performed based on
attempt is made to confirm the accuracy of the analytical
rotordynamic justifications only, mainly due to the costs
prediction beyond confirmation that the specification has been
involved. However, the Type I test does permit vibration level
met. Since the accuracy of the analytical prediction remains
checks at operating conditions, stall determination, impact of
largely in question, inferred information from the analytical
internal loading and deflections on dynamic behavior of
method should also be questioned.
individual components (seals and bearings) and a binary check
for stability (yes or no). Range testing is rarely done during the Vibration Demonstration tests
Type I test. Typically gas properties are held constant,
Mechanical Run
clearances are left at as-built values and alternate
configurations are not considered. Thus, while some aspects of Mechanical testing provides information related to the
rotordynamic behavior are tested, margins and robustness are critical speed location and some indication of that modes
left unchecked. The machines undergoing these tests leave behavior. The modal information is limited to only those
those factors to analytical studies whose accuracy may remain modes located below the maximum test speed achieved (trip
unchecked. Keep in mind, prediction of the stability condition speed in most cases.) Typically, this is only the 1st critical
(stable vs. unstable) of a particular machine is the first step in speed. The modal behavior information is restricted to the
developing a good analytical tool. However, it is not the only amplification factor which is sketchy at best. The amplification
step. As designs extend the operating or design experience, it factor can be highly sensitive to the acceleration/deceleration
becomes necessary to predict the stability threshold, separation rate and whose magnitude is not restricted by API. The test
margins and overall optimization of the design correctly, thus also validates the balance procedure’s effectiveness in meeting
the need for verification testing. the project vibration limit specification.
Design Verification Tests For certain low risk applications, this amount of information
is sufficient. What isn’t tested however can be significant. For
Unbalance Response
example, subcritical motor applications (incorrectly termed stiff
The likely first attempt to publish a verification test in an shaft) operate below the 1st critical. This critical speed can
industry standard was the unbalance response verification test have high amplification factors and can be damaging if the
published by API 617. The objective was simply to verify the separation margin is lost. Performing only a mechanical test
unbalance response prediction accuracy of the vendor’s will tell the user whether the mode is on or below the operating
rotordynamic analysis with regards to the machine’s unbalance speed. The amount of separation remains untested and can only
sensitivity within the operating speed range and the location of be inferred from the unverified analytical predictions. Thus,
the critical speeds (below trip speed.) The verification test the impact of support stiffness loss on the location of the mode
analyzes both the predictor tool and model employed. A in the field (due to structural or bearing clearance changes) can
methodology was refined over several editions of API 617 have significant risk associated with it.
within the limitations of the mechanical shop test.
String Test
Alternatively, performing this test in the balance bunker has
gained acceptance with the increase in at-speed balancing of String testing will provide the same type of rotordynamic
rotors and the freedom it permits in terms of weight placement information as obtained with the mechanical testing but at the
and additional measurement points. higher costs of assembling the entire (or high speed portion) of
the train. Information again limited to the critical speed
Stability
location and some indication of the modal behavior. Ignoring
Several methods have been developed analogous to the API other reasons to perform string testing (e.g. fit checks for trains
unbalance response verification test with the objective to verify being sent to remote portions of the globe), rotordynamic
the stability predictions of centrifugal compressors. As with justification for the test should be limited to the information
the unbalance response, the intent is to measure more than just supplied by the test, namely, dynamic behavior of the coupling
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
spacer(s) and rigid coupling effects on the dynamic behavior of • Balance bunker – Rotor, bearings, and pedestal
individual bodies. As noted in API 684 [10], a train lateral support impact on stability (bunker bearings if used) at
analysis should be requested for unusually long or heavy various speeds.
coupling spacers or when rigid couplings are used. In this case,
the correct boundary conditions at the shaft ends are obtained • Mechanical test (Vacuum) – Rotor, job bearings and
when the train is modeled (train in this case refers to the bodies casing support impact on “baseline” stability (or basic
on either side of the coupling(s) in question.) log decrement reflecting the bearings and rotor only)
at various speeds.
Testing of this train configuration should mimic the
analytical model to verify the behavior in question. With a • Performance test – Rotor, job bearings, casing support
rigid coupling, the rotordynamic behavior of the bodies rigidly and reduced aerodynamic and seal behavior impact on
coupled will be affected. For unusually long or heavy coupling stability for a limited speed range.
spacers, the dynamics of the spacer can only be accurately • Type I test – Stability level and margin at nearly the
modeled/tested with the hubs attached to the shafts. same operating conditions as expected in the field.
Full Load / Full Pressure (FLFP) Test The range of gas conditions, inlet and discharge
pressures and flow rates may be limited as a result of
Type I tests provide a stable vs. unstable behavior the test setup.
indication of the rotor to specific test conditions. The test also
reveals the change in lateral behavior of the measured modes Test Benefits
(typically the 1st critical speed as with the mechanical test) due The benefits derived from the rotordynamic testing can be
to internal loading of the compressor and application of gas identified for the two groups of testing. Vibration
pressures and densities close to the design values at the seals. demonstration testing provides the purchaser the following:
The latter introduces seal dynamics into the testing that is only
achieved during the FLFP test. This is important for both • Demonstration that vibration levels and critical speed
stability and synchronous behavior especially for machines separation margins (for those under the maximum test
incorporating hole pattern or honeycomb seals. The presence speed) specifications have been met – All tests
of subsynchronous vibration due to phenomena such as stall,
• A stable vs. unstable check is made for a specific test
surge or whirl may also be revealed during the FLFP test.
condition – Type I test
Design Verification Tests
• Proof of effectiveness of the balance procedure in
Unbalance Response meeting vibration level specifications – All Tests
The unbalance response verification test (URVT) provides a • Non-synchronous vibration levels examined – Type I
measurement of how well the analytical predictions match the test, to some degree all tests
vibration produced from a known unbalance. This in turn adds
confidence to the internal deflection, separation margin and Design verification extends those benefits to:
unbalance sensitivity calculations made from that analysis. On • Determining accuracy of unbalance response
the shop floor, the unbalance weight addition is typically calculation with regards to the unbalance sensitivity –
limited to the coupling. Some turbines and overhung machines Unbalance verification (shop floor)
have the ability to add internal weights or weights to the
overhung impeller. Optionally, the test can be performed in the • Determining predictive accuracy for mid-span
vendor’s balance bunker. The bunker permits more freedom in unbalances and deflections – Unbalance verification
terms of weight placement and measurement of shaft deflection (at-speed balance bunker)
at points other than the job’s proximity probe location (mid- • Reassurance that mid-span deflections are within
span, for example.) Of course, the analytical model needs to operating clearances for close clearance locations –
reflect the setup in the bunker; bunker bearings if used and Unbalance verification (higher for at-speed balance
support stiffness of the bunker pedestals. Since the intent is bunker)
verify the accuracy of the analytical predictions, these
differences in configuration should not affect reaching that • Verification of stability level prediction for system
conclusion. with no excitations (baseline stability) – Stability
verification (at-speed balance bunker and during the
Stability mechanical run test)
The stability verification (SVT) test provides confidence in
• Reassurance that rotordynamic model of shaft and
the analytical predictions regarding the stability level and
bearings is accurate – URVT and SVT
position relative to the stability threshold. As with the
unbalance response verification, options are available regarding • Robustness and optimization of machinery design –
the platform or test configuration to perform this test. The URVT and SVT
results of each platform can be summarized below regarding
the accuracy of the analytic method to predict: • Correctness of the stability model of impellers, annular
seals and other destabilizing mechanisms as well as
the effectiveness of any components which are

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

 intended to reduce the destabilizing effects (shunt particular test rig mounted on tilting pad journal bearings [12].
bypass, swirl brakes, damper seals) – Stability Measurement of the basic log decrement provides evidence as
verification (PTC Type I and Type II testing) to whether or not the machine’s rotor/bearing/support system
has the robustness to counteract the destabilizing mechanisms
The at-speed balance bunker can extend the benefits of the
that it will experience in the field. As shown in Figure 2, such
URVT by permitting weight placement at and measurement of
basic log decrement testing also provides the opportunity to
locations that are more sensitive to unbalance and the negative
assess competing tilting pad journal bearing models, a topic of
effects of high vibrations. For between bearing machinery,
much debate in the industry [11, 12].
critical close clearance locations that impact performance are
located at or near the mid-span where the first critical speed has
its peak deflection. In addition, unbalance creation is more
likely to occur at the mid-span due to deposits, erosion or
corrosion and is more likely to excite the 1st critical speed.
Unbalance placement at the coupling (typically the only readily
accessible location during the mechanical shop test) does not
excite the 1st critical speed significantly and provides minimal
information or assessment of the prediction accuracy of this
mode.
Typically, URVT on the shop floor is limited to the
coupling location. Journal probe vibration levels produced by
adding a 40W/N weight to the coupling is less than 0.25 mils
(6.35 µm) at MCS. Performing the verification test in the
bunker can permit measurement of the mid-span and journal
locations to a variety of applied unbalances. Direct
measurement of the relative displacements along the rotor
allows for closer scrutiny of the predicted shaft and bearing
dynamics as they relate to the amount of modal bending and
damping at the critical speed. It should be noted that the intent
of the test is to determine the vendor’s ability to predict the
unbalance response behavior of a model containing a shaft,
bearings and support structure. The results may not be
indicative of the actual behavior of the job machine and relies
on accurate modeling of the bunker pedestal dynamics.
However, the assurance gained from verification of the
predictive method in the bunker should carry over to the job
rotordynamic predictions. Figure 2) Base stability measurements versus speed
Vibration demonstration tests can reveal the presence of
instabilities. However, even if no problematic nonsynchronous RECOMMENDED PRACTICES OF ROTORDYNAMIC
vibrations are observed, the machine’s actual stability could be TESTING
very small (say less than the API design minimum requirement This section will focus on the rotordynamic testing of
of log decrement of 0.1) and close to the stability threshold. centrifugal compressors due to the complex dynamic behavior
This state of blindness with respect to the machine’s actual of this class of turbomachinery. Concepts developed in this
stability, as well as the significant uncertainties remaining in section can be applied to all types of machinery as they all
current modeling tools for stability prediction [11], can only be share certain aspects of rotating behavior. In addition, this
mitigated by directly measuring the machine’s log decrement section will focus on rotordynamics only. The tests listed
through a SVT. below, especially the non-verification testing, are also used to
Unlike the URVT, a SVT requires additional hardware in prove other aspects of the machine’s acceptability. These
the form on a temporarily, mounted shaker. While this is an aspects will not be discussed.
added complication, a SVT is often a cost and technically Vibration Demonstration Tests
effective alternative to the very expensive Type I test. Such
Type 1 testing is unlikely to receive project approval unless it is Mechanical Run Test
specified very early on during FEED. SVT can provide a Mechanical run test of the rotating equipment should be
valuable stability assessment of the machine during the other viewed as a minimal test to determine the rotordynamic
operations/tests that are often typically specified, namely, at- acceptability and should be considered for equipment that is
speed balancing, MRT or PTC 10 Type II performance tests. designed-for-purpose in contrast to equipment selected from a
When done in conjunction with high speed balance or the catalog.
MRT, an SVT can measure the machine’s basic log decrement.
Figure 2 shows the results of such testing versus speed for a

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

Test Procedure closure of the radial bearings or fit relaxation due to operation
at trip speed. Speed versus time for a typical mechanical test is
For most types of equipment, API standards include specific
shown on Figure 3.
procedures to follow during a mechanical test. Important test
factors related to the rotordynamic behavior of the equipment
are:
• Test speeds/duration
• Lube oil parameters (temperature, flow rate and
viscosity)
• Rotor/support configuration
Test Speeds/Duration
Operation of the compressor during the mechanical test
should include a warm up portion where the rotor speed is
incremented in 10 minute intervals carefully avoiding
exclusions zones of critical speeds and blade natural
frequencies. Following the warm up, operation at trip speed is
specified for 15 minutes followed by a non-interrupted 4 hour
run at maximum continuous speed. A coastdown from trip Figure 3) Speed vs. Time for a Typical Mechanical Test of a
follows the 4 hour run. Centrifugal Compressor
The warm up portion is included to ensure that the case and Lube Oil Parameters
rotor are given time to thermally expand gradually to avoid
creating unintended interferences between the two leading to Lube oil parameters should mimic that used for the intended
rub damage. The warm up portion also permits examination of application. Bearing flow, lube oil inlet temperature and
the rotor behavior at increasing speed intervals. Thus, faults viscosity should be within the specified operating ranges set by
can be detected at less energetic stages potentially avoiding the vendor for field operation. After stabilization during the 4
rotor/stator damage and project delays. Trip speed is included hour run, lube oil inlet temperature can be varied from the
to ensure that vibration levels (and overall operation) are minimum to maximum specified range to examine the effect on
acceptable at this speed. The 4 hour run portion is used to set vibration levels and bearing operation. Operation at the range
the thermal conditions of the system. Bearing temperatures and limits should be held until steady state conditions are achieved.
vibration levels are important factors to watch during the test. Rotor/Support Configuration
Stable levels of each parameter need to be reached during the
test. If any parameter shows signs of continual movement The rotor configuration should be as close to the operating
(increase or decrease), the test should be extended until stable condition as possible. The major concern centers around the
levels are achieved. If not, the test should be rejected. overhung weight associated with the coupling. Often
Following the 4 hour run, a coastdown from trip speed is mechanical testing is done at partial rated power levels.
performed. The coastdown is used to determine the overall and Smaller shop drivers need not have, nor in some cases could
synchronous behavior of the rotor/support system. This data accommodate, the larger couplings of the job. It is essential to
will also be used as the baseline for verification testing, if closely mimic the overhung weight of the job ½ coupling. This
performed. may require that a simulator be added to the drive assembly to
match the overhung moment. The opposite may be true for
Operation at trip speed during the mechanical run test is vendor’s smaller casings where the test coupling’s overhung
important for several reasons. First, running to trip speed moment is larger than the job coupling. Rotordynamic
increases centrifugal forces on the rotor which may relax predictions should be used to determine the impact of using the
interference fits and permit the rotor’s static shape to change. heavier overhung weights on the 2nd and 3rd critical speeds.
This may alter the balance state of the rotor. Second, reaching
trip speed is necessary for trip testing in the field. It is Note: For the cases where rotordynamic predictions are
convenient to reach these speeds during testing to identify any used to study the impact of configuration changes, it is strongly
potential problems. Finally, trip speed operation can help recommended that verification testing be performed. The
identify any critical speeds occurring just above maximum increased reliance on the rotordynamic model to predict
continuous speed that would otherwise not be seen on the test behavior that may not be fully tested necessitates that some
stand. verification of the rotordynamic prediction be undertaken.
Additional shutdown/startup transient operation can be Test Requirements
added at the beginning of the 4 hour test. When compared Focusing on the rotordynamic behavior, the requirements
against the coastdown following the 4 hour test, thermal for the mechanical test involve the overall and non-
transient behavior of the rotor can be examined and any synchronous vibration levels, the separation margins of
changes to the balance state of the rotor can be identified. This identified critical speeds, % change of vibration from MCS to
may prove useful in diagnosing Morton’s Effect, clearance trip speed, rubbing at close clearance locations and bearing
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
behavior in terms of temperature. API is frequently used to set The peak vibration with regards to the API vibration limit is
these limits. Overall vibration limit per API 617 [13] is defined taken in the operating speed range from minimum to MCS. For
as: a range of 7000 to 10000 rpm, our example case would report a
maximum overall vibration level of 0.5 mils (12.5 µm) pk-pk.
12000 (For this example, the overall vibration is assumed to equal the
𝑉𝑙 (𝑚𝑚𝑚𝑚 𝑝𝑝 − 𝑝𝑝) = 𝑚𝑚𝑚 �� , 1.0� synchronous vibration.)
𝑁

12000
𝑉𝑙 (µ𝑚 𝑝𝑝 − 𝑝𝑝) = 𝑚𝑚𝑚 �25� , 25�
𝑁

Where;
N = maximum continuous speed (rpm)
Additional requirements may include limits on the
amplification factor of critical speeds and time-dependent
vibration. API 684 [14] provides information on amplification
factors, separation margin requirements and factors influencing
vibrations beyond speed. Additional requirements as to the
thermal related transient maximum behavior may be imposed if
the Morton Effect phenomenon is likely. Figure 4) Bode Plot of Inboard Probe
Test Deliverables Rotordynamics Modeling
The deliverables of a typical mechanical test are: The rotordynamic predictions for the rotor should be
• Electronic recording of vibration and static data (i.e. available prior to the test. Since the model analysis itself will
bearing temperatures) not be tested, the results are reviewed for problem areas with
regards to the analysis and test requirements. Of concern:
• Overall vibration vs. time & speed (tabular or plot)
• Does the rotor meet sensitivity requirements? Per API
• Bode plot - synchronous vibration (mag & phase) vs. 617 [16], an unbalance application of twice the
speed residual (2 * 4W/N) should not produce probe levels
above the vibration limit. Sensitive rotors may not
• Vibration vs. frequency (at each speed and points
meet the vibration limits on test regardless of the
during the 4 hour run)
quality of the balance correction.
• Statement of acceptability
• Are there separation margin problems especially with
Several of the deliverables require vibration recording and the 2nd and 3rd critical speeds? To properly model and
analysis software/hardware be present during testing. predict the location of modes above the 1st critical
Discussions with the vendor should be held prior to testing to speed, users should verify that the overhung weight for
determine what the capabilities of the test facility are and if any the analysis matches that used for the test.
additional equipment is needed. Native files from the data Synchronous vibration at higher speeds should be
collection system, representing the baseline machine behavior, carefully examined for an indication of critical speed
may also prove useful for follow-up diagnostic work or to aide encroachment. Figure 5 shows the impact of a
field troubleshooting. In both cases, additional information potential mode just above 12,000 rpm on vibration
may be desired that was not examined or plotted during the magnitude and phase (circled on plot).
shop testing.
A bode plot is shown on Figure 4. For the X probe, the first
critical speed is easily identified at 5197 rpm with a peak
response of 0.72 mils (18.3 µm) 0-pk. The amplification factor
can be calculated by multiplying the peak response by 0.7071
(half power point – red dashed line) and locating the speeds
associated with that magnitude (green dashed lines). For this
5197
example, the amplification factor is ≈ (5570−4800) = 6.7. This
1
would require a separation margin of 17 ∗ �1 − �=
6.7−1.5
13.7% of minimum speed [15]. The amplification factor may
also be calculated from the slope of the phase angle in radians,
θ, at the peak response, Nc. This takes the form: 𝐴𝐴𝑝ℎ𝑎𝑠𝑒 =
𝑁𝑐
∗
∆𝜃
� ≈
5197
∗
4.712
= 6.4. Figure 5) Impact on Synchronous Vibration of Critical Speed Located
2 ∆𝑁 𝑁𝑐 2 1900 Above Operating Range
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
Test Knowledge/Verification Test Procedure
The knowledge gained from the mechanical test is basically A procedure similar to that of the Mechanical Test as
a pass/fail determination of the rotordynamic behavior. The described above is sufficient. For configuration #1, the full
vibration levels and separation margins (typically only four test should be run. For configuration #2, an abbreviated
concerning the 1st critical speed) are checked to show that they test can be used if the individual bodies were already
meet required levels. In addition, the presence of rubbing at mechanically run tested. (Unlike #1, individual body dynamics
close clearance locations (determined during the posttest are not altered when tested coupled with a long or heavy spacer
inspection) is used as a pass/fail performance of the test. The tube versus using a moment simulator of equal overhung
presence of non-synchronous vibration components may also moment.)
be an indicator of other mechanical faults (i.e., instability,
Test Requirements
mechanical looseness).
For configuration #1, the requirements should be identical
The rotordynamic predictions are not truly verified in a
to the Mechanical Test described above. For configuration #2,
mechanical test. In fact, the shaft vibration normally associated
additional requirements concerning the location of the coupling
with the quality of the rotordynamic behavior is more closely
natural frequency should be included incorporating vibration
related to the quality of the rotor balance performed.
measurements made at either end of the coupling. Peak
Additional checks of the prediction accuracy can be made by
responses at both of these locations could be an indication of a
incorporating other test requirements not required by API 617
coupling natural frequency interference.
7th Edition:
Test Deliverables
• Limit the discrepancy between predicted vs. measured
critical speed location No difference from the Mechanical Test.
st
o 1 critical speed is likely the only mode to be Rotordynamic Modeling
identified.
In either case, the rotordynamic model used for the behavior
o Provides a check of the shaft bending and support predictions should include the coupling and train bodies on
stiffness (stiffness is the primary factor in critical either side. For example, a train comprising a turbine rigidly
speed location). coupled to a generator, should be modeled in its entirety.
Similarly, a power turbine driving a reinjection compressor
• Limit the discrepancy between predicted vs. measured coupled with a long spacer should all be modeled as three
amplification factor (AF) separate rotors (PT, spacer and compressor) coupled together
o Amplification factor is an indicator of the modal by the flexible elements modeled as shaft elements or lateral
damping present and may vary end-to-end and spring elements. Figure 6 presents the rotordynamic for a
from X to Y probe. Care should be taken to rigidly coupled steam turbine / generator train.
compare the same probe from predicted to
measured.
o The AF can vary across measurement locations
due to runout and residual excitations in the shaft
not fully taken into account in the predictions.
Figure 6) Rotordynamic Model for a Steam Turbine / Generator Train
Beyond what is measured during the test, the rotordynamic
analysis is relied upon to predict the internal deflections of the Test Knowledge/Verification
rotor (marginally verified by the inspection for rubbing), More knowledge of the coupled train behavior is obtained
sensitivity to unbalance (not checked) and stability (not with the string test. However, the same statements regarding
checked). design verification can be made as with the Mechanical Test.
String Test Full Load / Full Pressure
With regards to rotordynamics, a string test is needed to Due to the relative cost of performing a full load / full
accurately determine behavior for the following configurations: pressure (FLFP) test, this option is typically limited to
1. A rigid coupling is used in the train – Individual body centrifugal compressors meeting all of the following criteria:
dynamics are affected by rigidly coupling them • Loss of production produces an unacceptable financial
together. The rotordynamic model and test impact
configuration should be performed with both bodies
coupled. • High pressure applications

2. There exists a long or heavy coupling spacer in the • Limited vendor experience with the design or
train – In this case, the dynamics of the coupling are of application
concern. To model the boundary conditions at the • Rotordynamic analysis reveals increased risks or
coupling hubs, the coupling should be tested installed concerns with the rotor stability
in the train.
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
Test Procedure operation in the field nor deterioration of the balance state from
erosion or deposits. An agreed level should take into account
For the FLFP test, the procedure should be developed in
both factors and fall somewhere in between.
concert with the rotordynamic analysis. The FLFP test is
intended to study the stability of the centrifugal pressure under Other requirements for the rotor and case vibration may
similar conditions to the field. By applying load and pressure, include:
destabilizing forces of the seals and impeller/shroud
interactions are introduced including seal clearance changes • Components of non-synchronous vibration to be less
due to internal deflections as a result of reaching full pressure. than 20% of the vibration limit or 0.2 mils (5 µm) p-p,
Operating points during the FLFP test should be determined, in whichever is less
part, by the rotordynamic analysis and reflect operating points • No stall related vibration components
of minimum stability. These could represent operation at MCS
near surge (highest differential pressure) or, in some cases, • No instability related vibration (associated with re-
partial speed towards stonewall. Since this test as normally excitation of the 1st natural frequency)
carried out is a “pass/fail” (e.g. the rotor is stable or not), test • Maximum housing vibration less than 0.1 in/sec
conditions should match as close as possible to the field (2.5 mm/sec)
conditions. The parameters of importance include gas MW,
power, suction and discharge pressure and temperature, speed • Limitations regarding thermal instability (Morton
and mass flow. If the exact gas composition cannot be tested, Effect) vibrations
some of these parameters will have to be compromised. The
Test Deliverables
rotordynamic model should be used to determine an appropriate
combination of factors to produce the maximum instability Deliverables are similar to the mechanical run test and
drivers or minimum log decrement. Miranda and de Noronha should include data for all purchased components tested (as
[17] propose modifications to the FLFP ASME Type I test to with the string test.) Increased emphasis is placed on the FFT
better assess the stability of centrifugal compression equipment. plots of shaft vibration during the test as this is the best
The modifications were intended to create conditions to submit indicator of instability, stall, whirl and other phenomena that
the rotor to instability mechanisms as near as possible to the produce non-synchronous vibrations. Performance data should
design conditions rather than reproduce similarity for be recorded during the test to confirm the input used to predict
performance evaluation. The conditions were developed with the seal and impeller dynamic behavior and aide further
the aid of the rotordynamic stability predictions. stability analysis if needed.
As with the Mechanical Test, the FLFP test should consist Rotordynamic Modeling
of a warm up portion where the speed is increased gradually to
When FLFP testing is selected, the rotordynamic model
permit stabilization and examination of the behavior at lower
should be expanded as necessary to conform to the Level II
speeds. This is followed by an extended run at MCS to ensure
stability analysis requirements of API 617 [18]. Given the cost,
thermal equilibrium of the entire machine is achieved. During
effort and reasons to perform FLFP testing, a Level I stability
this test phase, it is recommended that the operating curve at
analysis is insufficient to predict the behavior accurately. The
MCS is explored from the surge control line to the end of curve
Level II model will reflect changes in the stability level to MW,
(stonewall.) This operation may include four to five operating
gas pressures and temperatures, seal clearances and rotor speed
points and may include other speeds as highlighted by the
to the best of the vendor’s or purchaser’s analytic capabilities.
rotordynamic analysis. (Note: Other factors may dictate
The Level I model uses an empirical relationship that either
operation at other points as required, i.e. defining the surge line
estimates these effects or doesn’t take them into account at all
vs. speed, rated point defined at partial speed.)
when calculating the destabilizing forces present in the
Factors such as lube oil conditions and rotor assembly are machine.
expected to meet the field configuration and specified operating
Test Knowledge / Verification
ranges. When practical, the lube oil operating range should be
explored during the FLFP test. Lube oil inlet conditions impact The FLFP test will reveal the presence of instability, stall or
the dynamic behavior of bearings. As a critical factor in whirl for a prescribed set of operating conditions for the
determining the rotordynamic behavior, it is important to vary specific machine’s as-built conditions. The test is pass/fail as
these factors over the allowable ranges during testing. no measurement of the stability level or margin is included in
the test as described. Rotor stability at different gas
Test Requirements
compositions, other clearances within the tolerance range or
Test requirements for the FLFP test are defined by other operating points is determined by rotordynamic
agreement between the vendor and purchaser prior to the test predictions. The ability to operate successfully at these
and should be done at the contract stage. Holding the overall alternate points, which cannot be tested under all combinations,
vibration limit to the level specified for the mechanical run test depends on the stability margin (not measured by FLFP test) of
is impractical due to the additional rotor forces present during the machine.
the FLFP test. These include aerodynamic forces of the
impellers, stator-rotor interactions, seals forces and power
transmission forces. However, raising the limit to the vendor
recommended trip setting does not leave margin for off-design
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
Design Verification Testing failing the mechanism used to hold the weight in place.
Second, the method relied on larger vibration limits <2.0 mils
Such verification testing should be considered for the
(50 µm) p-p versus the current <1.0 mil (25 µm) p-p and good
following types of equipment:
balance correction practices to limit synchronous response on
• Special purpose (as defined by API) the test stand below 0.5 mil (12.5 µm) p-p. For this situation,
75% of the response would be attributable to the verification
• Loss of production produces an unacceptable financial
weight. At the time, this was better than no test at all.
impact
In 1997, Nicholas et al. [19] defined an improved procedure
• Services or applications with a history of bad actors to better correlate test stand vibrations to the analytic
(vibration related) predictions. This methodology was subsequently adopted by
• Critical service (as defined by the user) API 617 7th Edition and is explained in API 684 [20]. Their
method took advantage of vibration diagnostic equipment that
In addition, verification testing of either the unbalance permitted vector subtraction of recorded databases. The
response or stability should be considered when the procedure can be summarized as:
rotordynamic analysis demonstrates concerns or higher risks
associated with the application. The verification testing can be 1. Record the probe synchronous readings during
used to mitigate those risks when applied correctly. coastdown from trip speed following the four hour
mechanical run test – This represents the baseline
Unbalance Response Verification Test (URVT) vibration of the rotor
Test Procedure 2. Add the verification weight to the rotor – The method
The URVT test is basically a comparison of measured is general enough to accommodate weight placement
versus predicted vibration levels for the application of a known anywhere on the rotor
unbalance. The test is routinely performed following the four 3. Bring the rotor back to MCS and achieve steady state
hour mechanical run test. For compressors, the verification conditions (i.e. constant bearing temperature, vibration
weight is applied to the coupling flange. This is the only magnitude and phase) – Attempt to reproduce the
practical location available. For other machinery (e.g., steam operating condition of the machine at the conclusion
turbines, overhung single stage compressors), other locations of the four hour test in step #1. The sampling
may be available. Steam turbines may have field accessible frequency and speed increment should be identical to
balance planes and overhung compressors an impeller checknut that used in Step #1.
with a balance weight placement provision. The measured
response of the machine with the verification weight is 4. Record the probe synchronous readings during
compared against the analytical prediction using the same coastdown from trip speed following Step #3 – This
weight and location. While this is not a complete check of the represents the combined vibration of the rotor
analytic unbalance response accuracy, as it only compares the (verification weight and residual unbalance)
model’s prediction at the probe locations for one weight 5. Vectorially subtract the synchronous vibration
placement, it is an important first step in ensuring the accuracy database taken in Step #1 from that recorded in Step
of the model. #4 – The resulting data represents the response due to
It was recognized early on that an important aspect of the the verification weight placement
URVT test was to compensate for the residual unbalance in the 6. Compare the resultant data in Step #5 to the analytical
machine. This residual unbalance creates the synchronous predictions (incorporating the range of bearing
vibration witnessed during the mechanical test run and is clearances and oil inlet temperatures) – This is the test
present before and after the verification weight is applied to the for accuracy of the unbalance response predictions
rotor. The residual unbalance left in a rotor after balancing
(either low or at-speed balancing) is uncharacterized and, Graphically (and using vector math) the method can be
therefore, cannot be modeled. Thus, the analytical model will described as:
have only the verification weight as an excitation source for the ���⃗1 , (known) is taken at
• A baseline response reading, 𝑉
response.
one speed and one probe and is attributed to the
The initial attempt by API 617 6th Edition to compensate for ����⃗𝑟 , in the rotor (unknown).
residual unbalance, 𝑈
the residual unbalance was to apply a significant verification
weight to the coupling to raise the response to the vibration • The subsequent response, 𝑉 ���⃗2 , (known) taken after the
limit. The implication was that the majority of the response addition of the verification weight, 𝑈 ����⃗𝑣 , (known) to the
would be due to the verification weight placement. This had rotor is recorded. The net unbalance present in the
two important drawbacks (besides being analytically incorrect): rotor at this stage can be described as ����������������⃗
𝑈𝑟 + 𝑈𝑣 .
First, the amount of unbalance weight at the coupling needed to
raise the response to the vibration limit could reach unsafe
levels. Coupling flanges are not designed with the intent of
adding unbalance weight. Large rotating forces applied to the
coupling had a chance of failing the flange or, more probably,
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 • Performing the vector subtraction of the 1st reading
from the second yields the response, ���⃗ 𝑉𝑠 = ���⃗
𝑉2 − 𝑉 ���⃗1 .
URVT Vehicle
The accompanying vector math with the unbalance
Shop Floor (Mechanical Run Operating Speed Balance
state of the rotor, 𝑈 ����⃗2 − 𝑈
����⃗1 = ����������������⃗
𝑈𝑟 + 𝑈𝑣 − 𝑈 ����⃗𝑟 = ����⃗
𝑈𝑣 , Test) Bunker
demonstrates that the resultant response, 𝑉𝑠 , is due ���⃗
+ Test performed with the job + Increased flexibility for
only to the verification weight added to the rotor. configuration weight placement
This is shown on Figure 7. + Additional measurement
points possible
+ Additional verification of
the rotordynamic analysis
Table 1) Comparing Advantages of Each Test Vehicle

• As required by API 617 7th Edition, predictions for the

close clearance locations and unbalance sensitivity
should be adjusted as dictated by the verification
testing – For example, should the verification test
show the rotordynamic analysis is under predicting by
a factor of two, all response levels calculated for the
lateral analysis should then by multiplied by two and
sensitivity and close clearance requirements rechecked
Figure 7) Determination of Rotor Response due to Verification Weight for compliance.
All other aspects of the URVT should conform to the Depending on the vehicle, additional checks regarding
requirements of the mechanical run test in terms of oil inlet critical speed shape, amplification factor and modal response
temperatures, speeds and rotor support. can be performed.
Operating Speed Balance Bunker Option Test Deliverables
An option exists of where to perform the URVT. Operating The deliverables for the URVT should include:
speed balance bunkers (also referred to as at-speed or high • Rotordynamic analytical response for the setup,
speed balance bunkers) can be used and has advantages over weight placement and magnitude and measurement
the test floor. The advantages are associated with access to the locations used in the test. The response should be
rotor that running in the bunker presents. The operating speed carried out over the specified range of clearance and
balance bunker permits alternate locations for the verification oil inlet temperature to produce the largest variations
weight (i.e. mid-span, quarter-span), use of multiple weights in radial bearing dynamic coefficients. The analysis
and additional shaft vibration readings. The accessibility of the should identify the critical speed location (response
bunker permits additional verification of the rotordynamic peaks) and agreed upon unbalance response versus the
predictions, i.e. mid-span response relative to journal locations variation in bearing dynamic coefficients for the
for mid-span and quarter-span weight placements. verification weight.
There is one important complexity when considering
• Measured synchronous vibration data should be
performing the URVT test in the balance bunker; the rotor
plotted for the following conditions:
support stiffness in the bunker will be different than the actual
machine. Often the bunker bearings, while similar, are not o Coastdown following the 4 hour test plotted for the
identical to the job bearings. Different clearances, L/D ratio four proximity probe locations – This represents
and pad arc length and the existence of a levitation pocket will the baseline data for the machine (vibration as a
result of the residual unbalance and forces in the
alter the stiffness and damping coefficients. The support
rotor.)
stiffness of the pedestals will also be different than the job
machine. Bunker pedestals are intended to be flexible to permit o Coastdown following placement of the verification
adequate velocity measurements used in most bunkers’ balance weight and warmup at MCS (if needed) – This
software. The stiffness is a function of frequency and represents the vibration data for the residual +
verification weight.
inaccurate representation will produce significant differences
between predictions and measurements. This can lead to the • Vector subtraction of baseline data from the
erroneous conclusion that the model of the job compressor (not verification + residual data – This will be the rotor’s
using bunker bearings and pedestal stiffness) also suffers from measured response to the verification weight.
the same inaccuracy. Table 1 presents the advantages of each
option for URVT. • Comparison of the measured peak response speed
(typically the 1st critical speed) to the predicted range
– Discrepancies outside the permitted range (5% for
API 617) should be addressed by correcting the rotor
model (i.e., addressing the values used for support

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

 stiffness, bearing stiffness and/or shaft bending Notice that measured probe response with the verification
stiffness). weight does not resemble the predicted response, Fig. 8. This
is due to the residual unbalance response in the rotor.
• Comparison of the measured response magnitude Performing the vector subtraction of the two data files isolates
versus the predicted value at the required speed(s) – If the response due solely to the verification weight. This is
predicted value is less than the measured value, then compared against the predicted response on Fig 10.
the corrected analytical predictions based on the factor
of measured/predicted is supplied. Close clearance
and sensitivity requirements should be reviewed for
compliance.
• Other analytical vs. measured requirements as
specified in the test procedure.
Two examples of the URVT are provided to illustrate the
procedure. Both are taken from ref. [19]. The first example is
an 8 stage compressor. Figure 8 plots the predicted response
for a verification weight placed at the coupling location Figure 10) 8 Stage Compressor – Measured vs. Predicted Verification
Response
(40W/N magnitude) of the compressor for one drive end probe.
Figure 10 illustrates the need to remove the effect of
0.25 residual unbalance due to the relative insensitivity of the rotor
to coupling unbalance. Even at low amplitudes, ≈0.00015”, the
Min Stiff measured response falls within the maximum levels calculated
0.20
Max Stiff from the rotordynamic response especially at MCS where the
requirement was enforced. In addition, the peak response (1st
1X Amplit ude (mils p-p)

0.15 critical speed) falls within the range predicted by the analysis,
Figure 8. For this example, no correction of the lateral analysis
was needed.
0.10
The measured response of a three stage compressor’s
0.05
baseline and verification run is shown on Figure 11. In the
second example, the measured response to the coupling
verification weight is shown to exceed the predicted value at
0.00 MCS by 25%, Figure 12. The measured peak response of the
0 2000 4000 6000 8000 10000 12000 14000 16000
1st critical speed does occur within the range predicted by the
Rotor Speed (rpm)
analysis, Figure 11. For this case, the lateral analysis unbalance
Figure 8) Response Plot of Coupling Verification Weight for Min and response magnitudes at the close clearance locations and for
Max Bearing Stiffness 8W/N unbalance should be increased by a factor of 1.25. The
The corresponding probe response for the baseline and rub limits and sensitivity requirements should be rechecked for
verification weight added runs are plotted on Figure 9). A compliance.
drive-end probe was selected since the verification weight
(identical to that used in the analysis) was added to the coupling
and the drive-end would have the largest response. The shop
floor during the mechanical test run was used for the
verification test.

Figure 11) Measured Probe Response for 3 Stage Compressor

Figure 9) Measured Probe Synchronous Response for 8 Stage

Compressor
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 Stability Verification Testing (SVT)
SVT Methodology
Currently, there are no industry standards in place to guide
the end-user and OEM as to what methodologies and practices
should be used to accurately perform these SVTs. This tutorial
will attempt to provide some guidance in this area, specifically,
in the three fundamental elements of the testing process:
nonsynchronous excitation design, measurement process, and
Figure 12) 3 Stage Compressor – Measured vs. Predicted Verification damping ratio estimation. Since this last element dictates much
Response of the methodology requirements for the other two, our
discussion will begin with some guidelines regarding damping
Rotordynamic Modeling
ratio estimation.
As previously mentioned, the rotordynamic model for
SVT Damping Ratio Estimation
URVT needs to match the configuration of the test setup. If a
balance bunker is used, the support stiffness of the model It should be noted as to why the term “damping ratio
should include the dynamic effects of the bunker bearings and estimation” has been chosen instead of “log decrement
pedestals. The rotor model should also reflect the ½ weight of measurement.” This choice is meant to emphasize several
the test coupling (or bunker drive assembly) in its correct c.g. aspects:
location. The predictions should cover the minimum to
1. The fundamental modal parameter of interest is the
maximum bearing stiffness (as required by API 617 7th Edition)
damping ratio.
based on the specified ranges for the radial bearing clearances
and oil inlet temperatures. Finally, the analysis should specify 2. Estimation of the damping ratio involves a post-
the predicted ranges for the 1st critical speed and verification processing analysis of the measurement data. The
response magnitudes (additional information as necessary to resultant damping ratio is not measured directly, it is
determine compliance with other requirements.) only estimated.
Summarizing the aspects of the rotordynamic modeling 3. Logarithmic decrement is its own estimation technique
critical to this testing: for damping ratio.
• Support stiffness to represent the test vehicle Damping ratio estimation is a subset of modal parameter
identification. Since this identification occurs while the
o Design configuration for URVT performed during
machine is operating, such testing is often called operational
the mechanical run test on the shop floor
modal analysis which is a subset of the more general field of
o Reflecting the bearings and pedestals of the system identification. Identification of other modal parameters,
operating speed balance bunker such as natural frequency and mode shapes, has historically
received much greater attention. This is likely due to their
• Verification weight placement and magnitude importance with respect to resonance problems, a predominant
• Rotor vibration at the probes or other locations as used concern in every field whereas stability is not as much so.
during the URVT As mentioned earlier, the estimation involves a post-
o Include support motion if probes are mounted on processing analysis of the measurement data. While there are a
bunker pedestals multitude of estimation techniques available and continually
emerging, each technique basically involves a curve fitting of
Test Knowledge / Verification the data. Some are very simple and deal with just a single
The verification test provides a determination of the response channel, while others curve-fit multiple channels’ data
accuracy of the rotordynamic predictions in a quantitative at once. These techniques are being developed by various
fashion. Beyond the pass/fail observational nature of the communities, such as controls and speech processing, with each
vibration demonstration tests, the URVT demonstrates the technique designed to utilize time or frequency domain data.
predictive versus measured response discrepancy to a known It is beyond the scope of this tutorial to discuss all the
weight placed on the rotor. Performed correctly, the operating aspects of damping ratio estimation. However, it is important
speed bunker can provide additional accuracy assurances that machinery engineers understand some of the peculiar
regarding the modal response (i.e., mid-span versus journal challenges and the recommended practices. When conducting a
response ratio) and rotor sensitivity to locations other than the SVT on a turbomachine, the estimation process faces several
coupling 2. challenges when dealing with the measurement data:
• The data is typically from only a few vibration
2
Note: For some well balanced and well damped rotors, the 1st critical speed response locations.
may not be identifiable on the test stand as a clear peak in the vibration data and
will not be appreciably excited by a verification weight placed at the coupling.
Performing the test in the balance bunker enables mid-span weight placement
and greater excitation of the 1st critical speed for between bearing machines. probes, alternate unbalance weight locations) should be specified and agreed
All options regarding the use of the balance bunker for URVT (e.g., mid-span during contract development with the vendor.
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
 • The data contains additional response from the
significant presence of immeasurable, internal
excitations within the machine such as unbalance,
misalignment, aerodynamics, etc.
• The data likely contains contribution from multiple
modes, not just the mode of interest.
• The data is from a system whose dynamic
characteristics are not like those for static, nonrotating
structures.
Each of the above challenges influences what techniques
can be used for accurate stability estimation. Some techniques
rely on a large number of sensors, meaning they are not
appropriate here because response measurements are typically
only available at a few restricted locations on a machine. Other
techniques assume the system being tested is a static structure
whose dynamic properties are symmetric or self-adjoint.
Techniques relying on such an assumption are not appropriate
for a turbomachine where fluids within the machine create non-
symmetric cross-coupling in the system [21,22].
Perhaps the biggest estimation challenge is the fact that the
measurements from a rotor’s SVT often consists of significant Figure 13) Example Campbell Diagram
contributions from several modes, more than just solely the
mode of interest. Many structures possess very close natural To date, many previous SVT investigators have relied on
frequencies which can be difficult to identify through the the assumption that only the mode of interest, typically the first
estimation process. Sharing this high modal density forward mode, is participating in the response. It is easy to
characteristic, rotor systems possess pairs of sister modes, one understand why this assumption has historically been applied
forward whirling and one backward that are often in close by the industry. First, the measurements often only show one
proximity and share similar modal characteristics. peak in the frequency response functions. Second, the
nonsynchronous excitation was applied in the forward circular
The easiest way to observe the close proximity of these direction with the intent of only exciting the first forward
sister mode pairs is through a Campbell diagram. Figure 13 whirling mode. And finally, single degree of freedom (SDOF)
shows the Campbell diagram for a multistage centrifugal damping estimation techniques, such as amplification factor,
compressor. Typical for between-bearing machines, the first phase slope or mechanical log decrement, involve convenient
mode sisters remain very close in frequency as speed increases, and simple formulas that are familiar to every vibration
where the backward mode’s frequency is slightly less than the specialist.
forward mode. The second mode sisters become more and more
separated with speed, indicating the stronger influence of Unfortunately, even when forward circular excitation is
gyroscopics and support anisotropy for these sisters. The applied and only one peak is present in the measurements, the
influence of gyroscopics and support anisotropy is highly sister backward mode can be excited, reducing the accuracy of
dependent on the overall rotor system characteristics. For the SDOF damping estimators. Figure 14 illustrates this
example, unlike the between-bearing compressor, the first sister behavior for a simulated rotor system where the individual
modes for an overhung compressor will be very sensitive to modes’ response can be distinguished. Even though the forward
speed due to the overhung impeller’s strong gyroscopic circular excitation is being applied and only a single response
moment. peak is measured, the backward mode is excited. Because two
modes are responding, instead of just one, the peak is distorted
and, in this case, broadened, from that due to just a single
mode.

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

 There are a variety of powerful output-only identification
techniques available for OMA. Some rely on relatively long
recordings of vibration data for their estimation, while others,
such as MOBAR, were developed to handle vibration data over
a short duration.

While OMA is a promising approach, its strengths and

weaknesses with respect to stability verification testing have
yet to be fully identified. The amount of ambient excitation
present and the “color” of its frequency content, e.g., white,
pink, etc., can vary significantly within each machine and test
stand operating conditions (vacuum, FLFP, etc.). These
ambient excitation aspects, combined with variabilities in the
rotordynamic nature of different machines (wide variations in
stability and proximity of sister modes), all affect the signal-to-
noise ratio (SNR) and, thus, the reliability of damping ratio
estimation accuracy of the mode of interest, 1F.

Figure 14) Modal Participation from Nonsynchronous Forward, Because of the variability and evolving nature of OEMs’
Circular Excitation [23] adoption of estimation techniques, some end-user companies
will independently verify the SVT measurements using their
While the methodologies associated with SVT’s damping own trusted methodologies. This practice is directly analogous
estimation are still evolving, there are several ones, long to end-users’ independent verification of calculations for
considered good practice in the system identification rotordynamic predictions and aerodynamic performance
community, that have emerged as recommended practices: testing.
• Multiple degree of freedom (MDOF) techniques
provide the most reliably accurate estimates of
damping ratio without relying on any assumptions
about a particular machine’s rotordynamic
characteristics. To date, there are two MDOF
techniques which have demonstrated reliable accuracy
for stability verification testing: multiple output
backward autoregression for time domain response
measurements, and the prediction error method for
frequency response function measurements.
• To resolve the closely spaced sister modes of rotors,
MDOF techniques should be applied to multiple
output (MO) data sets consisting of response from
multiple locations along the rotor.
• Any candidate estimation technique should be
validated using simulated excitations and
measurements that follow the methodology planned
for shop testing but are taken from a simulated rotor
system where the eigenvalues and stability are known.
This validation should assess the potential for
asymmetry in the bearing system and various levels of
stability. While a simple rotor system can be used for
such validation, the most preferable choice is to use
the model of the machine to be tested.
Recently, some OEMs have been relying on ambient
excitations in conjunction with output-only identification
techniques for stability verification testing [37]. This approach,
known as operational modal analysis (OMA), has been applied
in some communities where either measurement or application
of an excitation mechanism is difficult or impossible. It is very
attractive for a machine’s SVT because no excitation device
needs to be installed.

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

SVT Nonsynchronous Excitation
Our SVT methodology discussion will next focus on
aspects regarding the nonsynchronous excitation that is
necessary to measure a mode’s stability. The design of both the
excitation device itself, and the signals that it applies to the
machine, are important elements of the SVT’s success.
Unless the subject machine is supported by radial active
magnetic bearings, the nonsynchronous excitation must
originate by some temporary means. The earliest forms of
excitation were from impacts made directly on the rotor
[24,25,26]. More recently, shakers were mounted externally on
machines’ bearing housings [4,27] with an example shown in
Figure 15.
Typically installed on the non-drive, outboard of the
machine, electromagnetic shakers have emerged as the most
popular device for nonsynchronous excitation. Two
methodologies have been used to temporarily attach the
Figure 16) Shaker with Bolted Shaft Extension [30]
shaker’s laminated rotor assembly, a bolted extension has been
used by several investigations [28,29,30] while others have
mounted a tapered sleeve [9,31]. Figures 16 and 17 illustrate
examples of these two attachment designs.

Figure 17) Shaker with Sleeve Mounted on Tapered Shaft End [9]
There are several key objectives that are important to the
successful design of the excitation device:
• Design for ease of installation and removal during
shop testing,
• Minimize alteration of the machine’s balance state and
rotordynamics when the device is installed,
• Provide sufficient force capacity and bandwidth to
excite the mode of interest when the machine is
operating at the test conditions.
When utilizing an electromagnetic shaker, its laminated
sleeve or shaft extension can add sufficient weight and inertia
to significantly alter the machine’s baseline dynamics.
Rotordynamic calculations should be conducted during the
shaker’s design process to examine the impact on the machine’s
rotordynamics, in particular, to ensure minimal impact on the
Figure 15) Shaker Mounted on Bearing Pedestal in Balance Bunker [4] mode of interest’s stability and frequency. If it is undesirable to
alter the test speeds, one may have to accept lower than desired
separation margins for other critical speeds. While it is vital to
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
protect the machine from damage during the testing, it should investigators the confidence to use single input excitation on
also be recognized by all parties that typical vibration industrial machines [9].
acceptance criteria are not applicable during a SVT. Regardless
of the exciter’s impact on the rotordynamics, it is a good
practice to design the shaker’s rotor assembly with a balance
correction plane as well as check balance the rotor assembly
with the shaker installed prior to testing. Low synchronous
vibrations help improve the quality of the SVT measurement
data and the resulting damping estimation.
Finally, the frequency content and direction of excitation
must be determined. Both must provide the type of
measurement data needed for the damping estimation technique
originally chosen. If time domain estimation techniques are
employed, the best signal-to-noise ratio (SNR) can be obtained
through a blocking test. A type of tuned-sinusoidal method
[32], blocking testing effectively tries to isolate a mode by
exciting at its natural frequency and in its predominant
direction. Several investigators have successfully applied this
excitation method [9,27,33] for stability verification testing of
rotor systems. Direction of the blocking excitation, forward or
backward whirling or along one axis, can be chosen to best
excite the mode of interest.
If a frequency domain estimation technique is chosen, the
required measurement data consists of frequency response
functions (FRF) across the frequency range containing the
mode of interest. Calculated using correlation functions that
consider noise in the system, a measured FRF has units of
response (displacement, velocity or acceleration) divided by
force. The frequency range is spanned by the excitation device Figure 18) Simulated Rotor FRF Measurements from Excitation in
using stepped sine, chirp or pseudorandom signals, with the only Horizontal Direction
final choice determined by the desired SNR and testing time.
Stepped sine is generally considered to have the best SNR, Mode 1F 1B
while other frequency signals can provide faster measurement Parameter ζ (%) ωd (cpm) ζ (%) ωd (cpm)
times. Actual 4.62 4110 8.56 3809
SIxMO 4.54 4111 8.39 3805
The direction of the applied excitation must be considered SIyMO 4.62 4111 9.02 3799
when the FRFs are being calculated during the measurement
process. When exciting in only one direction, such as along the Table 2) Modal Parameters Obtained from SIMO FRF
machine’s horizontal splitline or along one proximity probe’s Measurements of a Simulated Rotor System [33]
axis, the FRFs are easily calculated according to single input,
multiple output (SIMO) procedures. Contrary to popular As mentioned earlier, it is a popular choice to excite the
thinking, such SIMO testing along only a single axis is rotor in a forward circular direction. However, it is often not
sufficient to excite the first forward whirling mode and can recognized that exciting in a forward circular or any elliptical
provide accurate damping ratio estimates when used with an fashion essentially means that a multiple input, multiple outputs
appropriate MDOF frequency domain technique. (MIMO) test is being conducted. Whenever the excitation is
not along a single axis, but takes any form of elliptical shape,
Figure 18 presents the FRFs from a SIMO test conducted on
two forcing inputs are required creating a multiple input
a simulated rotor system with known stability. In this case,
situation. When multiple input testing is preferred, such as
horizontal excitation is applied at the inboard bearing and four
when using forward circular excitation, MDOF frequency
FRFs are obtained, one for each of the four bearing probes.
domain estimation techniques cannot be applied unless the
Noise has been added to the measurements to simulate real
measured FRFs are calculated from force and response data
world conditions. Table 2 compares the actual stability levels of
from two MIMO tests: one with forward circular input, and the
the first sister modes with that obtained using the SIMO
other backward circular input. Ewins [34], Maia [32] and
measurement data and a MDOF estimation technique. For
Pintelon [35] provide details on the practical and theoretical
either horizontal (SIxMO) or vertical (SIyMO) forcing, excellent
aspects of such multiple input testing.
accuracy is achieved not only for the primary mode of interest
(1F) but also its sister backward mode (1B). Vertical forcing SVT Measurement Process
provides a slightly more accurate damping estimate for the 1F
mode because the mode shape is more vertically oriented for Fair correlation with the predictions requires careful
this particular machine. Such performance has given attention during the measurement process. This means that the

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

operating parameters of the machine (speed, inlet pressure, etc.) dependent stiffness and damping characteristics, or using the
must be held within that assumed for the modeling predictions. traditional, whirl frequency independent model of
Oil supply temperature has proven to be an operating parameter synchronously reduced coefficients [11,12]. Current API
to closely monitor and control especially when trying to standards require the use of synchronously reduced coefficients
reconcile the difference between measurements and predictions for predictions. However, it is recommended that measurements
[9]. Without proper control of the bias currents, an should be correlated with predictions using both representations
electromagnetic shaker can impose a static force on the shaft. to help shed light on this ongoing debate. Comparing the
Therefore, the operating position of the journal bearings should correlations of the two representations is especially important
be also closely monitored. A significant shift in journal for base stability testing when using bearings with center pivots
operating position when the shaker is energized will alter the and low (<0.4) pad preloads.
bearings’ stiffness and damping characteristics, and thus, the
If an SVT is conducted as part of a performance test (e.g.
measured stability of the machine.
PTC 10 type I or II), the measurements should be correlated
When a frequency domain estimation technique is with API Level II stability predictions that include the
employed, accurate measurement of the FRFs is vital to the dynamics created by the machine’s internal seals and other
success of the SVT. Often overlooked by the turbomachinery components. Each manufacturer has its own methodology for
industry, there are two recommended practices originating from how these internal dynamics are modeled and analyzed. As
the modal testing community that should be adhered to with required by API Level II stability analysis procedures, this
respect to these FRF measurements: methodology should be explained and documented by the
manufacturer.
• Calculation of any frequency response function
requires measurement of the response outputs’ and the SVT Deliverables
forcing inputs’ magnitude and phase.
• For each test condition, the predicted range of stability
• The coherence of each FRF should be examined to (log decrement and damped natural frequency) for the
ensure the quality of the measurement around the machine with, and without, the presence of the shaker
mode of interest’s frequency. device.
It is often assumed that, once the current amplitudes are set • Description of the damping ratio estimation technique
in its control system, an electromagnetic shaker will produce a employed
force of constant amplitude and phase across the whole
frequency range. However, like other exciter devices where • Measured operating data for each stability test
“force drop-out” can occur, electromagnetic shakers can be condition, such as:
susceptible to irregularities in the amount of force that is o Speed
actually applied to the rotor when traversing a natural
frequency. A highly amplified natural frequency can create o Inlet and discharge pressure and temperatures
large vibration amplitudes at the shaker, resulting in significant o Molecular weight
differences between the actual currents being applied to the
shaker and those specified by the control system. o Oil supply pressure, temperature, and flow rate

Therefore, like any force transducer used in modal testing, it • When stability estimates are obtained from outputs’
is recommended that the electromagnetic shaker be calibrated time transient data,
to determine the relationship between the measured parameters o Sampling frequency
(such as flux or coil currents and rotor displacements) and the
forces applied to the rotor. Without measurement of the o Number of transients events recorded
shaker’s force during testing, the FRFs’ amplitudes at the mode o For each output location and operating condition,
of interest could be distorted, resulting in poor stability plot showing the measured time transient data
estimates.
• When stability estimates are obtained from frequency
SVT Rotordynamic Modeling response functions,
In addition to the issues discussed for URVT, there are o Description and records of the calibration of the
several additional modeling aspects that needed to be input force measurement
considered for stability verification testing. As mentioned
earlier, rotordynamic calculations should be conducted during o Number of averages taken, window type and
the shaker’s design process to examine its impact on the overlap percentage
machine’s behavior. This examination should include not only o Frequency resolution
the impact on the mode of interest’s stability and frequency, but
also the impact on other critical speeds and their separation o For each input/output location and operating
margins. condition,
When tilting pad journal bearings are employed, it is  Plot showing the final measured FRF
heavily debated whether the bearings should be represented (magnitude and phase) and the identified FRF
using their full coefficients which allow whirl frequency from the estimation technique
Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station
  Plot showing the final measured coherence SVT provides at least some verification concerning the
associated with the measured FRF reliability of the model’s stability predictions.
• Comparison of the measured stability (log decrement
CONCLUSIONS
and damped natural frequency) versus the predicted
range for each tested operating condition Rotordynamic testing is an effective and efficient tool,
when applied appropriately, to mitigate machinery risks. The
• Depending on the specifics of the acceptance criteria, tutorial presented the basis for determining the machinery
the resulting stability predictions from the corrected application risks. Test options available to address those risks,
analytical model should presented to determine test procedures, preparation for the test, knowledge gained from
acceptability of the design. each test and the testing benefits were discussed. The
SVT Requirements difference between vibration demonstration and design
verification testing was highlighted. The principle difference
Currently, there are no industrial standards in place for being that vibration demonstration tests assess the acceptability
guidance on what acceptance criteria should be applied for this of rotordynamic behavior in a pass/fail mode while design
type of design verification testing. Individual OEMs and end- verification testing is used to confirm the rotordynamic
users are developing their own acceptance criteria in the predictions. This verification provides confidence in
meantime. It is recommended that the criteria should have a extrapolating the design (by the vendor) and operation (by the
similar two-step evaluation process as that standardized for the purchaser) beyond the machine’s as-built and specific shop test
URVT: conditions. The recommended practices of performing the test
1. How well does the original rotordynamic model and options for vibration demonstration and verification testing
analysis predict the measured stability results? concluded the tutorial.
2. If the model has poor accuracy with respect to the NOMENCLATURE
measured stability results, does the machine still have
acceptable rotordynamic performance over the full Variables
range of design/operation after its model has been δ = Log decrement (dim)
corrected based on the test results? ζ = Damping ratio (%)
ωd = Damped natural frequency (L
When assessing the accuracy the stability measurements ����⃗1
𝑈 = Baseline unbalance (M-L)
and predictions, there is a key difference with the correlations
����⃗2
𝑈 = Unbalance with verification weight (M-L)
done during an URVT: the measured vibrational response
should not be under-predicted for the URVT, while the ����⃗𝑟
𝑈 = Residual unbalance (M-L)
measured log decrement should not be over-predicted in the ����⃗
𝑈𝑣 = Verification weight unbalance (M-L)
SVT. ���⃗
𝑉1 = Measured baseline vibration (L, L/T or L/T2)
The exact methodology used to correct the model should be ���⃗
𝑉2 = Measured vibration with verification weight (L, L/T
agreed upon, prior to testing, by the purchaser and OEM. Once or L/T2)
again, this is another area where manufacturers and end-users ���⃗
𝑉𝑠 = Vibration due to verification weight alone (L, L/T or
are developing their own methodologies. Pettinato et al. [9] L/T2)
applied two methods to correct the model and determine Acronyms
acceptability of a particular centrifugal compressor design. 1B = First backward rotor mode
One method applied a bias shift using the base stability 1F = First forward rotor mode
measurements, while the second method applied a slope AF = Amplification factor
correction based on measurements that included aerodynamic FEED = Front-end engineering design
excitations within the machine. FLFP = Full load, full pressure
FRF = Frequency response function
SVT Knowledge / Verification
MCS = Maximum continuous speed
Analogous to the knowledge obtained from the URVT, the MDOF = Multiple degree of freedom
stability verification test assesses the rotordynamic predictions’ MIMO = Multiple input, multiple output
accuracy to help verify a design’s stability characteristics MOBAR= Multiple output backward autoregression
beyond the pass/fail nature of the vibration demonstration tests. MRT = Mechanical run test
No nonsynchronous vibrations may be observed during a full OEM = Original equipment manufacturer
load, full pressure test. However, the machine’s stability level OMA = Operational modal analysis
(log decrement) and margin away from instability remain in PEM = Prediction error method
question. For the selected test conditions, the SVT yields some PTC = Power Test Code (ASME)
insights by providing a measurement of the stability level and, SDOF = Single degree of freedom
using this measurement in conjunction with the predictions, an SIMO = Single input, multiple output
estimate of the machine’s stability threshold. SVT = Stability verification test
URVT = Unbalance response verification test
Since we cannot test every situation, we must rely on the
accuracy of the model to design for these other situations. The

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

REFERENCES Practice 684 2nd Edition, August 2005, Sec 2.7.3, pp. 2-63
to 2-68 and Sec. 3.5.2, pp. 3-38 to 3-43.
1. C. Jackson, 1998, “Shop Testing – Is It Worth It?,” Orbit
Magazine, June, pp. 10-15. 15. Axial and Centrifugal Compressors and Expander-
Compressors for Petroleum, Chemical and Gas Industry
2. J. Heagney, 2011, Fundamentals of Project Management,
Services, API Standard 617 8th Edition, September 2014,
AMACON.
Part 1, Paragraph 4.8.2.9.
3. C. Carlson, 2012, Effective FMEA’s: Achieving Safe,
16. Axial and Centrifugal Compressors and Expander-
Reliable, and Economical Products and Processes Using
Compressors for Petroleum, Chemical and Gas Industry
Failure Mode and Effects Analysis, John Wiley & Sons,
Services, API Standard 617 8th Edition, September 2014,
Inc.
Part 1, Paragraph 4.8.2.10.
4. K. E. Atkins and R. X. Perez, 1992, “Assessing Rotor
17. M. A. Miranda and R. F. de Noronha, 2007, “ASME PTC
Stability Using Practical Test Procedures,” Proceedings of
10 Modified Test for Mechanical Assessment of
the Twenty-First Turbomachinery Symposium, Texas
Centrifugal Compressors,” Proceedings of the Thirty-Sixth
A&M University, pp. 151-158.
Turbomachinery Symposium, Texas A&M University, pp.
5. Axial and Centrifugal Compressors and Expander- 87-93.
Compressors for Petroleum, Chemical and Gas Industry
18. Axial and Centrifugal Compressors and Expander-
Services, API Standard 617 8th Edition, September 2014,
Compressors for Petroleum, Chemical and Gas Industry
Part 2, Paragraph 6.3.1.
Services, API Standard 617 8th Edition, September 2014,
6. Axial and Centrifugal Compressors and Expander- Part 1, Paragraphs 4.8.5 & 4.8.6.
Compressors for Petroleum, Chemical and Gas Industry
19. J. C. Nicholas, S. L. Edney, J. A. Kocur and J. F. Hustak,
Services, API Standard 617 8th Edition, September 2014,
1997, “Subtracting Residual Unbalance for Improved Test
Part 1, Paragraph 6.3.3.2.
Stand Vibration Correlation,” Proceedings of the Twenty-
7. Performance Test Code on Compressors and Expanders, Sixth Turbomachinery Symposium, Texas A&M
ASME, PTC 10-1997, Paragraph 3.2.1, p. 11. University, pp. 7-18.
8. Axial and Centrifugal Compressors and Expander- 20. API Standard Paragraphs Rotordynamic Tutorial: Lateral
Compressors for Petroleum, Chemical and Gas Industry Critical Speeds, Unbalance Response, Stability, Train
Services, API Standard 617 8th Edition, September 2014, Torsionals and Rotor Balancing, API Recommended
Part 1, Paragraph 4.8.3 & 4.8.4. Practice 684 2nd Edition, August 2005, Sec 2.9, pp. 2-97 to
2-104.
9. B. C. Pettinato, C. H. Cloud and R. S. Campos, 2010,
“Shop Acceptance Testing of Compressor Rotordynamic 21. Nordmann, R., 1982, “Modal parameter identification and
Stability and Theoretical Correlation,” Proceedings of the sensitivity analysis in rotating machinery,” Proceedings of
39th Turbomachinery Symposium, Texas A&M University, Conference on Rotordynamic Problems in Power Plants,
pp. 31-42. IFToMM, pp. 95-102.
10. API Standard Paragraphs Rotordynamic Tutorial: Lateral 22. Nordmann, R., 1984, “Identification of modal parameters
Critical Speeds, Unbalance Response, Stability, Train of an elastic rotor with oil film bearings,” ASME Journal of
Torsionals and Rotor Balancing, API Recommended Vibration, Acoustics, Stress , and Reliability in Design,
Practice 684, 2nd Edition, August 2005, Paragraph 2.3.8. Vol. 106, pp. 107-112.
11. J. A. Kocur, J. C. Nicholas and C. C. Lee, 2007, 23. Cloud, C. H., Maslen, E. H., and Barrett, L. E., 2009,
“Surveying Tilting Pad Journal Bearing and Gas Labyrinth “Damping ratio estimation techniques for rotordynamic
Seal Coefficients and Their Effects on Rotor Stability,” stability measurements,” ASME Journal of Engineering for
Proceedings of the Thirty-Sixth Turbomachinery Gas Turbines and Power, Vol. 131, No. 1, January, pp.
Symposium, Texas A&M University, pp. 1-10. 012504.
12. C. H. Cloud, E. H. Maslen and L. E. Barrett, 2012, “Rotor 24. Newkirk, B. L., 1924, “Shaft Whipping,” General Electric
Stability Estimation with Competing Tilting Pad Bearing Review, Vol. 27, March, pp. 169-178.
Models,” Mechanical Systems and Signal Processing, Vol.
25. Newkirk, B. L. and Lewis, J. F., 1956, “Oil film whirl—An
29, May, pp. 90-106.
investigation of disturbances due to oil films in journal
13. Axial and Centrifugal Compressors and Expander- bearings,” Transactions of the ASME, Vol. 78, pp. 21-27.
Compressors for Petroleum, Chemical and Gas Industry
26. Hagg, A. C., 1946, “The influence of oil-film journal
Services, API Standard 617 8th Edition, September 2014,
bearings on the stability of rotating machines,” ASME
Part 1, Paragraph 4.8.8.8.
Transactions, Journal of Applied Mechanics, Vol. 68, pp.
14. API Standard Paragraphs Rotordynamic Tutorial: Lateral A211-A220.
Critical Speeds, Unbalance Response, Stability, Train
Torsionals and Rotor Balancing, API Recommended

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station

27. Kanki, H., Fujii, H., Hizume, A., Ichimura, T., and
Yamamoto, T., 1986, “Solving nonsynchronous vibration
problems of large rotating machineries by exciting test in
actual operating condition,” Proceedings of the
International Conference on Rotordynamics, IFToMM, pp.
221-225.
28. Baumann, U., 1999, “Rotordynamic stability tests on high-
pressure radial compressors,” Proceedings of the Twenty-
Eighth Turbomachinery Symposium, Texas A&M
University, pp. 115-122.
29. Moore, J. J., Walker, S. T., and Kuzdzal, M. J., 2002,
“Rotordynamic stability measurement during full-load full-
pressure testing of a 6000 psi re-injection centrifugal
compressor,” Proceedings of the Thirty-First
Turbomachinery Symposium, Texas A&M University, pp.
29-38.
30. Bidaut, Y., Baumann, U., and Al-Harthy, S. M. H., 2009,
“Rotordynamic stability of a 9500 psi reinjection
centrifugal compressor equipped with a hole pattern seal—
measurement versus prediction taking into account the
operational boundary conditions,” Proceedings of the
Thirty-Eighth Turbomachinery Symposium, Texas A&M
University, pp. 251-259.
31. Takahashi, N., Magara, Y., Narita, M., and Miura, H.,
2012, “Rotordynamic evaluation of centrifugal compressor
using electromagnetic exciter,” ASME Journal of
Engineering for Gas Turbines and Power, Vol. 134, pp.
032505.
32. Maia, N. M. M. and Silva, J. M. M., eds., 1997,
Theoretical and Experimental Modal Analysis, Research
Studies Press Ltd..
33. Cloud, C. H., 2007, Stability of Rotors Supported by Tilting
Pad Journal Bearings, PhD dissertation, University of
Virginia.
34. Ewins, D. J., 2000, Modal Testing: Theory, Practice and
Application, Research Studies Press Ltd., 2nd Edition.
35. Pintelon, R. and Schoukens, J., 2001, System
Identification: A Frequency Domain Approach, IEEE
Press.
36. de Jongh, F., 2008, “The Synchronous Rotor Instability
Phenomenon – Morton Effect,” Proceedings of the Thirty-
Seventh Turbomachinery Symposium, Texas A&M
University, pp. 159-167.
37. Guglielmo, A., Mitaritonna, N., Catanzaro, M. and
Libraschi, M., 2014, “Full Load Stability Test on LNG
Compressor”, Proceedings of the ASME Turbo Expo 2014,
Dusseldorf, Germany, No. GT2014-25353.

ACKNOWLEDGEMENTS
The authors wish to thank ExxonMobil for its’ support and
Jim Byrne, Minhui He, Eric Maslen and José Vázquez from
BRG for their helpful suggestions.

Copyright© 2016 by Turbomachinery Laboratory, Texas A&M Engineering Experiment Station