VFD PDF

Table of Contents
ANALYSIS GUIDE FOR

VARIABLE FREQUENCY DRIVE OPERATED CENTRIFUGAL PUMPS
by
Thomas F. Kaiser
Lead Engineering Analyst
Sulzer Pumps (US) Inc.
Portland, Oregon
Richard H. Osman
Principal Product Engineer
Siemens LD-A
Pittsburgh, Pennsylvania
and
Ralph O. Dickau
Senior Engineering Specialist
Enbridge Pipelines
Edmonton, Alberta, Canada
Thomas F. Kaiser has been an Ralph O. Dickau is a Senior Engineering

Engineering Analyst with Sulzer Pumps Specialist with Enbridge Pipelines,
(US) Inc., in Portland, Oregon since 1996, in Edmonton, Alberta, Canada. His
leading the analysis team since 2001. In responsibilities include pumping equipment
this position he is responsible for performing specifications, selection, installation, and
stress, finite element, rotordynamic, and startup. He was team leader for several
vibration analyses for new and installed large pump and motor replacement and
centrifugal pump applications, calculation rerate programs on the pipeline system.
tool development, and seismic qualification Mr. Dickau is also responsible for
analyses. He began his career in 1994 troubleshooting pump and motor operating
with Sulzer Pumps Headquarters in Winterthur, Switzerland, as a problems. He has presented seminars on pump technology for
development engineer working in the field of rotordynamics. other pipeline companies through Enbridge Technology, has
Mr. Kaiser has B.S. and M.S. degrees (Mechanical Engineering) authored several technical papers, and has been with Enbridge for
from the Swiss Federal Institute of Technology (ETH), in 24 years.
Zurich, Switzerland. Mr. Dickau received a B.Sc. degree (Mechanical Engineering,
1978) at the University of Alberta and is a registered Professional
Engineer in the Province of Alberta.
Richard H. Osman is Principal
Product Engineer for Siemens Large
Drives-Applications (formerly Robicon), in ABSTRACT
Pittsburgh, Pennsylvania. He serves as
technical advisor to the company and works The use of variable frequency drives (VFD) in pumping
closely with the product development group applications with variable-duty requirements provides the user
at Robicon. He has previously worked with a variety of benefits, including potentially significant energy
for: Westinghouse Electric Corporation, savings and improved reliability achieved by means of speed
developing a variety of solid-state VFDs; reduction and avoiding part-flow operation. Energy savings are
Robicon Corporation, as Development primarily realized by running the equipment at high levels of
Engineer, where he designed special purpose thyristor DC-drives; efficiency and optimal operating speeds, matching the generated
Manager of AC-Drives Engineering at Robicon; and Technical pump head to the exact system requirements without the use of
Director of Heenan Drives Ltd.. He served for five years as energy consuming control valves. Running pumps at lower
Robicon’s representative to the NEMA adjustable-speed drives operating speeds and avoiding part-flow operation also positively
subcommittee and two years as Chairman. He was Director of influences component life and between maintenance intervals. The
Drives Engineering at Halmar Robicon Group, Vice-President of primary mechanical challenge of any VFD application is the wide
Integrated Product Development, and Senior Vice-President of continuous operating speed range. Excitation frequencies of fixed
Technology for High Voltage Engineering. speed applications miss most natural frequencies of the structure,
Mr. Osman received a BSEE degree (1965) from the Carnegie rotor, etc., and therefore potentially harmful resonance conditions
Institute of Technology. He has written a number of technical often do not occur. This is no longer the case with VFD applications,
papers on VFDs, is a registered Professional Engineer, and a where excitation frequencies become variable and the likelihood of
Senior Member of IEEE. encountering resonance conditions is greatly increased. Problems
81
Table of Contents
82 PROCEEDINGS OF THE TWENTY-FOURTH INTERNATIONAL PUMP USERS SYMPOSIUM • 2008
and failures in pumps and associated systems that are not caused (nonrotating) coordinate system. The corresponding definition of
by resonance are generally not VFD related and are therefore the unbalance force F in US units and SI units is given in Equations
not discussed. (3) and (4), respectively.
The paper gives an overview of medium-voltage VFD
technology as well as the main categories of resonance conditions
of concern with regard to mechanical vibrations of pump/motor
sets. The analytical and experimental identification of resonances
related to lateral rotor, torsional rotor, structural, and acoustic
dynamics are discussed in detail. The applicable set of analyses
and, if necessary, the corresponding appropriate corrective
measures, are designed to help ensure operation free of harmful
resonance conditions and problems caused by excessive
mechanical vibrations.
INTRODUCTION
Continuous operation under resonance conditions may result in
excessive equipment vibrations, reduced in-between maintenance
intervals, and premature equipment failure. Resonance conditions
with centrifugal pump applications can be divided into the four
categories of lateral rotordynamics, torsional rotordynamics, structural
dynamics, and acoustic resonance. Each of these categories requires
its own specific set of analyses and checks allowing up-front
identification of resonance conditions and corresponding corrective Geometric deviations between the individual impeller
action. Resonances may then be avoided, moved (to operating channels create a nonuniform pressure distribution at the
points where the resulting mechanical vibrations are acceptable), impeller outlet, which also rotates with rotor speed (1×). The
or be detuned altogether. These analyses can also help eliminate the resulting radial hydraulic force has much the same effect as
need for expensive factory string tests aimed at investigating mechanical unbalance and is therefore referred to as hydraulic
vibration performance. unbalance. Hydraulic unbalance increases with increasing
A basic understanding of the most commonly applied and flow-rate and usually exceeds mechanical unbalance by factors.
available variable frequency drive (VFD) technology and its rapid Unbalance affects lateral rotor and structural vibrations but not
development over the last few decades is helpful in the assessment torsional rotor vibrations.
of VFD-related vibration problems and the selection of the optimal
Self-Excited Vibration
set of analyses and checks.
The case studies presented in this paper give a detailed Self-excited vibration, also known as rotor instability, is most
illustration of the analysis procedures and methods that can be commonly associated with radial journal bearings, annular
employed in order to successfully identify resonance conditions seals, and hydraulic impeller-casing interaction. Self-excited
of concern with VFD applications. The same methods and tools vibration caused by lightly loaded cylindrical journal
can also be used to study the effect of design modifications bearings/guide bearings in vertical pump application are the
aimed at detuning resonances. The actual analysis work should most common cause of instability in centrifugal pumps. The
be carried out by individuals specifically trained for the task. On corresponding vibration frequency typically lies between 0.40
a broader level, the paper indicates the type of analyses and and 0.50 times running speed (subsynchronous vibration),
checks considered necessary as well as standard analyses that indicating a tangential mean fluid velocity cu inside the tight
may be omitted. The information presented may therefore be bearing clearance per Equation (5). The parameter R denotes the
used as an end-user guideline for selecting/purchasing of rotor radius at bearing location and ␻ is the angular shaft speed.
analysis support from the original equipment manufacturer
(OEM) or engineering consultants.
The rotordynamic software tools used for the case studies are
pump OEM in-house developments. Commercially available
rotordynamic software may be used instead. All structural analyses Pumps with excessively worn annular seals can show the same
were performed applying a general purpose finite element phenomenon with vibration frequencies in the 0.7 to 0.9 times
analysis software. running speed range (also above 1× running speed in case of
tangential fluid entry velocities > ␻R).
EXCITATION SOURCES AND AMPLIFIERS Instabilities are caused by the nonsymmetrical pressure distribution
This section describes the relevant excitation mechanisms and of the vibrating shaft, which creates a force component acting in
amplifiers of mechanical vibrations. the direction of the shaft orbit. This force feeds energy to the rotor
and thus the shaft orbital movement is accelerated. Instability
Mechanical and Hydraulic Unbalance occurs in case the energy put into the rotor exceeds the direct
Mechanical unbalance occurs when the mass centerline of a damping opposing the same vibration.
rotating component does not coincide with the shaft centerline. A Many vertical pump applications show a vibration component at
certain level of dissymmetry of the weight distribution is unavoidable or near 0.5× running speed in their amplitude spectrum (also
in rotating equipment. For the case of two-plane balancing, the referred to as oil whirl or bearing whirl). Instability usually only
unbalance measured in US customary units is defined in Equation occurs in case a structural or lateral rotor natural frequency is at or
(1). The factor K is a balance constant, W is the mass per balance near this 0.5× running speed frequency, changing the oil whirl into
plane (or journal), and N is the rotor speed. The SI unit equivalent an oil whip condition with potentially destructive vibration levels.
definition is shown in Equation (2) with the ISO Balance Quality In case the operating speed is increased after the onset of instability,
Grade G, rotor-mass m, and the angular speed of rotation ␻. The the vibration frequency will typically remain nearly constant,
unbalance force rotates with rotor speed (1×) and is therefore locked into the natural frequency of the structure or rotor as
a sinusoidal function of time when viewed in a stationary indicated in Figure 1.
Table of Contents
ANALYSIS GUIDE FOR 83
The main factors influencing the pressure pulsation magnitude
are the radial gap between the impeller outer diameter and the
volute/diffuser cutwater (B-gap), the percent of best efficient point
(BEP) operation, and the impeller outlet velocity u2.
B-Gap
The radial distance between the impeller vane trailing edge
and the volute cutwater or diffuser vane leading edge (B-gap)
heavily influences the pressure pulsation amplitudes. According to
investigations published in (Guelich and Bolleter, 1992), pressure
pulsation amplitudes decrease on average with a power of (⫺0.77)
on the relative radial gap as illustrated in Equation (7). For example,
a 2 percent B-gap will produce pressure pulsation amplitudes three
times higher than a 9 percent B-gap on an otherwise identical pump.
Figure 1. Waterfall Vibration Plot.
Measures against self-excited vibration include: Percent BEP Operation

In Figure 3, statistical data from 36 measurements of single and
• Reducing the mean tangential velocities in journal bearing and multistage pumps are plotted as dimensionless root mean squared
annular seal tight clearances. This may be achieved by means of
(rms) values for the frequency range of 1.25 to 20 times running
applying rough stator surface finish, swirl breaks at annular seal
speed, which covers vane-pass frequency. Pressure pulsations are
entry, stator-side honeycomb or hole patterns, smooth rotor surface
normalized according to Equations (8) and (9) for US units and SI
units, respectively. ∆PRMS is the rms value of the dimensional
finish, etc.
• Increasing bearing radial load by means of applying additional pressure pulsation measurement, ␳ is the fluid density, and u2 is
(intentional) misalignment in vertical pumps. the fluid velocity at impeller outlet. Flow is normalized as
shown in Equation (10) with Q being the effective flow and QBEP
• Avoiding critical speed situations between 0.5× running speed representing the best-efficiency flow. The curves displayed in
excitation and lateral or structural natural frequencies.
Figure 3 shows the strong dependency of pressure pulsations from
• Restoring design clearances in case of excessively worn the operating point with reference to percent of BEP operation
annular seals. (Guelich and Egger, 1992).
Self-excited vibrations affect lateral rotordynamics and structural

vibrations, not torsional rotordynamics.
Vane-Pass Pressure Pulsations

Vane-pass pressure pulsations are generated by the impingement
of the nonuniform impeller wake flow on the volute cutwater or
diffuser vane tips. Figure 2 depicts the nonuniform fluid velocity
profile at the impeller outlet. These pressure pulsations travel
through the system at the speed of sound of the pumpage. The
frequency of the vane-pass pressure pulsations, acting on the
stator, is proportional to the pump rotational speed N, the
impeller vane count z2, and multiples (n) thereof as illustrated
in Equation (6).
Figure 3. Pressure Pulsations Versus Percent of BEP Operation.
Impeller Outlet Velocity

Experience indicates that pressure pulsations in geometrically
similar pumps roughly increase with the square of the circumferential
speed as shown in Equations (11) and (12) for US units and SI
units, respectively:
Figure 2. Wake Flow at Impeller Outlet.

Table of Contents
The stagnation pressure ∆Pd is a measure of the unsteady

hydrodynamic load acting on the volute or diffuser vane.
Other Influencing Parameters

The linearized equation of motion consists of a stiffness-matrix
Parameters affecting the nonuniformity of the impeller wake K, damping-matrix D, mass-matrix M, and vectors of displacements
(x), velocity (x), and acceleration (x). The complex eigenvalue λX
flow and geometric parameters have an influence on the generation . ..
and amplitude of pressure pulsations. Among these parameters are: is defined in Equation (14). The imaginary component ␻X represents
• Thickness and form of the impeller vane trailing edge. the angular natural frequency. The eigenvalue determines the
corresponding natural frequency f (Equation [15]), modal damping
• Number of impeller vanes and combination of impeller vanes to value D (Equation [16]), and mode shape as shown in Figure 4. A
volute vanes/diffuser vanes (e.g., odd versus even number of mode with negative damping D represents an instable system. The
impeller vanes in double volute type pumps). rotor system is laterally stable when all significant modes provide
• Impeller specific speed. Pressure pulsations generally increase positive modal damping levels.
with increasing specific speed.
• Impeller geometry, particularly vane exit angle and shape of
volute cutwater (blunt versus hydraulically smooth).
• Staggering of impellers along shaft of multistage pumps and
staggering of the two halves of double suction impellers.
Vane pass pressure pulsations primarily affect structural vibrations

(e.g., bearing housing vibrations). A certain pressure pulsation
level will always be present in centrifugal pumps and does not need
to represent a problem. Excessive pressure pulsations can result in
high pump and piping vibrations, particularly in combination with
structural resonance or acoustic resonance. This may result in
vibration levels beyond alarm or shutdown and cause fatigue
failures in auxiliary piping, instrumentation, etc.
Other Excitation Sources

Other excitation sources include broadband hydraulic forces due
to recirculation, cavitation, and forces due to rotating stall.
Amplifiers
Vibration levels usually become excessive when amplified by
resonance. A resonance condition occurs when an excitation
frequency is within a few percent of a relevant natural frequency. In
that condition, the excitation force is acting again once the vibrating
component has come full cycle after the last “impact” by the force.
The excitation force and the vibration are synchronized, and the
vibration amplitude increases until limited by nonlinear effects.
With regard to mechanical vibrations in VFD operated centrifugal
pumps, resonance conditions can be divided into four categories:
structural resonance and torsional rotor resonance are typically
lowly damped and are likely to result in high levels of mechanical
vibration when properly excited. Lateral rotor resonances are in
some cases highly damped and operation on or near such a condition
(“critical speed”) may be perfectly acceptable. Acoustic resonance
conditions, amplifying mechanical vibrations via amplified pressure Figure 4. Lateral Mode Shapes and Mechanical Model.
pulsations, are usually only lowly damped. The various resonance
categories are discussed in detail in the following sections. The evaluation of the lateral rotordynamic behavior can either be
done by solving the homogeneous equation of motion (eigenvalue
LATERAL ROTORDYNAMICS calculation) or by specifying a set of excitation forces and subsequent
solution of the nonhomogeneous equation of motion (forced
General response analysis).
The damped lateral rotordynamic behavior of a centrifugal The eigenvalue approach and evaluation of results applying a
pump rotor is determined by the rotor geometry, the rotor mass and combined frequency-versus-damping-ratio criterion is further
inertia, and the interaction forces occurring between the rotor and discussed in this paper. This approach is less ambiguous compared
journal bearings, annular seals, and casing. Impeller wear rings, to a forced response analysis because it avoids the subjective
close-clearance bushings, and balance pistons are typical examples process of determining and applying excitation forces (typically a
of annular seals. Casing interaction occurs at impeller location, combination of mechanical and hydraulic unbalance loads).
between the wear rings in case of a closed impeller design, and is The results of a damped lateral rotordynamic analysis are best
generally destabilizing. These interaction forces are nonlinear but presented in the form of a Campbell diagram as illustrated in
may be linearized around a particular static rotor equilibrium Figure 5, plotting the natural frequencies and modal damping
position. Interaction forces vary with operating speed, pumpage factors versus pump operating speed. The intersection between
specific gravity and viscosity, load, state of wear, etc. Solving the a speed-dependent natural frequency line and the synchronous
linearized homogeneous Equation of Motion (13) results in a set speed excitation line is called a critical speed and represents a
of eigenvalues. resonance condition.
Table of Contents
the modal-damping-versus-frequency-separation acceptance
criterion illustrated in Figure 6 as binding. With a few exceptions
not discussed in this paper, rotor designs in violation should be
modified to meet this acceptance criterion. Design modifications
aimed at improving lateral rotordynamic stability can be divided in
two categories. A first category aims at increasing the frequency
separation margin between lateral modes and synchronous excitation
speed by means of increasing the rotor stiffness (Kxx) or reducing
the rotor mass (M). A second category of modifications intends to
increase modal damping.
Design modifications aimed at increasing lateral rotor
natural frequencies:
• Decreasing coupling overhung length and/or coupling weight in
case of overhung dominated modes (Kxx, M)
• Changing of impeller material from steel to aluminum (M)
Figure 5. Campbell Diagram.
• Increasing shaft size (Kxx)
A widely used eigenvalue acceptance criterion is defined in • Decreasing between-bearing span (Kxx)
Annex I of the API 610 standard (Eighth Edition, 1995; Ninth
Edition, 2003; Tenth Edition, 2004) and the ISO 13709 standard
• Tightening or restoring annular seal clearances (Kxx)
(2003), respectively. The combined frequency-versus-damping • Eliminating stator-side serrations applied to reduce leakage (Kxx)
criterion, depicted in Figure 6, is applied to each of the calculated • Applying stator-side circumferential grooves in balance pistons
lateral modes, limiting the evaluation to modes within a natural and center and throttle bushings. This reduces the adverse effect of
frequency range of zero to 2.2 times running speed. This frequency piston tilting onto the direct radial annular seal stiffness (Kxx).
range covers the most typical and significant rotor lateral excitation
forces including subsynchronous excitation, mechanical and • Changing from inline to back-to-back configuration, which
hydraulic unbalance and misalignment: reduces bearing spans and also adds damping at the center of the
pump (Kxx)
• Subsynchronous excitation (journal bearings): 0.4… 0.5× running
speed (typical) Design modifications aimed at increasing modal damping:
• Subsynchronous excitation (annular seals): 0.7… 0.9× running • Applying stator-side swirl breaks at the entrance of
speed (typical) impeller eye wear rings. This reduces the circumferential inlet
swirl, which in turn reduces destabilizing annular seal
• Rubbing: multiples of 0.5× cross-coupled stiffness.
• Mechanical unbalance: 1× running speed • Optimizing the journal bearing design. Journal bearings with
• Hydraulic unbalance: 1× running speed length-over-diameter ratios above one should be avoided. The
• Misalignment: 1× and 2× running speed destabilizing effect of cross-coupled journal bearing stiffness can
be reduced or eliminated by switching from cylindrical bearings to
multilobe or tilting-pad designs.
• Loading of vertical pump line-shaft bearings by means of
applying intentional misalignment between bearings and rotor
• Applying rough annular seal stator surface finish, stator-side
honeycomb or hole patterns, smooth rotor surface finish
In case a fixed speed centrifugal pump is converted to VFD

operation, the continuous VFD operating-speed range may already
be sufficiently analyzed and covered by the original fixed-speed
lateral rotordynamic analysis. Fixed-speed lateral analyses should
be carefully reviewed on a case-by-case basis before deciding
whether a new lateral analysis for the VFD operated application is
necessary or not.
Case Study—Standard Lateral

Figure 6. API 610 Lateral Rotordynamic Acceptance Criterion. Rotordynamic Analysis Procedure
The dynamic lateral stiffness and damping levels provided by
The API 610 (2004) lateral evaluation criterion also requires the journal bearing fluid film depend on the stiffness of the bearing
consideration of operating speeds outside of the defined pump housing/support structure itself. With most horizontal pumps, the
continuous operating speed range. A speed range of 25 percent of lowest bearing housing natural frequency is well above the first few
minimum continuous speed to 125 percent of maximum continuous lateral rotor bending mode natural frequencies. In these cases, the
operating speed needs to be investigated. corresponding journal bearing support stiffness can be considered
While, for cases in violation, the cited API standards allow as near-rigid and constant over the entire speed range of
proof of acceptability by means performing additional unbalance concern. The situation is entirely different with most vertical pump
forced response analysis, this approach is not recommended for applications. Vertical pumps are typically structurally flexible and
obvious reasons. An unbalance forced response analysis applies significant structural modes may appear at, near, or below
excitation sources at synchronous speed (1×), which cannot excite operating speed. This requires calculation of the dynamic bearing
subsynchronous modes. It is therefore recommended to consider support stiffness by means of performing harmonic response
Table of Contents
analyses of the entire pump and motor structure. Alternatively a

rotordynamic code capable of performing a combined rotor-structure
analysis may be employed.
The process of vertical pump lateral rotordynamic analysis is
explained using the example of a water transport booster pump.
The vertical turbine is VFD operated with a continuous operating
speed range of 257 to 514 rpm and driven by a 900 horsepower
asynchronous electric motor. The pump featuring a 42 inch
diameter discharge nozzle and 48 inch diameter columns generates
45 feet of differential head, moving 55,555 gpm while running at
514 rpm rated speed.
The forced harmonic response analyses were performed
applying a general purpose finite element software program using
beam elements for modeling of motor and pump structure. The
inertia and mass effects of the rotating element were considered
applying lumped mass elements attached to the structure at
bearing holder and impeller locations. Internal and surrounding
water effects were considered as additional mass effects applied
to the beam elements. The stiffness effect of the receiver-can
Figure 9. Amplitude Response at Upper Motor Bearing.
flange and gusseting was determined in separate static finite
element (FE) analyses and considered via spring elements in the
Figure 10 shows a Bode plot of the dynamic response at the
forced harmonic response model. Figure 7 depicts the calculated
suction-bell bearing due to harmonic excitation at the same location.
static receiver-can deflection due to unit-moment loading in the
Amplitude responses at approximately 2.5 Hz and 20.5 Hz correspond
horizontal direction.
to the first and third structural mode natural frequencies.
Figure 7. Calculated Receiver Can Flange Deflection.
Figure 8 shows the first three structural modes of the pump and
motor structure as a result of a modal structural FE analysis. The
modes occur at 2.6 Hz (156 cpm, first column mode), 14.2 Hz (852
cpm, first aboveground mode), and 20.5 Hz (1230 cpm, second
column mode), respectively. The dynamic bearing support stiffnesses
required for the lateral rotordynamic analysis are calculated in a
series of forced harmonic response analyses. Each of these
analyses applied a dynamic unit-load at a single bearing location
and varied the frequency of this harmonic load within a specified
frequency-range. Figure 9 illustrates the dynamic displacement
response at the upper motor bearing due to harmonic excitation
(unit force in lateral direction acting at the same location). The
response indicates an amplitude response peak at approximately 15
Hz, which corresponds to the second structural mode showing large
modal displacement in the aboveground portion of the machine.
Figure 10. Bode Plot of Response at Suction-Bell.
The calculated displacement curves UXX(f) are inverted to

derive the various dynamic stiffness curves KXX(f) as indicated in
Equation (17) and displayed in Figure 11.
Figure 8. First Three Pump Structural Mode Shapes.

Table of Contents
Case Study—Rotor Instability
Vertical pump bearings and bushings are typically only lightly
loaded, which may cause rotor instability problems. The phenomenon
is explained on the example of a vertical mixer pump intended
for VFD operation in slurry mixing service. The single-stage,
quad-volute pump, driven by a 200 horsepower asynchronous
motor, underwent an endurance test at the OEM factory. The pump
design features seven columns of 16 inch diameter, supporting the
2 inch diameter shaft at seven line-shaft bearing locations. The
bearing material is nickel-impregnated carbon mounted in bearing
holders that were initially loosely positioned between column
flanges. At startup, shaft vibration readings taken at the bottom seal
location showed two distinct peaks of both approximately 1.2 mils
peak-to-peak at 1× and 0.5× running speed frequency. The 50
percent running speed frequency represents “bearing whirl.” The
bearing whirl frequency decreased almost linearly over time
starting from 18.5 Hz, indicating wearing out carbon bushings. A
post-test inspection revealed that the diametrical bushing clearances
increased from an initial 7 mils design clearance to as much as 400
Figure 11. Dynamic Bearing Support Stiffness KXX(f) at Suction Bell. mils. The failed test triggered analyses and modal testing. Lowly
damped lateral rotor modes at approximately 50 percent running
The subsequent lateral rotordynamic analysis applied unbalance speed and structural column modes at 0.2, 4.5, 14.0, and 29.0 Hz
loads based on an ISO 1940-1 Grade G2.5 (equivalent to an API were calculated and support-structure modes at 10 and 26 Hz
balance grade of 15 W/N; IRD Balancing, 2007), evaluated at rated were measured by means of modal testing. Figure 14 illustrates
operating speed. The corresponding unbalance forces in the lateral the frequency and amplitude of column vibrations filtered to
forced response analysis are varied proportional to the square of synchronous speed and bearing whirl frequency, respectively.
the operating speed. The significant shaft response peaks lay The bearing whirl frequency decreases linearly with time,
outside of the continuous VFD operating speed range and are obviously because of continuous wear of the carbon bushings.
therefore not an operational concern. However, the analysis clearly After approximately 35 hours, the bearing whirl frequency started
proves that shaft amplitude response peaks can occur at the natural to lock into the third column natural frequency for a few hours
frequencies of the structure (refer to Figures 12 and 13 for details). (bearing whip condition), resulting in amplified vibration levels.
Figure 12. Lateral Shaft Response Amplitude at Upper Bearing.
Figure 14: Column Vibration and Natural Frequencies.
The pump was successfully retested after implementation of

design changes, which included fixation of the initially loose
Figure 13. Lateral Shaft Response Amplitude at Suction Bell. bearing holders, built-in misalignment between the bearing holders
Table of Contents
and the shaft, as well as structural modification of the pump • Some literature mentions 1× running speed as a torsional excitation
support structure. The successful retest did not show any significant frequency. However, shaft trains without gearboxes usually exhibit
bushing wear and the bearing whirl frequency remained constant only negligible 1× torque harmonics.
throughout the test.
• Shaft trains with a gearbox have additional steady-state torsional
Conclusions excitation sources that occur at 1× running speed (pitch cycle
runout), gear-mesh frequency (based on the number of gear teeth),
Horizontal Pumps and pinion-mesh frequency (based on the number of pinion teeth).
When converting a fixed speed application to VFD operation,
the original lateral rotordynamic analysis may already sufficiently The applicability of the above described excitation sources with
cover the new application. However, a careful review of the results regard to VFD operated shaft trains is further discussed.
of the original analysis and the new operating conditions is Similar to the recommended procedure for lateral rotordynamic
required on a case-by-case basis. The speed range above the analyses, the torsional rotordynamic behavior of a shaft train is
original fixed operating speed requires special attention because of best analyzed by means of a natural frequency analysis, subsequent
the increased power absorption and the increased likelihood of evaluation of resonance conditions, and, if necessary, performing of
encountering critical speeds, for example critical speeds with transient and/or steady-state forced response analyses. The standard
coupling overhung dominated modes. The primary excitation torsional natural frequency analysis is an undamped analysis. An
mechanism of concern is synchronous speed excitation due to undamped natural frequency analysis applies because virtually all
mechanical and hydraulic unbalance. API 610 Eighth through shaft trains only provide low levels of torsional damping. The
Tenth Editions (1995, 2003, 2004), Annex I, defines a reasonable damped and the corresponding undamped natural frequency do not
lateral rotordynamic acceptance criterion. significantly differ in case of low modal damping. Equation (18)
explains the relationship between the damped and undamped natural
Vertical Pumps frequency of a single degree of freedom system. In the example of
a 50.0 Hz undamped natural frequency and a 3 percent critical
A significant coupling between the dynamic behavior of the damping ratio, the damped natural frequency drops to only 49.98 Hz.
structure and lateral rotordynamics is often the case. The frequency
of lateral modes can coincide with structural natural frequencies,
resulting in amplified shaft vibrations. The correct assessment of a
vertical pump lateral rotordynamic behavior requires a combined
lateral-structural analysis covering the entire continuous VFD The results of a torsional natural frequency analysis are again
operating speed range. Significant structural natural frequencies best presented in the form of a Campbell diagram, plotting the
at 1× and 0.5× running speed should be avoided. The primary speed-independent natural frequency lines together with potential
excitation mechanisms of concern are synchronous speed excitation excitation frequencies. Forced response analyses apply forcing
(unbalance) and subsynchronous speed excitation around 50 functions at their corresponding load-induction locations. Startup
percent of running speed (bearing whirl). and short circuit excitation, for example, induce their load at the
In case a resonance condition cannot be detuned, it is also motor core. The outcome of a forced response analysis may be a
possible to lock out/program out a specific operating speed range plot of torsional stress along the entire shaft at a set constant
from continuous operation. operating speed. The forced response analysis can also be used to
TORSIONAL ROTORDYNAMICS calculate the torsional stress at a specific shaft location of interest,
varying the operating speed (frequency-sweep). In any case, the
General stress results can be further used in fatigue analyses in order to
evaluate the acceptability of the investigated resonance
Torsional vibration problems usually involve resonance conditions.
condition(s).
Torsional resonance occurs when a train torsional natural
frequency coincides with or is close to an applicable excitation Case Study—Standard Torsional
frequency. Due to the substantial torsional stiffness of both rigid
Rotordynamic Analysis Procedure
and flexible disc type couplings, the entire shaft train needs to
be analyzed. The torsional rotordynamic analysis procedure is demonstrated on
Shaft trains driven by electric motors are subject to transient and the example of an axially-split, horizontal, multistage barrel-type
steady-state excitation. There are two transient excitation sources pump in crude oil transfer service. The 10×10×13.5 CP 10 stage pump
of concern, which are startup and short-circuit fault condition. is VFD operated with a continuous operating speed range of 2000 to
Startup occurs at line frequency and short circuit excitation 3755 rpm. Figure 15 shows an isometric view of the pump assembly.
functions include one- and two-times line frequency components.
Potential steady-state excitation sources are:
• The 2× running speed excitation represents an interaction
between lateral and torsional vibration. In case of an elliptical,
i.e., noncircular, shaft orbit (1× lateral shaft vibration), the
forces necessary to keep the shaft on its orbit result in an
alternating moment acting on the shaft. This moment changes its
amplitude twice per revolution, and therefore represents a
dynamic torsional excitation source acting at 2× running
speed frequency.
• Vane-pass frequency excitation is another potential mechanism
but it was found to be significant only in case of impellers with a
low number of vanes and relatively thick vanes as applied in
sewage pumps.
• VFD operated pump trains are subject to VFD torque harmonics,
often occurring at multiples of 6× feed-frequency. Figure 15. Isometric View of Pump Assembly.
Table of Contents
The results of the natural frequency analysis are presented in the mode. The calculated maximum torsional stress is approximately 4.5
Campbell diagram displayed in Figure 16. Mode shapes and the psi (0.03 MPa). Stress responses as low as these are normal for torque
mechanical model are shown in Figure 17. The Campbell diagram harmonic excitation from modern, state-of-the-art VFDs.
includes 2× running speed excitation and the first three VFD
torque harmonic excitations at 6×, 12×, and 18× running speed.
Figure 18. Torsional Stress Response.
Steady-State VFD Torque Harmonic Excitation

Resonance conditions between VFD torque harmonics and
torsional modes are often unavoidable because of the large VFD
operating speed range. It is technically not feasible to detune these
types of critical speeds. The complete review of 43 torsional analyses
of VFD operated single and multistage horizontal centrifugal pumps
indicated a maximum pump shaft stress due to resonance with
VFD torque harmonics of 254 psi (1.75 MPa). The average shaft
stress response was 79 psi (0.54 MPa). The above-mentioned
stresses do not consider any stress concentration factors, which
would be a factor of approximately four at the most.
Once started and running in equilibrium on the utility, polyphase
Figure 16. Torsional Campbell Diagram. motors produce very little torque ripple (torque harmonics) at the
shaft. This assumes that the utility has negligible distortion and
unbalance, which is not always the case. Torque ripple increases
when a motor is VFD operated. When a perfect/ideal polyphase
motor is excited by polyphase sinusoidal currents, the angular
velocity of the stator magneto-motive force (MMF) will be
perfectly constant. The flux is also sinusoidally distributed around
the gap, and thus the angle between flux and MMF remains
constant and the torque is therefore a constant also. There are two
sources of torque ripple, which are nonsinusoidal currents from
the VFD, and motor winding effects. The motor windings are
discrete coils in discrete slots, so it is not possible to get a perfectly
distributed winding capable of producing a perfectly sinusoidal
airgap flux distributed around the gap. Generally these effects are
small, leading to motor-created torque ripple in the low single
percent range of rated torque. Some motor manufacturers even
deny that the motor can produce torque ripple by itself when
excited by pure sinusoidal current.
Any practical VFD powerful enough to run a pump cannot
operate as a linear amplifier, but must resort to simple “on” or
“off ” states of the power electronic’s switches. Because of this,
Figure 17. Torsional Mode Shapes. there will always be harmonics present in the output voltage and
current of the VFD, which is applied to the motor. Due to the
The natural frequency analysis revealed multiple critical speed three-pole output structure, the harmonics are all odd, with those
situations between torsional modes and steady-state VFD torque divisible by three being absent (a pole is the power electronic’s
harmonics. These resonance conditions have been further investigated block that creates the output to one phase of the motor). The lower
in damped forced response analyses applying torque excitation order harmonics (five, seven, 11, and 13) of the fundamental
functions provided by the VFD manufacturer. The applied dynamic output frequency cause the stator MMF angular velocity to
torque was 2.0 percent (sixth harmonic), 1.5 percent (12th), 2.8 percent fluctuate slightly. That causes the angle between flux and MMF to
(18th) and 1.4 percent (24th) of static torque. Figure 18 plots the vary with a characteristic pulsation frequency. The fifth and
calculated torsional stress due to 6× VFD torque harmonic excitation seventh current harmonics cause a torque ripple six-times the
at 1818 rpm operating speed, exciting the 181.8 Hz second torsional output frequency, and 11th and 13th harmonics cause 12th order
Table of Contents
harmonic torque ripple and so on. Of course the amplitude of the percentages of the corresponding static torque. The readings indicate
torque ripple is proportional to the amplitude of the current that the 1×, 2×, and 7× (vane-pass) dynamic torque values are mostly
harmonics. The poorest performing VFD in this respect is the less than 1 percent of static torque. In case of a 1 percent, one-times
load commutated inverter (LCI). This VFD creates a smooth direct dynamic torque applied at the first torsional natural frequency, the
current (DC) link current and switches it into the stator windings in calculated shaft stress at pump coupling hub location is only 116 psi
a prescribed sequence. The quasi-square wave current causes the (0.8 MPa). Conservatively only 1 percent of modal damping was
stator MMF to advance in steps of 60 degrees electrical at each applied in this steady-state forced response analysis. The same
commutation (a commutation is the transfer of current from one dynamic torque measurements did not indicate any significant peaks
power switch to the next one in sequence). An LCI with two parallel at 6× VFD torque harmonic frequency and multiples thereof.
circuits operating into a dual winding motor (12-pulse output) creates
3 to 8 percent torque ripple. Another popular circuit, the neutral point
clamped (NPC) circuit has five voltage levels from line-to-line and
uses pulse-width modulation (PWM) techniques. Depending on the
frequency of switching, the current is less distorted and the torque
ripple decreases into the vicinity of 1 percent to 3 percent. A third
architecture, the multilevel series-cell circuits, can further reduce the
current distortion such that the torque ripple decreases to 1 percent
of motor base rating. Multilevel series-cell circuits also apply PWM
technology. Figure 19 displays voltage waveforms produced by
LCI, NPC, and multilevel series-cell (PWM) VFD architectures. For
the LCI drive system, it is usual practice to perform a torsional
rotordynamic analysis of the drive train, investigating critical speeds
with LCI-VFD torque harmonics.
Figure 20. Dynamic Torque Measurements.
Startup and Short-Circuit Transient Excitations

While steady-state excitation sources may, in rare cases, result in
torsional problems, the primary concern with fixed-speed units lies
with transient excitations (startup, bus-transfer, and short-circuit).
Under resonance conditions, the dynamic torque levels associated
with (repeated) transient excitation events can result in excessive
shaft stresses, shaft breakage, or coupling failures. The maximum
dynamic stress amplitude in case of a startup or short circuit event
can be in excess of 10 times of maximum static torque (>10 per
unit). Torque in electric machines is often expressed as a per
unit (PU) value, derived by dividing the actual torque by the
corresponding rated (static) torque.
Startup
In contrast to direct-online (DOL) operation, VFDs always
incorporate current regulation and limiting functions. During
starting, the VFD output frequency starts at a very low value and
slowly increases, while the current is controlled to one PU or less.
Figure 19. Voltage Waveforms Generated by Different VFD The starting sequence assures that the machine flux is nearly
Architectures. nominal as torque is generated. Therefore the slip is always a small
value as the machine accelerates. This means that the torque is
The main technique to reduce torque ripple is to decrease the much smoother compared to direct-online starting. On the other
harmonic content of the output current. That is generally done by hand, when an induction machine is directly connected to the
increasing the PWM switching frequency, or adding additional utility (no VFD), there is a large inrush current, on the order of 5
voltage levels to the output voltage, or both. For the cases where this PU. Initially, the rotor flux is zero, but the stator flux assumes its
is not possible, application of filters is an alternative option. It is normal amplitude and angular velocity as the utility voltage is
common practice to equip VFDs with an L-C (inductor-capacitor) applied. So, due to the large currents, there are very large torque
filter on the output to attenuate some harmonic currents in order to pulsations as the rotor flux slips by the stator flux until equilibrium
reduce motor torque ripples. This technique is used for LCI and is attained. The key difference in the VFD operation is that it can
NPC VFDs. The disadvantage is that the filter causes extra expense, adjust both frequency and voltage simultaneously, thus allowing the
increased size, and poorer efficiency. For drive systems with high current and the slip to be carefully controlled during all situations.
sensitivity to torque ripple, the multilevel series cell designs offer It is worthwhile to note that thyristor-based solid-state starters can
an advantage. control current during starting but not frequency, so they are not
able to produce nearly as smooth a start as a VFD.
One-Times Running Speed
and Vane-Pass Frequency Excitation Bus Transfer
Dynamic torque measurements taken at the coupling spacer of the A particularly bad event for transient torques is bus transfer. In
previously described barrel-type transfer pump are presented some applications, the motor is switched rapidly from one utility
in Figure 20. The dynamic torque amplitude data are shown as source to another. During the open interval, the motor flux does not
Table of Contents
decay much, but the rotor position begins to fall behind where the STRUCTURAL DYNAMICS
stator flux will be when reconnected. If reconnection occurs at the
wrong time, the inrush currents react with already present flux to General
create a very, very large torque transient. This must be avoided. If The correct use of analytical and experimental techniques is
a VFD is present, it will trip when the utility voltage goes to zero, essential to identifying and solving structural resonance
then smoothly restart when the voltage is restored. problems. Most centrifugal pumps and their supporting system
are too complex to be analyzed with simplified hand-calculation
Short-Circuit methods. These simple techniques are not discussed in this paper.
One feature common to all VFDs (both medium voltage [MV] The presented case studies apply finite element analysis (FEA),
and low voltage [LV]) is a means of limiting short-circuit current the most popular analytical method in use. When choosing the
at the output. This is included to prevent a short-circuit in the motor appropriate modeling technique and combined with the
or its cables from damaging the VFD. There are a number of ways necessary level of detail, including application of the correct
to accomplish this. In the current-fed circuits (LCI and symmetrical boundary conditions, FEA can be highly accurate and useful in
gate controlled thyristor [SGCT] based), the direct current link identifying and resolving resonance conditions. This section also
filter is an inductor that filters the DC link current. If an external discusses modal analysis as an experimental technique used to
short-circuit occurs, the fault current is strictly limited by the identify structural natural frequencies, mode shapes, and
inductor in the short-term, allowing time to phase back the line-side recording of operational deflection shapes (ODS). The most
converter and thus extinguish the current. The voltage-fed drives accurate and useful results are often obtained when combining
have a more difficult problem, as the DC link capacitor will try to analytical and experimental methods.
discharge through the output devices and into the short circuit. In Horizontal pump structural vibration problems primarily occur
the case of those VFDs using insulated gate biploar transistors at 1× rotational frequency and at vane-pass frequency. Vibration
(IGBTs) the devices are fast enough to turn off when the overcurrent problems at rotational speed frequency typically involve support
is detected, before any damage is done. Modern IGBTs are structure (baseplate, etc.) modes combined with rigid-body motion
inherently current limited to about 3 to 4 PU current, and thus will of the pump casing. The second problem set, with vibrations
limit the current even if an overcurrent function does not occur. occurring at vane pass frequency, usually involves bearing
The voltage-fed drives using IGCTs have a fast detector to initiate housings and sometimes auxiliary piping. Vertical pump structural
a turn-off of the devices if the current gets too high. There is also vibration issues typically occur at running speed frequency or at
a small inductor to limit the rate of rise of current to give the subsynchronous frequencies, often around 50 percent of rotational
detector time to function. In all cases, the response time is short, speed. In most cases structural vibrations do not become excessive
from tens to a hundred microseconds. Accidental short circuits unless amplified by resonance. Structural resonance occurs
in the motor terminal box are so common that VFDs must have when the natural frequency of a structure coincides with an
short-circuit protection. From the motor point of view, if something applicable excitation frequency like bearing whirl/instability
fails in the VFD, such as a shorted device, the universal response (0.5×), mechanical and hydraulic unbalance (1×), or pressure
of all VFDs is to turn off all the switching devices, disconnecting pulsations at vane-pass frequency.
the motor so it does not see a short circuit on its terminals. A single Structural modes are only lowly damped. Resonance conditions
shorted device can be prevented from causing large fault currents are therefore likely to result in excessive structural vibration
since the other devices can successfully interrupt the circuit. levels if properly excited. In case of VFD applications with wide
operating speed ranges, chances for resonance conditions to be
Conclusions present are high. It is recommended to predict resonance conditions
Drive protection and slow-start characteristics of VFDs effectively by means of FEA or modal testing before the pump is installed in
and reliably prevent high torsional stress situations in case of short the field. Identified resonance conditions should be detuned or
circuit fault conditions and motor startup. These worst case torsional avoided. A first case study explains the process of
excitation mechanisms are therefore eliminated when applying identifying and resolving a horizontal pump baseplate structural
a VFD. With the exception of LCI-type VFDs, the steady-state resonance problem.
torque harmonic excitation levels induced by state-of-the-art VFD
Baseplate Vibration Problems
technology results in low and acceptable torsional shaft stresses.
Other potential torsional excitation mechanisms such as 1× and 2× A multistage, barrel-type 6×8×10.5 CP 13 stage pump
running speed and vane-pass frequency excitation also generally operating at 3580 rpm (59.7 cpm) fixed speed in boiler feed-water
result in acceptably low torsional stresses. service, exhibited excessive horizontal bearing housing and
It is therefore suggested to consider shaft trains supplied with casing vibrations. Unfiltered peak vibration levels in the
state-of-the-art VFD technology as torsionally safe and to omit horizontal direction of 1.80 in/sec on the inboard and 1.38 in/sec
torsional rotordynamic analysis on such shaft trains. on the outboard bearing housing were measured. The corresponding
The only exceptions from this rule are LCI-type VFDs where a peak vibration levels filtered to 1× rotational frequency were
steady-state torsional analysis investigating the effect of VFD torque 1.60 in/sec and 1.13 in/sec, respectively. Pump vibrations in the
harmonics is recommended and, possibly, shaft trains including vertical direction as well as all motor structural vibrations were
gearboxes. A standard torsional rotordynamic acceptance criterion acceptably low. A modal finite element analysis simulating a
requires a minimum of 10 percent frequency separation margin grouted baseplate and rigid pump-casing was performed. Figure
between torsional modes and applicable excitation frequencies. In case 21 presents a plot of a calculated baseplate mode shape with a
of violation of this criterion, detuning of the natural frequency of 62.2 Hz natural frequency. The pump-side is structurally entirely
concern is an option for fixed speed applications only. There, the most decoupled from the motor-side because of the grouted condition
commonly applied detuning method implements changes to the of the baseplate. The mode shape characterized by inboard and
coupling torsional stiffness, designed at providing the desired 10 outboard pedestals moving in phase with each other was
percent separation margin. In case of VFD applications with wide confirmed by modal testing. An impact test performed on
continuous operating speed ranges, critical speed situations are both bearing housings revealed amplitude peaks at 62.5 Hz,
virtually unavoidable, and it is recommended to prove the acceptability confirming structural resonance as the root cause of the
of the rotor design by means of forced response analysis investigating vibration problem. The result of an impact test performed on the
the resonance conditions of concern. Chances are high that the inboard bearing housing in the horizontal direction is presented
steady-state resonance conditions produce acceptably low stresses. in Figure 22.
Table of Contents
Figure 23. Structurally Modified Baseplate and CP Pump.
Impact tests performed after the modifications indicated that the

Figure 21. Baseplate Mode Shape. baseplate natural frequency of concern increased by 8 percent to
67.5 Hz. Unfiltered peak vibration levels dropped by approximately
90 percent to 0.177 in/sec on the inboard bearing housing and
0.112 in/sec on the outboard bearing housing, respectively. Pump
vibrations in the vertical direction and all motor vibrations
remained low.
The identification of baseplate structural resonances applying
FEA is fairly easy and straightforward. The major advantage of
the FEA approach is the possibility to perform a reasonably
accurate, low-cost analysis in the early phases of a project. Design
changes can be implemented before the actual structure is built.
Another advantage is the ability to test out design variations and
thus optimize the baseplate design on the computer, avoiding
modifications of the actually built structure. The most accurate
results are usually achieved when combining FEA with modal
testing. This approach allows tuning of the FE model to the actually
measured natural frequencies and comparison of mode shapes.
Obviously a combination of methods is only possible in case the
structure is already built and installed.
Not all baseplate structural resonance problems should be
overtuned, i.e., detuned by means of reinforcing the structure. In
some cases it may be more practical to undertune by adding weight
(e.g., filling pedestals with sand or steel pellets, replacing an
existing bearing housing with a heavier one, etc.) or cautiously
weakening the structure.
Experimental Testing
Two experimental testing methods are shortly discussed. Modal
testing allows identification of structural natural frequencies and
Figure 22. Bearing Housing Impact Test. mode shapes. This type of testing is performed on the standstill
machine. The only forces acting are the ones induced by an
The resonance problem could have been resolved by means instrumented hammer or by a shaker. A hammer impact excites the
of reducing the excitation forces or by detuning the structural structure over a wide frequency range. This allows measurement of
resonance. The first approach was not recommended in this natural frequencies within a desired frequency range. A shaker will
case because of the acceptably low vibration levels measured in excite the structure at a single, specified frequency and is typically
the vertical direction. In case of a structural resonance with a applied in a frequency-sweep mode, scanning the structure for
supersynchronous natural frequency, detuning is best done by structural resonances. The other experimental testing method,
increasing the stiffness and therefore increasing the structural operational deflection shape, records structural motion forced by
natural frequency. Various detuning options were analyzed and operating loads. ODS measurements illustrate and quantify the
optimized with FEA. The following is a list of implemented parts of the structure that are really moving at a particular
design modifications, depicted in Figure 23: frequency due to effectively present excitation. Refer to
• Extended the existing C 6 inch × 8.2 lb channels down to the Richardson (1997) for further information on ODS.
deckplate and added further support in the axial direction by welding Impact testing is a popular, easy to perform modal testing
on a triangular shaped gusset plate. The two existing C-channels method that allows measuring of structural natural frequencies.
connect the individual pedestals in the horizontal direction. It can be done quickly and rather inexpensively applying an
instrumented hammer, a dual-channel fast Fourier transform
• Reinforced each pedestal with two short-height outside (FFT) analyzer with postprocessing software, and an accelerometer
structural rips probe. Structural resonances may be identified by means of a
• Added additional side-rail gusset plates single-channel impact test measuring the response of structure
Table of Contents
to a nearby impact. However, the positive identification of a Bearing Housing Vibration Problems
resonance condition requires the use of a dual-channel FFT Horizontal pump bearing housing natural frequencies are
analyzer. One channel is used to record the impact force and the typically well above synchronous speed and are therefore not
second channel records the response, typically measured by an excitable by 1× unbalance excitation. Bearing housing vibration
accelerometer probe mounted to the structure under investigation. problems often occur at vane pass frequency or multiples thereof.
The three individual diagrams in Figure 22 represent amplitude As usual with resonance problems, resolution can be achieved by
response, cross-channel phase, and cross-channel coherence reducing the excitation source or by detuning the resonance
plots. A resonance condition is not just defined by a peak in the condition. Practical means of reducing the excitation source are
amplitude response spectrum but also by a phase shift. For a described in the previous section excitation sources and amplifiers.
single degree of freedom system, the resonance frequency This section focuses on the identification of bearing housing
(natural frequency of the system) is accompanied by a 180 natural frequencies and applicable detuning methods. It is possible
degree phase shift. In case of real structures with multiple to calculate bearing housing natural frequencies using FEA but the
modes, the phase shift will be less than 180 degrees but still process is cumbersome and time-intensive. The casing stiffness
identifiable. Coherence is a measure of how much of the and the rotor mass acting on the bearing both influence bearing
response energy originates from the impact. Valid impact tests housing dynamics. These influencing factors and the need of an
have coherence levels close to 100 percent at the resonance accurate 3D computer aided design (CAD) model of the housing
frequency. A peak in the amplitude spectrum could be caused itself make the upfront calculation of bearing housing natural
by other nearby operating equipment. In such a case, the frequencies rather expensive and unpractical.
amplitude peak does not represent a resonance, the phase Bearing housing natural frequencies do not significantly change
spectrum is likely to show no significant phase change, and between test and field installation. Bearing housing impact tests on
the coherence would be significantly less than one at the a test-stand need to be performed on the fully assembled pump,
corresponding frequency. For further details on modal testing, securely mounted onto test pedestals or the actual job baseplate. It
refer to Ewins (1986). is recommended to impact test bearing housings on a test-stand
Baseplates Mounted Onto Flexible Support Structures and avoid resonance situations that could potentially be excited by
pressure pulsations at vane-pass frequency. However, there is no
Special attention needs to be given to baseplates mounted onto direct correlation between a measured bearing housing natural
flexible skids or offshore platforms. The flexible substructure acts frequency and the corresponding bearing housing vibrations that
as an additional spring in series with the baseplate stiffness and occur when operating the equipment at the resonant frequency.
therefore influences the baseplate natural frequencies. The case of In case a pump runs at or near best efficiency point and sufficient
a 14×14×18.5D MSDD two stage pump mounted on a flexible skid B-gap was applied, the resulting pressure pulsation levels may not
demonstrates the influence of the flexible support structure. The cause a vibration issue even when operating at resonance conditions.
VFD operated pump is in crude oil service with a continuous In order to avoid potential bearing housing vibration problems it
operating speed range of 2500 to 3960 rpm, which corresponds to is recommended to detune identified bearing housing resonances.
a 1× synchronous speed excitation range of 37.5 and 72.6 Hz. The The following detuning methods have been successfully applied:
finite element analysis was used to optimize the baseplate and skid
design in order to keep the VFD operating speed range free from • Reducing of the stiffness of the bearing housing to bearing
bracket joint by means of relocating the lower pair of housing bolts
structural resonance conditions that could be excited by unbalance
as close to the upper pair as possible. Applied to outboard bearing
(1×). Figure 24 illustrates the solid model of the baseplate and
housings, this very effective modification often drops the vertical
flexible skid. The first six structural natural frequencies are listed
bearing housing natural frequency by 100 to 150 Hz.
in Table 1. The middle column lists the natural frequencies for the
case of a rigid skid and the column to the right states the same data • Alternatively, a stiffness reduction may also be achieved by
for the case of a flexible substructure. means of slotting of bearing brackets aimed at reducing the contact
area between housing and bracket.
• Changing from a 180 to a 360 degree bearing housing attachment
• Adding weight to the housing (bolted-on or cast-in)
• Providing additional structural support (bracing)
In many cases it is perfectly acceptable to not fully detune a
resonance condition but to move it to lower frequencies, excitable
at lower operating speeds. Equation (11) indicates that pressure
pulsation levels are proportional to the square of the speed.
Vertical Pump Structural Vibration Problems
Figure 24. Solid Model of Baseplate and Skid. The process of vertical pump structural analysis is described in
the “LATERAL ROTORDYNAMICS” section. Structural modes
Table 1. Calculated Structural Natural Frequencies. are lowly damped and resonance situations with 1× and 0.5×
synchronous speed should be avoided. Both types of resonance
can result in excessive structural vibrations. Detuning options
include adding mass, applying seismic supports, and stiffening of
driver-stands, columns, mounting plates, or sole plates, etc.
Conclusions
Structural resonance situations with 1× running speed frequency
and vane-pass frequency should be avoided in case of horizontal
The case demonstrates the significant influence of the flexible pump applications. With vertical pumps, excitations of concern
support structure and the usefulness of FEA in the design process. are at 0.5× and 1× running speed frequency. Structural natural
Table of Contents
frequencies and mode shapes can be identified applying FEA

and/or modal testing. In case structural resonances cannot be
detuned, alternative options aimed at ensuring acceptable operation
and vibration performance include:
• Locking out/programming out of specific operating speeds from
continuous operation
• Detuning of a resonance to a natural frequency excitable by Figure 26. Incidental and Reflected Waves at Channel Reduction.
lower rpm operation. The excitation levels at a lower rpm are likely
to be reduced because pressure pulsation amplitudes and unbalance
forces both are proportional to the square of speed. A 50 percent diameter reduction in a cylindrical channel causes
36 percent of the power of the incident pressure wave to be
• Reducing the excitation magnitude by means of balancing, reflected. In case of an 80 percent reduction (DB = 0.2*DA), 85
increasing B-gap, etc. percent of the incident power is reflected. Similarly, an expansion
where the pipe diameter doubles, 36 percent of the power of the
ACOUSTIC RESONANCE incident pressure wave is reflected.
General A common and most important example of acoustic
resonance in centrifugal pumps are long crossover (long
Acoustic resonance can occur in suction and discharge piping, crossunder) acoustic resonance in multistage, volute-type
in the long crossover of multistage pumps, and other fluid filled pumps of opposed impeller design. The hydraulic passage
channels. The phenomenon takes place when the return of a length (fluid-filled channel length) extends from the long
reflected pressure wave, generated by a periodic excitation, crossover volute lip, through the long crossover and receiving
coincides with the generation of the next pressure pulse. In that impeller to the next short crossover volute lip. Vane-pass
case, a standing wave forms inside the fluid filled channel. Standing frequency pressure pulsations generated by the long crossover
waves are acoustic eigenvalues in pipes or piping systems. Acoustic entry stage impeller are the driving force.
resonance occurs when an acoustic eigenfrequency coincides The wavelength λ is a function of the fluid speed of sound c
with an excitation frequency, for example pressure pulsations and the excitation frequency f and is defined in Equation (21).
at vane-pass frequency. The channel length may be equal to The fluid speed of sound is a function of the fluid density ␳ and
quarter, half, or full wavelengths and multiples thereof as the adiabatic (or isentropic) bulk modulus KS as indicated in
illustrated in Figure 25. Equation (22).
The bulk modulus is a measure of the fluid’s resistance to

compression. Fluid volume is a function of the applied pressure,
the fluid temperature and compressibility, and the initial volume V0
under atmospheric pressure. Temperature rises under compression,
Figure 25. Standing Wave-Forms. which further increases the pressure. Slow compression of a
fluid, allowing full dissipation of the generated heat, leads to the
The channel length is determined by a reflective condition at measurement of an isothermal (constant temperature) bulk
the end of the channel opposite to the induction of the pressure modulus. An adiabatic (or isentropic) bulk modulus is measured
wave. Reflective conditions exist at channel reductions, valves, when compressing the fluid rapidly and measuring pressure due to
etc. The reflection at a channel reduction is best explained compression and thermal expansion. The adiabatic bulk modulus
with acoustic impedance theory. Acoustic impedance Z is a should be used for this consideration. The bulk modulus is defined
function of the fluid density ␳, the fluid speed of sound c, and as the slope of the volume versus pressure curve as illustrated in
the channel cross-sectional area A as defined in Equation (19). Figure 27. The slope may be defined as a secant, drawing a line
Acoustic impedance is a measure of how much a cross-section from the origin (V0) to a point on the curve or it can be defined as
of a fluid-filled channel resists fluid motion when subject a tangent at a specified point on the curve, leading to the definition
to pressure. of the secant bulk modulus KS shown in Equation (23) and the
tangent bulk modulus KT in Equation (24), respectively. Figure 27
illustrates the two definitions.
The ratio of reflected to incidental power of a wave hitting a

channel reduction is defined in Equation (20). A sketch of a
channel reduction is shown in Figure 26.
Table of Contents
seven-vane impeller at long crossover entry stage. The figure also
indicates the increased chance of encountering acoustic resonance
when converting from fixed speed operation to VFD operation.
Figure 27. Secant and Tangential Fluid Bulk Modulus. Figure 28. Acoustic Resonance Diagram (7×).
The stiffness of the pipe or channel structure also influences the The AR diagram shows resonance conditions associated with the
fluid speed of sound. For a cylindrical channel of diameter D, wall three products. The diesel and kerosene resonances occur close to
thickness t, and made out of a material with Young’s modulus of 3560 rpm operating speed, at approximately 3625 rpm and 3495
elasticity E and density ␳, the effective speed of sound c1 is defined rpm, respectively. This explains the high vane-pass vibration levels
in Equation (25). when pumping these two products.
A first corrective measure applying a 30 degree angle cut-back
to both long crossover/crossunder volute lips, resulted in a scissoring
effect between the impeller vane twist angle and the volute
cutwater. This measure has a tendency to “smear-out” the pressure
pulse and also increases the B-gap between impeller and volute. As
a result, the rms vane-pass vibration levels dropped from 0.66
Table 2 states approximate speed of sound values for some in/sec to 0.45 in/sec for diesel pumped at comparable operating
commonly pumped fluids. The values for diesel, kerosene, and gasoline conditions. This drop in vibration level, achieved by means of
can vary, mainly because of differences in chemical composition. reducing the excitation forces, was not sufficient and the acoustic
resonance condition needed to be detuned. The AR diagram for the
Table 2. Approximate Fluid Speed of Sound Values. case of a five-vane impeller installed at the long crossover entry
stage is presented in Figure 29. The reduced excitation frequency
resolved the resonance condition for all products and over the
entire continuous VFD operating speed range.
The available (system) damping that limits pressure pulsations is

typically low. Structural vibrations caused by acoustic resonance
can therefore easily reach damaging levels. In case of long
crossover acoustic resonance, bearing housing rms vibrations at
vane-pass frequency in excess of 1.0 in/sec (25.4 mm/sec) have
been measured.
Case Study—Multistage Pump

Long Crossover Acoustic Resonance
The identification and cure of long crossover acoustic resonance
are explained on the example of a 6×8×11J MSD2 multistage Figure 29. Acoustic Resonance Diagram (5×).
pump in pipeline service, pumping batches of diesel, kerosene, and
The vane-pass vibration levels measured in the vertical direction
gasoline. The pump initially operated at 3560 rpm and was later
of the inboard bearing housing dropped to acceptable levels after
converted to VFD operation with an operating speed range of 2000
implementing this second corrective measure. The gasoline levels
to 3960 rpm. The pump was fitted with seven-vane impellers.
were 0.005 in/sec rms 5× and 0.013 in/sec rms 7×, kerosene showed
Operation at 3560 rpm indicated highly product dependant
0.007 in/sec rms 5× and 0.019 in/sec rms 7× levels, and diesel was
vane-pass vibration levels in the vertical direction of the inboard
at 0.007 in/sec rms 5× and 0.016 in/sec rms 7×, respectively.
bearing housing. The corresponding filtered (vane-pass frequency)
rms vibration levels were 0.06 in/sec for gasoline, 0.42 in/sec for Corrective Measures
kerosene, and 0.66 in/sec for diesel, indicating a resonance condition.
The theoretical identification of long crossover acoustic The preferred corrective measures are the ones that detune
resonance is best done applying an acoustic resonance (AR) acoustic resonance conditions, providing a minimum of 10 to 20
diagram, plotting the hydraulic passage length together with the percent of separation margin between wavelength curves and the
frequency dependant wavelength curves. Acoustic resonance hydraulic passage length line. The following detuning features
occurs at the intersection between these lines. The AR diagram is apply for long crossover acoustic resonance:
the acoustic resonance equivalent of a Campbell diagram used in • Changing of the vane count of the impeller feeding the long
lateral and torsional rotordynamics. Figure 28 presents the AR crossover. This changes the excitation frequency and therefore the
diagram for the above described pipeline pump and the case of a wavelength λ.
Table of Contents
• Rearranging the destaging arrangement in case of destaged Despite the diversity of power circuits, there are two common
pump. Destaging the impeller at long crossover exit increases the properties of these drives:
hydraulic passage length.
• All of them accept commonly available AC input power of fixed
• Locking out of speed ranges from continuous operation voltage and frequency and, through switching power conversion,
create an output of suitable characteristics to operate a particular
In case of large continuous operating speed ranges, a resonance type of electric machine, which means that they are machine
condition near maximum speed may be eliminated at the price specific. The authors will limit their discussion to those suitable for
of creating a new resonance at the lower end of the speed range. induction and synchronous motors.
This approach may be perfectly acceptable because pressure
pulsation levels are proportional to the square of the speed (refer to • All of them are based on solid-state switching devices. The
development of new devices is the most important driver of
Equation [11]). VFD technology.
If detuning is not feasible, the applicable excitation levels can be
reduced by means of implementing the measures listed below. These Figure 30 shows the basic structure of most common AC drives.
modifications do not eliminate the resonance condition and some There is an input conversion circuit that converts the utility power
vane pass vibrations may still exceed the maximum allowable levels. into DC, and then an output inversion stage that changes the DC
These corrective measures only apply to the impeller feeding the long back into variable AC.
crossover because it is the driver of long crossover acoustic resonance.
• Cutting back of the crossover and crossunder volute lips in a 30
degree to 45 degree angle, resulting in a “scissoring” effect with
impeller vane twist angle. This, together with increasing the B-gap,
is the most effective measure reducing excitation levels.
• Minimizing the long crossover volute cutwater thicknesses,
providing a “bullet-nose” type profile
• Minimizing the impeller vane outlet thicknesses of the long
crossover entry stage impeller by means of overfiling
• Reduction of the shroud thickness at the impeller periphery by
means of chamfering
Figure 30. General Structure of a VFD.
• Providing the long crossover entry stage impeller with extra trim
(increasing the B-gap) Applying VFDs offers a variety of benefits:
• Operation close to BEP/avoiding of part-flow operation • Energy savings where variable flow control is required; variable
Conclusions speed operation replaces wasteful throttling controls
Acoustic resonance in pumps can cause high bearing housing • Optimization of rotating equipment performance
vibrations and also adversely affect casings and baseplates. These • Elimination of gears or other power transmission devices by
amplified structural vibrations can result in auxiliary piping, matching the base speed of the motor to the driven load
instrumentation, and bolting fatigue failures. Acoustic resonance is
also a source of excessive structureborne noise. The likelihood of • Automation of process control by using the VFD as the final
control element, leading to more efficient part-load operation
encountering acoustic resonance is greatly increased when opening
up the continuous operating speed range by applying a VFD. • Reduction of the rating and cost of the electrical distribution
Chances of encountering resonance conditions are further system by eliminating motor starting inrush
increased when pumping multiple products with different speeds
of sound. Upfront identification of potential acoustic resonance • Extension of motor, bearing, and seal life
condition is recommended and fairly easy when applying an AR • Reduction of noise and environmental impact. Electric drives are
diagram. The main difficulties are the identification of the correct clean, nonpolluting, quiet, efficient, and relatively easy to repair.
hydraulic passage length and the applicable fluid speed of sound
value. If necessary, the pump OEM can provide the passage length. Development of Semiconductor Switching Devices
In case of a multistage opposed-impeller centrifugal pump, the VFDs became practical after the invention of the thyristor in
applicable hydraulic passage length is well defined and the 1957. Table 3 shows a comparison of the properties of devices
acoustic resonance check can be performed in the tendering or commonly in use today.
design stage of an order. The channel length in suction and
discharge piping may be less clearly defined and special care is Table 3. Semiconductor Switching Devices.
required when calculating lengths.
It is recommended to perform a long crossover acoustic resonance
check in all cases of VFD operated multistage opposed-impeller
centrifugal pumps applying an AR diagram. Resonance conditions
should be detuned if at all possible.
MEDIUM VOLTAGE
VARIABLE FREQUENCY DRIVES
General The thyristor (silicon controlled rectifier type, SCR) is a
This section presents technical background on commercially four-layer (one thin disc of silicon with four layers of different
available VFDs suitable for operating an induction or synchronous, properties) semiconductor device that has many of the properties
standard polyphase alternating current (AC) motor at medium of an ideal switch. It has low leakage current in the off-state, a
voltage (2300 V to 13.8 kV). small voltage drop in the on-state, and takes only a small signal to
Table of Contents
initiate conduction (power gains of over one million are common). currently in use. While the type of semiconductor device should
When applied properly, the thyristor will last indefinitely. After its not be the most important issue to a user, in general the newer
introduction, the current and voltage ratings increased rapidly. devices provide better drive performance.
Today it has substantially higher power capability than any other
solid-state device but no longer dominates power conversion in Drive Control Technology
the medium and higher power ranges. The major drawback of the Parallel to the development of power switching devices, there
thyristor is that it cannot be turned off by a gate signal, but the have been very significant advances in hardware and software
main current must be interrupted in order for it to regain the for controlling VFDs. These controls are a mixture of analog
blocking state. The inconvenience of having to commutate (turn and digital signal processing. The advent of integrated circuit
off) the thyristor in its main circuit at a very high energy level operational amplifiers and integrated circuit logic families made
has encouraged the development of other related devices as dramatic reductions in size and cost of the drive control
power switches. possible, while permitting more sophisticated and complex
Transistors predate thyristors, but their use as high-power control algorithms without a reliability penalty. These developments
switches was relatively restricted (compared with thyristors) until occurred during the 1965 to 1975 period. Further consolidation
the ratings reached 50 A and 1000 V in the same device, during the of the control circuits occurred after that as large-scale integrated
early 1980s. These devices are three-layer semiconductors that circuits (LSI) became available. In fact, the pulse-width modulation
exhibit linear behavior but are used only in saturation (fully turned control technique was not practical until the appearance of LSI
on) or totally off. In order to reduce the base drive requirements, circuits because of the immense amount of combinatorial logic
most transistors used in VFDs are Darlington types (a tandem required. Clearly, the most significant advance in drive control has
arrangement in which there are two transistors, the first one been the introduction of microprocessors into drive control
feeding the base input of the second). Even so, they have higher circuits. The introduction of cheap and powerful microprocessors
conduction losses and greater drive power requirements than continues to expand the capability of drive controls. A modern
thyristors. Nevertheless, because they can be turned on or off drive should have most of these features. The performance
quickly via base signals, transistors quickly displaced thyristors in enhancements include:
lower drive ratings, and were once widely used in pulse-width
modulated inverters. They in turn were displaced by insulated gate • More elaborate and detailed diagnostics owing to the ability to
store data relating to drive internal variables, such as current,
bipolar transistors in the late 1980s. The IGBT is a combination speed, firing angle, etc. The ability to signal to the user if a
of a power bipolar transistor and a metal-oxide semiconductor component has failed.
field-effect transistor (MOSFET) that combines the best properties
of both devices. A most attractive feature is the very high control • The ability to communicate both ways over industry standard
input impedance that permits them to be driven directly from lower protocols with the user’s central computers about drive status.
power logic sources. Their power handling capability has increased • The ability to make drive tuning adjustments via keypads with
dramatically and they are now viable alternatives to thyristors and parameters such as loop gains, ramp rates, and current limits stored
integrated gate controlled thyristors in the largest drive ratings. in memory rather than potentiometer settings.
It has long been possible to modify thyristors to permit them to
be turned off by a negative gate signal. These devices are four-layer • Self-tuning and self-commissioning drive controls.
types and are called gate-turn-off thyristors, or simply GTOs. • More adept techniques to overcome power circuit nonlinearities.
These devices have been around since at least 1965, but only in
the mid 1980s did their ratings increase to high power levels. AC Variable Frequency Drives
Present-day GTOs have about the same forward drop as a The effect of new solid-state switching devices was very significant
Darlington transistor (twice that of a conventional thyristor). GTOs on AC variable speed drives, and it shows no sign of stopping. AC
require a much more powerful gate drive, particularly for turn-off, drives are machine specific and more complex than DC drives,
but the lack of external commutation circuit requirements gives mostly because of the simplicity of the AC machine. Solid-state
them an advantage over thyristors. GTOs are available at much VFDs have been developed and marketed for wound-rotor
higher voltage and current ratings than power transistors. Unlike induction motors (WRIMs), cage-type induction motors (IM), and
transistors, once a GTO has been turned on or off with a gate pulse, synchronous motors.
it is not necessary to continue the gate signal because of the Historically, WRIM-based VFDs were in common use long
internal positive feedback mechanism inherent in four-layer before solid-state electronics. These drives operate on the
devices. Unfortunately, high cost and very large switching losses principle of deliberately creating high-slip conditions in the
restricted the use of GTOs to only those applications in which machine and then disposing of the large rotor power that results.
space and weight were at a premium. In 1997 the IGCT was This is done by varying the effective resistance seen by the rotor
introduced. This is very similar in construction to the GTO, but a windings. The WRIM is the most expensive AC machine. This
new method of turn-off and special metallurgy has resulted in a has made WRIM-based variable speed drives noncompetitive as
device considerably better than the GTO in forward drop and compared with cage induction motor drives or load commutated
switching losses. It has a sibling, the symmetrical gate controlled inverters using synchronous machines. The slip (s) is defined as
thyristor (SGCT), which has similar properties but can also block the ratio of the difference between the stator flux angular
reverse voltage, which the IGCT cannot. Voltage-fed circuits use velocity and the rotor angular velocity, to the stator flux
IGCTs while current-fed circuits need SGCTs. angular velocity.
Today the thyristor, IGBT, SGCT, and IGCT form the technological
base on which the solid-state variable speed drive industry rests.
There are other device technologies and enhancements in various
stages of development that may or may not become significant
depending on their cost and availability in large current ratings (> Thus when a machine is operating at a speed less than that of the
50 A and a voltage rating of at least 1000 V). These include: trench stator flux (determined by the stator voltage frequency and the pole
gate construction for IGBTs, silicon carbide semiconductors, number), it is slipping. That is not necessarily bad, but of the total
variants of the four layer switch such as the MTO (MOS turn-off power flowing across the air-gap into the rotor, the fraction s is lost
thyristor), and MCT (MOS controlled thyristor). New switches are as heat in the rotor, and (1-s) comes out of the shaft as useful
expected to come along and significantly improve on the devices mechanical power.
Table of Contents
Induction Motor Variable Speed Drives 120 degrees square pulses of current separated by 60 degree
Because the squirrel cage induction motor is the least expensive, intervals of zero current. But, since there must be a capacitor bank
least complex, and most rugged electric machine, great effort has at the output, the motor current has improved waveform due to the
gone into drive development to exploit the machine’s superior capacitors absorbing some of the harmonics. Furthermore, this
qualities. Owing to its very simplicity, it is the least amenable to type of drive generally also uses a form of pulse-width-modulation.
The 120 degree output current pulses are subdivided into several
variable speed operation. Since it has only one electrical input
narrower width pulses interspersed with short intervals of zero
port, the drive must control both flux and torque simultaneously
current. This technique suppresses the lowest-order (current)
through this single input. As there is no access to the rotor, the
harmonics (five, seven, etc.) Since every switching causes an
power dissipation there raises its temperature and therefore very
energy loss in the output devices, fast switches are needed in order
low-slip operation is essential. Induction motor VFDs in the
to support a high switching frequency without excessive losses in
past have had the greatest diversity of power circuits. Today, for
the output switches. Due to the output capacitor filter, acoustic
induction motor MV drives rated 2300 V and up, there are
noise in the motor is not an issue.
essentially only two main choices:
• The IGBT/IGCT voltage-fed pulse-width modulated drive. In this Current-Fed Versus Voltage Fed Circuits—
type of voltage source inverter both the frequency and amplitude The Two Basic Topologies
are controlled by the output switches alone. A representative circuit
The terms voltage-fed and current-fed refer to the two basic
based on IGBTs is displayed in Figure 31. This shows the
VFD strategies of applying power to the motor. In Europe, these
conceptual design, but as indicated is used for VFDs operating at
are called voltage-impressed and current-impressed, which is a
less than 1000 V. The input converter is a diode bridge so that the
clearer description.
DC link operates at a fixed unregulated voltage. The diode front
In voltage-fed circuits, the output of the inverter is a voltage,
end gives close to unity power factor, independent of load and
usually the DC link voltage. The motor and its load determine the
speed. This type of drive is called pulse-width-modulated because current that flows. The inverter does not care what the current is
the output voltage waveform is synthesized from constant (within limits). Usually, these drives have diode rectifiers on the
amplitude, variable-width pulses at a modest (200 to 1200 Hz) input. The main DC link filter is a capacitor.
frequency so that a sinusoidal output is simulated. The lower In current-fed circuits, the output of the inverter is a current,
(voltage) harmonics (five, seven, 11, 13, 17, 19, etc.) are not usually the DC link current. The motor and its load determine the
present in modern PWM drives. One advantage is smooth torque, voltage. The inverter does not care what the voltage is. Usually
low output harmonic currents, and no cogging (cogging is visible these VFDs have a thyristor converter input stage, and the DC link
jerky rotation at low speed caused by large harmonic currents). element is an inductor.
Although this approach eliminates the phase control requirements Table 4 states a summary comparison of the properties of the
of the thyristor converter, it requires fast output switches. Since two types of VFDs.
every switching event causes an energy loss in the output devices,
fast switches are needed in order to support a high switching
Table 4. Comparison of Current-Fed and Voltage-Fed Drives.
frequency without excessive losses in the output switches.
Occasionally, high-frequency switching may cause objectionable
acoustic noise in the motor, but that has been overcome with
special modulation techniques and higher switching frequencies
(3000 Hz and up). There are IGBT PWM VFDs on the market
today in the range of 1 kW to 1 MW at 460, 600, and 690 VAC
rated input. As with all voltage source inverters fed by diode
rectifiers, regeneration of power to the line is not possible. The
voltage-fed PWM MV drive is available in two subtypes: The
neutral-point clamped circuit, and the series cell multilevel circuit.
Medium-Voltage Variable Frequency Drives

For drives rated 2300 VAC and above on the output, there are a
number of choices of design of both current- and voltage-fed types:
• The load commutated inverter
• The current-fed SGCT inverter
• The neutral-point-clamped inverter
• The multilevel series cell VFD
The Load Commutated Inverter
As illustrated in Figure 32, the load-commutated inverter is
Figure 31. Low-Voltage PWM Drive Illustrating Voltage-Fed based on a synchronous machine. All the thyristors are naturally
PWM Technology. commutated (natural commutation is the turn-off process of a
• The SGCT based current-fed pulse width modulated drives thyristor when a sinusoidal voltage source in the circuit applies
(refer to Figure 33 for example). In this type of current source reverse voltage to the device), because the back electromagnetic
inverter only the frequency is controlled by the output switches. field (EMF) of the machine commutates the machine-side
The input converter is a thyristor bridge so that the DC link converter. The machine-side converter operates exactly like the
operates at a variable controlled voltage, and with the link inductor, line-side converter, except the phase back angle is about 150
functions as an adjustable current source. The thyristor front end degrees. The machine naturally applies reverse voltage to an
gives a variable input power factor, dependent on load and speed. off-going device before the next thyristor is gated. This imposes
Ordinarily, this drive would produce an output like the LCI, that is some special design criteria on the synchronous motor, which
Table of Contents
includes operation at a substantial leading power factor over the inductance during the transfer of current. The choice of
speed range, requirement for sufficient leakage inductance to limit capacitor is determined by the permissible ring-up voltage
the thyristor di/dt, and ability to withstand harmonic currents in the during commutation. All current-fed VFDs with self-commutating
damper windings. The LCI uses two thyristor bridges, one on the switches have the need of a “buffering” capacitor between the
line-side and one on the machine-side. The requirement for the impressed current of the inverter and the inductance of the
machine to operate with a leading power factor, typically 90 motor. Furthermore, if the capacitor bank exceeds 0.2 PU, the
percent, requires somewhat more field excitation and a special possibility of self-excitation of the motor exists, necessitating a
exciter compared with that normally applied to synchronous motor. contactor between machine and drive. Voltage-fed circuits do
This also results in a 10 percent reduction in the torque for a given not require these elements, because one can arbitrarily change
current. The machine-side devices are fired exactly synchronized the voltage across the leakage inductance.
with the rotation of the machine, so as to maintain constant torque
angle and constant commutation margin (the time available to turn
off a thyristor). This is done either by rotor position feedback or by
phase control circuits driven by the machine terminal voltage. Only
resistor-capacitor (RC) networks for voltage sharing are necessary.
The output current is very similar in shape to the input current, i.e.,
quasi square-wave, which means a substantial harmonic component.
The harmonic currents cause extra losses in the damper bars, and
they give rise to very significant torque pulsations. The drive is not
self-starting due to the low machine voltage at low speeds that
cannot commutate the thyristors. Therefore, the drive is started by
interrupting the DC link current with the line-side converter in order
to commutate the inverter thyristors. The line-side converter is
regulated to control torque. A choke is used between converters Figure 33. SGCT Current-Fed MV VFD with 12-Pulse Input.
to smooth the link current. Load commutated inverters came into
commercial use about 1975 and are used mainly on very large Since the capacitor bank provides some filtering of the output
medium voltage drives (1 MW to 100 MW). At these power levels, current, the quasi-square wave from the SGCTs does not reach
multiple series devices are employed (typically four at 4000 V the motor. Most of the high-frequency components are absorbed
input), and conversion takes place directly at 2.4 or 4 kV or higher. by the capacitors. Further motor current improvements are made
The efficiency is excellent, and reliability has been very good. by harmonic elimination switching patterns for the SGCTs. At
Although they are capable of regeneration, LCls are rarely used in low frequencies, many pulses per cycle are possible and
four quadrant applications because of the difficulty in commutating harmonic elimination is quite effective. But the SGCT
at very low speeds where the machine voltage is negligible. frequency limit of a few hundred Hertz restricts harmonic
Operation above line frequency is straightforward. elimination at rated frequency to the fifth and maybe the
seventh harmonic.
This frequency limit comes about due to the nature of the
SGCT turn-off (and to a lesser extent, turn-on) mechanism. The
device is turned off by applying a negative current to gate that is
larger than the anode current, over a period of a few microseconds
and interrupting the regenerative turn-on mechanism. There is a
short period during which the device experiences extremely
high internal power dissipation. This is the source of switching
losses that have to be considered in the thermal management of
the device.
The SGCT gate driver, in addition to providing a turn-on
pulse comparable to the thyristor driver, needs to be able to
deliver a peak negative current of more than 100 percent of the
main current in order to turn off the device. Thus, the SGCT
Figure 32. Load-Commutated Inverter with 12-Pulse Input. driver has a peak VA rating of two to three orders of magnitude
higher than that for a thyristor, and perhaps 10 times the average
Despite the need for special synchronous motors, the LCI drive power requirement. This is an important factor in that all
has been very successful, particularly in very large sizes, where only the gate power must be delivered to a circuit floating at
thyristors can provide the necessary current and voltage ratings. medium-voltage potential.
Also, high speed LCIs have been built. Now that self-commutated
VFDs are available, the LCI is becoming less popular. The Neutral-Point-Clamped Inverter
The Current-Fed SGCT Inverter Figure 34 illustrates such a circuit, the neutral-point-clamped
inverter. There have been many VFDs of this type applied at 3300
Another medium-voltage bridge inverter circuit is displayed V output with 4.5 kV GTOs, but the circuit has only recently been
in Figure 33. Here the output devices are SGCTs (two or three extended to 4 kV probably due to the improved properties of the
6.5 kV units per leg will be required) that can be turned off (as IGCT. In the newer versions of this drive, the GTOs are replaced
well as on) via the gate. There must be a capacitor filter on the with IGCTs. These devices are similar in construction to a GTO,
output rated about 0.3 PU kVAR. Since the motor appears to be but are turned off quickly (one microsecond) by drawing all the
a voltage source behind the leakage reactance, it is not possible anode current out through the gate, so that the turn off gain is unity.
to commutate the current between motor phases without a This requires a higher current gate driver, but lower average
voltage to change the current in the leakage inductance. When power requirements since the turn-off time is so short. The main
an SGCT turns off, there must still be a path for the current improvement is that the IGCT can operate with a very small or no
trapped in the motor leakage inductance, which is provided by a snubber (resistor-capacitor [RC] electrical circuit suppressing
capacitor bank. The capacitors resonate with the motor leakage electrical voltage transients).
Table of Contents
Figure 34. Neutral-Point Clamped MV VFD-IGBT Version with 12-

Pulse Input.
Figure 35. Multilevel Series Cell PWM Drive.
In this 4 kV output design, the total DC link voltage is 6 kV,
with a midpoint established at the center of the capacitor filter.
Each leg of the bridge consists of two 6.5 kV IGCTs in series.
There are diodes in reverse across each IGCT to permit motor
current to flow back to the link, and still more diodes (same
voltage rating as the IGCTs) connecting the midpoints of the
inverter legs back to the midpoint of the DC link. The total device
count is 12 IGCTs and 18 diodes. The neutral-point-clamped
inverter offers several advantages in those cases where series
devices would be necessary anyway. First, the clamping diodes
permit another voltage level, the DC link midpoint, at the
output. This cuts the voltage step seen by the motor in half,
and more important, creates another degree of freedom in
eliminating output harmonics. Also, the clamping diode positively
limits the voltage across any one device to half the link voltage,
enforcing voltage sharing without additional RC networks. Figure 36. Power Conversion Module Used in Series-Cell
Since the switching devices in this circuit are never subjected Multilevel VFD.
to reverse voltage, it is preferable to use asymmetrical devices
in which the lack of reverse blocking capability is traded off for Each cell generates the same AC output. The fundamentals are
lower conduction and switching losses (SGCTs are IGCTs that equal in magnitude and in phase, but the carrier frequency is
have reverse voltage blocking capability, as required by the staggered among the cells in a particular phase. Although an
current-fed circuit). individual cell operates at 600 Hz (or more), the effective
Device protection during a short circuit can be a problem, as switching frequency at the motor is twice the carrier frequency
the IGCT can carry almost unlimited fault current like a times the number of cells per output phase. This low switching
thyristor. Unlike the current-fed circuits where fault current is frequency and the excellent high frequency characteristics of the
limited, in the voltage-fed circuit, the DC link capacitor can IGBT has the advantage that the IGBT switching losses are very
source very large fault currents in the event of a short or a small. The devices can switch well above rated current without
commutation failure. Protection schemes generally attempt to the need for snubbers, which also helps in maintaining excellent
detect the onset of fault current and turn off the devices before efficiency. Waveform quality is unaffected by speed or load. For
it grows beyond the safe turn-off level. It is also possible to use the five cell/phase VFD, there are 10 930 V steps between the
the NPC topology with IGBTs as the switching devices. negative and positive peaks. With this technique, the concern for
Although IGBTs are currently available at 6.5 kV, the IGBT high dv/dt on the motor windings is largely avoided.
NPC is more often constructed of series IGBTs rated 3300 V A major advantage of the IGBT over all other power switches is
to reach 4 kV output. The concept of NPC can be extended to the extremely low gate power required. The peak power is about
M-level inverters, although the number of diodes grows rapidly. 5 W with an average of much less than 1 W. This dramatically
Since each device is topologically unique, adding redundant simplifies the delivery of gate power compared to the GTO/IGCT.
devices would require twice as many, instead of just one more. Although there are more active devices in the perfect harmony (60
IGBTs and 60 diodes in the inverter sections) than in the other
The Multilevel Series-Cell Inverter circuits, the elimination of snubbers, voltage sharing networks, and
The patented series cell arrangement of Figures 35 and 36, high power gate drivers compensates for the additional switching
also known as the perfect harmony drive, addresses the devices. The type of IGBTs employed are third and fourth generation
previously mentioned design issues in a unique way. Since there isolated base modules, generally the same mature product as
are no devices in series, only series cells, the problem of those found in 460 VAC and 690 VAC LV PWM drives, and are
voltage sharing among series devices is not present. The also used in traction applications. The IGBTs are protected by
rectifier diodes and the IGBTs are both closely coupled to the an out-of-saturation detector circuit that augments the built-in
DC link capacitor in the cell and thus cannot be exposed to current limiting behavior. Since the cells are assembled into a
more than the bus voltage, regardless of the load behavior. nonconducting framework and are electrically floating, the
Since there is no DC link choke, a voltage transient on the AC mounting and cooling of the IGBTs are no more complex than in
mains is converted into a current pulse by the relatively high a low voltage PWM drive. It is possible to put redundant cells in
leakage reactance of the transformer secondary, and does not the string, and also to operate at reduced output with one or
add to the voltage seen by the diodes. more cells inoperative.
Table of Contents
Comparison of Medium-Voltage Motor Drives 3. Are there special requirements on ride-through? (Ride-through
All the types of drives mentioned above are capable of providing occurs when there is a utility line disturbance, e.g., a sag, swell, or
highly reliable operation at a justifiable cost, and have been proven transient, and the VFD continues to operate without tripping.)
in service. They all have efficiencies above 95 percent. The most What happens if the drive trips and restarts on a line disturbance?
significant differences among them have to do with power quality, Is such a condition acceptable to the driven load or process?
that is, how close to a sinewave is the input current, and how well
does the output resemble the sinusoidal utility voltage. Table 5 4. What is the input power factor (PF) requirement (either from the
compares them on a number of different factors. Note that the utility or the user)?
voltage-fed drives have an advantage in input harmonics and power
factor, and the drives, which do not use thyristors, have a wider
• Voltage-fed VFDs typically have a good PF at any load and
speed without correction.
speed range.
• Current-fed VFDs may require correction capacitors or an
active front end.
Table 5. Comparison of Medium Voltage VFDs.
5. What are the input harmonic current limits imposed by the utility
or the user? IEEE 519 (1992) or other?
• What is the utility system impedance (short circuit current
or MVA)? These data are needed to perform harmonic calculations.
The VFD supplier can perform them, but needs to have valid
source impedance data.
• Are there requirements for maximum voltage drop or
flicker? Flicker is covered by IEEE 519 (1992) but is treated
differently from harmonic distortion. This may be an issue for
loads that fluctuate over a two to ten second range. It is not usually
a problem with a pump or fan load.
• Is this an “islanded” supply, like an oil platform or
water treatment plant on emergency power? If so, what are the
characteristics of the generators and the power system impedance
on which to base distortion calculations?
VFD Application Checklist
• What is the unbalance and distortion of the customer’s
This section lists issues that should be considered in any VFD utility? One needs to understand that a VFD’s harmonic performance
application. It is not meant to provide a detailed discussion, but is adversely affected by preexisting distortion or unbalance. VFD
rather to raise awareness on the part of the user or buyer as to manufacturers specify power factor and harmonics at the rated
necessary information and things to watch out for. power of the VFD on the basis of an undistorted and balanced
utility supply.
Line-Side Issues
• Below 500 kW, there are situations in which a six-pulse
1. What is the required motor power rating and the VFD input input is acceptable, but above 1 MW, the minimum pulse number
voltage and frequency? is 12, and 18 pulse or more may be needed in some cases.
• If the power rating is less than 500 kW, the use of a • Using the same pulse number in a diode rectifier as
low-voltage VFD, i.e., 380 to 690 V rated input, should be considered. compared to a thyristor rectifier does not give the same harmonic
They are significantly cheaper then MV drives in the low results. In the thyristor rectifier, the higher order harmonics are
power range. larger due to the higher di/dt values present in thyristor circuits.
• Oversizing the motor and VFD may lead to poorer 6. Is there a requirement for input switch-gear with the VFD, such
harmonic and power factor performance of the VFD, as well as
lower efficiency. as a disconnect switch?
• Rated motor power is attained at its rated voltage and 7. Is there a requirement for input protection instruments?
frequency. If the load reaches rated motor torque at less than rated Typically, VFDs have built in overcurrent protection on both
speed, then rated power will not be delivered as the motor and VFD line-side and machine-side, as well as undervoltage protection, but
will have reached rated current, before rated voltage is attained. some users may want more functions.
This can be a difficulty with pumping systems where the head is
not exactly known. If the full torque is not attained until somewhat 8. Are there large utility power factor correction capacitors nearby?
above rated frequency, then the VFD and motor can deliver rated
power by operating in the constant power range. • The drive may trip when they are switched on, due to the
disturbance caused by connecting uncharged capacitors to the
2. What is the required tolerance on voltage and frequency? utility supply.
• This especially applies to power systems that are fed by a • For drives with significant input harmonics, the capacitors
generator, and may have more significant fluctuations. may resonate with the source inductance at a harmonic frequency.
This creates large (undesirable) voltages at the harmonic
• Typically, a VFD will be able to provide rated output within frequency, and sometimes requires adding detuning reactors in
the range of 95 percent to 110 percent of nominal input voltage, series with the capacitors.
with a plus or minus 5 percent tolerance on frequency.
• Drive manufacturers always assume that there will be no
• Some VFDs can produce less than rated output at power factor correction capacitors on the motor, as they may
significantly less than rated input voltage. VFD performance needs interfere with the drive operation. PF correction capacitors shall
to be coordinated with the user’s requirements. not be put on the output of a VFD!
Table of Contents
• It is permissible to put surge capacitors on an MV motor • Large VFDs are frequently water-cooled with a deionized
fed by a VFD, but it is not necessary. water loop inside. In that case, there is a water-to-water heat exchanger
to user cooling water, or a water-to-air heat exchanger located outside.
9. Is there a dedicated stepdown transformer feeding the VFD input?
This is called a drive isolation transformer, or DIT. If so, the effect 9. Is the installation in a seismic zone that requires certification? If
of the VFD input harmonics on the transformer must be considered. so which standards apply? Usually, as a minimum, foundation bolts
must be designed to prevent the drive enclosure from toppling over
• If there is a transformer integral to the VFD, or a DIT feeding due to the effect of lateral acceleration forces.
the VFD, one should check to make sure that the user’s utility can
withstand the inrush of the transformer when it is energized.
Saturation inrush events can be as large a 10 to 12 PU current. 10. Are there shock and vibration requirements? Some VFDs are
designed to withstand shock and vibration, but not all.
Environmental and Physical Issues
11. Is this a marine application aboard ship? If so, which standards
1. Is there adequate space for the VFD or is space limited? If so, how?
apply? Some but not all VFDs are designed to marine standards,
Medium-voltage VFDs are big due to the need for clearances inside.
such as ABS and Lloyds.
2. What is the requirement for protection class of the VFD
12. Is there a customer requirement for acoustic noise?
enclosure (e.g., IP22, NEMA 12, etc.)?
• The most common protection classes for VFDs are NEMA 13. What is the customer’s requirement for efficiency?
1/IP22 (finger-proof and drip-proof), and NEMA 12/IP54, which
is finger-proof, drip-proof, and dust-tight. These protection classes • In evaluating efficiency, all the equipment associated with the
VFD installation must be included. This includes power electronics,
are not intended for outdoor applications, but must be placed in an
input transformer, input filters, output filters, and auxiliaries.
indoor electrical room.
• For application requiring an outdoor location of the VFD, a • The difference in efficiency among modern VFDs is very
small compared to the energy savings to be attained by eliminating
special weatherproof enclosure is usually available in which the
valves and dampers as a flow control method.
VFD is located and is equipped with ambient control.
3. What is the user’s requirement for maximum ambient temperature 14. Is the VFD location a hazardous or potentially hazardous
and humidity? area (classified area per NEC)? VFD enclosures are almost never
explosion-proof.
• The industrial standard for maximum ambient is 104⬚F
(40⬚C) and 95 percent relative humidity, noncondensing. 15. Does the VFD go into a mine, nuclear power plant or other
• Higher ambient temperatures can sometimes be accommodated strictly regulated location? There may be special standards and
by derating the current output of the VFD, if it is directly air-cooled. regulations that apply, which rule out standard VFD designs.
4. What is the altitude of the installation? High altitudes require 16. Is third-party certification required (like UL, CE, or CSA)? In
thermal derating and there is also a concern for the reduced some applications, local codes require such a certification for the VFD.
dielectric strength of the air.
Machine-Side Issues
•A rule of thumb is to derate air-cooled equipment by 1
1. What is the rated motor power, voltage, full-load current,
percent per every 328 ft (100 m) above 3281 ft (1000 m) altitude.
and frequency?
5. Are there corrosive gases present (like hydrogen sulfide, • The VFD has to have a current rating at least as large as the
chlorine, ozone, etc.)? These are common in water-treatment plants, motor full-load amps.
and all electronic equipment can be damaged by concentrations
above a few parts per billion (ppb). • The motor has to be chosen for the required maximum
torque, not power. Motor power is nominally defined at the base
frequency and drops with speed.
6. Is there conductive dust in the environment (e.g., coal dust)?
Conductive dust will encroach on the creep spacing between 2. Is extended speed operation required; what is the maximum
live parts, and it must be kept out of the VFD and any frequency (extended speed is an operating regime above rated
electrical equipment. frequency where the voltage is held constant, but the frequency is
increased)? This is an application issue for the motor, as they are
7. Is the application at a seacoast environment? That location not usually guaranteed to operate satisfactorily above rated speed,
aggravates the humidity and corrosive gas conditions. but they can be designed for such duty.
8. How will the drive losses be dissipated into the local environment? 3. What are the motor parameters? Synchronous, induction, or
Modern VFDs are very efficient, and the larger the VFD, the wound rotor? The VFD manufacturer should be able to obtain the
more efficient. parameters to check for VFD motor compatibility.
• Typical efficiencies for VFDs above 1 MW are above 96
percent including the transformer (efficiency should not be a 4. Is the VFD going to be used with a motor that is smaller or larger
significant factor in choosing among different VFDs). than the VFD power rating? In some drives this creates a component
sizing issue and the VFD manufacturer should be advised.
• Transformer losses must be considered in case of an indoor
installation. Often, VFD manufacturers state efficiency without
5. Is this a retrofit VFD onto an existing motor? Is the VFD
the transformer.
compatible with an existing motor?
• Is there adequate cooling-air or air-conditioning for an • Is the insulation system capable of handling the peak voltages?
air-cooled unit? The building has to be able to absorb the losses
continuously, or draw in fresh air from outside and exhaust heated air. • Is the insulation system capable of handling the dv/dt?
Table of Contents
• Is the insulation system capable of handling common-mode • In general, one should avoid opening contactors to operating
voltage, i.e., voltage on the windings with respect to earth? power electronics so that current interruption transients are minimized.
6. In case of a synchronous motor application 14. Is there a requirement for motor/cable insulation system
monitoring equipment? These systems may not work on VFDs. The
• In a synchronous motor, an excitation source is necessary for partial discharge detection devices may not work either, due to
the motor field. This has to be considered in the design of the VFD.
the high frequency content of the VFD output.
• What kind of excitation is necessary?
• Brush-type, brushless with DC stator exciter (common 15. Is there a requirement for motor protection instruments? Some
on motors for direct-on-line applications), or brushless with AC commercially available motor protection instruments do not work
stator exciter except at 50 or 60 Hz.
• Who is to supply the excitation power supply?
16. Is the motor location a hazardous or potentially hazardous area
• What is the voltage and frequency of the auxiliary power (classified area)?
supply that will feed the exciter power supply?
17. Bearing currents have been an issue in some LV motors and
• Is this an LCI retrofit? In some cases load commutated
drives. The most conservative practice for MV VFD applications is
inverters have become too old to be effectively supported, and may be
replaced with a more modern drive without need to change the motor. to use an insulated bearing on the nondrive end, and apply a shaft
grounding device on the drive end.
7. Is synchronous transfer required (In synchronous transfer, the Load Issues
VFD brings the motor up to the speed corresponding to the utility
frequency, and phase synchronizes the motor voltages to the utility. 1. What is the speed-torque curve of the application? Is it a
Then there is a contactor that closes to connect the motor to the centrifugal load or constant torque, or something else? What is the
utility, after which the VFD contactor opens.)? If so, one needs to normal operating speed range?
make sure that there is a supplier for the switch-gear and the master • This is important for the proper coordination of VFD, motor,
programmable logic controller (PLC) to coordinate the transfer. and load. Motors are frequently self-cooled by means of an internal
fan, and therefore cooling is less efficient at lower operating speeds.
8. What is the customer requirement for torque pulsations, if any?
Some applications are very sensitive to torque pulsations as they • Centrifugal loads, in which the torque is proportional to the
may excite a shaft-train torsional natural frequency. In these cases, square of the speed, almost never require special cooling provisions
a special VFD may be necessary. The VFD manufacturer can in the motor.
describe the frequency and amplitude of torque pulsations for use • Constant torque loads, where the torque is a constant
in a torsional rotordynamic analysis. independent of speed, will need special cooling provisions at
operating speeds less than 1 PU.
9. Is this a single motor or multiple motor application? If different
motors are to be used on a single drive (as in a test stand), the range • The VFD is generally force-cooled by air or water, so it
does not matter what kind of load is applied.
of power and voltage needs to be specified. It can be difficult to
tune a VFD for a wide range of motors.
2. How much starting torque is required?
10. Is there a stepup transformer on the output of the VFD? This • All motors and rotating loads with sleeve bearings have a
presents special difficulties for the VFD and for the transformer. breakaway torque, which is the torque required to overcome the
bearing friction when there is no oil film.
• It is not a recommended configuration.
• What amount of starting torque is required? This is a key • Some applications require very large starting torque, such
as SAG mills and conveyors. In these cases, extra current may be
variable in designing the transformer.
needed from the VFD to get the torque.
11. Is this a long cable (over 3281 ft [1000 m]) application? Does • Starting torque is a complicated issue because the drive
the customer have a requirement for voltage distortion, dv/dt, and controls do not work as well at zero speed. It needs to be considered
peak voltage at the motor? If so, what are the cable parameters that PU torque from the motor cannot be greater than the PU
and length? current supplying it. Of course it is always possible to get less if the
control is incorrectly tuned.
12. Is there a customer requirement for distortion at the motor?
3. Are there overloads applied? What is the peak overload torque, and
• The motor leakage inductance filters the voltage such
the speed at which it is applied? How often does the overload occur?
that significantly less high-frequency current flows into the
motor. Because of this there is more voltage distortion than • There are two kinds of overload ratings in the VFD world:
current distortion. 150 percent for one minute and 110 percent for one minute. The
latter is used for easy to start loads (centrifugal), and the former for
• It is the distortion of the current that is important to the hard to start applications.
motor performance.
• The power electronic devices have to be sized to handle the
13. Is there a requirement for switch-gear between the VFD and the peak value of current, as well as the average power losses.
motor? It is important to assure that it is properly specified and that Therefore, a power electronics apparatus does not have the
it is interlocked with the VFD. overload capacity of an electric machine or transformer.
• An output contactor should close before the VFD run 4. Is there a gearbox in the drive train? The effect of pulsating
request is issued. torque may cause the gears to wear prematurely. In this case,
• An output contactor should not open before the VFD has one needs to be extra careful about torque pulsations and
stopped passing current to the motor. possible resonances.
Table of Contents
5. Is the application a rapid pulsating load, such as a reciprocating 1. Excessive lateral shaft vibrations can cause mechanical seal
compressor? If so, what is the torque versus crank angle curve? failures, damage to journal bearing, touching/rubbing/wear in
This may have an effect on VFD tuning and sizing. annular seals, high bearing housing vibrations (vibration transferred
from shaft to journal bearing), noise, etc.
6. Does this application require regeneration (regeneration is when
the motor acts like a generator and continuously returns electrical 2. In case a fixed speed horizontal pump is converted to VFD
energy to the VFD, taking it from the kinetic energy of the load. operation, the existing fixed-speed lateral analysis may sufficiently
The energy is then returned to the utility source)? If so how much cover the continuous VFD operating speed range. This needs to be
and for how long? carefully assessed on a case-by-case basis.
• Most voltage-source VFDs cannot regenerate energy to 3. Baseplate resonance can cause excessive bearing housing and
the utility.
casing vibration. This could result in fatigue failures of piping,
• Some current-fed VFDs like LCIs can continuously regenerate. bolts, etc., as well as coupling failures.
7. Does this application require dynamic, occasional, or emergency 4. Not all amplitude peaks revealed by impact testing represent
braking? If so, how much and for how long (Dynamic braking is excitable structural modes and the vane-pass excitation levels may
when the motor acts like a generator and continuously returns be low, not resulting in excessive vibration levels (e.g., large B-gap,
electrical energy to the VFD, taking it from the kinetic energy of operation close to BEP, operation at low power levels).
the load. The energy is then dissipated in a resistor, and not
returned to the utility.)? NOMENCLATURE
• Some voltage-source VFDs can be equipped with a A = Area; ampere
dynamic braking package that permits this kind of application. c, c1 = Speed of sound
cu = Circumferential fluid velocity
8. Is this a cyclic load? What are the peak torques and duty cycles? D = Critical damping ratio (--)
DA, DB = Pipe diameter
• Make sure that the rms current rating of the motor and VFD Da = Annular seal diameter
are not exceeded. D3 = Volute lip diameter
D2 = Impeller outer diameter
9. Does the customer require a special factory test regime, or a dv/dt = Rate of rise of applied voltage
back-to-back test? These tests are time-consuming and expensive. di/dt = Rate of rise of applied current
CONCLUSIONS E = Young’s modulus of elasticity
F = Unbalance force ([lbf] or [N])
The application of VFDs in pumping applications provides the user f = Frequency (Hz)
with a number of benefits. However, large continuous VFD operating G = ISO balance quality grade (mm/sec)
speed ranges greatly increase the risk of encountering resonance K = Balance constant
conditions of some kind. While centrifugal pump operating problems Kxx = Stiffness
and failures can be caused by a variety of sources, only Resonance KS = Adiabatic (isentropic) bulk modulus
problems are directly related to VFD operation. The four resonance M, m = Mass (kg)
categories of concern are associated with lateral rotordynamics, N = Rotor speed (rpm)
torsional rotordynamics, structural dynamics, and acoustic resonance. n = Integer (--)
Each of these categories has its own recommended analyses and ∆P = Pressure differential ([psi] or [Pa])
checks (to identify resonances), detuning methods, and risks if ∆P* = Dimensionless pressure differential (--)
not acted upon. Operation at lowly damped resonance conditions q* = Normalized flow (--)
can result in excessive mechanical vibration and related failures. R = Radius
Table 6 summarizes the main phenomena, analysis methods and s = Slip
recommended corrective action in case of resonance. The preceding rms = Root mean square
text should be consulted for more detailed and complete information. t = Wall thickness; time (sec)
The appropriate set of analyses and checks can reliably replace a U = Unbalance ([oz in] or [kg m])
factory string-test aimed at identifying resonance conditions/ensuring Uxx = Displacement
acceptable vibration performance. u2 = Fluid velocity at impeller outlet ([ft/sec] or [m/sec])
V = Volt
Table 6. Summary of Resonance Issues, Analyses, and Corrective Action. W = Mass (lbm)
Z = Acoustic impedance
z2 = Impeller vane count (--)
Q = Flow rate ([gpm] or [m3/sec])
␣x = Eigenvalue real-part
λ = Eigenvalue (–); wavelength ([m] or [in])
␳ = Density ([lb/ft3] or ([kg/m3])
␸ = Phase (rad)
␻ = Angular speed of rotation (rad/sec)
␻R = Rotor angular velocity (rad/sec)
␻S = Stator flux angular velocity (rad/sec)
␻x = Angular natural frequency (rad/sec)
List of Acronyms
AC = Alternating current
DC = Direct current
DIT = Drive isolation transformer
Table of Contents
EMF = Electromagnetic field Lateral Critical and Train Torsional Analysis and Rotor
GTO = Gate-turn-off (thyristor) Balancing,” First Edition, American Petroleum Institute,
IGBT = Insulated gate bipolar transistor Washington, D.C.
IGCT = Integrated gate controlled thyristor
ANSI/HI Standard 2.1 2.5, 1994, “American National Standard for
IM = Induction motor
Vertical Pumps—For Nomenclature, Definitions, Applications
kV = Kilovolts
and Operation,” Hydraulic Institute, Parsippany, New Jersey.
kVAR = Kilovolt-amperes reactive
KW = Kilowatt Bedford, B. D. and Hoft, R. G., 1964, Principles of Inverter
LSI = Large-scale integrated (circuit) Circuits, New York, New York: John Wiley & Sons, (ISBN-10:
LCI = Load commutated inverter 0471061344).
LV = Low voltage
Bose, B. K., 1981, Adjustable Speed AC Drive Systems, New York,
MCT = MOS controlled thyristor
New York: John Wiley & Sons.
MMF = Magnetomotive force
MOS = Metal-oxide semiconductor Brichant, F., 1984, Force-Commutated Inverters, New York, New
MOSFET= Metal-oxide semiconductor field-effect transistor York: McGraw-Hill Publishing Company (ISBN-10:
MTO = MOS turn-off thyristor 0070077134).
MV = Medium voltage Cornman, R. E., 1986, “Analytical and Experimental Techniques
MVA = Megavolt ampere for Solving Pump Structural Resonance Problems,” Third
MW = Megawatt International Pump Users Symposium, Turbomachinery
NPC = Neutral point clamped Laboratory, Texas A&M University, College Station, Texas,
ppb = Parts per billion pp. 27-32.
PF = Power factor
PLC = Programmable logic controller Dickau, R. and Pardo, C., 2004, “Centrifugal Pumps in Heated
PU = Per unit (ratio of actual motor torque to rated (static) Bitumen Pipeline Service,” Twenty-First International Pump
motor torque) Users Symposium, Turbomachinery Laboratory, Texas A&M
PWM = Pulse width modulation University, College Station, Texas, pp.10-17.
RC = Resistor-capacitor Ehrich, F. F., 1992, Handbook of Rotordynamics, New York, New
SCR = Silicon controlled rectifier York: McGraw-Hill Publishing Company (ISBN 0-07-019330-4).
SGCT = Symmetrical gate controlled thyristor
WRIM = Wound-rotor induction motor Europump, May 2004, “Variable Speed Pumping—A Guide To
VFD = Variable frequency drive Successful Applications,” Brussels, Belgium.
VA = Voltage-ampere Graff, K. F., 1975, Wave Motion in Elastic Solids, New York, New
VAC = Volts of alternating current York: Dover Publications (ISBN 0-486-66745-6).
VAR = Voltampere reactive
Ghandi, S. K., 1977, Semiconductor Power Devices, New York,
REFERENCES New York: John K. Wiley & Sons.
API Standard 610, Eighth Edition , 1995; Ninth Edition, 2003; Guelich, J. F., 1999, Kreiselpumpen—Ein Handbuch fuer
Tenth Edition, 2004, “Centrifugal Pumps for Petroleum, Entwicklung, Anlagenbau und Betrieb, Berlin, Germany:
Petrochemical and Natural Gas Industries,” American Springer-Verlag.
Petroleum Institute, Washington, D.C..
IEC Standard 60529, 1999, “Degrees of Protection Provided by
Ewins, D. J., 1986, Modal Testing: Theory and Practice, Baldock, Enclosures (IP Code)”, Edition 2.1.
England: Research Studies Press.
Kosow, I. L., 1973, Control of Electric Machines, Englewood
Guelich, J. F. and Bolleter, U., 1992, “Pressure Pulsations in Cliffs, New Jersey: Prentice-Hall.
Centrifugal Pumps,” ASME Journal Vibration Acoustics.
Lienau, W., 2005, “Identification of Bearing Housing Vibrations of
Guelich, J. F. and Egger, R., March 1992, “Part Load Flow and a Large Barrel Type Injection Pump and Their Analysis by
Hydraulic Stability of Centrifugal Pumps,” EPRI report Means of the Finite Element Method,” Twenty-Second
TR-100219. International Pump Users Symposium, Turbomachinery
IEEE Standard 519, 1992, “Recommended Practices and Laboratory, Texas A&M University, College Station, Texas,
Requirements for Harmonic Control in Electrical Power pp. 1-9.
Systems,” The Institute for Electrical and Electronic Engineers, Marks’ Standard Handbook for Mechanical Engineers, Tenth
Piscataway New Jersey. Edition, 1996, E. A. Avallone, T. Baumeister III, Editors, New
IRD Balancing Technical Paper 1, 2007, “Balance Quality York, New York: McGraw Hill (ISBN 0-07-004997-1).
Requirements of Rigid Rotors—The Practical Application of Marscher, W. D. and Campbell J. S., 1998, “Methods of
ISO 1940/1,” Entek IRD Balancing, Columbus, Ohio. Investigation and Solution of Stress, Vibration, and Noise
ISO Standard 13709, 2003, “Centrifugal Pumps for Petroleum, Problems in Pumps,” Fifteenth International Pump Users
Petrochemical and Natural Gas Industries,” International Symposium, Turbomachinery Laboratory, Texas A&M
Organization for Standardization, Geneva, Switzerland. University, College Station, Texas, pp. 143-156.
Richardson, M. H., March 1997, “Is it Mode Shape, or an NEMA Standards Publication 250, 1997, “Enclosures for
Operating Deflection Shape?” Sound and Vibration Magazine, Electrical Equipment,” National Electrical Manufacturers
Bay Village, Ohio. Association, Rosslyn, Virginia.
Pelly, B. R., 1971, Thyristor Phase-Controlled Converters and
BIBLIOGRAPHY Cycloconverters, New York, New York: John K. Wiley & Sons.
API Standard 684, 1999, “Tutorial on the API Standard Paragraphs Rao, S. S., 1995, Mechanical Vibrations, Third Edition, Addison-
Covering Rotor Dynamics and Balancing: An Introduction to: Wesley Publishing Company (ISBN 0-201-52686-7).
Table of Contents
Robinett, F. L., Guelich, J. F., and Kaiser, T., 1999, “Vane Pass Scoles, G. J., 1981, Handbook of Rectifier Circuits, New York,
Vibration—Source, Assessment, and Correction—A Practical New York: John K. Wiley & Sons.
Guide for Centrifugal Pumps,” Sixteenth International Pump Sen, P. C., 1981, Thyristor DC Drives, New York, New York: John
Users Symposium, Turbomachinery Laboratory, Texas A&M K. Wiley & Sons.
University, College Station, Texas, pp. 121-138.
Vance, J. M., 1988, Rotordynamics of Turbomachinery, New York,
Schaefer, J., 1965, Rectifier Circuits: Theory and Design, New New York: John K. Wiley & Sons (ISBN 0-471-80258-1).
York, New York: John K. Wiley & Sons.
Schwartz, R. E. and Nelson R. M., 1984, “Acoustic Resonance ACKNOWLEDGEMENTS
Phenomena in High Energy Variable Speed Centrifugal
Pumps,” First International Pump Symposium, Turbomachinery The authors would like to thank the managements of Sulzer
Laboratory, Texas A&M University, College Station, Texas, Pumps Inc. and Siemens Energy & Automation Inc. for permission
pp. 23-28. to publish this paper.

VFD PDF

Uploaded by

Copyright:

Available Formats

VFD PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

VFD PDF

Uploaded by

Copyright:

Available Formats

Table of Contents

ANALYSIS GUIDE FOR

Thomas F. Kaiser has been an Ralph O. Dickau is a Senior Engineering

Figure 1. Waterfall Vibration Plot.

Measures against self-excited vibration include: Percent BEP Operation

Self-excited vibrations affect lateral rotordynamics and structural

Vane-Pass Pressure Pulsations

Figure 3. Pressure Pulsations Versus Percent of BEP Operation.

Impeller Outlet Velocity

Figure 2. Wake Flow at Impeller Outlet.

The stagnation pressure ∆Pd is a measure of the unsteady

Other Influencing Parameters

Vane pass pressure pulsations primarily affect structural vibrations

Other Excitation Sources

In case a fixed speed centrifugal pump is converted to VFD

Case Study—Standard Lateral

analyses of the entire pump and motor structure. Alternatively a

Figure 7. Calculated Receiver Can Flange Deflection.

Figure 10. Bode Plot of Response at Suction-Bell.

The calculated displacement curves UXX(f) are inverted to

Figure 8. First Three Pump Structural Mode Shapes.

Figure 12. Lateral Shaft Response Amplitude at Upper Bearing.

Figure 14: Column Vibration and Natural Frequencies.

The pump was successfully retested after implementation of

Figure 18. Torsional Stress Response.

Steady-State VFD Torque Harmonic Excitation

Figure 20. Dynamic Torque Measurements.

Startup and Short-Circuit Transient Excitations

Figure 23. Structurally Modified Baseplate and CP Pump.

Impact tests performed after the modifications indicated that the

Vertical Pump Structural Vibration Problems

frequencies and mode shapes can be identified applying FEA

The bulk modulus is a measure of the fluid’s resistance to

The ratio of reflected to incidental power of a wave hitting a

The available (system) damping that limits pressure pulsations is

Case Study—Multistage Pump

Medium-Voltage Variable Frequency Drives

Figure 34. Neutral-Point Clamped MV VFD-IGBT Version with 12-

You might also like