CHAPTER 2

LITERATURE SURVEY

2.0 Introduction
'l'here are many aspects of vibration monitoring and analysis techniques followed
by their applicability to various cases like time and frequency domain, shaft and casing
vibrations, anti friction and sleeve bearings, general diagnostic procedures applicable to
machines like fans, pumps, hydrolsteadgas turbines, generators, etc. In this chapter
an overview of relevant literature covering the above aspects of rotating machinery
in general and power plant equipment in particular is presented.

2.1 Vibration Analysis Techniques

Practical aspects of vibration parameters, units of measurement, transducer selection
and its utility in each application, types of presentation of vibration parameters, analysis
techniques involved, interpretation of vibration patterns etc., are explained in simpler
terms by Jackson[l], Catlin[2], Bloch[3] and Collacott[4]

A review of the techniques and application of mechanical signature analysis for

the purpose of testing, monitoring, diagnostics and system identification done up to 1982
is presented in the works of Volin[S] and Hundal[6]. Volin[5] covered almost all the
aspects of machine vibration monitoring like standards and criteria, trending,
measurements and instruments, data analysis, gear vibration analysis, pumps and
valves, other general rotating machinery, ground vehicles, quality assurance and
case studies. Hundal[6] reviewed the development of various analytical techniques like
random decrement technique, time domain averaging, cepstrum analysis, shock
spectrum technique, adaptive noise canceling, etc. Detection and early diagnosis of
potential failures of rotating machinery is covered in detail by Bosmans[7].
. .

The aspects of transducer selection for vibration

computer based machinery health monitoring system etc are
The perfo-ce evaluation of advanced signature analysis
Ponseele et.al,p] with two typical sampling methods.
2.2. Shaft Vibration Monitoring
The utility of vibration vector monitoring is demonstrated by Humes[lO]. It was
observed that monitoring of both vibration amplitude and phase is more et'fective in
diagnosing machine faults and predicting impending failures than monitoring vibration
amplitude alone. The importance and evaluation of shaft centre line and related
parameters such as attitude angle and eccentricity ratio is explained by Pryor[] 1] for
problems like bearing wear, bearing degradation. external and internal preloads etc.,
using two perpendicular proximity probes. The compensation of data when the probes
are not installed at 90 degrees apart is given by Bently[l?].

The utility of shaft centre line position for information about the condition of oil
film in a bearing, bearing stability, alignment of the shaft bearing system and radial
forces on the shaft is explained by Bently[l3]. Misalignment and shaft crack-related
phase relationship for 1X and 2X vibration components of rotor response are detailed
by Muszynska[141.

A summary of the results of research on fluid flow-generated rotor instabilities

in fluid handling rotating machines(such as in bearings, seals or a stator case) is given
by Muszynska and Bently[lS] discussing the phenomena and cures of self-excited
vibrations like whirl and whip and prediction of threshold of stability. The
significance of shaft centre position monitoring and typical orbits are discussed by
Mandke et,al,[l6]. The advantages of polar plots are demonstrated by Bently[l7].
The utility of phase monitoring for the identification of faults like bent shaft,
misalignment, looseness, critical speed, resonance etc., is explained by Chhaya[l8]

The use and interpretation of vibration measurements primarily for rotors running on
journal bearings is done by Morton et.al,[l9] with emphasis on location of vibration
transducers for rotor and bearing vibration, signal processing and classification of rotor
dynamic phenomena like stable/unstable vibration and interpretation. An
experimental analysis technique to diagnose sleeve bearing problems and to verify its
design is presented by Shakweer[ZO] with a case history. A theoretical analysis is
done by Bhave et.a1,[21] to show that the excitation forces due to misalignment of s h a h
is at twice the rotating frequency.
12
 An on line computer system for phase-angle monitoring in turboyenerators is
discussed by Wang[22] underlying the necessity of two vibration and one reference
signal probes for true phase angle monitoring. Notes on the analytical investigation of
rotor motion, imposed vibration, self-excited vibration, journal bearing whirl,
flow excitation by working fluid, rotating friction, unsymmetrical rotor,
interaction between transverse and torsional modes of vibration have been given by
Smith[23] to help in the recognition of the causes of rotor vibration in turbo machinery.

2.3 Condition Monitoring And Fault Diagnostic Systems

Xiuzhu et.al[24] presents a vibration monitoring and diagnostic system based on
pattern recognition with its structure and functions including on-line application of system
to common faults like oil whip, radial dynamic rubbing and misalignment. An
optimum maintenance approach using condition monitoring and fault diagnosis is
presented by Taylor [25] and Haddad[26] giving the four levels of vibration
measurement and analysis for condition monitoring and fault diagnosis.

Vibration monitoring is one of the techniques for identification of impending and

incipient failure condition and forms an important pan in the development of systems
and methods for real-time condition monitoring of mechanical machinery[27]. The role of
software in predictive maintenance is shown by Carey[l8] with features like enveloping
for discrete fault alarms, curve-fitting of trend data, alarm prediction and fault
diagnostics to identify the specific fault based on theoretical calculations for the
particular machine configuration.

The concept of predictive maintenance is debated by Scheithe[29] with relative merits

and demerits of damage related and time related condition maintenance. The
economics of condition monitoring methods is illustrated by Neale et.al.[30] with total
conceivable gross annual savings from condition monitoring. The host of rotating
machinery vibration problems and symptoms with probability distributions is
documented by Sohre[31], Jackson[l] and Collacott[4]. An evaluation of rotating
machinery vibration by means of fixed and portable systems is given by
Cochard[32]. Predictive maintenance strategy with application of computers to condition
monitoring is discussed by Tranter[33].

A production line dedicated rotating machinery analysis is discussed by Middleton[34]

and Ikeuchi et.a1,[35] for the vibration and noise testing of variable speed machine with
order analysis and illustrated by the example of a gear box noise testing system in the
former and marine diesel engine Campbell diagram in the latter. Serridge[36] gives
ten crucial concepts behind trustworthy fault detection in machine condition
monitoring. The development of an integral PC-based predictive maintenance expert
system for automated machine diagnostics is described by Petersen[37] which
automatically scans vibration spectral patterns and determines the source and probable
causes of vibration, with recommendations for corrective action.

An integrated machine vibration data collection and analysis system is explained by

Stouppe et.a1,[38] where the vibration data could be transmitted from different places
across a whole country to a central computer using a modem over standard telephone
lines and networking. Without human interruption, the computer system processes the
vibration measurements to identify machinery which are in a condition of alarm or are
trending towards such a condition along with a summary list of potential problems and
recommendations through the use of a diagnostic expert system. Bates[39] deals with
alarm envelope and order tracking and cost effectiveness of the order of 75% in
maintenance.

2.4 Anti Friction Bearings

The bearing characteristics for rolling element versus fluid film bearings are vividly
described by Harker et.al, [40] in respect to their principle of operation, relative
motion, life, damping, shock absorbing qualities, vibration transmission etc. Frequency
domain interpretation of vibration signals, causes of premature failure and stages of
failure are explained. Different techniques of bearing condition monitoring are
discussed by Bannister [41] with respective severity criteria, standards, failure zones ,
and illustrations. The derivation of the equations of motion for roller bearing
components is made available by Palmgren [42]. The crest factor principle is well
explained in [41],[43],[49]. The concepts of bearing failure, beating frequency
formulae, merits of overall vibration level monitoring and constant percentage
bandwidth comparison are presented in [43].

An advanced vibration analysis technique known as High Frequency Resonance

Technique (HFRT) or envelope analysis wherein early diagnosis of the condition of
rolling-element bearings is possible, is described in detail
in[41,43,44,45,46.47]. While analog envelope analysis is presented in the other
references mentioned above, Joelle et,a1.[47] deal with the digital envelope analysis
where the spectrum is zoomed at the resonant frequencies followed by the Hilbert
transformation. Theoretical background of Hilbert transform technique in machine
diagnostics is given by Randa11[48] with illustrations for roller element bearings.
The effect of speed variation does not appear to cause large changes according to
Darlow et.a1,[44, 451 establishing the fact that once a resonance has been selected
as an information carrier, it can be used over the entire operating speed range.

Another high frequency technique known as the shock pulse measurement (SPM)
[41, 49, SO, 51, 52, 53, 54, 561 consists of the measurement and analysis of the
maximum value of high frequency component of the high pass band filtered signal to
get the damped transient. Typical bearing condition criteria on SPM are given in [41]
and [50]. As described by Howard (491 this shock pulse (i) is a sharp rise short duration
pulse of energy related directly to the severity of the contact in the bearing element
contact ellipse @)depends on sound propagation (iii)is attenuated (14dB or 80%) at
mechanical interfaces - thereby enabling bearing condition assessment by measuring rate
of occurrence of shock emissions in relation to amplitude.

Typical shock emission profiles of good and damaged bearings and SPM charts for
commercial instruments are dealt by Howard [49] and Brown [SO] respectively.
Some experimental observations and results are illustrated by Smith[51] comparing
acceleration, shock pulse. acoustic emission and jerk measurements for slightly damaged
bearings at medium to low speeds. High frequency detection is discussed in [52] where
peak detection is recommended due to the fact that it includes total
impact spikes generated by the flaw. The technical
monitoring procedure, selection of pick up points etc., is
 The superiority of an eddy current system is established by Spencer
et.a1,[54]. Low speed bearing monitoring using acoustic emission transducer is
illustrated by McFadden et.al. [55] in which the base strain sensitivity of the
transducer has been utilized. Improper lubrication which constitutes 43% of the sources
of bearing failure is explained in [41] and [56]. Allenby[56] explains the
importance of diametrical clearances and circumferential clearances in SPM technique.
Bannister[41] describes the assessment of lubrication condition using SPM. thereby the
detection of lubricant starvation and bearing condition.

Bearing analysis using a fiber optic probe that focuses directly at the outer race of the
bearing is presented by Phillips[57] where the reflected light is calibrated to obtain
the changing gap.

The condition monitoring of fans with roller element bearings using the low
frequency, high frequency and the shock pulse technique together with a dual
frequency approach is discussed with numerous examples in [58]. The effects of
geometrical imperfections like inner ring waviness, varying roller element diameter
is dealt in detail by Sunnersjo [59] for a lightly loaded radial bearing with positive
clearance and moderate speeds. McFadden[60] made an effort to extend the model for
the high-frequency vibration produced by a single point defect on the inner race of a
rolling element bearing under radial load to describe the vibration produced by
multiple point defects. The model incorporates the effects of bearing geometry,
speed, load distribution, transfer function and the decay of vibration.

Ball bearings with artificially introduced defects have been used to study the
practical limit of defect detectability by Miyach et a1[61]. Life tests are also made and
results indicated that in most cases bearing damage initiation can be detected early
enough before catastrophic failure occurred, by monitoring the time history of
bearing vibration. This m;ts so because the growth speed of flaking was not usually very
fast. Jixi et.al,[62] proposed a detection method directly by analyzing the vibration
signal to detect the quality of the contact surface for ball bearings.

Computer simulation and application of the cepstral analysis to the ball bearing fault
diagnosis by Jae et.al,[63] show that power cepstral analysis is a very effective
17
method to distinguish between direct and reflected waves. This method has been found
to be better than other conventional methods based on RMS and Kurtosis values and
probability density function(PDF)

The adaptive noise canceling technique(ANC) is described by Bannister[41]

and Chaturvedi et.a1,[64], the former containing the principle in detail and the latter the
procedure, illustrations and experiments. This method consists of filtering the
background noise out and is a non-coherent technique in which use is made of an
auxiliary or a reference input derived from one or more sensors located at points in
the noise field where the signal is very weak or undetectable. This input is
adaptively filtered and subtracted from a primary input containing both signal and noise
resulting in an attenuation of the background noise.

A brief review of the parametric spectral estimation method as well as a report on

some preliminary work that compares with the FFT technique when applied to low
speed rolling element bearing is presented by Mechefske[65]. Various alarm indicating
algorithms are described by Peterson et.a1,[66] including overall, envelope and
selective frequency band alarming with applications to general sleeve bearing and
roller bearing machinery, four stage centrifugal and rotary screw compressor,
precision spindle and steam turbine.

Van dyke[67] developed an algorithm where bearing fault related peaks in machinery
vibration signatures are identified in order to quantify rolling contact bearing fault
with several examples of automated expert diagnostic system using cepstrum analysis. A
technique of approximating rolling element bearing frequencies based on a data base of
over 2000 bearings from six different manufacturers is discussed by Schiltz[68], the
input being the rotating speed and number of rolling elements. A unique necessity of
incipient failure detection of roller bearings in rocket engine (for NASA)handling
liquid oxygen is reportcd by Jolly[69]. Acoustic emission technique is found
most effective technique for incipient failure detection and corrective
Undetected, these failures might lead to major fin due to friction arising o
minor fault . lianfang et.al.[lO] have shown that the combination of
demodulated resonance or Kurtosis analysis technique is an effective
. . .
bearing vibration signal extraction and fault detecti,on as illustrated by a case study of
a piston pump. Use of logarithmic amplitude scale is recommended by
Archambault[71] for trending overall velocity and acceleration levels, introducing a
new bearing severity chart based on crest factor in dB.

Sound intensity technique using two microphones is used for defect detection in rolling
element bearings by Tandon et.a1,[72]. Envelope method gives better results for
incipient defects than advanced stage defects as demonstrated by experimental
testing conducted by Amato et.a1.[73] on bearings having simple and multiple defects.
Grey-system theory of forecasting where known information is used to predict the
unknown (used effectively in economics and agricultural problems), is applied by Kuhnell
et.a1,[74] with illustrations in the prognosis of roller element bearing defects such as fan
roller bearing and also in vibration acceleration of a rail road diesel engine.

2.5 Fans And Compressors

A pictorial guide to the interpretation of frequency spectra is given by Bickel[75] with
illustrations for fan acoustic spectrums for 1/1 octave, 1/3 octave and narrow band
analysis. Effects of common issues like change of foundation, transducer location and
speed variation are clearly depicted. An overview of condition monitoring methods
with emphasis on industrial fans is elaborately given by Cory[76]. Development of a
PC-based off-line expert system for evaluation of turbomachinery response as applicable
to a centrifbgal compressor is reviewed by Kirk et.a1,[77] with a unique reasoning system
to evaluate 'yes', 'no' and 'don't know' response in the inquiry session.

Vibration testing for diagnosing the problems in centrifigal and axial flow fan
units is described by Cory[78] giving acceptance standards for different fan sizes and
the frequencies that can help to diagnose typical fan problems in bearings, belts, blades,
gear boxes, motors etc.

2.6 Motors And Pumps

Development in pump condition monitoring is summarized by Murray et.al,[79] with
special emphasis on microprocessor based vibration condition monitoring and analysis.
Pump f a i l w mode forecasting through the use of an integrated diagnostic
methodology(1DM) is presented by Silva 1801 giving information on condition
monitoring, system diagnosis, failure prognosis and prescriptive action to provide
means to detect failure rnodes(insitu), failure state. time to failure and maintenance
(if needed). Induction motor vibration diagnosis with techniques to identify machinery
andlor magnetic(e1ectric) faults is given by Maxwell[8l]. Vibration monitoring in
circulating water pump for diagnostics. failure analysis and remedies together with
vibration criteria is discussed by Arora(821. Applications to low speed machinery
and some case studies of gear tooth spalling for a jack shaft and looseness and bent shaft
for a steam turbine driven boiler feed pump are given by Taylor[25]. A computer based
predictive maintenance system for periodic vibration monitoring on multi-stage Boiler
feed pumps is presented by Chen[83].

2.7, Hydro Turbine Generators

Development of a condition monitoring system for hydro turbine generators, which
uses non contact and seismic probes for shaft and bracket vibration monitoring and
incorporates features such as trend display, frequency analysis, alarm display, has
been reported by Eriksson [84]. Different types of forces act on various components of a
hydro turbine generator set, resulting in deformation of the individual components and in
the vibration of the set as a whole according to Vladislavlev[85]. The forces causing
vibrations in a hydro set can be mechanical, hydraulic and electrical, which can be again
subdivided into steady state and transient types. The influence of different design
parameters on the critical speed of a hydro turbine generator is described by Walter [86]
wherein the actual critical speed finally is decided by not only the rotor and guide bearing
design but also the stiffness of the bracket, the stiffness of the connection between the
bracket and the foundation, the stiffness of the foundation itself, etc. Use of rotor
dynamic models in predictive maintenance is described by Bettig et.al[87] in which
variables characterizing the state of a deterioration mechanism are determined by trending
from on-line measurements and also simulation is performed to predict either the machine
life or the maintenance period.

2.8 Steam Turbines And Generators

Kuo[88] describes a microcomputer-based vibration diagnostic system for steam
turbines and generators consisting of on line data acquisition, monitoring the alarm
levels, trending, diagnosing and guiding to rectify the most common fault like
unbalance. Besides illustrating the system function and system structure, a diagnostic
chart is also presented for detecting the cause of vibration from the observed
phenomena.

Off-line monitoring of large turbine generators in Central Electricity Generation

Board(CEGB) is described by Thomas[89] with examples of the detection and diagnosis
of a failed turbine blade and cracked generator rotor. The advanced system based on
vibration measurement for various rotating machinery of a nuclear power plant is
described by Higuchi et.al.[90].

An expert system approach has been dealt by John Scheibel et.a1,[91] that
automates the diagnosis of a transverse rotor crack in a steam turbine and other
common vibration fault signatures. An integrated diagnostic system to define
malfunction conditions of a turbogenerator is illustrated by Diana et.a1.[92].

A review of the vibration monitoring and diagnostic system and measurement

techniques applied to a modem 2000MW power station is illustrated by Brown[93]
containing main turbo-alternator vibration measurement system. Application of an
expert system to rotating machinery health monitoring is featured by Hill et.a1[94],
giving the key elements of an expert system, its functional designs and overview, design
consideration and knowledge engineering establishment.

An Electrical Power Research Institute(EPR1) perspective of fossil power plant

applications of expert system is made available by Valverda et.a1,[95] wherein along
with the other performance projects, electrical generator monitoring system and turbine
condition monitoring system are featured. The knowledge base of turbine condition
monitoring contains about 150 rules and diagnostic strategies directed towards
seven major fault types which are attributed to 26 specified mechanical failure cases.

Another turbo generator monitoring and diagnostic expert system is developed by

Kato et.a1,[96] with features like (a) diagram-based domain-specific knowledge
representation models to describe the large volume of expertise for knowledge
acquisition and for reasoning explanation and (b)the diagnosis reasoning process
containing real time diagnosis and detailed diagnosis.
 Two methods for on line monitoring using Mobile Diagnostic Centre(MDC) are
presented by Barger[97]. While one is the evaluation of flow-related excitation of
turbine stationary components, the other is concerned with analysis and evaluation of
cracked generator.

A Microcom~uter based rotating Machinery vibration Monitoring and Diagnosis

system (MMMD-11) for a 200 MW turbogenerator in a power plant in China is
sumeyed by Wen-HU et.a1.[98] and Xu Min[99]. A newly developed 'fuzzy diagnosis
technique' is used in the system in which thc malfunctions and symptoms are
indicated by their membership functions. The malfunction such as bending of the
rotor due to the temperature difference and rubbing, loss of parts, misalignment, oil
film whirling and whipping etc., can be diagnosed in this system. An approximate
method to calculate the behaviour of a crack in the presence of a thermal field, in a
general rotor system is presented by Diana et.a1,[100] assuming a simplified stress
distribution around the cracked area.

2.9 Gas Turbines And Generators

An expert system approach to fault diagnosis is described by White[lOI] based
on a fault matrix model using a set of rules and Baye's algorithm for evaluation of
results. Sum and difference frequencies in vibration of high speed rotating machinery
is illustrated and validated with spectral analysis by Ehrich[l02] and Eshleman[l03].
An assessment of gas turbine vibration monitoring is made by Lifson et.al,[l04] with
guidelines for proper transducer selection, severity criteria, economics of the monitoring
systems, data base analysis giving the dominant vibration related outages. The
advantages of modem vibration analysis in condition monitoring are listed by Neil
Baines[lOS] giving the falselmissed alarm trade off cycle with energy level
monitors. The paper by Eisenmann[l06] addresses itself to the description of shaft
and casing vibration ::sponse characteristics of large, single shaft. industrial gas
turbines with two typical examples.
2.10 Case Studies
A systematic machine condition monitoring, containing 6000 measuring points for a
paper mill machinery is presented by Sullivan[l07]. Case studies are discussed by
Sullivan[lO8] consisting of looseness of foundation of a piston compressor and an
electro-lift motor, unbalance and misalignment at a paper mill, structural problem in a
fan and a screw compressor with successful corrections in all cases.

A monitoring system which avoided unscheduled breakdowns of various types of

machines in a petrochemical plant is described by Brown et.al.[l091. Test-bed
monitoring in automotive and aircraft industry is featured by Brackenboro[llO]
dealing with complete vibration test analysis. Machine condition monitoring using
vibration analysis in an iron-ore mine is made available by Brown et.al,[lI 11.

Some case histories in vibration monitoring are given by Wong[l12] concerning sub-
synchronous and out-of-balance vibration in gas turbine and compressor trains,
excessive vibration and bearing failure in a centrifugal pump motor, vibration in a
compressor train, rotor blade fixing in a steam turbine etc. Low speed machine
vibration monitoring as found in the mining industry is shown by Penter[ll3]. A case
study of vibration of an utility set where the bearing problem was confirmed by the
presence of predominant sub harmonic(<O.SX) component is presented by Panwalkar
et.al,[llrl].

The basics of practical vibration along with various rotating machinery faults
and successful case studies are reported by Rao[l 151. Saavedra[ll6] narrates the
experience from fault of critical rotating machines containing motor, centrihgal
pump, epicyclic gear box, Pelton wheel turbine and a vacuum pump. Jones[ll7]
presents some case histories in machinery vibration and acoustic emission analysis
covering operating problems and solution for 50,000 RPM cooling fan with
aerodynamic induced vibrations. A review of various rotor balancing methods and the
application of a minicomputer for multi plane balancing of 20KVA micro alternator
and 70MW turbo generator rotor using least square approximation is given by
Ramakrishna et.al,[ll8]. An overview of vibration analysis and condition monitoring
techniques are crisply described by Gardiner[ll9] and Mathew[l20] respectively.
Common vibration problems and procedures of diagnosis in hydro turbine generators is
explained by Sengupta[lZl]. The cumulative sum charting procedure can be used to
establish a knowledge base of fault data for Artificial Neural Network (ANN)
implementation based on historical parameter measurements according to Hott and
Griffiths 11221. A W can provide accurate, incipient fault diagnosis of machinery based
on real industrial data corresponding to historical faults.

2.1 1 Present Work

The above literature still warrants the necessity of an integrated approach for on-line
vibration condition monitoring, analysis, assessment and diagnostic guidance systems
based on detailed vibration signal analysis techniques. correlation analysis between
vibration and critical operating parameters, complete rotor dynamic characterization of the
targeted machinery and finally access of information to all the key personnel even at
remote locations, through an intranet-web site. The aim of this thesis work is to devise,
develop and implement such systems on a desk top computer with add-on data
acquisition cards. The systems and methodologies are built for on-line vibration
condition assessment and diagnostic guidance not only based on vibration signal analysis
but also with the aid of operating parameter-vibration correlation analysis and rotor
dynamic characteristics. Further integration is achieved to cater to the company's multi
product environment by customization of systems to hydro turbine generator, steam
turbine generator, boiler feed pumps and industrial fans and building an intranet web site
containing the diagnostic guidance methodologies for all the above.

The salient features of the present work are

3 On-line vibration monitoring, analysis, and condition assessment based on
vibration signal analysis in time and spectral domain, spectral distribution,
spectral trending etc along with multi plane balancing routines based on
influence coefficient method, for insitu balancing.
> Custornization to hydro turbine generators, steam turbine generators and
boiler feed pumps.
9 Correlation analysis between critical operating parameters and triaxial
vibration behaviour and generation of sorted correlation bar charts
P Rotor dynamic characteristics support with critical speed and mode shape
analysis, unbalance response characteristics and parametric studies.
P Intranet web site for interactive guidance software to aid in ma1 function
diagnosis based on general mechanical fault diagnosis, correlation analysis
and rotor dynamic characterization