PICKING UPBAD VIBRATIONS
Machinery condition monitoring isgaining increasing general accept-ance throughout industry as ameans of achieving reduced main-
tenance costs, increased machine reliability andincreased safety standards. The correct andefficient application of condition monitoring torotating machinery should enable plant andmaintenance engineers to predict the major eventsaffecting machine's mechanical condition (iebearing failures) and plan corrective action inadvance, thus reducing or avoiding the amount ofroutine maintenance work done at fixed intervals.
An effective condition monitoring programmewill usually offer significant economic advantagesover routine or breakdown maintenance by givingearly warning of maintenance requirements sothey can be incorporated into a scheduled workprogramme to reduce down-time and lostproduction. The condition monitoring developmentgroup at Acoustic Technology is engaged indeveloping and applying techniques for effectivemachinery condition monitoring. Pilot projectshave been undertaken to assess the effectivenessof different condition monitoring techniques indetecting early warning of failures in machineryused in arduous environments.
The approach taken was to use vibrationmeasurements whenever possible as the basiccondition data, as these could easily be taken onsite using portable instrumentation with themachines in operation. Techniques such as oilparticle analysis were considered, but the majorityof machines encountered had grease lubricatedroll ing element bearings. Temperaturemeasurements taken on slow speed conveyorbearings proved generally inconclusive. Othertechniques such as motor current spectral analysisand gearbox spectrum analysis were initially usedas condition indicators but simpler techniquesproved just as powerful. The vibration measureprogramme was therefore specifically aimed atidentifying defects in the rotating elements of themachines and damage to the rolling elementbearings.
Overall vibration velocity levels measured onthe bearing housings of the machines combinedwith frequency spectra in the range of 0 to 1 kHzwere used as the prime indicator of defects in therotating elements of the machines. Vibrationspectra was obtained at the start of the project andthen additional frequency spectra recorded on amonthly basis, where the overall vibration levelhad increased significantly during the previous
Roger Smith, Marketing
Director of Acoustic
Technology, describes theapplication oj various condition
monitoring techniques tomachines in arduous service.
month. Although simple in nature, this basictechnique quite clearly demonstrated its ability toidentify a range of defects such as couplingimbalance, loose fitting bearings and loose holdingdown bolts.
The identification of defects in rolling elementbearings by means of vibration techniques is morecomplex, and beset by two main problems in thetype of machinery under consideration. Firstly, thebearing noise is often masked by low frequencyvibration of the machine. Secondly, where there isconsiderable impact noise the spurious peaks can
be misinterpreted on some instruments as beingdue to bearing damage. Three different techniqueshave been investigated for rolling element bearingcondition assessment: shock pulse measure-ments, Kurtosis measurements and envelopeanalysis.
The shock pulse method made use of ahand-held battery powered meter with a probe heldagainst the bearing cap. Where there is a defect inthe internal surface of a rolling element bearing,the vibration signal from the bearing will contain aseries of peaks generated each time contact withthe defect is made. In shock pulse measurementsthe peaks from a damaged bearing excite a highfrequency accelerometer within the probe (at32 kHz). The gain required to just detect the peakis recorded together with the gain required to givea continuous stream of peaks (called the peak andcarpet levels). This data can be trended, and insome cases the onset of bearing faults identified.
Kurtosis, which is defined as the fourth momentof the frequency spectrum is, in reality, a measureof the peakiness of the spectrum distribution.
The meter used was a battery powered devicethat filters bands of bearing noise and then countsthe number and levels of peak signals in a giventime. The meter gives an indication of bearingcondition, and also numerical results can betrended. Both the shock pulse and Kurtosis
. Defect
Loose Holding Down Bolts
Inadequately Lubricated Bearings
UnbaTas^ ' .
Damaged. Bearings Requiring Replacement
$h*af£/Be>arit$g. Looseness
Drive B.elt Faults
Misalignment
Hifh Structural Vibration/Strain
Gearbox Faults
Number
18
14
8
14
4
2
4
4
2
Individual machine defects encountered by a monitoring programme
MANUFACTURING ENGINEER FEBRUARY 1990
methods were found not to be suitable for theslower speed machinery, and also were signifi-cantly affected by impact noise from conveyors,shakers and vibrating screens.
To overcome these particular problems the useof envelope analysis was investigated. This is basedupon frequency analysis of the envelope of the timesignal. A time signal's envelope is the amplitudemodulation of that signal. The envelope of a vibrationtime signal obtained from defective bearings willcontain discrete peaks with a periodicity determinedby the location of the fault. Spectrum analysis of theenvelope signal will produce a harmonic series witha fundamental frequency determined by thecharacteristic bearing frequency (stationary fault),or, in the case of rotating faults, the spectrum willalso contain sidebands due to the amplitudemodulation within the envelope signal. The followingenvelope measurement technique was employed.First the acceleration signal from the vibration meterwas recorded on a tape recorder with a flat highfrequency response. Then, in the laboratory, thesignal was filtered into three frequency bandsextending from 2.5 kHz to 20 kHz. The signal wasprocessed in this way to remove the low frequencyvibration (1x and 2x) associated with rotatingmachinery which can otherwise mask bearingvibration. An enveloping circuit was then used todemodulate the signal to reveal the low frequencymodulation components caused by bearing defects.The demodulated time signal was split into itsspectral components using a real time analyser. Thecharacteristic bearing defect frequencies could thenbe used to identify any peaks present in thefrequency spectrum. This technique was successfulin detecting, for example, cage faults on thebearings of grading screens and shakers. The faultsdetected in this way revealed early damage due tobrinelling, the onset of damage due to corrosion ofan inner race and poor lubrication. In all cases thefaults were detected with enough time to planremedial action.
In projects of this type the initial period isgenerally used to identify basic engineering faultson the plant which have to be rectified beforemeaningful condition monitoring can begin inearnest. Such problems include inadequatesupport structures, poor alignment and imbalance.On plants where machinery is of the reciprocatingtype, loose holding down bolts were a commonfault (see table). Prompt corrective action for thistype of fault prevents possible damage to bearingsand the rotors themselves. In terms of thecondition monitoring techniques employed, theoverall vibration level measurement, whilst basic inconcept, still proved to be a powerful tool in thedetection of general machinery faults. Shock pulseand Kurtosis measurements were successful inindicating rolling element bearing condition onsome machinery, but only the enveloped analysisproved successful on the slower speed machineswith inherent high background noise levels.For more information on Acoustic Tech-nology circle ME47
COMPACT ANALYSISOF VIBRATIONA COMBINATION of sophisticated analysis ofresults with a compact size and weight makes theVibrospect machine vibration analyser a powerfultool for both product development and routinemaintenance.
Available from INA Bearing Co Ltd of SuttonColdfield, the Vibrospect FFT is more than avibration measuring tool, it also includes in theoperating menu a two-plane balancing program.
Menu-driven software permits the user to selecta wide range of test modes. Results are displayedon a large LCD or graphic display screen or abuilt-in printer can provide a permanent record.The unit weighs 6 kg and is battery powered.
Routine measurement for monitoring a largenumber of machines can be made by using thememory facility which can store data on up to 128elements. For specific bearing analysis, the unit isprogrammed to differentiate between failure of theinner race, outer race or bore and roller failure.
For vibration measuring and analysis, 11frequency ranges can be selected from 0-10 Hz upto 1-20 Hz. The measured vibration profile can bedisplayed over a time base to monitor whether thesignal is constant or fluctuates over time. Vibrationcan be analysed either by velocity, acceleration,displacement or by enveloping.
By using the so-called waterfall diagram, anychanges in the vibration pattern can beemphasised. Here a series of 16 spectra are
recorded over a pre-determined time base. Theseare displayed in a 3-D graphic mode, withamplitude and frequency in the X and Y axes andtime displayed along the Z axis.
Irregular vibration patterns can be recorded byusing the transient measurement mode. This canalso be used to quantify the resonant frequency ofcomponent or structures. This is particularlyrelevant to the R&D environment.For more information circle ME48
TRANSFORMER HEALTHCONDITION MONITORING techniques have beendeveloped to cover a broad range of industrialequipment items, and some of these techniqueshave been applied to transformers in an attempt toassess their current state of health. Thermographyand low frequency vibration have enjoyed somesuccess, but do not offer a specific diagnosis andprediction of future performance. Oil analysis alonemay indicate the need for an oil change, but littlemore.
A substantial body of research within theCentral Electricity Generating Board (CEGB) led tothe development of a technique based on theanalysis of gases dissolved in the oil which doesindicate both current transformer health, and willpredict the future performance of the unit. Thistechnique is now available to industry, and in itsearly months of release has been taken up by wellover 200 operators and 700 transformers.
. Called Transcheck, the service requires that asmall sample of oil is taken from the transformeron a routine basis (perhaps once or twice a year).The sample is analysed for dissolved gases,
acidity, moisture, solids content and dielectricstrength, and the results submitted for diagnosisusing the conclusions of the research, which havebeen included in a computerised diagnosissystem.
The diagnosis indicates both the currentcondition of the transformer, and when more thanthree samples have been taken over a period,offers a prediction of the future performance of thetransformer. Results of each sample analysis anddiagnosis are reported in detail within 48 hours tothe operator, together with recommendations foraction as necessary. All records and results aremaintained as a part of the service, furtherreducing the operator's workload.
The technique is shown to detect 95% oftransformer faults, both mechanical, electrical andthermal, in advance of equipment trip or failure.The technique provides operators with a reliablemeans of assessing equipment condition in anarea which is often extremely important to plantand factory operation.For more information circle ME49
MANUFACTURING ENGINEER FEBRUARY 1990
SPC IS A WAY OFLIFE AT AVON AMES
Computerised statistical process con-trol has been so successful incontrolling the grinding of rubberrollers at Melksham-based Avon
Ames that for more than a year no roller has beenrejected because it falls outside manufacturingdimensional tolerance.
And those tolerances are tight. In their finalgrinding, the photocopier rollers, manufactured bythe company that is part of Avon IndustrialPolymers - itself a subsidiary of Avon Rubber -must be finished to within 10 microns. And that ona roller 310 mm long and 30 mm in diameter.
Since Cadar computerised SPC was introducedmore than a year ago, the reject rate for grindingerrors has been cut from around 7% to zero. That isan achievement even for a company that throughtotal quality management is striving for higherstandards all the time.
Avon Ames manufacturing success story hasbeen built partly on the intrinsic benefits ofcomputerised SPC and partly on the way in whichit has been implemented, according to Tony Giles,the company's quality control manager.
He says: "When implementing SPC it isimportant to look closely at the process you aretrying to control right at the outset. Almostcertainly, you will not want to monitor all aspects ofit all of the time, but certain aspects most of thetime. Setting up an SPC system that concentrateson the really critical factors builds your confidencein your process."
SPC has become something of a way of life atAvon Ames, as part of the company's total qualitymanagement approach. Its use in making thephotocopier rollers is an example of its allpervading influence on the quality of the finalproduct.
It starts with the steel cores of the rollers thatare shipped in from outside suppliers. When a newbatch of cores arrives a sample is measured andthe results SPC-charted as part of goods inwardcontrol. At least 10 critical measurements aretaken on each core including five measurements atdifferent points along its length in order to checkfor the uniformity of its diameter.
The end journals are also measured and the TIRfigure - based on a theoretical straight linethrough the centre of the core - is calculated inorder to check for straightness. Currently, AvonAmes charts these SPC results manually, but Gilesdoes not rule out switching to computerised SPCat some stage in the future.
After passing through various manufacturingprocesses, which includes covering the steel coreswith rubber, the final, and critical stage, is thegrinding with the results of the final grindcontrolled by an SPC computer workstation. Thatterminal is close to the action down on the factoryfloor installed in its own cubicle near to thegrinding machines.
The workstation, based on Olivetti hardware,was supplied by Cadar Measurement and ControlSystems, the Sheffield-based quality and process
management specialist. It runs Cadar's ownvariable measurement software package and somespecially written software, also produced by Cadar.
Rollers are measured using a non-contact lasermounted in a specially designed jig and linked tothe Cadar workstation by means of a PLC. Thelaser moves down each roller and takes 19diameter measurements. These are automaticallyfed into the Cadar workstation and recorded onscreen either immediately or at the end of thatroller's measurements.
The results are recorded on screen in two ways.Just five of the dimensions are critical and these arehighlighted in a statistical table. In addition, all 19results are charted graphically, using a speciallydesigned scattergraph so that the operator gets avisual indication of what the roller looks like.
Giles says: "The screens have been set up togive the optimum performance. The operator getsto the information he wants quickly by usingsimple menu commands." Part of that informationis an XR chart of the results of a batch of rollers'measurements. This gives an indication of how theprocess is working and drift can also be spottedand corrected at an early stage.
Avon Ames' commitment to SPC is just part ofthe company's dedication to total quality manage-ment. That philosophy is practically seen in theway in which the company sets up multi-discipli-nary teams to tackle production problems.
Says Giles: "We take the view that everybodymakes a contribution to quality at the end of theday. And a guy who's running a machine will knowmore about how it works than anyone else."
It is an approach that has kept Avon Ames atthe front of the market for its specialised rubberproducts. But Giles and other members of AvonAmes management team recognised that strivingfor quality is a way of business life. So SPC is likelyto play an even larger part in the company's totalquality management in the future.For more information on Cadar circle ME50
GRAPHICMONITORINGDISPLAY
PHILIPS INDUSTRIAL AUTOMATION has added apowerful new data acquisition and monitoringsystem to its wide range of condition monitoringsystems for plant installations. The company isalready well established in machine monitoring forvibration, speed, displacement, pressure, tempera-ture etc and the new data acquisition systemprovides a convenient means of displaying plantparameters with powerful built-in colour graphicsand animation.
Designated the PR3000 series, the new systemhas been specifically developed for use in
conjunction with Philips RMS700 turbo-machinemonitoring system, but is ideal for supervisingwhole plant such as automotive production lines,refuse incinerators, large scale chemical pro-cesses and glass makers.
By using a distributed signal gatheringarchitecture, the PR3000 system can monitor theoutputs of up to 1000 transducers, locatedhundreds of metres apart. The new PR3020 isdesigned to sort and process incoming data from amachine monitoring system or other processtransducers. This data can then either be archivedfor the compilation of shift reports, used to initiateother processes, or used to signal fault conditionsto the operator.
Analogue signal monitors (ASMs) are used togather data locally from up to 76 input channelseach. They then transmit this data at high speed tothe system management computer, through afactory wide RS422 network. This configurationgreatly reduces the length of wiring required, and
fail-safe circuitry ensures communication evenafter a break-down.
Each ASM has a built-in real-time multi-taskingtransmitter, processing unit, reduction trip alarmand an intelligent standard interface. ASMs acceptinputs from all conventional sensor types,including temperature, vibration, voltage, flow,current, on-off, displacement, pressure, force,strain, noise, weight and power sensors. Linear-isation and scaling capabilities are provided.
A key feature of the system is the advanced datacompression and display capability. A colourgraphics system can be used to display clearlycurrent, historical and trend information. Mimicdiagrams, bar graphs and trend curves can assistthe operator in assessing the current situation.These can be supplemented with animationfacilities and colour changes to draw attention toalarm conditions. All data processed in this way isalso available in hard-copy form.For more information circle ME51
MANUFACTURING ENGINEER FEBRUARY 1990