Enveloping for Bearing Analysis Robert M. Jones, SKF Condition Monitoring, San Diego, California ‘The use of enveloping or demodulation techniques in ma- chinery analysis instramentation allows a user to detect fault signals in rotating element boarings during their operating his- tory. This article describes the technique and provides many examples which compare the results of various measurements fom the same bearings In the continuing effort to improve methods of gathoring data todetermine the condition of rotating equipment, another sig- zal processing technique has been incorporated into some por- table data collectors. Known by either the term “enveloping” ‘or “demodulation,” Uh technique allows tho user to detoct fault ‘Signals much earlier than previous techniques permitted, “The short answer to hovt this is done is thatthe signal pro- cessing mathematically adds up all the harmonics of the fun ‘damental froquoncy that are located inside a given filter range land precents the information in the FFT frequency range thet the user has selected. That is the short answar, now here isthe hhow as explained by a mechanical engineer. Most machinory vibration generates harmonics. If tho dy namic rango of the data collectors large enough, these signals ‘can be seen for soveral orders of the fundamental. If the dy- rname range is small, the signals are buried in the noise floor 1 referred to as “being down in the mud.” Even with ‘dynamic rango display, the harmonics still disappear ‘within a short span and cannot be seen. The trick isto capture the harmonics without including the fundamental frequency. To do this most efficiently, SKF has {neorporated into thor data collector four permanently sot digi tal filters and two methods of measurement, selected by the user depending on the frequency and vibration units of inter. fest. The frequency ranges ofthese filters are: 1) 5 to 100 Hz; 2) 50 to 1000 Hz; 3) 500 to 10,000 Hz; and 4) 5,000 to 40,000 Hz The question thon is which flte to use for which application. ‘The two methods of measurement are "Enveloped Accelera- tion” and Enveloped Velocity.” Please note that to use eavel ‘oped velocity with the SKF data collector you must use a ve locity sonsor. Thore is no intogration function and if an rometer is used in conjunction with the “Enveloped setup. the data callected will be completely errone- Tho filtor setup is best explained by using soveral actual ‘examples, Ifthe area of interest isthe bearing on a slow rote ing kiln for example, then the Ball Pass Frequency Outer Race {BPFO) and the Ball Pass Frequency Innor Race (BPFI) are de- termined, Let's assume the BPFO i 2.53 Hz and the BPFT is 4.32, He. The important thing to remember is that the solection of the filter is determined by the fundamental frequency of intor fect and that this froquency must be bolow the low frequency ‘cutoff of the filter. In this eat, since the first iter starts at 5 Hz and both of our fundamental frequencies of interost are be- low 5 He, the first filter would be selected. For another ox ample, assume we have an 1800 RPM motor-with a bearing that has a BPFO of 107.6 Hi. The filtr selected would be the third since the first filter would never see the harmonics because they are above the filter range and the second filtor would in clude the 107.6 Hz fundamental signal. The foutth filter may also capture and display the information ifthe fault is genes ating strong harmonics. However, for analysis, the filter lected should be the frst available above tha fundamental fe quency of interest. ‘What the filters do isto capture a time block of the signal the block size being the specific Hlter range. This block of data {is thon processed by high pase filtering, rectification, low pass filtering and then processing into a frequency spectrum. The result of this “black magic” is that the repetitive signals ere ‘enhancod ond the nonropetitive signals are degraded. ‘This procedure can be followed by using our example of a damaged bearing with a BPFO of 107.6 He and the third filter. ‘The fundamental of 107.6 Hz plus the second (215.2 Hz), third (922.8 Ha) and fourth (430.4 He) harmonics aro removed by the filer that starts at 500 Hz So, beginning with the fifth harmonic to the one just below 10,000 Hz, the energy in these harmon- {esis added together. Ifthe bearing is on a pump that has cavi- tation, a random nonrepetitive noise, then the cavitation sig nals are dograded and dopressod. This allows the enhanced baring fault signal to bo soon above this mathematically de- pressed noise floor. Assuming that the FFT is set for & Fg, f 4400 Hz, then the display would show tho enhanced fundamen- tal at 107.6 aad the second and third harmonics. I is very im- portant to remember that these are enhanced signals and the ‘amplitudes shown are e summation of the enesgy in the filter band. Because of the summation, the amplitudes are false and donot compere with any normal acceleration signal. However, the signals are trendable. Since the same filter band is used fach time, any increase in damage to the equipment will in- ‘tease the signal strength in that band. As the signal increasod ver time, it will show up as an increasing overall amplitude trond or the specific frequency of intorost may be trended, Le. 107.6 Ha. Although this analysis technique works very well for bear ings, it also provides eerly warning of other faults such as in ‘gear boxes, misalignment, and looseness. Any type of machin- fry vibration that generates harmonics can be examined, Bearing Fault Stages ‘The past problem with a “ono time look" ata suspected bear ing fault war that one never knew for sure in what stage of damage the bearing was operating. By the time you were posi- tive the bearing was in trouble, you wore also in trouble be- ‘cause you hadn't told anyone soon enough to get it fixed in an orderly maner. Using enveloping in conjunction with SEE technology is possible to now divide the Life ofa bearing into four stages. And the enveloped FFT clearly defines these stages for you, The following examplos will show each stage. ‘Stage ono is a good boaring. Ifthe bearing is heavily loaded the BPFO frequency may be seon. This is generated by the rol ing elements going into and out of the load zone. At one in stant in time there are three elements in the zone and in the next instant there are four followed In the next instant with throe again, This changing support inthe load zone causes the outer race of the bearing to flex microscopically, generating. nergy at the BPFO. The difference in a heavily loaded boar- ing and a damaged bearing is the lack of harmonics. As the hearing dogredes, harmonics of the damage frequency will be- gin to show in the FFT. For this reason, when doing troubli Shooting, it is important to set the F,,,_ $0 that two oF thre harmonics of the boaring frequencies ean be captured. In like ‘manner, if examining a gear box, the Fy, Should be set to cap- ture two oF iree harmonics ofthe gear mesh frequency. tage two 's a boaring with some woar as noted by the har- monies. There is no reason to change the bearing at this point Infact, have soon bearings pulled at this stage and the only damage apparent is pinpoint spalling in the races. However, ts the harmonics inerease in amplitude it may be prudent to increas the frequency of data collection. Bearing degradation {s usually linear for w period of time and can be trended, but ts the lifetime ends it will bocome nonlinear In stage three the damage in the bearing te becoming termi nal. In the FFT the fundamental frequency and the harmonics(nated Aang Trad Ramin ‘igure 1, Vibration time history ofa flawed bearing Figure 3 Bnveloped aocleretion spectrum ofa flawed bearing. Fegan it | Figute 2. Velocity spectrum ofa flawed bearing will bogin to show sidebands of bearing shaft rotating s For example, using our BPFO of 107.6 Hz on a shaft turning 1800 RPM or 30 He, tho sidobands will beat 77.6 Hz and 197.0, Hz (107.6 plus and minus 30) and the second harmonic will have sideband at 185.2 Hz and 245.2 He (215.2 plus and mi ‘nus 30), Further progression of tho damage will genarate ad ditional sidebands at 2x running spood (47.6 and 167.6). Now you have overlapping frequencies and it becomes a mess to Figure out. But beware, the bearing isin the last days ofits life and should be changed as soon as possible. ‘in addition to the spectrum information, the overall magni- tudes will give you a clue tothe condition of the bearing. Us- {ng the third filter, stage one willbe from 0.1 o 0.4 gand stage two from 0.4 to 4g, Stage three is from 4 to 10 g and stage four is a measurement of over 10 g.Ifit isa vory slow speed bear- ing using the first or second filtor, those magnitude limits are areatly reduced since the slower speeds do not generate as, ‘much energy with the same size flaw. A general rule of thumb ‘would be to reduce the magnitudes forthe first filter by a fa {or of 100 and the second filtar by a factor of 10. In the same tnanner a high speed bearing using the fourth filter would have the magnitudes increased by a factor of 4. These ranges are ‘subject to modification as more ompirical data are collected and correlated with what we see when the bearing is taken apart, Nevar make a bad bearing call using overall magnitude ‘only. Check the frequency spectrum forthe energy content and where it is being generated, “To dotermine stage four, it is necessary to incorpo analysis technique into the process. SEE (Spectral Emitted Energy] technology, which is built into the data collector Clusive with SKF, analyzes the bearing condition usin ‘oped acoustical energy generated by the rolling elements over rolling foreign particles in tho lubrication. These particles ould be from lubrication contamination or from particles Sspalled from the races. In any event, if SEE technology is used fand a signal is detected, the beating is probably in stage four ‘which says that failure may occur very soon and the machine Should be shut down immediately. I there Is no SEE signal because the rolling elements and races or particles are not ‘making contact through the oil film, then the damage analy reverts to stage three and the bearing should be monitored Closely and changed as soon as possible. SEE signals appear in the FFT at the same frequency as any of the bearing fault Figure 4. SEE spectrum of flawed bearing. frequencies or gear mesh frequencies when examining a gear box In our previous example, ifthe boaring bad damage inthe futur race and the rollers and outer race were making metal to total contact, the SEE FPT would display an energy source at 107 6 He, ‘One note on the observed frequencies. The boaring frequency calculations are for a perfect bearing, In production, this per- fect bearing doesn't exis. So therefore, the variations that oc cur will cause some variations in the actual frequency ob Served. In addition, an older bearing will have worn and changed the cloarances it had when new. Bearing Fault Diagnoses “The following spectrums will illustrate bearings in various stages with comparisons to standard velocity spectra. Although ‘we don’t usually look in the time domain except when we are Testing a gear box, the bearing flaw signals will be there ifthe flav is large enough. Usually there is so much data captured inthe time domain that it defies interprotation. In Figure 1 the inner race flaw shows up as each roller passes over the flaw. ‘Tho markers show the Interval tobe 0.00748 sec which equal the frequency of 193.5 Hz, The beering was damaged during installation. Figure 2 ie vibration signature of the same boar ing. an SEF 6201, We do see the BPFO in the velocity spectrum since the damage was extensive, but there are no harmonics fan the amplitude is low. Figuro 3 isthe same data processed ‘using enveloped accoloration. Note the difference from Figure 2. All tho nonropetitive signals are depressed and the BPFO with harmonics is very clear. Finally, in Figure 4 wo will look ft SEE spectrum, SEE unite are expressed as either SEES or SES units, In hindsight we should have sot the F,., 10 600 Hz. Ths would have displayed the additional harmonics shown in Figure 3 ‘a looking at these four spectrums one «night think that these moasurements aren't necessary. However, when it co time to make x decision on whether to shut down a machine ani work on it, use ell the tools you have to help make those ‘decisions, The better our information, the better our decisions. Figure 5 isthe spectrum of an MRC bearing that had poorly stored. It had lain on the shelf for seven years in an ‘acidic environment. The acid had etched the balls and the side bf he race. When the bearing was installed, the flaw in the race {did not meke contact with the bas, but the balls generated the ‘ ryity spectrum of a damaged boariag. Mh a | tid led | Figure 6. Enveloped acceleration spectrum of the damaged bearing. BSF (Ball Spin Frequency) st 05.0 He as it rotated. Thore was no dofect signal showing in the velocity spectrum. It only ap- ears in the enveloped signal as shown in Figure 6. The am plitudos are very low but we could hoar the flawed balls click {ng as the machine ran Figure 7 is the spectrum of a bearing where "+ Bos compatie 35 floppy disk = Zoom & 20 ke real time + Narrow band, 1/1 ‘and 13 octave analysis E:z Operation ‘9.4 Color depley + DOS Comeate 3.5° floppy disk + Buttin printer 25-1,600 ines + Zoom & S2ktt2 + Order tracking + Realtime ootave 2 Ge1B & Centronics printer interface ONO SOKKI TECHNOLOGY INC. F-4200, Circle 110 Figure 6, Acceleration specrum of flawed bearing. damaged sometime during its life, The owner expressed dis pleasure at the high amplitude of vibration and the analysis Showed the cage flaw. Further searching by sorial number found the bearing to be over 20 years old! One additional prob: Jem here wa thatthe eighth harmonic of Fundamental Train Frequency (PTF) overlaid the third harmonic of the rotating “ynamicrange + Builtinhigh speed thermal pntor + RS-252GP.18 + Ughtweight (8 bs.) wn bacdt LCD display CF-1200 Pen) ‘Builtin sensor power supply = Time, specirum and overall display + AcIDCrBatey operation + PCMCIAV 2010 ‘memory and cata Storage + Lightweight (7109) with back LCD display 2171 Executive Drive, Suite 400, Addison, IL 60101, (800) 922-7174 F-5200, circle 111, $R-5300, Circle 112 (CF-1200, Circle 113,ws Bently Nevada Corporation ote 4996 Seminar Schedule oie ate Course Description Location iad eee tee wei apt PE ee ae so tet ay ee Figure 8. Velocity specirum of flawed bearing, ‘Ann 23-26 Machinery Diagnostics May 7-10. Survey ef Machinery Monitoring Mortal Gusbee May 14-17 Machinery Diagnoses May 28-91 « Survey of Machinery Monitoring nian. June 11-14» Survey of Machinery Monitring Figure 10, Bnvloped acceleration spectrum of flowed bearing = In the United Sites call ol oe 1-800227 5514 ext. 9682 17 Water Steel Minden, Moved 83423, Circle 114 on Inquiry Card sn - Diagnose Figure 11. SEE spoctrum of «flawed bearing : speed. Not a good situation and the bearing was changed. Af- and balance i | anywhere? sor manufacturer who wanted proof that we could detect bear- 7. C71) ere. ! ing flavs. Before we arrived, he had placed flaws in ono of tho six bearings and asked us to find thon, Wo did. Figures isthe Scceloration spectrom, The fundamental BPFO doesn't show up but th harmonies do, This would alert you that something ‘was going on, but most poople don't take routine bearing mo $urement in accalration. Heres one additional note on bearing troubleshooting. Most bearings will have their BPFO end BPET at non integers ofr. tating speed. In other words otating speed is 1 and the BPFO will bo at 3 Sex or4.73», This ie valuable clue when looking Ma machine and you have no idea what bearings are installed Figure 9 isa velocity spectrum, the method most people use and there Is sbeotely nothing hee that wot cause one to ‘tony. Ofthe sx bearings io the unit, the own didnot all tvhich on hed the far, but because ofthe diferent sizes, the bearing law frequoncies areal diferent turned out the one with a BPFO of 400 tz was the bad one, sol've gone ahead and Iarked that frequency. Figure 10s the same mesurementpro- Coneed using accsleration enveloping. Ths clearly shows the @ BPFO that was not visible in the velocity spectrum. And, to show how far out the harmonics are visible, we have set the Fa, to 10k, not the usual! Finally, these data were pro- SCHENCK Sed ing Se technalog Shown a Fgue heh NT_OF SCHENCK CORPORA om otha the SEE signal le preeen whch indctes motto RERSE MIS ee NEeHT Cirele 145 on Inquiry Card6th Annual Short Course on FLUID-FILM / ROLLING BEARING TECHNOLOGIES & ROTORDYNAMIC INTERACTIONS May 21-23, 1996 ‘me course will focus on bearings, the vital tribological elenents of rotating machinery, beginning with their fundamental principles of operation through computer evaluation of their operational performance characteristics fand Limitations. Fluid-film & rolling element bearing design, application, selection, and influence on the dynamic behavior of machinery will be reviewed and demonstrated. Workshops SLL provide participants with hands-on ‘exercises in problew solving utilizing Btate-of the-art computer software Call today or visit our home page for details! (Pq RBTS, 1100 Fast Hector Street Tels (610) 825-7370 ELEM conshohsten PA 96K USA Fas(610) 825-4762 Inttp://www.netaxs.com/~ rbts! BBO Uni) machine monitor. $.006-TREBEL-2 - SCHENCK GGG tN cenama bens Citele 117 on Inquiry Card Figure 14, Acceleration spectrum of gear box beorng metal contect. Note that the amplitudes are very low so the contact is vory light “The next four spectrums are from very slow speed bearings. This application is ideal for enveloping. Thoro is no intogra tion and therefore no amplification ofthe low frequency noise present in ell data collectors. The Fy, can be set at 0 Hz and fhe spectrum will not have the typical "ski slope” low fre quency roll-off ‘Those spoctrams are from a conveyer drive gear box. The bottom bearing has|« history of failure and the customer was required to keep three days inventory an hand to maintain his “justin tims” shipments ifthe boaring filed. By boing able to know the condition of these bearings, their repair could be scheduled end the expensive products in inventory could be teduced. The rotating speed of the shaft for this unit i 8.3 RPM. ‘The BPFO for the first thro spectrums is 1.7-1.8 Hz and for the fourth, a different bearing, itis 1.4 Hz. The BPFO varied because of speed changes while the different readings were be ing taken. Figure 12 is the boaring procossod with enveloped acceleration. With the BPFO at 2.8 Hz, there are twe harmon- les displayed. But, since there are no sidebands of running speed apparent, the boarings were diagnosed as having some Tight damage but it wasn’t panic time yet. Figure 13 s normal velocity spectrum, Notice there is absolutely nothing to indicate bearing damage at the BPFO of 1.7 Hz. Figure 14 shows the eame bearing using conventional acceleration pro- cessing. No damage was noted oven at the extremely low sig nal level, Bist Figure 12 clearly shows the BPFO with harmon ies. Our advice was to keep running the machine but increasethea al her| Feuengy “Enveloped acceleration spectrum of another geor box boar Figure 16. enveloped acceleration spacteum of «damaged laboratory fest Dearing ‘igure 17. Eneoloped ecceleration spectrum of defective gear box bearing. the observations to once per wook rather than once a month. Considering the'ritical and expansive nature ofthis operation, this is « bearing that will be changed at the next opportunity ‘The next example, Figute 15, isthe other bearing in the same ‘gearbox. It is heavily loaded and just starting to show some harmonics. itis difficult to obtain examples of stage three damage. Ifthe observer is paying attention a all, then you can both feel and hhear that something is not right. The bearing example in Fig. ture 16 is from «laboratory test where we damaged the bearing, {in order to get the sidebands. You can sev thatthe frequency gap between tho sidebands and the fundamental i equal to the Fotating spood of the sbaft, 40 Hz. These same sidebands are about the harmonics also. The final exemple, shown in Figure 117,18 gear box that was in a stage three condition. The owner twas a litle ely about it. He had asked us for a dame of tho en- veloping procoss and how it could bo usod in a route setup. When we set up the Fy, we did not know the gear mesh fr quency and the owner didn’t know, so we selected an Fag, of 41000 Hz, This is fine for routine route data collection. Iwo hhad been allowed to go back and take data in a troubleshoot ing mode, we would have selected a higher Fy, probably about 2000 Hs, We barely caught the tocond harinonte at 900, Hi but the sidebande are very clear at 25 He, the shaft rotating spood. When we told the owner he had a serious problem with this gear box, ho just smiled and said, “I know, I just wanted to.oo if your date collector was any good.” It is, SouNp Ano aRATONFEBRUARY 1286 Piezoelectric Sensors and Assemblies ‘Whatever your sensor application, apability to produce ic components, attach s We've built sensor assemblies for sane medical sansors including blood a fl range of testing at qualty levels oer optic X¥ positioning, ga leak Gotectors, nertal guidance systems, a stignmert, ouch eoteen encore. wiration montering and vareus pressure monitors and dopplar ow (Our assembly capably also includes tquivalnt to mi spec standards, ineuding environmental eyeing Morgan Matroc Inc ad Morgan Mat 4 at 11444 Fo 244d. Beet aa panes MATROC See MATROC ‘irclo 118 on Inquiry Card Perfect balancing of rotors from BCR Vole TOLLFREE ¥.000-TREBEL2 © SCHENCK GQG5E8 A UNIT OF SCHENCK CORPORATION 535 Ac Seal, Deer Pak, NY 11729, “elite}242-4010 Fex516)202077 Circle 119 on Inquiry Gardim epphicakimn : Macksnery Beall monitor ong A meclinerg Can be hocked fer ts altK by meatar neg she vrbved2en cand. Tham, perpen ey FET to tee ata. Machinery Health Monitoring—an Application of Spectrum Analysis BEARING INSTABILITY MISALIGNMENT he L AA Ip be oolighn i Fig. 8.9 Examples of spectral exceedances above “windowed” spectrum. Courtesy Dymac.