THE MODEL 682A05 BEARING FAULT DETECTOR (U.S. Patent No. 6,889,553) A New Approach for Predicting Catastrophic Machine Failure

Transcription

1 TN4 THE MODEL 682A5 BEARING FAULT DETECTOR (U.S. Ptent No. 6,889,553) A New Approch for Predicting Ctstrophic Mchine Filure by Jmes Robinson Emerson Process Mngement Mitchell Illig IMI Sensors Thoms Brown NSK Corportion Ry Limburg PCB Piezotronics Abstrct Detecting mechnicl fults in berings nd mchinery hs long been recognized s being importnt for preventing ctstrophic filure nd effective mintennce plnning. The humn senses of sound nd touch were the first mechnisms used to detect mchinery problems. Electronic sensors hve since offered the bility to feel nd listen to mchinery with more precision, t more loctions, nd over more time thn ws ever before possible. Interprettion of the electronic signls delivered by these sensors hs provided the mintennce engineer with the dignostic informtion necessry to pinpoint bering fults, thus enbling more efficient nd predictble mintennce effort. However, skilled nd trined personnel hve been required to effectively interpret this dignostic informtion. As electronic sensors hve become more sophisticted, so too hve the dignostic techniques, leding to the bility of erlier detection of filures with less required skill. PekVue dt nlysis offers proven technique for the erly detection of high frequency, impctrelted filures, such s bering or ger fults due to wer, loss of lubriction, nd contmintion. This method provides mesure of the true pek ccelertion, t high frequencies, which gives n indiction of impending filure. Trending PekVue dt cn provide n indiction of problem tht cn be further dignosed using spectrl mesurements. The cquisition nd trending of PekVue dt hs required the knowledge nd use of expensive nd sophisticted vibrtion dt collection equipment, until now. The Model 682A5 Bering Fult Detector provides high frequency PekVue dt s 4 to 2 ma output signl tht cn be monitored with conventionl process monitoring equipment, such s DCS, PLC, or SCADA system. Furthermore, the unit provides second 4 to 2 ma output signl proportionl to overll, low frequency vibrtion. This low frequency signl provides n indiction of mchine running speed fults such s imblnce, mislignment, nd looseness. An nlog output signl is provided for dignostic, spectrl mesurements. The fmilir DIN ril pckge instlls conveniently longside other process signl conditioners. The Model 682A5 Bering Fult Detector uses proven, PekVue erly detection methodology, requires less opertor trining, works with existing process monitoring equipment, nd offers the dvntge of 24/7 monitoring.

2 . Introduction IMI Sensors (A Division of PCB Piezotronics, Inc.) Model 682A5 Bering Fult Detector module is DIN ril signl conditioner tht interfces to n ICP ccelerometer. The device converts the ccelerometer signl into two industry stndrd 4 to 2 ma output signls. The first 4 to 2 ma output is linerly representtive of the overll vibrtion level in velocity or ccelertion units. This overll velocity vibrtion level is cquired over Hz bndwidth, which is sensitive to mchine fults such s imblnce, mislignment, nd others tht mnifest themselves t lower frequencies round running speed or hrmonics thereof. The second 4 to 2 ma output is representtive of the high frequency (> Hz) pek g level. Mny bering problems re ccompnied by short durtion energy fults (generlly ctegorized s stress wves), which re detectble by n ccelerometer locted in the proximity of the fult. Stress wve ctivity ccompnies fults such s impcting, ftiguing, nd friction, which occur t frequencies bove Hz. Experience hs shown the mplitude (s mesured with n ccelerometer) of the stress wves provide relible indictor of the severity of the fult. The relibility is further enhnced with continuous monitoring of the stress wves for which the Model 682A5 is designed to ccomplish. The Model 682A5 observes these stress wves over time period sufficient to incorporte minimum of 6 revolutions of the mchine being monitored. A logrithmic scle is used for the pek glevel to ccommodte dynmic rnge sufficient to the glevel chnge from smooth running mchine to mchine with serious fults. This pper discusses cse studies estblishing the correltion of the pek glevels with fult detection nd severity ssessment bsed on the pek vlue (PekVue) nlysis methodology introduced by Emerson Process Mngement 2, CSI in 997. The 682A5 is then presented with controlled test results nd compred to dt obtined with PekVue. 2. Bckground The methodology employed in the Model 682A5 Bering Fult Detector is bsed on experience gined from PekVue pplictions. With PekVue, spectrl dt is employed to identify the specific component fults nd the time wveform is used to detect fults nd ssist in the ssessment of the severity of the fults. A brief description of the PekVue methodology follows:. Select nlysis bndwidth sufficient to include the nticipted mximum fult frequency with few hrmonics. 2. Select high pss filter greter or equl to the mximum frequency selected in the previous step. The signl being nlyzed is pssed through this filter. 3. Select the number of lines to be employed in spectrl nlysis to ensure the cpture of dt over 5 or more revolutions of the mchine being nlyzed. 4. The PekVue time block of dt consists of bsolute pek vlues from the time wveform observed over ech time increment within the dt block. 5. When the PekVue time block of dt is filled, spectrl nlysis on the time wveform cn be performed. The pek glevel within the time block is sved for trending nd fult severity ssessment. From ech PekVue nlysis time block, single pek vlue is obtined for trending. Tht pek vlue is the mximum bsolute glevel observed over severl revolutions (5+ recommended) of the mchine being monitored nd is the foundtion of the 682A5 methodology.

3 Results demonstrting the correltion of the pek glevel from PekVue to fult presence nd severity for three cses re presented. The first is for n ccelerometer mounted on the input shft bering housing of lrge roller mill gerbox turning t 893 RPM. The second is for n ccelerometer mounted on the bering housing over the output shft of pinion stnd gerbox turning t 5 RPM. The third cse presents the correltion of the pek glevel to lubriction fult in lrge mchine turning t RPM. Cse # : Roller Mill Gerbox The PekVue time wveform for 5 revolutions of the input shft on the roller mill gerbox is presented in Figure. The nlysis bndwidth ws set t 4 Hz nd the time increments for ech dt point in the PekVue time wveform is.977 msec. The number of dt points in the PekVue time wveform is 24. s G n i n o i t r e l e c c A IGB Roller Mill Pek Vue RMPV G2 Gerbox Point 2 Route Wveform 6Feb2 2:43:59 (PkVueHP Hz) RMS = 2.87 LOAD =. RPM= 893. (4.89 Hz) PK(+) = 5.9 CRESTF= 3.92 DCof = RevolutionNumber Figure PekVue time wveform on input shft to roller mill gerbox t 89 RPM Ech of the 24 dt points in the PekVue time dt block contins the bsolute pek vlue tht ws detected over ech sequentil.977 msec time increment. For the entire time dt block (which includes 5 revs of the shft) the pek vlue is 5g s, which is the prmeter recommended for trending when crrying out PekVue nlysis. The trended vlues for this mesurement point (input on roller mill gerbox) over 3 dy time period re presented in Figure 2. Note tht the lert nd lrm levels re those recommended for the bering fult detector s described lter in this pper.

4 Pek g Trend Dt For Mesurement Point G Pek G's Alert 4 Fult D y s Figure 2 Trended pek glevel from input shft to roller mill gerbox t 89 RPM The PekVue spectrl dt (computed from the PekVue time dt presented in Figure ) positively identified the fult s n inner rce fult. Bsed on the pek glevels, decision ws mde to replce the bering round dy 275 in Figure 2 (glevels round 8 g s). The pek g level returned well below the lert level (see Figure 2) post bering replcement. A picture of the removed bering showing the inner rce fult is presented in Figure 3. Figure 3 Dmged bering from Roller Mill Gerbox

5 Cse # 2: Pinion Stnd Gerbox The trended pek glevel over n 8 dy time period from mesurement point on the bering housing of the output shft (turning t 5 RPM) of pinion stnd gerbox is presented in Figure 4. The lert/lrm levels shown on Figure 4 re extrcted from the recommended bselines for the Mchine Fult Detector. The trended prmeter exceeded the lert level on September 23, 999, nd ws replced following the lst dt point (My 25, 2) in Figure 4. G n i m f e v W RFPD HOT MILLF5 PINION STAND 2 HMF5PINSTDP4H BOT BRG OUTPUT SIDE HORIZ.PKVUE Trend Disply of TIME WAVEFORM Bseline Vlue:.9424 Dte: 6Mr98 k e P x M 3 2 FAULT ALERT WEAK SIDE Dys: 6Mr98 To 25My Dte: 25My Time: 9:38:4 Ampl: Figure 4 Trended pek glevel from output shft t 5 RPM of pinion stnd gerbox. A picture of the defective bering is presented in Figure 5. Clerly this bering ws ner ctstrophic filure. Following bering replcement, the pek glevel returned to levels experienced in July/August 998. Figure 5 Defective Inner Rce from output shft turning t 5 RPM of pinion stnd gerbox.

6 Cse # 3: Lrge Slow Speed Mchine The third cse is for lrge mchine turning round RPM where fult in this cse is lck of lubriction. Referring to Tble (next section), the recommended lert nd lrm levels re. nd.27 g s respectively for this speed mchine. The trended pek glevel from this mchine is presented in Figure 6. The pek glevel obtined on October 2, 22 (.4g s) is greter thn the recommended lert level. Following the reding cquired on October 2, 22, the mchine ws shutdown nd tken through mjor overhul. The mchine ws strted up gin round Mrch, 23 nd the pek glevel ws mesured to be.52g s (well bove the lrm level of.27g s). A second (post rebuild) reding ws cquired on Mrch 2, 23 yielding pek glevel of.73g s. On Mrch 25, 23, smll mount of grese ws dded to the bering resulting in n immedite decrese in the glevel reding to.32g s. A postulte ws then dvnced tht the bering ws clened out during rebuild but ws not repcked. Sufficient grese ws then dded to pck the bering with resultnt decrese in the pek glevel to the prerebuild glevel of round.2g s. G n i m r o f e v W k P k P HIGH POLYMERIZER AGITATOR X96. 3rdX3 PVAXIAL VESSEL EAST BRGr M FAULT r 2 M.4 ALERT t 2 c r O.2g M u A 2. WARNING Dys: 4Aug2 To 3Mr3 3 r M Trend Disply of Wveform PP Bseline Vlue:.88 Dte: 8Nov98 Dte: 3Mr3 Time: 3:8:56 Ampl:.8 Figure 6 Trended pek glevel from polymerizer gittor t RPM

7 3. Model 682A5 Bering Fult Detector Methodology The two primry outputs from IMI Sensor s Bering Fult Detector (682A5) re:. 4 to 2 ma linerly proportionl to the RMS or Pek overll of the input from the ccelerometer in either g or ips units to 2 ma proportionl (on log scle) to the bsolute pek glevel of the high frequency (greter thn KHz) component of the input from the ccelerometer in gunits. As shown in Flow Digrm, the RMS or Pek overll vlue is obtined with True RMS/DC converter. The output of the RMS/DC converter is the input to V/I converter providing the 42 ma output signl (proportionl to the RMS/pek overll). The input to the RMS/DC converter is either the output of the ccelerometer (if gunits re being used) or the output of n nlog integrtor scled to convert g s to ips units (if ips units re being used). The user selects the full scle reding through dip switch selections within the Bering Fult Detector Module. For the pek glevel output, the signl from the ccelerometer is routed through high pss filter set t either KHz or 5 KHz (dip switch selectble). The output from the high pss filter is ) full wve rectified, 2) routed through log mplifier, nd 3) routed to pek follower circuit which cptures the pek vlue over preset time intervl (fctory setting with defult intervl of seven seconds). At the conclusion of ech present time intervl, the pek hold signl is ccepted s the module output for the next time increment. The pek hold circuit is rpidly zeroed out, nd continues gthering pek vlues over the next time increment (the output lgs the input by one time increment). The requirement on the time increment must include minimum of six revolutions (5 is desirble) of the mchine being monitored. Therefore, the fctory set 7 second increment is useble for mchinery turning s low s 5 RPM. Experience with PekVue hs demonstrted tht it is not unusul to experience glevels of 5+ g s with some fults nd t the sme time, tht sme signl cn be less thn.2 g s in the bsence of fults. Therefore it is desirble to ccommodte 6+dB dynmic rnge in the output. Thus, the ccelerometer signl is routed through log mplifier followed by pek hold module (hold time set by fctory). The V/I module providing the 4 to 2 ma output follows the pek hold module. A typicl Bering Fult Detector output of g s versus ma is presented in Figure 7, which demonstrtes tht liner reltionship exists over 6 db dynmic rnge. Flow Digrm

8 Figure 7 Trnsltion curve for 4 to 2 ma output of Bering Fult Detector to input g level The pek glevel (observed over 6+ revolutions) is the prmeter used to identify the presence of fult nd estblish the severity of the fult. The output of the Bering Fult Detector provides the bility to monitor/trend the pek glevel on 24/7 bsis. When fult ppers nd progressively increses in severity, the pek glevel will correspondingly trend upwrd. Experience from PekVue enbles the bility to estblish generic lert nd lrm levels (bsed on the speed of the mchine), which cn be used s guidelines. The recommended generic lert nd lrm levels re presented in Tble. The generic levels cn be reset for ech individul mesurement point once bselines re estblished. Speed Rnge (RPM) Alert Limit (Pek glevel) Alrm Limit (pek glevel) less thn Tble Recommended Alert nd Alrm Levels (Generic)

9 4. Test Results Severl tests under controlled conditions hve been crried out to estblish tht the pek glevel from the Bering Fult Detector is pproximtely the sme obtined using PekVue nd tht the pek glevel exhibits significnt upwrd trend s the severity level increses. To estblish similrity in the pek glevel from the Bering Fult Detector to tht obtined from PekVue, the test conducted ws simply to obtin dt from both devices on severl berings. To estblish n upwrd trend with severity level, bering fults were purposely introduced. The fulty berings were then run under 5% dynmic lod (bout five times typicl lods) until ctstrophic filure is observed. The pek glevel redings obtined from the Bering Fult Detector module nd from the PekVue methodology t pproximtely the sme time for severl test cses re presented in Tble 2. Clerly the two methodologies re in greement reltive to the pek glevel results. PEAK GLEVEL (G'S) Mchine Speed (RPM) Bering Fult Detector PekVue Tble 2 Comprison between PekVue nd Bering Fult Detector At the NSK Bering Test Fcilities 3, fults were introduced into specific bering components by n electric dischrge mrking pen. The berings were then loded t 5% dynmic lod (bout 5 times tht experienced normlly) nd run until ctstrophic filure occurred. Ctstrophic filure ws determined by one of the following: ) Shft Lockup 2) Temperture exceeds preset level 3) A trip point set from shock level (SPM technology) ws exceeded. The output of the Bering Fult Detector module ws recorded over the test period. Some tests rn pproximtely 7 dys prior to filure while others filed in pproximtely 7 hours (no consistent pttern ws observed). There were seprte bering tests crried out. A defect ws formed on six of the ten berings (two ech on the bll, outer rce, nd inner rce) with n electric dischrge mrking pen. No defects were introduced on one bering. Two berings were under lubricted nd one over lubricted. Ech of the berings ws operted t 5% of dynmic lod rting (bout 5 times the norml lod) nd rn either until filure occurred or terminted becuse of elpsed time. Running speed for ll berings ws 8 RPM. The lert nd lrm levels for berings running t 8 RPM would be set t 5 g s nd 9 g s respectively (see Tble ). However, these test berings were running under extreme lods nd hence re not expected to fll under the sme lrm/lert levels. Selected grphicl results re presented below.

10 The results from one of the two defective berings (bll defect) re presented in Figure 8. The durtion of the test ws pproximtely 3.2 dys with the finl excursion lsting bout 5 hours. Note how the glevels were rpidly incresing t the end of the test (terminted by either temperture or high shock levels). Bll Defect Run 5 4 g's Time (5 Units = 6.7 hrs) Figure 8 Output from defective bll test run t NSK under 5% dynmic lod until termintion. A grphicl result from one of the two berings with n inner rce fult is presented in Figure 9. The durtion of the test ws bout 3.5 hours. The excursion t the end of the test lsted bout.5 hours. If n lrm/lert hd been set, it would not hve been exceeded until the finl excursion. ID Defect Run g's Time ( Units = 33.3 hrs) Figure 9 Output from defective inner rce test run t NSK under 5% dynmic lod until filure.

11 Grphicl dt (g s versus time) from one of the two under lubricted test berings is presented in Figure. The durtion of the test ws bout 6.5 dys (terminted becuse of time). A probble lrm/lert level ws exceeded bout 4 hours into the test. There were excursions bove probble lrm level throughout the test ccompnied by n upwrd trend in the bseline, but there ws not n indiction tht the bering ws ner ctstrophic filure. Under Lubricted Run g's Time ( Units = 33.3 hrs) Figure Output from n under lubricted bering test run t NSK under 5% dynmic lod until terminted (no filure). 5. Conclusion The Model 682A5 Bering Fult Detector provides n output, which is representtive of the pek glevel experienced by n ccelerometer for the high frequency (frequencies greter thn the fctory set high pss filter) components of the signl. The glevel from the Bering Fult Detector module ws demonstrted to be essentilly the sme s tht obtined from the PekVue methodology. The pek glevel hs proven, through longterm experiences with PekVue, to be both relible detector of the presence of fult nd relible indictor of the severity of the fult. Since the output of the Bering Fult Detector is essentilly the sme s the pek glevel from PekVue, it follows tht the output of the Bering Fult Detector will yield dt for relible fult detection nd provide ssistnce in estblishing the severity of the fult. Tests were crried out t n NSK Corportion bering test fcility where intentionl fults were introduced into berings run under 5% rted dynmic lod until either ctstrophic filure occurred or llotted time for the test expired. In every cse where ctstrophic filure occurred, there were significnt increses in the pek glevel (output of the Bering Fult Detector) over time period preceding the filure. IMI Sensors 2 Emerson Process Mngement, CSI 3 NSK Corportion 3425 Wlden Avenue 835 Innovtion Drive 42 Goss Rod Depew, NY 443 USA Knoxville, TN USA Ann Arbor, MI 485 USA