Sleeve Bearing Fault Diagnosis And Classification - WSEAS
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El AroussiSleeve Bearing Fault Diagnosis and ClassificationYASSINE ELYASSAMI, KHALID BENJELLOUNElectrical department, LA2IMohamamdia engineer’s school, Mohammed V UniversityAgdal, Rabat, Moroccoelyassami@yahoo.fr; bkhalid@emi.ac.maMOHAMED EL AROUSSIElectrical department, S2C2SR-LASIHassania School of public worksKm 7, Casablanca, MoroccoMoha387@yahoo.frAbstract: Sleeve bearing is a bearing without any rotating element but with a sliding component, it isan expensive component in special machinery elements. Their faults can damage other principalmachinery parts like shafts and cause very important production lost and high maintenance cost. Becausethose bearings are a special case just a few researchers studied detection and classification of sleevebearing faults by data vibration analysis.In this paper we present a diagnosis and a classification methodologies applied to different kind of sleevebearing damages. We develop a two-lobe bore sleeve bearing vibration database from a set of largeinduction machine equipment, then we use temporary and frequency process to diagnose faults, in thefinal step we classify those sleeve bearings by different methods based on entropy extraction and faultclassifiers fusion.Keywords: Condition monitoring, Machine vibration, Diagnosis, Fault classification, Sleeve bearing.1. Introduction:A sleeve bearing also known as a plain bearingor journal bearing is a bearing in which a shaftrotates freely in a supporting metal sleeve orshell with a layer of oil or grease separating thetwo parts due to fluid dynamic effects. Journalbearings are used to support high radial loadsand are used for low to high speeds. Typicalapplications include large milling systems,engine crankshafts, gearboxes, and shaftbearing supports. There are five basic types ofjournal bearings: plain cylindrical bore, lemonshape (two-lobe bore), four-lobe bore, four padtilting pad, and five pad tilting pad bearing [1].In this paper the second type two-lobe bore (seeFig.1) is considered.Fig. 1: Sleeve bearing lemon shape with two oil ringE-ISSN: 2224-3488163Volume 12, 2016
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El AroussiThis paper addresses fault diagnosis andautomated classification of sleeve bearings withdifferent defects types. The feature extraction isdone with entropy methods, the temporal andfrequency analysis are used for sleeve bearingdiagnosis and the fault classification isperformed with many classifiers.In Section 2, the adopted approach -is exposed.Section 3 presents signal vibration acquisition,treatment and classification based featureextraction, while section 4 gives an account ofobtained results. Finally, the conclusion ispresented in section 5.Induction machines can be purchased withantifriction or sleeve bearings. The sleevebearing is in many cases the unique bearing typeto utilize on larger motors and generators inexcess of 2000 Hp (Horse power) and for somespecial design requirements on medium motors(ships, draglines ). In general sleeve bearingmotors are significantly more expensive thanthe anti-friction bearings but this may not be assignificant when the total life cycle cost is takeninto account. Theoretically, sleeve bearingshave infinite life and they guaranty moreprocess reliability [1]. The sleeve bearingsoperate under the principal of hydrodynamiclubrication. As the shaft rotates, it builds up awedge of oil between the shaft and bearing byusing oil ring (see Figure1). Flood lubricationcan be used where additional cooling is requiredas a redundant feature.The sleeve bearing faults detection anddiagnosis was studied using vibration analysis[2, 3], the spectral methods are developed toperform this diagnosis [4, 5]. The statisticalapproach was recently used for journal bearingfault classification by several techniques such asfisher linear discriminant, K-nearest neighborand support vector machine [6, 7, 8].2. Proposed fault detection processThe bearing diagnosis proposed method is a treesteps process (see Fig. 2). The first step isvibration signal measurement, the second stepis a time-frequency analysis and the third stepdeals with fault classification using featureextraction. We develop a vibration databasemeasured by accelerometers, then we treat theinput signal by filtering and interpreting graphsin time and frequency domain, after that weextract entropy features for each sleevebearing/direction, and finally we detect the bestclassification using learned models and someclassifier algorithms.Fig. 2: Sleeve bearing fault classificationE-ISSN: 2224-3488164Volume 12, 2016
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El AroussiMachine (SVM) applies structural riskminimization learning method for regressionand classification [13, 14], The PrincipleComponent Analysis (PCA) is using for reducedata dimension and complexity based on atransformation of some possibly correlatedvariables into a smaller number of uncorrelatedvariables to perform a satisfying accuracy [15,16].We considered four sleeve bearing defaultstates: Normal, Friction fault, Excessive wearfault, and abnormal lubricating fault. Eachclassifier estimates bearing state with arecognition accuracy rate. The entropyextraction theory estimates the complexity oftime series through the comparison ofneighboring values as detailed in [9, 10]. Thisextraction is characterized by 51-parameterfamily corresponding to probability density,asymmetry and concentration measurement.Those parameters can completely represent thestatistical signal characteristics. We use them infault diagnosis of sleeve bearings and we hopethat they can give a good separation betweendifferent classes. We use as traditionalstatistical technique classifiers, the LinearDiscriminant Analysis (LDA) as lineardimension reduction method by performing inparallel the maximization of the between-classscatter and the minimization of the within-classscatter [11, 12], the multi-class Support Vector3. Experimental phase3. 1 Sleeve bearing vibration acquisition:The experiment was carried on 26 sleevebearings. The used induction motors andgenerators are three phases 400 Hp to 2250 Hpwith 1000 rpm and full load running asoperating conditions. Vibration signals of thosesleeve bearings are obtained by magnetic baseaccelerometers mounted on horizontal andvertical direction and taking from bearinghousing.Pedestal mounted sleeve bearingFlange mounted sleeve bearingFig. 3: Sleeve bearing acceleration acquisitionE-ISSN: 2224-3488165Volume 12, 2016
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El AroussiIn this paper we study and diagnose the sleevebearingvibrationintwolargemotors/generators with different mountedmodel [17] (see Fig. 3):3. 2 Sleeve bearing vibration diagnosis:The diagnosis of sleeve bearings ischaracterized by the specifications [19] listed inFig. 4.Temporal signal formNoisy spectrumNoisy signalFriction defectExcessive Clearancedefect Pedestal mounted sleeve bearings are usedfor the first set of motors/generators (850Hpto 2250Hp). Motors/generators with integralpedestal bearings are as easy to mount andalign as motors/generators with flangemounted bearings. Separate pedestal bearingare mounted on a common base frame.Spectral signal formAbnormallubricating Flange mounted sleeve bearings are used forthe first set of motors/generators and they aremounted on the end-shields of themotor/generator (400Hp to 1225Hp).The vibration data was collected from sleevebearing housing trough two piezoelectricaccelerometers ICP 603C01 with frequencyrange up to 10 kHz and a 1mv/m/sec2sensitivity. The signals are transferred to thevibration module NI 9234 as inputs then the NIcontroller CompactRIO- 9022 records andcommunicates complete data to the computer byexploiting LABVIEW software [18]. Thecapture duration for each sleeve bearing is set to30 seconds with 25.6 kHz sampling rate. Weapply a low pass band filter Fc 500 Hz to allsignals and we analyze the vibration data in bothtime and frequency domains to detectcharacterized bearing defects.Fig. 4: Sleeve bearing excessive clearance, frictionand lubricating defectsThe vibration captured signals by horizontal andvertical accelerometers are analyzed for defectdiagnosis. Then sleeve bearing can be classifiedafter applying some algorithms and methods tothe feature extraction data. Four different states:(1) Sleeve bearing N 1: Normal healthy, (2)Sleeve bearing N 2: Friction defect, (3) Sleevebearing N 3: Excessive clearance defect, (4)Sleeve bearing N 4: Abnormal lubricating andfriction defects. Fig. 5 and Fig. 6 presentacceleration and spectrum based horizontalaccelerometer data for those different bearingsclasses.Accelerations (g)Sleeve bearing N 1:Normal healthyHistograms (g)4SG1H2.5SG1Hx 100.320.20.11.501-0.1-0.20.5-0.310.08E-ISSN: e 12, 20160.25
WSEAS TRANSACTIONS on SIGNAL PROCESSINGSleeve bearing N 2 :Friction defectYassine Elyassami, Khalid Benjelloun, Mohamed El Aroussi44.50.8x eve bearing N 3:Excessive 00006000-0.14000-0.22000-0.30-0.410Sleeve bearing N 4:Abnormallubricating andfriction Fig. 5: Sleeve bearings Acceleration/Histogram based horizontal dataSleeve bearing N 1:Normal healthy2500200015001000500X: 33.33Y: 272.30Sleeve bearing N 2 :Friction defect50X: 99.97Y: 192.7100X: 199.9Y: 184.61502002503003504001000X: 200Y: 829.9900800700600X: 300Y: 477500X: 100Y: 357.24003002001000Sleeve bearing N 3:Excessive 250020001500X: 199.9Y: 10291000X: 299.8Y: 626.2X: 99.93Y: 611.4500X: 16.67Y: 47.710Sleeve bearing N 4:Abnormallubricating andfriction defects501001502002503001000900X: 299.9Y: 766.9X: 199.9Y: 880.5800700600500X: 99.97Y: 323.7400300200100050100150200250300Fig. 6: Sleeve bearing FFT Spectrums based horizontal dataE-ISSN: 2224-3488167Volume 12, 20160.5
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El Aroussiinfluenced by the chosen classifier. We usedifferent amount of vibration data learning:10%, 20% and 30% windows of training weretaken for each sleeve bearings in the horizontaland vertical axis to calculate the statisticalfeatures. Then we use them as input to differentclassifier process based entropy extraction(LDA, PCA and SVM). For the testing purposethe vibration data is split into two sets, the firstcomposed by 10 to 30% of the total inputs isused for training, and the second one composedby 90 to 70% is used for validation and testingclassification using LDA, PCA, SVMalgorithms based entropy extraction. The resultof SVM classification was satisfactory (seeTable 1). Then, in Table 2, we expose theclassification results for the studied sleevebearings.According to figures 5 and 6, we can see that theacceleration and FFT spectrum depend to thebearing defect, the healthy normal bearingpresents a good temporal signal with somepeaks in his FFT spectrum (in the naturalfrequency of structure at 50Hz and itsharmonics), in the lubricating defect case wecan detect a noisy acceleration spectrum, thefriction and excessive clearance defects werecharacterized by the temporal accelerationforms and by some spectrum peaks withmedium or high level.4. Results and DiscussionsWe present in Table 1 a comparison betweenseveral classifications methods using entropyextraction feature as input, the recognitionaccuracy rate with same experimental data isTable 1: Recognition sleeve bearing classification accuracy based accelerometer dataAmount 8,54%96,87%71,43%89,29%92,86%97,62%94,05%100%Table 2: Recognition accuracy based flange sleeve bearing classificationClassNormal healthyClass 2Friction defectClass 3Excessive clearancedefectClass 4Abnormallubricating, andHVfrictiondefectsVHVHV100100000000S. Bearing 2b001001000000S. Bearing 3b000010010000S. Bearing 4b000000100100S. Bearing 1a001001000000S. Bearing 2a000010090010S. Bearing 3a0006,670010093,33S. Bearing 4a001001000000S. Bearing 5a000010010000S. Bearing 6a000086,6796,6713,333,33S. Bearing 7a000010010000S. Bearing 8a0020063,3396,6716,673,33S. Bearing 9a001001000000S. Bearing 10a001001000000S. Bearing 11a000096,671003,330S. Bearing 12a000093,3393,336,676,67S. Bearing 13a000093,3396,676,673,33S. Bearing 14a009086,676,6713 ,333 ,330Pedestal mounted sleeve bearingHS. Bearing 1bE-ISSN: 2224-3488168Volume 12, 2016
Flange mounted sleevebearingWSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El AroussiS. Bearing 15a001001000000S. Bearing 5b000086,679013,3310S. Bearing 6b000010093,3306,67S. Bearing 7b000096,671003,330S. Bearing 8b000093,3396,676,673,33S. Bearing 9b000093,331006,670S. Bearing 10b000093,331006,670S. Bearing 11b000076,679023,3310In table 2, recognition accuracies classificationresults were acceptable.[5] Ioannis Tsiafis, K.-D. Bouzakis, GrigorisTsolis, Thomas Xenos“Spectral methodsassessment in journal bearing fault detectionapplications, The 3r d International Conferenceon diagnosis and prediction in mechanicalengineering systems, May 31 – June 1, 2012,Galati, Romania5. Conclusion:In this study we diagnosed sleeve bearings andwe classified them in four different fault classesfor both pedestal and flange mounted bearings,one healthy and three with different defectstype. The analysis of the vibration accelerationsignal, obtained using horizontal and verticalaccelerometers, was used to detect the specificdamage. The treatment of the feature extractionwas, then, applied to characterize each classwith acceptable accuracy recognition for the 26studied sleeve bearings.It was seen that the percentage of correctclassification was between 92 and 100% for ourapproach based entropy extraction and SVMclassifier algorithm.[6] A. Moosavian, H. Ahmadi andA.Tabatabaeefar“Journal-bearingfaultdetection based on vibration analysis usingfeature selection and classification techniques”,Elixir Control Engg. 49, 2012[7] A. Moosavian, H. Ahmadi andA.Tabatabaeefar “Fault diagnosis of mainengine Journal bearing based on vibrationanalysis using fisher linear discriminant, Knearest neighbor and support vector machine”,journal of vibroengineering, volume 14, 2012[8] Byungchul Jeon, Joonha Jung, Byeng D.Youn, Yeonwhan Kim, and Yong-Chae Bae“Statistical Approach to Diagnostic Rules forVarious Malfunctions of Journal BearingSystem Using Fisher Discriminant Analysis”European Conference Of The Prognostics AndHealth Management Society 2014References:[1] William R. Finley Mark M. HodowanecSleeve vs. Anti-friction bearings: selection ofthe optimal bearing for induction motors PaperNo. PCIC 2001-33 Siemens Energy &Automation, Inc.[2] Parno Raharjo, “An investigation of surfacevibration, airborne sound and acoustic emissioncharacteristics of a journal bearing for earlyfault detection and diagnosis” A thesissubmitted to The University of HuddersfieldMay 2013[9] Yan, R.; Liu, Y.; Gao, R.X. Permutationentropy: A nonlinear statistical measure forstatus characterization of rotary machines.Mech. Syst. Signal Process. 2011, 29, 474–484.[10] Bandt, C.; Pompe, B. Permutation entropy:A natural complexity measure for time series.Phys. Rev. Lett. 2002, 88, 174102–1–174102–4.[3] N. Dileep, K. Anusha, C. Satyaprathik, B.Kartheek,K.Ravikumar,“ConditionMonitoring of FD-FAN Using VibrationAnalysis”, International Journal of EmergingTechnology and Advanced Engineering,Volume 3, Issue 1, January 2013.[11] R. Duda, P. Hart, and D. Stork, PatternClassification, Wiley-interscience, New York,2001.[12] K. Fukunaga, Introduction to StatisticalPattern Recognition, Academic Press, seconded., 1990.[4] Narendiranath Babu, Manvel Raj andLakshmanan, “High Frequency AccelerationEnvelope Power Spectrum for Fault Diagnosison Journal Bearing using DEWESOFT”,Research Journal of Applied Sciences,Engineering and Technology 8(10), 2014.E-ISSN: 2224-3488[13] Manikandan, J., Venkataramani, B.: Studyand evaluation of a multi-class SVM omputing 73, 1676–1685 (2010)169Volume 12, 2016
WSEAS TRANSACTIONS on SIGNAL PROCESSINGYassine Elyassami, Khalid Benjelloun, Mohamed El Aroussi[14] Debnath, R., Takahidel, N., Takahashi, H.:A decision based one-against-one method formulti-class support vector machine. PatternAnalysis & Applications 7, 164–175 (2004)[15] Nimityongskul, S., Kammer, D.C.:Frequency domain model reduction based onprincipal component analysis. MechanicalSystems and Signal Processing 24, 41–51(2010)[16] Serviere, C., Fabry, P.: Principalcomponent analysis and blind source separationof modulated sources for electro-mechanicalsystems diagnostic. Mechanical Systems andSignal Processing 19, 1293–1311 (2005)[17] ABB / AMZ Synchronous Motors EN 082005[18] Yassine Elyassami, K. Benjelloun, M. ElAroussi: Bearing Fault Diagnosis andClassification Based on KDA and Alpha-StableFusion. Contemporary Engineering Sciences,Vol. 9, 2016, no. 10, 453 – 465.[19] Dave Felt, “Understanding BearingVibration Frequencies”, EASA 2003E-ISSN: 2224-3488170Volume 12, 2016
Keywords: Condition monitoring, Machine vibration, Diagnosis, Fault classification, Sleeve bearing. 1. Introduction: A sleeve bearing also known as a plain bearing or journal bearing is a bearing in which a shaft rotates freely in a supporting metal sleeve or shell with a layer of oil or grease separating the two parts due to fluid dynamic effects.
cooper bearing craft du bearing dual vee economy bushing edt elges fag bearing federal bronze fk bearing frantz bearing fyh bearing general bearing ggb bearing tech graphalloy hepco hiwin hub city hudson bearing hyatt
absent on the stationary sleeve surface. They found that such an interfacial slippage severely deterio-rates the performance of the bearing. In a hydrodynamic journal bearing, the sleeve is normally made of non-steel materials such as bronze compounds and plastics. The shear strength on the interface between the fluid and the sleeve may be
in the research and application of fault diagnosis. In this paper, some notions and the basic principles for the unanticipated fault detection and diagnosis are given. A general process model applied to the diagnosis for the unanticipated fault is designed, by adopting a three-layer progressive structure,
CDS G3 Fault List (Numerical Order) Fault codes may be classified as sticky or not sticky: Type of fault Method to clear Not sticky Clears immediately after the fault is resolved Sticky Requires a key cycle (off and on) after the fault is resolved to clear. CDS G3 Fault Tables 90-8M0086113 SEPTEMBER 2013 Page 2G-3
In this paper, the calculation of double line to ground fault on 66/33 kV transmission line is emphasized. Figure 3. Double-Line-to-Ground Fault 2.4 Three-Phase Fault In this case, falling tower, failure of equipment or . ground fault, fault current on phase a is I a 0. Fault voltages at phase b and c are: V b .
Capacitors 5 – 6 Fault Finding & Testing Diodes,Varistors, EMC capacitors & Recifiers 7 – 10 Fault Finding & Testing Rotors 11 – 12 Fault Finding & Testing Stators 13 – 14 Fault Finding & Testing DC Welders 15 – 20 Fault Finding & Testing 3 Phase Alternators 21 – 26 Fault Finding & Testing
illustrated in Figure 3. This is a fault occurring with phase A to ground fault at 8 km measured from the sending bus as depicted in Figure 1. The fault signals generated using ATP/EMTP are interfaced to the MATLAB for the fault detection algorithm. (a) Sending end (b) Receiving end Figure 3. Example of ATP/EMTP simulated fault signals for AG fault
KHADER M, P G ASST IN COMMERCE GHSS BRAMMAKUNDAM VILLUPURAM DT – 94863 35786 Ït thlif¡fhf¡ ru¡nf‰¿ brš»wh . Ït thlif¡fhfnth mšyJ Ïytrkhfnth ru¡nf‰¿ bršth . Ït jh‹ V‰¿ bršY« bghU fS¡F¡ fh ÕL jUeuhf brašgL»wh . ÏtuJ ftd¡ Fiwthnyh mšyJ Ãw têænyh ru¡FfS¡F V‰gL« nrj« mšyJ ÏH ÉF bghJ ru¡nf‰¿ bghW gh»wh . jå ru¡nf .
