University of Huersfiel Repository Towsyfyan, Hossein, Hassin, Osama, Gu, Fengshou an Ball, Anrew Characterization of Acoustic Emissions from Mechanical Seals for Fault Detection Original Citation Towsyfyan, Hossein, Hassin, Osama, Gu, Fengshou an Ball, Anrew () Characterization of Acoustic Emissions from Mechanical Seals for Fault Detection. In: The 5r Annual Conference of The British Institute of Non Destructive Testing September Manchester., 9 September, The Palace Hotel, Manchester, UK. This version is available at http://eprints.hu.ac.uk/58/ The University Repository is a igital collection of the research output of the University, available on Open Access. Copyright an Moral Rights for the items on this site are retaine by the iniviual author an/or other copyright owners. Users may access full items free of charge; copies of full text items generally can be reprouce, isplaye or performe an given to thir parties in any format or meium for personal research or stuy, eucational or not for profit purposes without prior permission or charge, provie: The authors, title an full bibliographic etails is creite in any copy; A hyperlink an/or URL is inclue for the original metaata page; an The content is not change in any way. For more information, incluing our policy an submission proceure, please contact the Repository Team at: E.mailbox@hu.ac.uk. http://eprints.hu.ac.uk/
Characterization of Acoustic Emissions from Mechanical Seals for Fault Detection Abstract Hossein Towsyfyan, Osama Hassin, Fengshou Gu an Anrew Ball School of Computing an Engineering, University of Huersfiel, Queens gate, Huersfiel, HD DH, UK Telephone: 7557 E-mail: Hossein.Towsyfyan@hu.ac.uk The application of high-frequency Acoustic Emissions (AE) for mechanical seals iagnosis is gaining acceptance as a useful complimentary tool. This paper investigates the AE characteristics of mechanical seals uner ifferent rotational spee an flui pressure (loa) for evelop a more comprehensive monitoring metho. A theoretical relationship between friction in asperity contact an energy of AE signals is evelope in present work. This moel emonstrates a clear correlation between AE Root Mean Square (RMS) value an sliing spee, contact loa an number of contact asperities. To benchmark the propose moel, a mechanical seal test rig was employe for collecting AE signals uner ifferent operating conitions. Then, the collecte ata was processe using time omain an frequency omain analysis methos to suppressing noise interferences from mechanical system for extracting reliably the AE signals from mechanical seals. The results reveal the potential of AE technology an ata analysis metho applie in this work for monitoring the contact conition of mechanical seals, which will be vital for eveloping a comprehensive monitoring systems an supporting the optimal esign an operation of mechanical seals.. Introuction Mechanical seals are one of the most important types of shaft seals which are foun on rotating machinery an can protect against leakage across much higher pressure ifferences than the other seals []. The seal is mae between the very smooth, very flat faces of two rings, one is attache to an rotates with the shaft, an the other is attache to the housing an is stationary. A schematic illustration of the components of a mechanical seal is shown in Figure. Rotating face Stationary face. Shaft Flange Contact area Figure. Schematic illustration of mechanical face seals [] Since AE signals carry information about the etails of micro-amage process, a significant amount of research amount of research has been reporte on the conition
monitoring of mechanical seals using AE technology[]. for instance, lingar et al.[], Hisakao et al.[5] an Hase et al.[6] investigate the correlation between AE signal characteristics such as AE count, count rate an amplitue with sliing contact. During conition monitoring of mechanical seals, toay's AE systems are able to process thousans of AE events per secon an recor them to ata storage. A number of stuies have investigate the use of the time-omain anfrequency-omain techniques in processing the AE signals from mechanical seals an showe that using these techniques, it is possible to process AE signals using sophisticate computing methos [7-8]. Analysing the AE signal in the frequency omain coul characterise the AE signals especially uner noisy conitions, however it is often ifficult to ientify effective features for fault iagnosis in the frequency spectra ue to the limitations mainly cause by the nature of AE events such as high frequency, attenuation, ispersion, multiple reflections an the non -linear character of AE signal uring its propagation [9]. Holstein et al. [8] claime that analysis in the frequency omain was investigate in the research on fault iagnosis of mechanical seal. However, they faile to present the analysis results to prove the effectiveness of this metho. Boness an Mcbrie [] performe a comprehensive stuy involve the measurement of AE signals in sliing contact an reporte that impact of friction surface at microscopic level (asperity contact) was the main source of acoustic emissions uring pure sliing friction. They also propose an empirical moel which presents AE RMS value in terms of wear volume. Miettinen an Siekkinen [] stuie the AE response to the sliing contact behaviour of a mechanical seal on a centrifugal pump uner ifferent working conitions. They reporte the possibility of etecting leakage, ry running an cavitation in face seals by measuring the RMS value of AE signal. In subsequent stuies, Mba et al. confirme that the RMS value of AE signals can be use to monitor seal conition [7]. In another stuy, Fan et al. evelope a mathematical AE moel base on elastic asperity contact an pointe out that the level of AE measurement epens on the sliing spee, the loa supporte by contact, the number of asperity contact, an surface topographic characteristics among others []. In this paper, a mathematical moel of the AE generate in frictional asperity contact is evelope to establish a relationship between AE RMS value an working parameters of seals (rotational spee, loa an number of asperities in contact). Furthermore, the AE characteristic of mechanical seals are presente in the time omain an the frequency omain firstly an then AE feature parameters such as RMS an kurtosis values are successfully explore in associating with flui pressure (loa) an rotational spee of shaft.. Mathematical moel concept. Moelling friction The basic iea behin this work is moelling the friction between the asperities which is assume to be the main sources of AE in mechanical seals. In general, the tangential contact friction between a pair of asperities in contact can be efine by []: F = τ A () where, A an τ are the area of one asperity in contact an shear stress at the asperity contact respectively. The coefficient of friction f of a single asperity can be expresse as:
f = τ p () where, P is the normal pressure in a single asperity contact. Substituting τ from Equation into Equation an rearranging the integral yiels to: F = τ A = f p A = f W () Where W is the normal loa. Consiering the Greenwoo an Williamson moels [] for the contact of real surfaces, the probability of making contact at any selecte asperity can be expresse as (see Figure): p ( z) = f ( z) () If the number of asperities per unit area is D, the expecte number of contacts in any unit area is: n = D f ( z) (5) Contact area Smooth surface Reference plane in the rough surface Figure. Greenwoo an Williamson for contact moel Base on Hertz theory the maximum eflection in the contact area can be expresse as: W / δ = ( ) (6) E R. Friction energy release rate moel The frictional work one by friction force F on a point that moves a sliing istance S in the irection of tangential sliing contact is the prouct: U iae = Fs (7) Now we calculate sliing istance S in asperity contact base on Figure. In this Figure, a is the Hertzian raius of the asperity contact circle given by: a W R / ( ) E = (8)
Figure. The concept of isplacement in asperity contact Thus sliing istance S in asperity contact can be expresse as: / W R i / S = a + a = ( ) = ( δ R ) (9) E Substituting F from Equation () an extracting s from Equation (9), Equation (7) can be rearrange as: U iae f E E Fs W W = f W = () = WR Equation () can be re arrange in terms of maximum eflection as follow: U iae E f W R R δ = () Since δ = Z, the mean frictional work of one asperity contact is: U iae E = f R W ( z ) f ( z) f ( z) The total frictional work in the asperity contacts U AE can be expresse as: U AE A n U iae () = () where A is the apparent contact area; n is the number of contacts in unit area given by Equation (5).Thus, the total frictional energy can be expresse as: U AE = A E f R W ( z ) f ( z) n () f ( z) Substituting n from Equation (5) into Equation () will result in: E U AE = A D f W ( z ) f ( z) R The total time in frictional contact can be calculate as: (5) S a ( δ R ) / t = = = (6) v v v where v is sliing spee. Usingδ = z, the mean friction contact time is:
t = R / ( z ) / f ( z) (7) v f ( z) Now we efine the total number of asperity contacts between the two surfaces as: N = A D f ( z) (8) Diviing Equation (5) by Equation(7), an efining all constant parameters as K, the acoustic energy release rate can be expresse as: U = Nv ( / W ) (9) Supposing that a portion (k f ) of the energy release as a result of friction converts to AE pulses an the gain of the AE measurement system is (kg) we have : U ( / AE = K f K g N v K W ) (). Frictional AE Moel Base on etaile iscussion foun in the work of Fan et al. [], the RMS value of the AE signal excite by the contact friction can be expresse as: V = RU () rms AE where R is the electrical resistance of the AE measuring circuit. Substituting Equation () into Equation () an efining all constants with K, V rms in frictional asperity contact can be expresse as: V KN v W / rms = () Base on the Equation (), the RMS value of AE signal increases with loa, rotational spee an number of asperities in contact.. Experimental Test Rig an Test Mechanical Seal The test rig is shown in Figure. The riving power is a. kw, -phase AC electric inuctions motor an the rive shaft can be run at ifferent spees up to a maximum of 8 rpm. Two John Crane Type 68 MP pusher cartrige seals an a stainless steel tube forme a pressurize chamber. An auxiliary circulating system, was connecte with the chamber to pressurize the working flui (plain water in this research) an take away the heat generate by the friction of mechanical seal. Two WD S/N FQ6 AE sensors (ch an ch) with an operating frequency range from khz to MHz were employe to obtain the AE signals, allowing high frequency events ue to asperity contacts to be monitore. The transucers were place irectly on the cartrige of the non-riven en (NDE) seal as shown in Figure (right). The signal from AE sensor is amplifie an acquire by a MHz high spee ata acquisition system with 6 bit resolution. 5
AE Ch Bearings AE sensors Motor DE seal NDE seal AE Ch Figure. Layout of the test rig (left) an position of AE transucers (right). Experimental Proceure In the inustrial applications, the AE signal is create not only by the contact of the mechanical seal faces, but also from other sources, such as the bearings, electrical motor an electromagnetic noises. In orer to monitor the sliing contact of the mechanical seals, it is necessary to istinguish the AE signals generate by the seal faces from the backgroun noises, which requires an appropriate experimental proceure an signal processing metho. The experimental stuy in this work inclue four comparative programs: manual operation of seal (MOS), running with sensor on seal (RSS), running with sensor on flange (RSF), an test uner ifferent spees an pressures (TDSP). In the MOS experiment, the shaft was turne manually to generate a sliing of seal faces. The purpose to carry out this test was to relate the acoustic emission event to the sliing contact of faces. In aition, MOS an RSS were use to compare the ifference between AE signals picke up uner static an running conition. The RSF refers to running the test on conition that the AE sensors were mounte far from seal faces as shown in Figure5. As can be seen, in RSF experiment, the position of AE sensors is far from the sealing area, consequently it is expecte that no AE signal woul be picke up from the sealing interface ue to the quicker attenuation of higher frequency signal. Purpose to carry out this test was to evaluate the noises. AE Ch AE Ch Figure 5. Position of the sensors on flange Both the RSS an RSF tests were carrie out uner bar (seale pressure) an at 9 rpm to allow the seal faces to come into contact. These tests woul be compare to evaluate the proper frequency ban of frictional AE signal an eliminate the noise signals.the aim of TDSP program was to establish a correlation between AE activity 6
with rotational spee an loa. This was accomplishe by controlle incremental loa at constant spee. The TDSP experiment consiste of five ifferent spees: 9,,5,8 an rpm, at each spee there were three ifferent loas (seale pressure): bar, 5bar an 8bar. To evaluate the repeatability of the tests, each test has been run for three times. Through all the tests the seals were coole with plain water. The tests were performe uner near isothermal conitions an starts save at a temperature aroun 8 C, so that the effect of temperature on the acoustic emission responses can be reuce. 5. Results an Discussion 5. Frictional AE Generation During this test the AE ata acquisition system was kept on working an the shaft of the test rig was turne manually. AE events were clearly observe in the signals picke up by both AE sensors. Since it is impossible to generate noise by just turning the shaft, the AE was likely cause by the rubbing of seal faces rather than by the noise sources of the test rig. As an example, Figure 6 shows the typical raw AE signals aroun an event. It can be observe that only backgroun noises were recore before an after the turning of the shaft. This proves that the signals from the AE sensors are the acoustic emissions cause by the rubbing of seal faces. When the shaft of the test rig was turne, the amplitue of AE signals from the both sensors increase significantly an their magnitues were similar..5 AE event ch(blue) an ch (re).5 -.5 - Backgroun noise -.5 5 5 5 Figure 6. The raw AE signal in the MOS experiment 5. Ientification of the AE Signals from Faces To insight into the AE signals from seal faces, analysis in the frequency omain was carrie out on experimental ata an the results are shown in Figure 7-9. The AE signal from RSS test coincies very well with the signals in the MOS test. It can be seen that the frequency peaks is approximately the same but the amplitue of the former was higher than that of the latter. Compare the MOS test, the RSS test generates more backgroun noises an burst type emissions. In RSF, the AE energy mainly concentrate in two frequency bans whose center frequencies were about 5 an khz, as shown in Figure 9. In RSS, however, in aition to the above two frequency bans, there were some higher frequency bans whose center frequencies were about 7
,7,7 an 5 khz. Base on our comparison of AE spectra, It can be conclue that the frequency of AE from the seal faces was concentrate in six frequency bans at center frequencies of about 5,,,7,7 an 5 khz, an that the first two lower frequency bans also existe in all AE signals an signals from other sources of the test rig. These results le us to consier the frequency ban aroun from to 5 khz to be the specific ban from the seal faces. To eliminate the noise signals, we use a igital ban-pass filter of the frequency ban 5 khz. RMS, skewness an kurtosis of the filtere signals were calculate to express the AE energy generate by seal faces an to characterize AE signals from mechanical seals. x -.9.8.7.6.5.... 5 6 Frquency(kHz) Figure 7. The AE spectrum in MOS test for ch 6 x - 5 5 6 Frquency(kHz) Figure 8. The AE spectra in RSS test for ch 6 x - 5 5 6 Frquency(kHz) Figure 9. The AE spectra in RSF test for ch 8
5. AE Waveform Figure shows the amplitue of the AE time omain signals of the NDE mechanical seal when operate at 5 rpm rotational spee with three loas; bar, 5 bar an 8 bar. Three tests were performe. As can be seen, the AE signals in three tests show no significant ifference in each column. It can be conclue that the repeatability of AE signal was goo, furthermore in each row, when the loa increases (from right to left) the amplitue of AE signal increases. ch ch Test uner 8bar-5 rpm Test uner 5bar-5 rpm Test uner bar-5 rpm - - - 5 Test uner 8bar-5 rpm 5 Test uner 5bar-5 rpm 5 Test uner bar-5 rpm - - - 5 Test uner 8bar-5 rpm 5 Test uner 5bar-5 rpm 5 Test uner bar-5 rpm - - - 5 5 5 5. AE spectrum Figure. AE signal for NDE mechanical seal Figure shows the spectrum of AE row signals shown in Figure. As can be seen in each row, when the loa increases (from right to left) the peak value also increases ue to more asperity contact which clearly shows the effect of loa on the frictional AE signals. ch ch 6 x - Test uner 8bar-5 rpm 6 x - Test uner 5bar-5 rpm 6 x - Test uner bar-5 rpm 5 6 5 6 5 6 6 x - Test uner 8bar-5 rpm 6 x - Test uner 5bar-5 rpm 6 x - Test uner bar-5 rpm 5 6 5 6 5 6 6 x - Test uner 8bar-5 rpm 6 x - Test uner 5bar-5 rpm 6 x - Test uner bar-5 rpm 5 6 Frquency(kHz) 5 6 Frquency(kHz) 5 6 Frquency(kHz) Figure. AE spectrum for NDE mechanical seal 9
5.5 Spee an Loa Characteristics The correlation between rotational spee an AE RMS values for NDE mechanical seal is presente in Figure. Base on the propose moel in section, the relationship between AE RMS value an contact loa is a square root tren of (W / ), however the AE RMS value is proportional to the square root of sliing spee. From here, it can be conclue that the contact loa has less influence on AE RMS value compare with sliing spee. As can be seen in Figure, the AE RMS value increase slightly with loa (the curves are very close to each other) an is higher for the higher loa conitions. Furthermore, the AE RMS value increase incrementally with the sliing spee. As can be seen the slope of each curve is relatively high however the curves have the same tren which gives goo evience that the effect of sliing spee is more than contact loa. These results are in goo agreement with the propose moel.. AE RMS value in ch RMS value (mv) RMS value (mv).. 8bar. 5bar bar 8 6 8 Spee(rpm) AE RMS value in ch.5.. 8bar. 5bar bar. 8 6 8 Spee(rpm) Figure. AE RMS value an rotational spee for NDE Mechanical seal 5.6 The variation of AE Kurtosis uner ifferent operating conitions Figure gives kurtosis of AE signals measure uring TDSP experiment. It is note that the value shown in this Figure is the ifference between the kurtosis of measure AE signal an Gaussian istribution. As iscusse in Section, the height istribution of asperities is an important property to escribe the characteristic of rough surface. As it is assume that acoustic emissions only generate at the asperity contacts, the istribution of AE signal will be an inication of istribution of asperity contacts. Hence if contact only occurs at some peak asperities, the istribution of AE signal has a kurtosis greater than. The more asperities contact each other, the more kurtosis of the AE signal generate is close to to achieve Gaussian istribution. As can be seen in Figure, for the experiments conucte at low spees (9 rpm), kurtosis peake uner 5 bar an roppe to the lowest value uner bar. When the seale pressure was 8 bar, kurtosis value lays in between the 5 bar an bar. Since kurtosis at low spees varie ranomly with the seale pressure, it oes not support the results presente by mathematical moel that the AE activity increases with number of asperity contacts. A
plausible explanation is that at low spees only the hyrostatic lubrication activate an kurtosis is not sensitive enough to the change of asperity contacts uner hyrostatic lubrication conitions. However, Kurtosis of the AE signals measure at spees higher than rpm only varie in a narrow range aroun zero. This is because of the hyroynamic effect at higher spee, which helpe to establish goo an stable lubricant films in the range of teste pressure. These results prove the effect of number of contact asperities on istribution of AE signal as presente in mathematical moel. Kurtosis Value Kurtosis value in ch 8bar 5bar bar Kurtosis Value - 8 6 8 Spee(rpm) Kurtosis value in ch 8bar 5bar bar - 8 6 8 Spee(rpm) Figure. AE Kurtosis value an rotational spee for NDE Mechanical seal 6. Conclusion This paper was combine with two establishe signal processing methos: time omain an frequency omain analysis in an effort to prouce a powerful combination that in common can istinguish AE feature parameter such as RMS an kurtosis in mechanical seals. Base on obtaine results AE characteristics are irectly correlate to the loa (seale pressure) an rotational spee. It successfully emonstrate the AE energy level is higher for the higher loa an spee conitions ue to higher rate of frictional energy release from asperity contact. The variation of Kurtosis value shows the effect of number of contact asperities on istribution of AE signal. These results are in goo agreement with mathematical moel evelope in this work. Base on the propose moel, AE RMS value increases with flui pressure (loa), rotational spee an number of asperities in contact. The results reveal the promising potential of AE technology an ata analysis metho applie in this work for monitoring the contact conition of mechanical seals an may help in the esign an operation of seals.
References. Lebeck, Alan O., Principles an Design of Mechanical Face Seals, Wiley- Interscience Pulication: John Wiley & Sons, Inc: NJ, (99).. Lokesh A., Gupta an Dimitrios Peroulis, " Wireless Temperature Sensor for Conition Monitoring of Mechanical Seals", Proceeings of the n European Microwave Conference, Amsteram, -7.().. Towsyfyan, Hossein., Gu, Fengshu., Ball, Anrew., A review of mechanical seals tribology an conition monitoring. In: Proceeings of Computing an Engineering Annual Researchers' Conference: CEARC'. University of Huersfiel, Huersfiel, -5, ().. S, Lingar., C,W, Yu., C,F, Yau., Sliing wear stuies using acoustic emission, Wear (99), 6 6 597 6. 5. T, Hisakao., T, Warashina., Relationship between friction an wear properties an acoustic emission characteristics: iron pin on harene bearing steel isk, Wear 6 (998) 7. 6. A, Hase., M, Waa., H. Mishina., The relationship between acoustic emissions an wear particles for repeate ry rubbing, Wear 65 (8) 8 89. 7. Mba, D., Roberts, T., Taheri, E., Application of acoustic emission technology for etecting the onset an uration of contact in liqui lubricate mechanical seals,(6),insight, Vol.8,No.8,86 87 8. HOLENSTEIN, A. P., S.INNOTEC an B., F. Diagnosis of Mechanical seals in large pumps, Sealing Technology, Vol., (996), 9-. 9. - Fan, Y, "The Conition Monitoring of Mechanical Seals Using Acoustic Emissions", PhD thesis,university of Manchester, Faculty of Engineering an Physical Sciences 6.. R.J. Boness, S.L. McBrie, Ahesive an abrasive wear stuies using acoustic emission techniques, Wear,Vol. 9, (99) 5.. J. Miettinen, V. Siekkinen, Acoustic emission in monitoring sliing contact behaviour, Wear, (995),8 8, 897 9.. Yibo Fan, Fengshou Gu, Anrew Ball, Moelling acoustic emissions generate by sliing friction, Wear,Vol. 68, () 8 85..Haral, Lubbinge., ON THE LUBRICATION OF MECHANICAL FACE SEALS, PhD thesis, The University of twente, January 999..J,A, Greenwoo.,J,B,P,Williamson., Contact of nominally flat surfaces, Proceeing of the Royal Society of Lonon Series A 95, (966) 9.