ECNDT 6 - Poster 7 Pulse-echo Ultrasoic NDE of Adhesive Bods i Automotive Assembly Roma Gr. MAEV, Sergey TITOV, Uiversity of Widsor, Widsor, Caada Abstract. Recetly, adhesive bodig techology has begu to play a more promiet role i automotive idustry. Nodestructive evaluatio of the adhesive joits is a challegig task for may reasos: access to the bod is oly available from oe side, the acoustic impedace mismatch betwee steel ad the adhesive material is very large, ad the thickess of the adhesive layer may be sigificatly varied. May ultrasoic techiques for NDE of adhesive bods have bee proposed, icludig time ad spectral domai methods. Geerally speakig, however, the lateral resolutio of these techiques is ot high eough to facilitate the proper evaluatio of automotive joits whose widths are typically 5 mm. To icrease this resolutio i the pulse echo mode, we employed flat, highly damped ultrasoic trasducers havig a elemet size of approximately 3-6 mm. I this case, the small trasducer diameter causes strog diffractio i the ultrasoic wave as well as mode coversio at the iterfaces betwee the layers of the joit, resultig i a very complex received waveform. To detect disbods withi the joit, the echo sigal is compared with a referece waveform that was obtaied from the first metal sheet outside of the joit. I the case of disbodig at the frot metal-adhesive iterface, the received ad referece waveforms are very close. Disbods at the rear adhesive-metal iterface cause a phase iversio of the reflected respose; this ca be estimated by referece waveform subtractio. The proposed techique has bee tested usig a set of steel ad alumium samples with the varyig adhesive layer thickesses. Itroductio I recet years, adhesive bodig has become a itegral part of automotive productio techology. As such, the durability of adhesive joits is a essetial elemet cotributig to the overall stregth ad safety of the automobile. For this reaso, it is ecessary that quality assessmet be icorporated directly withi the productio process. The umerous advatages of ultrasoic o-destructive evaluatio methods make it perfectly suited for this applicatio. The body of a typical car is comprised of puched metal sheets with may joits cosistig of two or more of these sheets (.7 mm thickess) with layers of adhesive material betwee them. Thickess of the adhesive ca vary betwee. ad mm. Ultrasoic testig of these adhesive joits is difficult for two primary reasos: () the high acoustic impedace mismatch betwee the layers, ad () the high atteuatio of ultrasoud iside the adhesive layer itself. Most research i this area is focused o the detectio of void disbods at both the first metal/adhesive iterface (closest to the trasducer) ad the secod adhesive/metal iterface (furthest from the trasducer). Due to the great impedace mismatch betwee the adhesive ad the metal (especially steel) there are strog, log lastig reverberatios i the first sheet that effectively mask the small echoes that result from the secod iterface. May ultrasoic techiques have bee proposed for NDT of the automotive adhesive joits []. Methods based o the decay rate of reverberatios i the first sheet ad
measuremet of the reflectio coefficiet [], ufortuately, ca oly be used for the detectio of disbods at the first iterface. To assess the itegrity of the adhesive joit as a whole, a techique has bee proposed i which the first through thickess low frequecy resoace of the joit is measured [3]. Despite the arrowig of the ultrasoud beam by a specially desiged collimator, low frequecy resoace ecessarily implies a restrictio i the lateral resolutio of these techiques a attribute that is highly desirable i the evaluatio of typical automotive joits. A higher degree of resolutio could be potetially obtaied usig commo pulse echo techique. To reduce the amplitude of the reverberatios i the first sheet ad, i tur, to detect the phase of the echo from the secod iterface, digital sigal processig algorithms based o iverse filtratio have bee developed [4 6]. However, the iverse procedure requires the exact waveform similarity of successive echoes. However, the iverse procedure requires the waveform to simultaeously resemble that of successive echoes. This requiremet is ot satisfied i cases of strog acoustic beam divergece ad mode coversio at the iterfaces betwee the layers. To extract the echoes from the secod iterface ad estimate the reflectio coefficiet at the first iterface, we propose a processig algorithm based o compariso of the output waveform with a previously recorded referece respose from the first metal sheet outside of the joit.. Simplified wave model I this paper, we cosider the pulse echo NDT of adhesive joits with flat iterfaces ad ukow adhesive layer thickesses where oly sigle side access is permitted. The classificatio of typical defects is preseted i Fig.. I short, the method should distiguish three mai cases: good joits (case C), void disbods at the first metal/adhesive iterface or a altogether absece of adhesive material (case A), ad void disbods at the secod metal/adhesive iterface or absece of the cotact betwee the adhesive material ad the secod metal sheet (case B). Trasducer Metal sheets Adhesive A B C Figure. Types of defects: A disbod at the first iterface, B disbod at the secod iterface, C good joit. Amplitude calculatios for multiple echoes received from multilayer structures have already bee published (see, for example, [7]). Eve i the case of a two-layer structure, the echo patter ca be rather complicated. I this study, the most proouced set of pulses was selected. Oe set of pulses, a, is produced as the wave reverberates i the first metal sheet, where is a umber of the pulse, =, Aother set, b, is geerated as the wave experieces a oe-fold propagatio withi the adhesive layer, ad is reflected from the secod iterface. There are also sets of pulses resultig from several reflectios withi the adhesive material. Due to strog atteuatio, however, their amplitudes are cosiderably dimiished. I additio, although there are umerous pulses geerated from the propagatio of the wave
i the secod metal sheet, due to low trasmissio, the amplitudes of the partial pulses are eve smaller. As a result, they usually produce a oise like output waveform. Lettig A be the amplitude of the mai bag pulse, the amplitudes of the reverberatios i the first layer are: T a A γ R ( R R η ) = K = () T K = A, γ R = R R η () where T, R represet the trasmissio ad reflectio coefficiets at the trasducer/first sheet iterface, ad η is the atteuatio coefficiet i the first sheet. The reflectio coefficiet at the first metal/adhesive iterface R is egative; moreover, i the case of disbod at the first iterface R =. The correspodig time delays of these pulses with respect to the mai bag pulse are τ. Let h be the amplitude of a pulse passed oe time i the adhesive layer ad times i the metal layer: T T h ( η ) = = a R R K γ γ (3) R R γ = R η R (4) T ad R represet the trasmissio coefficiet at the first iterface ad reflectio coefficiet at the secod iterface, respectively, ad η is the atteuatio coefficiet i the adhesive layer. For a good joit R >, whereas, i the case of disbod at the secod adhesive/metal iterface R =. The time delays of these partial pulses are τ + τ. Thus, partial pulses with amplitude h have the same time delay, ad due to costructive iterferece, the resultig amplitude b is: b T h = K R = γ γ (5) The amplitudes a, ad b, calculated for the steel epoxy adhesive joit, are preseted i Fig.. The acoustical properties of the materials are listed i Table where the atteuatio coefficiets used i the calculatios were η =, ad η =.7. The output waveform ca be represeted as a superpositio of several resposes associated with the reflectios at differet iterfaces of the joit. I case of a sigle metal sheet, the output waveform s (t) cosists of reverberatig pulses havig amplitudes a (solid lie i the left graph, Fig. ). If the metal sheet is i cotact with the adhesive, the respose s (t) produced by reflectio at the first metal/adhesive iterface has amplitudes a (dashed lie). Reflectio at the secod iterface gives the respose s (t) which cosists of the pulses with amplitudes b (doted lie). 3
Table. Material properties. Material Desity, kg/m 3 Velocity, m/s Impedace, 6 kg m - s - Polystyree (trasducer) 34 56.47 Steel, 589 77 45.4 Alumium 63 7 7.6 Adhesive (epoxy based) 5 3.65.9.9.8.8.7.7.6.6.5.5.4.4.3.3.... 3 4 5 6 7 8 9 3 4 5 6 7 8 9 3 4 5 6 7 8 9 3 4 5 6 7 8 9 Figure. Calculated amplitudes of the resposes: left graph a, a (solid ad dashed lies), b (doted lie); right graph a (solid lie), ad a + b (doted lie). The output waveform ca be estimated as a sum of the resposes s(t)= s (t) + s (t) ad umerous small pulses produced by multiple reflectios iside the adhesive layer ad the secod metal sheet. The polarity of the respose s (t) depeds o the impedace ratio at the secod iterface ad ca be used to detect disbod betwee the adhesive layer ad the secod metal sheet.. Sigal processig. The first metal/adhesive iterface To evaluate the first iterface, usually the differece betwee the decay rates of amplitudes is used []. This differece is quite otable for the resposes s (t), ad s (t), where s (t) is a respose obtaied for a semi-ifiite adhesive. I practice however, this discrimiatio should be udertake betwee s (t) ad s(t)= s (t) + s (t). The behavior of the amplitudes of the pulses s(t) depeds with time delay τ, ad due to costructive iterferece, it may be very close to the decay of s (t) as it is show i the right graph of Fig.. To detect disbod at the first iterface, we propose to compare the waveform W (measured at the poit of iterest) with the referece waveform S ref that was previously recorded outside of the joit. The estimated deviatio of the differece R=W S ref ca be used as a measure of sigal similarity, as it is show i Fig. 3. 4
Measuremet of the referece waveform S ref i the area of a sample outside of the joit Measuremet of the waveform W i the area uder the test Subtractio R=W S ref Evaluatio of deviatio r = R r > threshold r < threshold Adhesive at first iterface Disbod at first iterface Figure 3. Sigal processig algorithm for disbad detectio at the first iterface.. The secod adhesive/metal iterface To detect disbods at the rear adhesive/secod metal sheet iterface, we propose to compare the measured waveform W with the referece waveform S`ref, which simulates the reverberatios i the first metal sheet i the case of a semi ifiite adhesive layer (Fig. 4). The waveform S`ref ca be calculated based o the experimetal waveform S ref by itroducig a additioal dampig. Therefore, the differece W S`ref is a estimatio of the respose s (t), where determiig the phase allows a coclusio to be geerated regardig disbod at the secod iterface. Geeratio of the referece waveform S`ref Measuremet of the waveform W for the sample uder test Subtractio S W S`ref (evaluatio of the respose S) Estimatio of the phase of the respose S (decisio block) o-iverted S Disbod at secod iterface iverted S Good joit Figure 4. Sigal processig algorithm for disbad detectio at the secod iterface 5
3. Experimet 3. Experimetal setup To cofirm the proposed method, sets of steel ad alumiium samples of types A, B, C (Fig. ) were prepared. The thickesses of the metal sheets were i the rage of.7. mm, ad the thickess of the adhesive layer gradually varied alog the specimes from. up to mm. A Paametrics V99 removable delay lie trasducer havig a 5 MHz cetral frequecy ad a 4 mm elemet size was used i the experimets. 3. The first adhesive/metal iterface To detect disbod at the first iterface, the waveform W was compared with the referece waveform S ref. The waveforms measured for the A, B, ad C samples are show i Figs. 5 ad 6 as dashed lies. The deviatio of R=W S ref is very small for the sigle sheet, whereas the presece of the adhesive causes its otable icrease. Figure 5. Waveform W (dashed lie), W R ref (solid lie), obtaied for mm steel sheet. 3.3 The secod adhesive/metal iterface Fig. 7 shows the results of the sigal processig followig the algorithm preseted i Fig. 4. Substatio of the referece S`ref from the waveform W effectively suppresses the reverberatios i the first metal sheet s (t) ad reveals residual resposes. Amog these resposes, s (t) is domiat. The amplitudes of s (t) reach maximal values at =4, which is i agreemet with the theoretical model (Fig. ). The time delay τ betwee the pulses of the resposes s (t) ad s (t) is proportioal to the thickess of the adhesive layer. The phase of s (t) is o iverted for the disboded secod iterface, whereas a good cotact at this iterface causes the phase iversio. 6
Figure 6. Waveform W (dashed lie), W R ref (solid lie), obtaied for a sample with disbod at the first iterface (upper graph), ad fully boded (lower graph). The thickess of the adhesive is.6 mm. Figure 7. Waveform W (dashed lie), W R`ref (solid lie), obtaied for a sample with disbod at the secod iterface (upper graph), ad fully boded (lower graph). The thickesses of the adhesive are.6 ad.5 mm. 7
4. Coclusio The proposed NDT method of the adhesive joits has bee successfully tested for all combiatios of steel (.7,.,.5, ad. mm) ad alumiium (., ad.5 mm) adhered, ad for all thickesses of the adhesive layer withi the rage of.. mm. Refereces [] Adams R.D., Drikwater B.W. Nodestructive testig of adhesively boded joits. NDT E It., 997, v. 3 (), p. 93 98. [] Goglio L., Rosetto M. Ultrasoic testig of adhesive bods of thi metal sheets. NDT E It., 999; v. 3 (6), p. 33 33. [3] Robiso A.M., Drikwater B.W., Alli J. Dry coupled low frequecy ultrasoic wheel probes: applicatio to adhesive bod ispectio. NDT E It., 3 v. 36 (), p. 7 36. [4] Freematle R.J., Charllis R.E., ad White J.D.H. A Z trasform techique for thi layer reverberatio cacellatio applied to ultrasoic NDT of adhered structures. IEE colloquium. Digest No. 994/. 7/ 7/4, 994. [5] Charllis R.E., Freematle R.J., White J.D.H., ad Wilkiso G.P. Ultrasoic compressio wave NDT of adhered metal lap joits of ucertai dimesios. Isight, 995, v. 37 (), p. 954 963. [6] Charllis R.E., Freematle R.J., Wilkiso G.P., ad White J.D.H. Compressio wave NDE of adhered metal lap joits: ucertaities ad echo feature extractio. Ultrasoics 996, v. 34, p. 35 39. [7] Krautkramer J., ad Krautkramer H. Ultrasoic Testig of Materials. Spriger-Verlag, N.Y., 969, p. 53 55. 8