LEAKY PLATE WAVE INSPECTION OF BIAXIAL COMPOSITES Richrd W. Mrtin University of Dyton Reserch Institute Dyton, Ohio, 45469-0001 Dle E. Chimenti Mterils Lortory Wright Ptterson AFB, Ohio, 45433-6533 INTRODUCTION The utiliztion of leky plte wves in scnning rrngement hs exhiited improved reliility nd incresed sensitivity to importnt defects in unidirectionl mteril [1,2,3]. The ppliction of frequency modultion to the usul tone urst signl used to generte plte wves hs lso een shown to enhnce defect discrimintion [2,3]. In the current work, leky plte wve techniques hve een pplied to the inspection of ixilly lminted grphite-epoxy composites. Test smples hving 8, 16, nd 24 plies, respectively, re studied. Test specimens cont in severl types of defects - simulted delmintions, porosity, nd ply cuts. In ddition, series of impct-dmged smples re exmined to study the method's sensitivity to this type of delmintion. All simulted defects were detected, nd comprisons with conventionl norml-incidence C-scn mesurements hve shown tht the plte wve technique is more sensitive to oth porosity nd ply cuts, consistent with our oservtions on unixil composites [3]. Novel gting methods hve een pplied to the plte wve spectr to improve defect detection in ixil composites. LEAKY PLATE WAVE GENERATION Leky plte wves re generted y em of ultrsound propgt ing to the smple nd striking the plte t n incident ngle. If conditions of em width nd frequency re fvorle, the rerdited field ppers distorted nd displced in the direction of propgtion of the plte wve. In our ppliction trnsmitter excitt ion is provided y swept frequency tone urst which is generted y frequency modult ing the RF tone urst with tringulr wveform t udio rtes in the rnge of 5-20 Hz. Previous pulictions descrie modulted tone-urst leky wve genertion in greter detil [1,2,3]. 1663
DATA ACQUISITION SYSTEM DESCRIPTION The incident ultrsonic em consists of series of swept frequency tone ursts through the rnge of 0.8 to 4.5 MHz [3). The received signl is mplified, video-detected, nd smpled with gted integrt ing mplifier. The experimentl output is then time-vrying signl, most of whose power is concentrted t n udio frequency equl to twice the product of the modultion frequency with the numer of plte modes sutended y the modultion ndwidth. The resulting signl of plte wve modes is then ndpss filtered t 700 Hz on the high frequency end for nti-lising nd t 25 Hz on the low frequency end to remove the fundmentl of the 56 ms sweep period nd D.C. components from the signl. Removl of the D.C. nd low frequency «20Hz) components from the received signl is required to prevent degrdtion of the resolution of the spectrl clcultions descried in lter prgrphs. The plte wve signls re digitized over the entire 56 ms sweep per iod t rte of 2064 smples per second with totl of 128 points recorded. Dt re then sent to microcomputer for processing nd imge genertion [3,4). TEST SAMPLES The chrcterized smples utilized in this study re 6 ixil grphite/epoxy lmintes of 8, 16, nd 24 plies in thickness constructed from AS4/350l type mteril. The stcking sequences re [0)/90 2 ], [0 2 /90 2 ],nd [0 2 /90 2 ]3,respectively. The size of ecn smple is 3 inches wi~~ y 4 inches inslength. Simulted defects re emedded in smples of three different thicknesses. Delmintions re simulted with 0.5 nd 0.25 inch dimeter teflon wfers (0.5 mii thick) nd porosity is simulted with very light scttering of 40 micron dimeter microlloons. There re lso single-lyer ply cuts. Three smples of different thicknesses hve lso een sujected to low velocity impcts with 12.7 mm dimeter stinless steel II on pendulum impcter to produce true delmintions. PLATE WAVE DATA SPECTRAL PROCESSING While the detected RF envelope of the received signl shows minim t vrious frequencies, some chrcteristic of this signl is needed to distinguish etween defects nd defect-free mteril. Our method is to utilize spectrl processing. Dt processing softwre first clcultes 256 point Fst Fourier Trnsform (FFT) on the filtered low frequency detected RF envelope of the received signl. In this cse only the 0-500 Hz frequency rnge of the FFT spectrum is of interest. In the generl cse, the spectrl frequency rnge of interest depends upon the modultion frequency introduced y the sweep, the RF frequency rnge nd the thickness of the pnel under test. The medin frequency of the re under the Fourier mplitude spectrum is computed for ech smpling point on the pnel using integrtion techniques. This medin frequency mesures the overll frequency shifts in the mode dt s mode structure chnges when pssing over defects in the pnel. High medin frequencies re generted in defect-free mteril s the plte wve propgtes through the entire thickness of the pnel (mny modes), while lower medin frequencies re generted over defect re of the pnel ecuse the pnel thickness is effectively decresed nd produces fewer modes. Medin frequency thus provides mesure of reltive defect depth. Figures 1, 2 nd 3 depict these stges in signl processing. In ech figure, the () plots re of the detected RF envelope which cn e viewed s either mplitude versus frequency or mplitude versus time dt. The () plots re the result of pplying 25-700 Hz ndpss filter plus Hmming window to the signl 1664
in (), while (c) is the FFT of wveform (). Note the shift towrd lower frequencies in the FFT spectr (c) progressing from Figure 1 to Figure 3 (effectively thicker to thinner mteril). The medin frequency t ech mesured point (x,y) on the smple i5 stored s one dt vlue in C-scn imge. An dvntge of using the medin frequency technique is tht no prior knowledge of defects or their resulting frequency responses is required to produce high qulity imge. Also, medin frequency spectrl processing provides wide dynmic rnge of response to defects while preserving sensitivity to smll or wek flws. The depth informtion is c curte s long s the frequency difference representing defects close to the sme depth is equl to or greter thn the resolution of the FFT. In the cse presented, the FFT resolution is 8.9 Hz, ut it could e improved y utilizing ny of severl methods. A previous puliction descries leky plte wve spectrl processing in greter detil [3). TEST RESULTS Figure 4 shows C-scn imge of 24-ply test pnel with simulted defects. Dt for the imge ws generted y pplying the medin frequency signl processing technique to the leky plte wve dt cquired from the pnel. The column of defects on the left is locted etween lyers 8 nd 9 in the pnel while the column of defects on the right is etween lyers 16 nd 17. In ech column, the top defect is 0.5 inch dimeter delmintion, the second from top is 0.25 inch dimeter delmintion, the third is 0.5 inch dimeter re of porosity, while the ottom defect is 0.5 inch wide ply cut with 0.125 inch gp. A pir of 2.5 MHz unfocused trnsducers with 0. 375 inch dimeter elements re utilized to cquire dt. The incident ngle of the em to the pnel is 15 degrees from surfce norml, nd the swept frequency tone urst covers the rnge of 0. 8-4.4 MHz in 56 ms sweep time (17.9 Hz modultion). In the imge of Fig. 4, incresing medin frequency is proportionl to defect depth. Progressively drker shdes of gry in the imge represent greter defect depth. Thus, the two simulted delmintions on the left re shown s white ecuse they re closer to the entry surfce of the pnel while the deeper delmintions on the right pper s drker shdes of gry. The porosity nd ply cuts do not conform to the sme gry scle ecuse only prtil signl reflections re possile from these structures due to the width of the em. Defect-free res of the pnel re lck ecuse they represent the mximum pnel thickness, nd thus the highest medin frequency. Figure 4 shows C-scn imge of the sme pnel utilized for comprison purposes. Dt for this imge ws cquired with 10 MHz, 4- inch focused trnsducer with 0.25 inch dimeter element t norml incidence to the pnel. Softwre gt ing nd front surfce trcking were used to cquire the imge dt [5). Not li defects re shown (especilly porosity nd ply cuts) nd of course no depth informt ion is indicted. Figs. 1, 2 nd 3 depict representtive wveforms recorded to generte the medin frequency imge of Fig. 4. The medin frequency for the defect-free FFT of Fig. lc is 128 Hz, while medin frequency for the 1yer 8 delmintion of Fig. 2c is 112 Hz nd for the lyer 16 delmintion of Fig. 3c is 73 Hz. The most notle difference etween these wveforms nd the previously reported wveforms otined with unidirectionl mteril [3) is the incresed wveform complexity; however, the medin frequency lgorithm still correctly encodes depth informtion. Another finding of this investigtion is tht the em direction should e prlle1 to the orientt ion of the top surfce plies to optimize mode structure nd therefore the defect contrst potentil. Incident em ngles etween 10 1665
TlME (ms) o 56 1.0 O.S 50 B 0.5!3 O :> g ~ 25 O -0.5 0.8 4.4 FREOUENCY (MHz) () () 100 200 300 FREQUENCY (Hz) (C) 400 500 Fig. 1. () Plot of plte wve modes of defect-free re showing envelope of detected ultrsonic signl. () Bndpss filtered nd windowed version of signl in (). (c) FFT spectrum of the signl shown in (). O 1.0 TIME (ms) 56 0.5 50!3 o 05. > CI)!:J g O ~ 25 0.8 4.4 FREOUENCY (MHz) -0.5 100 200 300 FREOUENCY (Hz) 400 500 () () (e) Fig. 2. () Plot of plte wve modes of lower level defect showing envelope of detected ultrsonic signl. () Bndpss filtered nd windowed version of signl in (). (c) FFT spectrum of the signl shown in (). TIME (ms) 56 O 1.0 f------' 0.5!3 o 05. > o 0.5 0.8 4.4 FREOUENCY (MHz) Fig. 3. () () 100 200 300 FREOUENCY (Hz) (e) 400 500 () Plot of plte wve modes of n upper level defect showing envelope of detected ultrsonic signl. () Bndpss filtered nd windowed version of signl in (). (c) FFT spectrum of the signl shown in (). 1666
Fig. 4. () C-scn imge of 24-ply pnel utilizing medin frequency, proportionl to defect depth. Drk res re defect-free: gry scle correltes with distnce of defect from surfce. () Norml incidence C-scn of sme pnel gted on ck surfce. nd 30 degrees hve een investigted. Acceptle mode structure ws generted only t ngles close to 15 degrees. Figures S nd S re lso imges of the 24-lyer pnel ut hve een cquired y mesuring the depth of specific nulls in the detected sweptfrequency RF envelope wveform. Two softwre gtes, ech of which detects the minimum dt vlue within the gte, hve een utilized. Their loction nd typicl mplit~des t two specific pnel res re indicted y the position nd heights of the rectngulr gtes shown in Fig. 6. Minimum vlues detected in gte #1 result in low mplitudes tht selectively imge the deeper delmintions (white res) in Fig. S, while minimum vlues detected in gte #2 (Fig. 6) result in high mplitudes (drk res) for li defect re s nd low mplitudes for defect-free mteril (Fig. S). The null in gte #2 is t frequency chrcteristic of defect-free mteril. The contrst nd definition of defect res in Figs. S nd S re superior to those in the medin frequency imge of Fig. 4, ut the resulting signl hs no depth informt ion encoded. A medin-frequency proces sed imge of n 8-ply thick pnel which hs een impcted with n energy of 1 joule is shown in Figure 7. The RF sweep frequency rnge is 1.2 to 4.0 MHz in 56 ms time period. Delmintions from n impct generlly re smll in plies ner the impct surfce, ut grow in re outwrd nd downwrd from the impct nd overlp ech other in successively deeper plies. This morphology is verified y performing softwre 'gted norml incidence C-scn with gtes set t the depth of ech ply interfce within the mteril. Figure 7 is tken from such scn nd is n imge from gte on the ck surfce of the smple 1667
O ~ Fig. 5. () Imge formed y mesuring minim in gte #1. Deeper delmintions re selectively imged in white. () Imge formed y mesuring minim in gte #2. All defects re imged in drker shdes of gry. FREOUENCY (MH ~ L- -LL- ~ -J 43 13 43 FREOUENCY ( MH~ Fig. 6. Positions nd mplitudes of two m1n1mum vlue softwre gtes on the leky plte wve mode spectrum. () lower level defect. () upper level defect. showing the cumultive dmge of ll levels of delmintions. A 10 MHz focused trnsducer hs een utilized. Figure 7 shows two res of dmge ner the top surfce (white res), nd then lrger re s of dmge t greter depths within the mteril, shown y incresingly drker shdes of gry. Blck indictes undmged mteril. Figure 7 verifies the dmged re, ut encodes no depth informtion. The wide coustic em nd lrger x,y step size of Figure 7 result in lower sptil resolution thn the imge in Figure 7. Note tht the leky plte wve medin-frequency processing of rel delmintion (Fig. 7) is much etter thn tht of the simulted delmintion in Figure 4. The improvement is pprently due to greter coustic impednce mismtch in the rel delmintion. 1668
Fig. 7. () C-scn imge of 8-ply ilopct dmged pnel utilizing medin frequency processing of the plte wve mode spectrum. () Norml incidence C-scn of sme pnel gted on the ck surfce. An energy of 6 joules hs een utilized to impct 24-ply thick pnel. A norml incidence 10 MHz C-scn is performed with softwre gte t every ply interfce nd t the ck surfce to verify delmintion dmged res. The ck surfce gted imge is shown in Figure 8. Delmintions re generted in oth the O nd 90 degree ply directions nd they enlrge in re outwrd nd downwrd from the impct site. A leky plte wve scn is performed utilizing frequency sweep rnge of 0.8-4.4 MHz in 56 ms period. A very wide softwre gte is then plced on the received RF envelope of the leky plte wves. This gte will detect the minimum mplitude vlue tht flls within the time intervl of the gte. The gte covers the sweep over the time of out 10ms (1.4 MHz) to 44ms (3.6MHz). The resulting imge is shown in Figure 8 nd hs the interesting effect of outlining the edges of the mny levels of delmintions. The effect might e cused y very deep null eing creted when the coustic em is on the edge of delmintion nd prtilly covering oth delmintions. Although this technique does not provide depth informt ion, reltive depth of the delmintion re s cn e esily inferred y prior knowledge of the mnner in which impct delmintions re formed. The width. nd loction of the wide gte is not criticl, ut experimenttion hs shown tht the est imges re produced with gte width of 0.5 to 0.75 of the sweep time nd nerly centered in the sweep nd. CONCLUSIONS Swept frequency leky plte wve techniques hve een shown to provide relile detection of simulted delmintions, porosity, nd ply cuts in ixil lmintes. Also, the extent nd reltive depth of delmintions due to impct dmge cn e determined. Spectrl processing techniques void the prolems ssocited with trcking the mplitude of specific null ecuse the entire mode signl is utilized. Medin frequency processing of the mode spectr provides good generl purpose processing technique for defect detection nd reltive depth determintion, since medin frequency is relted to depth. Specific minimum vlue gt ing techniques hve lso een shown to provide importnt informtion nd defect detection cpilities. 1669
Fig. 8. () Norml incidence C-scn of sme pnel gted on the ck surfce. () C-scn imge of 24-ply impct dmged pnel utilizing wide minimum vlue softwre gte on the leky plte wve modes. ACKNOWLEDGEMENTS This work ws performed t the Air Force Mterils Lortory t Wright-Ptterson AFB, Ohio nd supported in prt under Contrct Numer F33615-86-C-5016 nd y the AF Office of Scientific Reserch. The uthors would like to thnk Mr. Mrk Ruddell for his cre in performing the experiments descried in this pper. REFERENCES 1. D.E. Chimenti nd Y. Br-Cohen, 1985 IEEE Ultrsonic Proceedings, B.R. McAvoy, ed., New York, (1986), p 1028 2. D.E. Chimenti nd A.H. Nyfeh, J. Appl. Phys. 58, (1985), 4531. 3. R. W. Mrtin nd D. E. Chimenti, "Signl Processing of Leky Lm Wve Dt for Defect Imging in Composite Lmintes", Reviewof progress in ONDE, VoI. 6A, D. O. Thompson nd D. E. Chimenti, eds., Plenum Press, New York, 1987, pp 815-824. 4. R.W. Mrtin nd R.J. Andrews, "Bcksctter B-scn Imges of Defects in Composites", Review of Progress in ONDE, VoI. 5B, D. O. Thompson nd D. E. Chimenti, eds., Plenum Press, New York, 1986, pp 1189-1198. 5. B. G. Frock, R. W. Mrtin, T. J. Morn, nd K. O. Shimmin, "Imging of Impct Dmge in Composite Mterils", Review of Progress in ONDE, VoI. 7B, D. O. Thompson nd D. E. Chimenti, eds., Plenum Press, New York, 1988, pp 1093-1099. 1670