ULTRASONIC SIGNAL ANALYSIS OF COMPOSITE STRUCTURES USING THE ENTIRE WAVEFORM William D. Brsey Oak Ridge Y-12 Plant* Martin Marietta Energy Systems, Inc. Oak Ridge, TN 37831 INTRODUCTION Cmpsite structures are ften desirable fr their strength and eight characteristics. Since cmpsites are nt as ell characterized mechanically as metallic r ceramic structures, much rk has been perfrmed t btain that characterizatin and t develp methds f determining the mechanical prperties f a cmpsite nndestructively. Mst f the rk in the literature t date has been perfrmed n nnenclsed structures n hich severa! prmising nndestructive evaluatin {NDE) techniques have been used t predict the integrity f the cmpsite. 1 ~ An increased use f cmpsites in enclsed gemetries has led t the desire t transfer and further develp the mst prmising NDE techniques used n nnenclsed gemetries t enclsed gemetries. 5 6 As part f the further develpment f ultrasnic inspectin techniques, a high-speed data-acquisitin system has been assembled hich digitizes an entire RF signal avefrm at each pint n a Kevlar-epxy cmpsite structure. The cmplete data set f ultrasnic A-scans is then available fr analysis f particular features hich might rdinarily have been verlked hen nly the maximum amplitude ls recrded frm data in a preset gate. The system prvides a three-dimensinal vie f the data in hich either the XY, YZ, r XZ planes can be displayed using pseud clr r gray scale t indicate the maximum signal amplitude in agate set under prgram cntrl. Depth infrmatin can be displayed n the XY data plane, and varius image display techniques can be implemented t enhance certain material r defect cnditins. EXPERIMENTAL CONFIGURATION The equipment used in data cllectin included a turntable, real-time clck, cmputer, CAMAC delay mdule, CAMAC transient recrder, pulser receiver, and ultrasnic transducer. Data ere cllected as a digitized avefrm frm a cmpsite specimen. A 7.5-MHz transducer as used in pulse-ech mde t bth transmit and receive ultrasnic signals. The turntable as specially mdified ith a cmputer cntrlled transducer psitiner hich alled inspectin f the cmpsite test specimens ver * Operated fr the U.S. Department f Energy by Martin Marietta Energy Systems, Inc., under cntract DE-AC05-840R21400. 1247
mst f the surface. The specimen as rtated n the turntable fr a preset number f degrees, at hich time the real-time clck sent an interrupt t the cmputer t begin taking data. The cmputer then signaled the delay mdule t start the pulser receiver. After 63 us, hich crrespnded t the ater path distance beteen the transducer and the specimen, the delay mdule signaled the 100-MHz transient recrder t digitize and stre the avefrm int the cmputer. Data ere transferred t disk cartridges fr permanent strage and pst-acquisitin data analysis. This sequence as repeated fr each f the pints at hich data ere cllected. A schematic representatin f the experimental cnfiguratin is shn in figure 1. (a) PHOTOGRAPH OF DATA ACOUISITION SYSTEM CAM AC TRANSIENT RECORDER 1 CA MAC DELAY MODULE l 1 TRANSDUCER PULSER COMPOSITE ) RECE IVER r-- SPECIMEN REAL TURNTABLE COMPUTER TIMECLOCK (b) SCHEMATIC OF DATA ACOUISITION SYSTEM 1 / Fig. 1. Data-acquisitin system fr three-dimensinal ultrasnic data. 1248
Ultrasnic Inspectin Ultrasnic inspectin f the Kevlar-epxy cmpsite inding as perfrmed ith pulse-ech ultrasnic analysis using a transducer nrmal t the cmpsite surface. Cuplant beteen cmpsite surface and transducer as prvided by a ater stream fling thrugh a bubbler. The bubbler as ffset rrm the cmpsite apprximately 0.15 mm fr a cnstant laminarfl f ater ver the cmpsite. The ater used as demineralized ith all bubbles remved. A zer degree mark, hich had been used fr previus inspectins, 6 as used fr aligning the sphere ith the transducer. An alignment and vacuum hlding fixture as used t psitin the specimen n the turntable and hld it in place. Inspectin f the cmpsite as perfrmed frm 4 suth latitude t 70 nrth latitude in a single scan. By cnventin, the nrth ple f the spherical specimen as lcated at the stern in the mandrel. During the scan, the transducer as incremented 1 each revlutin f the turntable. The test specimen as then inverted and anther scan as made frm 4 nrth latitude t 85 suth latitude. Less.surface area culd be scanned in the nrthern hemisphere because the stern at the nrth ple interfered ith the transducer ater bubbler at lcatins abve 70 nrth latitude. Data ere cllected at 1 intervals arund the specimen at each latitude. Seven micrsecnds f data ere cllected at each lcatin n the specimen hich crrespnded t 700 digitized data pints fr each ultrasnic avefrm. Every seventh pint as stred n the disk cartridge s that each avefrm had 100 pints. The effective frequency f the stred digitized data as 14.3 MHz since each avefrm as digitized by the transient recrder at 100 MHz. Each datum as stred as a byte ith a value beteen O and 127 units. Data Analysis and Display The data cnsisted f avefrms 100 pints in length taken every degree arund the specimen fr 360 at either 90 latitudes fr a suth ple scan r 75 latitudes fr a nrth ple scan. Therefre, the data files fr each specimen ttaled 5.94 millin data values. T display this quantity f data, the data set as cnsidered as a cube here the X-axis crrespnded t lngitude crdinates n the specimen, the Y-axis crrespnded t latitude crdinates n the specimen, and the Z-axis crrespnded t pints n the avefrm. When either the XY, XZ, r YZ data planes ere displayed, the third axis f data as gated and held cnstant. The XY data plane display crrespnds t an rdinary c-scan ith lngitude and latitude as the X and Y axes. T avefrm pints ere chsen t indicate a gated regin f interest. Clr r gray levels in the display crrespnded t the maximum signal amplitude beteen the t chsen pints. This display culd als be pltted in spherical crdinates using axnmetric prjectin. Figure 2 shs a c-scan f a cmpsite specimen in hich tefln-shim-simulated delaminatins had been included. The gated regin fr this C-scan included avefrm pints 50 thrugh 70. A linear amplitude plt culd be btained fr a given lngitude and latitude as shn in figure 2. The linear amplitude plt f lngitudes fr the chsen latitude as dran alng the X-axis at the bttm f the c-scan display, hile the linear amplitude plt f latitudes at the chsen lngitude as dran alng the Y-axis at the right side f the C-scan display. The avefrm fr the pint at hich the chsen lngitude and latitude crssed as als displayed alng the left side f the c-scan. 1249
~ ;:: s LONGITUDE Fig. 2. Linear amplitude plts n lngitude vs latitude C-scan fr thr ee lines f data in X, Y, and Z planes. The YZ data plane display crrespnds t a B-scan shn as avefrm vs latitude. T lngitudes ere chsen t indicate a gated regin f interest. If the t pints cincided, then the lngitude as fixed at a single value. Clr r gray levels in the display crrespnded t the maximum signal amplitude ver the chsen range f lngitudes. Figure 3 shs a avefrm vs latitude linear amplitude plt f a specimen ith imbedded tefln-shim-simulated delaminatins. The linear amplitude plt as made at a fixed lngitude f 270, and as prduced fr a chsen avefrm and latitude t graphically display tefln-shim signal amplitudes. The avefrm amplitude plt as dran alng the X-axis fr a chsen latitude at the bttm f the B-scan avefrm vs latitude display, and the latitude amplitude plt as dran alng the Y-axis fr a chsen avefrm pint at the right side f the B-scan avefrm vs latltude display. The data values for each lngltude crrespnding t the intersectin f the chsen avefrm and latltude are displayed at the left slde f the B-scan. :::>!::: lu..j Q. :::> ::;; t- <{ i=..j <{ <{..J z <:> iii WAVEFORM Fig. 3. Linear amplitude plt s n avefrm vs latitude B- scan fr thr ee l i ne s f data i n Z, Y, and X pl anes. 1250
The XZ data plane display crrespnds t a B-scan shn as lngitude vs avefrm. T latitudes ere chsen t indicate a gated regin f interest. If the t pints cincided, then the latitude as fixed at a single value. Clr r gray levels in the display crrespnded t the maximum signal amplitude ver the chsen range f latitudes. Figure 4 shs a lngitude vs avefrm linear amplitude plt fr a cmpsite specimen cntaining tefln-shim-simulated delaminatins fr hich the latitude range as fixed at 1 abve the specimen equatr. The linear amplitude plt as prduced fr a chsen lngitude and avefrm t display tefln shim signal amplitudes graphically. The lngitude amplitude plt as dran alng the X-axis fr a chsen avefrm pint at the bttm f the B-scan lngitude vs avefrm display. The avefrm amplitude plt as dran alng the Y-axis fr a chsen lngitude n the right side f the B scan display. An amplitude plt fr the data values f all latitudes crrespnding t the intersectin f the chsen lngitude and avefrm is displayed at the left side f the B-scan. The clr r gray scale depth plt display is a lngitude vs latitude plt in hich the clr r gray levels indicate the depth f the maximum signal amplitude ithin a chsen gated regin f interest. The display culd als be pltted in spherical crdinates using axnmetric prjectin. Figure 5 shs a gray scale depth plt f a cmpsite specimen cntaining tefln-shim-simulated delaminatins. The gated regin fr the depth plt included avefrm pints 40 thrugh 65. A histgram f data values culd be shn fr the depth plt indicating the relative number f data pints at each depth. A linear depth plt culd be btained fr a given lngitude and latitude. The linear depth plt f lngitudes.fr a chsen latitude as dran alng the X-axis at the bttm f the depth plt display. A linear depth plt f latitudes fr a chsen lngitude as dran alng the Y-axis n the right f the depth plt display. The avefrm fr the pint at hich the chsen lngitude and lati tude intersected as displayed n the left side f the depth plt. FRONT SURF ACE :::::>... :2..J a: 0.. :::::> :2... u.. t=..j >..J z 3: {!) (/) "'.. LONGITU DE Fig. 4. Linear amplitude plts n lngitude vs avefrm B-scan fr three l i nes f data in X, Z, and Y planes. 1251
J: 1- c.. (.!) z 1- tii ~ <1: <1:...1 a: u z ::> LONGITUDE Fig. 5. Linear amplitude plts n lngitude vs latitude depth plt fr three lines f data in X, Y, and Z planes. Several image-enhancement techniques ere used t islate areas f interest ithin a display. These techniques included histgram equalizatin, arbitrary clr scale and gray scale alteratins, and abslute-value clr scales. In additin, the histgram plt culd be scaled t enhance regins f interest fr either amplitude r depth plts. A histgram f data values as displayed at the right side f the plt. The histgram shed the relative number f each data value ithin the plt. Data values ranged beteen O and 127. Histgram equalizatin divided up the data values such that apprximately equal percentages f data ere assigned t each f 16 available clrs r gray shades. This as a particularly useful technique hen mst f the data ere ithin a narr range f values hich crrespnded t nly a fe clrs. Arbitrary clr r gray scale alteratins ere achieved by assigning each f the 128 data values a separate clr hich culd be changed individually r in grups. The changes ere rade thrugh the use f multiple preset clr maps in cnjunctin ith data value selectin rrm the histgram display. This technique as mst useful hen certain data features culd be enhanced by highlighting a particular set f data values. Lg and inverse lg clr scales ere used t enhance either the l r high value amplitude r depth data. Abslute-value clr scales ere used as an image enhancement because f the biplar nature f the RF avefrm data. In this clr scheme, the t extremes f the data value scale ere given the same clr r gray level. The clr scales then rked in frm the t extremes t the center data value. Therefre, an equally psitive r negative datum uld have the same clr. When vieing data in the XZ r YZ planes, this technique alled the extraneus undulatins f the avefrm t be verlked and nly the magnitude f the signal tbe cnsidered. The base amplitude f the avefrm using this technique as set t zer, instead f the 65 used in the nrmal clr scale, hich added t the ease f vieing the data. Figure 6 demnstrates the difference beteen the nrmal and abslute value clr scales. 1252
::J!:::..J "- ::;!..J z!:2 V'J......... y... -.. j.. 71l 4 se.... 32........ - - e. a....... l. 4 l 4 3 l l.. ~) NORMALCOLORSCALE. TEFLON..... SH IM-----+-'.\' BACK _...:r SUR FACE...-- 128 ::J!:::..J "- ::;!..J z!:2 V'J 128 ::;! a: (!J 1- V'J I (b) ABSOLUTE VAL VE COLOR SCALE. Fig. 6. Abslute-value clr scale signal amplitude B-scan plt f tefln-shim data. 1253
The histgram plt culd be scaled t display the height either as a lg r linear functin f the number f data ith a particular data value. Data values crrespnded t either signal amplitude r depth, depending n the plt type. The idth culd be displayed and scaled as a linear, lg, r inverse lg functin f data values. Thrugh these scaling techniques, small perturbatins n the histgram culd be islated. These regins are ften apprpriate candidates fr arbitrary clr-scale alteratins hich enhance regins f interest n the plt. DISCUSSION OF RESULTS The XZ data plane plt in figure 4 clearly shs the ut-f-rund f the specimen by the sine-ave appearance f the frnt surface. This lack f specimen cncentricity hindered analysis f the specimen by alling either the frnt r back surface f the avefrm t alternately enter the gated regin during a prtin f the scan. Tefln-shim-simulated delaminatins ere detected in the cmpsite specimen at varying depths including ne 20 mils abve the back surface interface f the cmpsite and aluminum mandrel. Delaminatins clse the back surface are ften undetected hen a preset gate is used during data acquisitin. By digitizing the entire avefrm, it as pssible t find defects clse t the frnt r back surfaces, hich therise might have been missed. Varying degrees f epxy matrix ere detected in a specimen und ith sme dry fiber. The ultrasnic signal becmes extremely attenuated at the depth at hich significant lss f matrix ccurs. Near the equatr, here the fiber as und dry fr half f the thickness, the signal attenuated very quickly. At apprximately 30-40 frm the equatr, reflectins frm internal surfaces are detectable near the middle f the inding. The back surface as detectable near the plar regin, indicating the epxy matrix had permeated the entire thickness. A YZ data plane B-scan as used t determine the change in signal attenuatin frm the equatr t the suth ple, hich indicated the degree f matrix penetratin. Macrsphere ceramic vids ere detected using the B-scan data displayed in the XZ plane. C-scan plts f amplitude r depth data did nt readily distinguish the simulated vids frm the surrunding indingpattern nise. These ceramic spheres ere beteen 30 and 80 mils hich as difficult t distinguish frm fiber verlap regins in the cmpsite. REFERENCES 1. E. G. Henneke II, K. L. Reifsnider, and W. W. Stinchcmb, Thermgraphy -- An NDI Methd Fr Damage Detectin, Jurnal f Metals, Vl. 31, N. 9 (1979), pp. 11-15. 2. A. Vary and R. F:-Lark, Crrelatin f Fiber Cmpsite Tensile Strength With Ultrasnic Stress Wave Factr, Jurnal f Testing and Evaluatin, Vl. 7, N. 4 (1979), pp. 185-191. 3. S. S. Lee and J. H. Willlams, Jr., Stress-Wave Attenuatin in Thin Structures by Ultrasnic Thrugh Transmissin, Jurnal f Nndestructive Evaluatin, Vl. 1, N. 4 (1980), pp. 277-285. 4. R. A. Blake, Ultrasnic Image Histgram Evaluatin and Enhancement, IEEE 1983 Prceedings f the 6th Annual Micr-Delcn Cnference, Neark, Delaare (1983). 1254
,. D. M. Byd and B. W. Maxfield, Hlgraphic Nndestructive Evaluatin f Spherical Kevlar/Epxy Pressure Vessels, UCRL-84280, Larence Livermre Natinal Labratry, University f Califrnia, May 23, 1979. 6. W. D. Brsey, Ultrasnic Analysis f Spherical Cmpsite Test Specimens, Cmpsites Science and Technlgy, Vl. 24, (1985), pp. 161-178. 1255