ULTRASONIC TRANSDUCER ATTENUATION BEHAVIOR OF FIBER GLASS COMPOSITE LAMINATES (FGCL) THROUGH SIGNAL FILTERING APPROACH

Transcription

1 ULTRASONIC TRANSDUCER ATTENUATION BEHAVIOR OF FIBER GLASS COMPOSITE LAMINATES (FGCL) THROUGH SIGNAL FILTERING APPROACH M. F. Mahmod 1,2, Elmi Abu Bakar 3 ad M. Z. Mohamed Pauzi 1 1 Departmetal School of Mechaical Egieerig, Uiversiti Sais Malaysia, Nibog Tebal, Pulau Piag, Malaysia 2 Faculty of Mechaical Egieerig ad Maufacturig, Uiversiti Tu Hussei O Malaysia, Parit Raja, Batu Pahat, Johor, Malaysia 3 Departmetal School of Aerospace Egieerig, Uiversiti Sais Malaysia, Nibog Tebal, Pulau Piag, Malaysia mfaisal@uthm.edu.my ABSTRACT I ultrasoic o-destructive tests, trasducer is oe of the most part eed to determie carefully. The atteuatio behaviour of trasducer will impact overall ultrasoic measuremet accuracy base o sigal processig aalysis. I ormal applicatios, ultrasoic ispector relies o trasducer calibratio result produced by the maufacturer where there are o doubts to questio the accuracy of measuremet results. However, the atteuatio behaviour of trasducer ca be defied based o sigal-to-oise ratio (SNR) value. I this paper, the atteuatio behaviour of ultrasoic trasducer was ivestigated based o SNR value through sigal filterig approach. Sice the detectio of flaw i composite lamiates usig ultrasoic o-destructive testig (NDT) approach is highly complex due to oise occurreces, several types of filter were applied ad compared each other i order to propose the suggested filter base o SNR result. A 2.25 MHz sigle crystal immersio trasducer is used to perform ultrasoic scaig for composite lamiates material which is thickess up to 7.4 mm with scaig rate 7.50 mm/sec uder lab coditio. As a compariso result, whe applyig discrete wavelet trasform (DWT) de-oisig approach, SNR was ehaced ad caused defect detectio was easily idetified. Keywords: No-destructive testig, ultrasoic testig, atteuatio behaviour, sigal processig, sigal-to-oise ratio. INTRODUCTION The Ultrasoic o-destructive testig (NDT) are highly demaded ispectio techique for composite lamiates material. The ability to detect iterlayer defects i a thicker pael without complex setup caused this techique become more popular tha others ispectio techique such as x-ray, thermography ad shearography. I geeral practice, ultrasoic ispector used various types of trasducer for ultrasoic ispectio depedig o dimesio of specime, material properties ad scaig coditio. I geeral, the higher the frequecy of trasducer used, the shorter the wavelegth ad suitable for homogeous thi plate material ispectio such as sheet metal or weldig joiig quality ispectio. I commo applicatio, ultrasoic ispector relies o the calibratio result which is produced by maufacturer or routie maiteace schedule ad also with selfexperiece i order to determie the performace of trasducer beig used. This practice however isigificat for uexperieced ispector who really depedig o measurable data. I the past few years, several research has bee performed to improve the detectio accuracy of ultrasoic pulse-echo ispectio. Some researcher focused o electroic circuit based icluded amplificatio, filterig ad variable gai parameter [1] while aother researcher focus o developig a better sigal processig approach to ease the defect detectio process [2] ad [3]. However, these approaches are complex ad more coveiet for research based task. I order to determie the performace of ultrasoic trasducer for detectio accuracy, Howard et al. was measure the sesitivity of sigal gathered usig SNR approach respected with various types of trasducer parameter [4]. Meawhile, certai defect of composite lamiates also applicable to detect sigificatly by usig air-coupled o-cotact trasducer but limit to some dimesio [5]. Sice the SNR approach used as oe of measurable parameters to determie the ispectio quality of ultrasoic measuremet, several works has bee doe icluded pipelie ispectio [6], alumiium buffer rods i molte bath [7] ad i composite material ispectio [8]. However, pre-processig work like sigal filterig ad smoothig are required to ehace the SNR for better flaw sigal detectio. Several studies o pre-processig work for SNR ehacemet has bee doe lately usig 2 dimesioal aalytic wavelet thresholdig method [9], time frequecy eighbourhood frequecy [10] ad wavelet de-oisig approach [11] ad [12]. I this paper, the atteuatio behaviour of both ultrasoic sigle crystal cotact ad ultrasoic immersio trasducer was ivestigated usig three sigal filterig approaches which is smoothig, Butterworth filter ad Discrete Wavelet Trasform (DWT). The selectio of Butterworth filter ad DWT are based o previous research as doe by B. Kirthika et al. [13] ad Vaclav Matz [14] where wavelet de-oisig are claimed applicable to reduce high frequecy oises. Through this approaches, the sigal acquisitio was performed usig ultrasoic pulse-echo testig for fibre glass composite lamiates pael. This experimet was coducted uder ambiet temperature ad all procedure were followed ASTM E1065 stadard [15]. The oise occurreces i acquired sigal were decreased the propagated ultrasoic eergy from the trasducer. Thus, the comparative study of several sigal filterig approach was coducted to defie the better sigal filterig through SNR calculatio. 1286

2 As a result, the selectio of sigal filterig is importat where differet SNR results will impact overall study of atteuatio behaviour for ultrasoic trasducer. METHOD APPROACH a) Overall flow chart I this study, the overall research is divided ito two phases which is sigal pre-processig ad sigal atteuatio aalysis as show i Figure-1 below. Sigal pre-processig phase ivolved sigal acquisitio of a ultrasoic pulse-echo testig, sigal filterig approach icluded smoothig, Butterworth filter ad discrete wavelet trasform (Daubechies packet). The followig phase is sigal atteuatio aalysis where the SNR was calculated before evaluated to fialize the result. geerated high voltage electrical pulse ad the drive the trasducer to geerate high frequecy ultrasoic eergy. This ultrasoic eergy fially propagates through the specime i the form of wave sigal. If the discotiuity occurs alog the wave path, part of ultrasoic eergy will reflect back from the flaw surface. Cosequetly, the reflected sigal from flaw surface as kow as echo will trasform back ito a electrical sigal that fially is displayed o a scree. Durig this ivestigatio, there are 9 sets of ultrasoic testig data was recorded based o type of defect ad locatio alog the wave path. A averagig of SNR calculatio will be performed based o several testig data are performed i ext research step. Figure-2. Ultrasoic testig setup. Figure-1. Coverall process flow. b) Sigal acquisitio I this research step, two differet type of ultrasoic pulse-echo trasducer was used to acquire ultrasoic A-sca sigal from composite lamiates material. The trasducers used were 5 MHz ultrasoic sigle crystal cotact trasducer ad 2.25 MHz sigle crystal immersio trasducer. The purpose to use differet type of trasducers frequecy is to ivestigate whether differet frequecy idicate sigificat SNR result. There are 24 layer-by-layer of fiber glass pre-preg sheets stacked together to form FGCL pael with 7.4 mm thick. Durig ultrasoic testig, the FGCL pael was fully immersed ito the water as show i Figure-2. Pulse receiver c) Sigal pre-processig After sigal acquisitio, the ext research step was sigal filterig as a part of sigal pre-processig phase. I this step, three types of sigal filterig has bee applied which is smoothig, Butterworth filter ad discrete wavelet trasform (DWT). Through smoothig algorithm, each data poit was modified as kow as shift ad multiply techique where idividual poit that higher tha the immediate adjacet poit are reduced while the lower poit tha immediate adjacet poit are icreased. This repeatig process will smoother the sigal by reducig the oise. I desigig the Butterworth filter, the trasfer fuctio coefficiets of a th-order low pass digital Butterworth filter ad ormalized cut-off frequecy, are required. I this study, the author proposed third-order Butterworth low pass filter with ormalized cut-off frequecy is 0.15 rad/samples due to better oise filterig ad sigal feature as compare to higher order. However, the icreasig of ormalized cut-off frequecy will icrease the oise of sigal. Aother sigal filterig approach used i this study is discrete wavelet trasform (DWT). Through this sigal de-oisig techique, oredudat restoratio was produced ad provided better spatial ad spectral localizatio of sigal formatio. Hece, DWT result multilevel decompositio where the sigal was decomposed i approximatio ad detail coefficiet at each level as show i Figure-3 below. I this study, the author proposed the wavelet basis Daubechies3 (db3) up to level 3 for sigal decompositio 1287

3 because the sigal will loss more features if decompose greater tha level 3. the peak sigal, Ps is take before the mea ad stadard deviatio of A-sca sigal extracted. The maximum value of oise ca be calculated usig + where is a appropriately chose costat. SNR ca be expressed as (2) below. Ps SNR ( K ) (2) I this experimet, filtered A-sca sigal i full half display as show i Figure-5 below is used i order to make the calculatio of SNR easily. Peak of sigal, P s, mea sigal, ad stadard deviatio, as labelled are determied usig MATLAB calculatio. Figure-3. DWT tree structure for approximatio ad detail coefficiet. d) Sigal atteuatio aalysis I order to determie ultrasoic atteuatio sigal profile, several set of ultrasoic A-sca raw RF-sigal alog sigle path regio-of-iterest (ROI) has bee recorded. Figure-4 show raw RF-sigal before covert to full half display for better aalysis. P s Figure-5. Ultrasoic atteuatio wave sigal. P SNR ca be expressed as (3) below where the subtractio of peak sigal, P s ad mea, is divided with multiplicatio of costat, K ad stadard deviatio,. SNR P K s (3) Figure-4. Ultrasoic A-sca raw RF-sigal. Several defiitio of SNR i ultrasoic idustries have bee itroduce where the applicatio is to idetify quatity useful such as detectability of flaw, idetify the trasducer performace ad wave sigal study. However, i this experimetal, the author ted to use oe of commoly use which is based o sigle trasducer locatio. The peak of the sigal which is deoted as P s is divided by a estimated of peak of oise, P to form the SNR as (1) below. P P s SNR (1) Statistical approach also ca be apply to estimate a maximum oise value. A ultrasoic A-sca sigal alog ROI which cotai the sigal ad oise is use. The, Compariso betwee filtered ad o-filtered SNR associated with differet type of defect are discussed i result ad aalysis. RESULT AND ANALYSIS I order to ivestigate a ultrasoic trasducer atteuatio behavior, the author compared SNR value based o differet type of sigal filterig approach. Nie repeatig measuremets alog oe directio of a ROI was performed i order to esure the reliability of result take. Besides, compariso betwee both immersio trasducer ad cotact trasducer was doe. Ultrasoic sigal graph i Figure-6 shows compariso betwee ufiltered ad filtered sigals usig various sigal filterig approach. The detail compariso usig SNR calculatio are discussed i the followig sectio. 1288

4 Figure-6. Filtered ad o-filtered ultrasoic atteuatio wave sigal. a) SNR betwee sigal filterig ad type of defect I this sectio, the selectio of sigal filterig approach was idetified. Through SNR result, the differece betwee sigal filterig are sigificatly differetiated. Table-1 shows the compariso SNR value i various sigal filterig with differet size of defect. Third level DWT result highest umber of SNR i all size of defect while Butterworth filter ad smoothig algorithm demostrate less sigificat as compared to ufiltered sigal. I additio, the lower the level of DWT de-oisig, the lower the SNR value achieved. It is because, the process of oise removal are respectively with the higher level of DWT sigal de-oisig approach. immersio ad cotact trasducer is slightly high from 4.62 to 9.81 percet due to higher frequecy used. Moreover, third level of DWT decompositio remais better sigal de-oisig approach because were resulted higher SNR value amog others sigal de-oisig approach. Table-2. SNR of defect i various sigal filter. Table-1. SNR of differet sigal filterig. b) SNR betwee sigal filterig for differet trasducer Sice the selectio of sigal filterig approach was idetified by comparig SNR value, a atteuatio behavior of immersio trasducer was studied through SNR of differet trasducer used. Based o Table-2 below, a ultrasoic cotact trasducer demostrated higher SNR umber i every type of sigal filterig almost 36 percet. The differece of SNR betwee CONCLUSIONS The ivestigatio of ultrasoic trasducer atteuatio behavior is demostrated usig sigal filterig approach. The atteuatio measuremet was performed by comparig the frot ad back surface echoes sigal usig pulse-echo mode. Ultrasoic sigle crystal cotact trasducer idicated highest SNR as compared with immersio trasducer up to 9.81 percet due to higher frequecy used. Thus, the SNR icreased with the icreasig of trasducer frequecy. As sigal filterig compariso, DWT de-oisig approach ca be cosidered as oe of the best oise removal approaches i this study. However, the selectio of mother wavelet still required further determiatio for other sigal processig applicatio i order to ivestigate a ultrasoic trasducer. 1289

5 ACKNOWLEDGEMENTS The author would like to thak the Miistry of Educatio (MOE) Malaysia for fudig support through FRGS research grat (203/PMEKANIK/ ), PRGS research grat (203/PAERO/ ), USM Iovative Seed Fud (1001/PAERO/AUP100236) ad awardig him UTHM Academic Staff Traiig Scheme. REFERENCES [1] W. Zhag, P. Zhu, J. Shi, ad Y. Che Research o ultrasoic echo sigal detectio system. Iteratioal Coferece o Computer Applicatio ad System Modelig, Proceedigs. 8(1): [2] T. Boude ad a H. Trasform Sigal processig methods for materials defects detectio. IEEE Iteratioal Ultrasoic Symposium [3] Beammar, R. Drai, ad A. Guessoum Ultrasoic flaw detectio usig threshold modified S- trasform. Ultrasoics. 54(2): [4] P. J. Howard, D. C. Copley, ad R. S. Gilmore A Sigal-to-Noise Ratio Compariso of Ultrasoic Trasducers for C-Sca Imagig. Review of Progress i Quatitative Nodestructive Evaluatio. 14(1): [5] M. F. Mahmod, Elmi Abu Bakar, A. R. Othma, A. R. Ramzi, ad CY Goh Fiber Glass Composite Lamiates (FGCL) Measuremet usig 3 Axis Pulse Echo Scaig Uit. Jural Tekologi. 1(76): pp [10] G. Liao ad D. Liu Time frequecy eighbourhood sigals de-oise method i ultrasoic ispectio. Iteratioal Coferece o Measurig Techology ad Mechatroics Automatio, ICMTMA : pp [11] V. Matz, R. Smid, S. Starma, ad M. Kreidl Sigal-to-oise ratio ehacemet based o wavelet filterig i ultrasoic testig. Ultrasoics. 49(8): pp [12] J. L. Sa Emeterio ad M. a Rodriguez-Heradez Wavelet deoisig of ultrasoic A-scas for detectio of weak sigals. 19th Iteratioal Coferece o Systems, Sigals ad Image Processig, IWSSIP. pp [13] Kirthika, P. Malathi, C.L. Yashwati Sivakumari ad P. Sudharsa A Comparative aalysis of Deoisig Techiques i ultrasoud B mode images. Iteratioal Joural of Advaced Research i Computer ad Commuicatio Egieerig. 3(1): pp [14] Vaclay Matz, Radislav Smid, Staislav Starma ad Marcel Kreidl Sigal-to-oise ratio ehacemet based o wavelet filterig i ultrasoic testig. Ultrasoic. 49(8): pp [15] A America Natioal Stadard (ASTM E1065) Stadard Guide for Evaluatig Characteristic of Ultrasoic Search Uits. 100 Barr Harbor Drive, West Coshohocke, PA , Uited States. pp [6] R. Challis, V. Ivcheko, ad R. Al-Lashi Ultrasoic atteuatio measuremets at very high SNR: Correlatio, iformatio theory ad performace. Joural of Physics: Coferece Series. 457(1): pp [7] H. Viumdal ad S. Mylvagaam Ehacig sigal to oise ratio by fie-tuig tapers of cladded/ucladded buffer rods i ultrasoic time domai reflectometry i smelters. Ultrasoics. 54(3): pp [8] Z. Talebhaghighi, F. Bazzazi, ad A. Sadr Desig ad simulatio of ultrasoic deoisig algorithm usig wavelet trasform ad ICA. The 2d Iteratioal Coferece o Computer ad Automatio Egieerig (ICCAE). 1: pp [9] M. R. Hoseii, M. J. Zuo, ad X. Wag Deoisig ultrasoic pulse-echo sigal usig twodimesioal aalytic wavelet thresholdig. Joural of the Iteratioal Measuremet Cofederatio. 45(3): pp