eee DEFECT DETECTION IN RAIL Application and potential of guided wave rail inspection JL Rose, Md Aviali and W-J Song Three applications of guided waves for rail inspection in the railroad industry are considered: fixed sensors on rail fa guided wave rail inspection ear: and @ sensoron-irain system. Basie elements of guided wave analysis, reflection dnd transmission principles, penetration power, modes of ‘nergy induction and reception, sample experiments on air transducers, cut rail, successes to date and future directions are discussed. Dr Joaph Rose i the Pau Morrow Dr Reve weceived tis PRD fom Dreel University 1989. He recived varity of ewan anda fiat nth 1995 Discover ‘wands for tecoloncal onvaion in Avarion dnd Aerospace for he devclpet of hand Iai probe for ageing aircraft mapecton Hea tothe rcpt of Universy Faculty Scholar Medal for acievemen in Engnerng in 1996, wn outtnng research warn 1997, and apron research 3002 e wa ato the ASNT Mehl Honour Ltr ecient in 200 ‘Dv Rove ts author of over to pte, our ta books nd over 400 ances on lvesonc NDE, wave mechanic medical ulrasund,othexine Trnding, concrete Inspection. pipe and abi nection, and composite inate inspection. Textbooks lace Basle Piyscs in Diagnostic Urasound fon Wie & Sons I, NewYork, 1979, and Ulrasonic Waves in Sold Medi Conbraige Unoery Pres, 1989. He has served as this 815 years as o project manager with the &, a BR Dr Woncsoon Song. i working os @ Post. tetrad eacarch essrite tn the Pe State Uirasonie Lab He recived @ PAD degre fro the Deparment of Mechanical Engineering ‘Norhettom University ir December 2000 ond BBS degre from the Diparinen of Mechanical in Febvary, eis neesed conducting rescerch both experiments ond anciycal conpaston, on CEH mui layer sinactres aug uresoni ued wes, ‘Aadivess for correspondence: 411 E Bart & EnginecringSiencesBuldin, Dept of Engineering Sciences and Mechance, Pen State Universi Univeriy Park #416902, USA Tle +1 (814) 863.8026; Pax: +1 (814) 8/64: E-mails Geng Insight Vot 44 No @ June 2002 Introduction “The widespread applications of guided wave inspection technology tare outlined in the literature. Basie physics and mechanics ff guided wave propagation can be found in further studies fand recent work on guided wave inspection of ral is reported elsewhere! "Atal @ natural ave-guide, Energy is tapped inthe rail it propagates, Wave interference ceurs asthe wave modes convert fant ounce back an forth until various wave packets of energy fate formed. Basic principles of dispersion curve and wave strc ‘analysis ypicl of those discussed in the literature” are applicable but effects now become three-dimensional. Wave propagation is ‘more complex. Certain modes and frequencies may’ not be able 10 inspect a fail in all lations, head, web, base for example. One possible wave siuctre from hundreds of possiblities is illustrated In Figure 1 for displacements w, ¥, and w, in the x y, and z tivections respectively, For the sample result shown, defects in the head or base could be detected reasonably well, but with llmost no detection possiblity inthe web, Hostever, for another mode and frequency, some sensitivity in diferent areas would be Dbtained. A combination of thocsical and experimental analysis isl be necessary to cary forward development of complete rail inspection protocol 4 Figure 1. Three-dimensional wave structure distribution possibilty along slices of ral for @ particular mode R. and Frequoncy F “Thice possibilities exist for guided wave rail inspection. They ace considered in this paper, each of which are briefly discussed below Fixed sensors 0° ral IF it were possible to place sensors permanently om a rl located for example every to of thee miles, it would be possible 1 implement an inspection for broken rail ina very aight-forward fashion, An activator could be used, possibly tonsisting of an impact device or a 20, 40 ot even 60 KHz ‘xeitation source, whereby the simple techniques of pulse echo fnd through transmission could be used to isolate broken ral Situations, The tain itself could be used as a source, since tltrasonie energy is induced int the nil rom the moving ain,a Guided wave rail inspection car is possible o develop puided wave rail inspection car that could travel up to say 30 miles per hour along a rail with ‘ulrasonie transducers mounted say 30, 40, oF 80 feet apart, whereby energy can be induced into the rail at one ead and received at the other end. A variety of different transducer arrangements could be considered, include laser vibrometery EMAT technology, or air transducers. Some sample reslis ane presented inthis work tat ilstrate how these techniques fare being developed. Some signal procesing is required to Sepate the tain motion noise from the acta induced and received ulsasonic signals eaveling inthe sail. The natura wave filtering characteristis of the rail and specific induced kinds of energy would make this possible ‘Sonsors on a train ‘An interesting idea would be 10 include sensors on a tein (Ope example wood be to we laser vibrometery or air probes that could actually see defects far ahead of the train as it ‘moves dawn te tack at some distance, say two, thee, oF Four niles away. Ulrasonie energy and vibrational patterns would propagate forward from the moving tain and reflected signals ‘would modify the patterns recorded bythe transducer if defects ‘were encountered. Using basie physics and mechanies, some sjnal processing and pattem recognition, i would be possible to identity broken rail and eral situations yng ahead ofthe teai's locomotive. A feasibility sty is reporced in one paper". Harmonic wave induction into the rail is posible via Hereian contact loading between the wheel and Ue el, which serves as 4 moving line source. This eoupled with the combelke sleeper spacing, wheel spacing. rail lengths, et. impars a large range ‘of harmonic motions into the rail. There are obviously many Timitations to the metod that would have wo be overcome, for example rail wen-outs, switches, road exossings,theemite welds in a structure and so on. An examination of various ‘wave propagation and vibration pattems indicates that these Timitations could possibly be overcome (2) Broken rll Figure 2. Gu ‘of mode conve 38 Some basic aspects of wave propagation in & rail that ‘would prodnce cern reception pater are illustrated briny in Figures 2,3, and 4. Asan vleasonie wave impinges on a defect for an inerface between two material, tere is wave reflection and transmission that occurs at that anomaly because of guided ‘wave properties and boundary conditions at the anomaly. Many ‘modes can be reflected and many transmitted. Constant frequency (ertica line) analysis on a dispersion curve identifies the modes ‘that could contribute transmitted or reflected energy. Each made travels with ts own group velocity and amplitude depending on the defects that are encountered, Natural solutions ofthe boundary conditions associated with the wave propagation problem show this. Bot finite element and boundary element techniques are boing used to study tis reflection and transmission problem in a vate of diferent structive and fr a varicty of diffrent detect shapes and sizes (see Figure 5, for example). Figure 2 shows a schematic ofthis wave reflection and transmission possibility If we go further and look at Figure 3 as an example, imagine & tex coming dovn the wack and imparting energy ito the ral. there were no break n the ral, the energy received by the sensor fixed on the train would remain roughly the stme with respect {0 the random variations occurring asthe train travels down the sali sai» — al Settor!—Senor2 Pak | ———— oil MN |) Pulse echo response i) Through-ransmission response Figure 3. Projected sensor time history of elastic wave eneray as tain approaches sections of instrumented rail with no break {nd also instrumented ral with a break Figure 4. Possible wave modulation in wavetorm from a ‘break’ ‘versus ‘no break Insight Vol 44 No 6 tune 2002rai. On the other hand, i reflection occurs, tsuperimposes on to the sending energy, easing all sorts of variations manifesting themselves as modulations ofthe waveform being examined {lor txample, time-varying amplitude, ete.) This scenario is lustated in Figures 3 and 4 Cable ay Preumatic ‘Sysser Piston ‘ } Testa Figure 5 Railimpactor device is Figure 6. Sample received RF wavetorm and frequency spectra from 8 Source 2243 faway Penetration power studies A number of differen techniques wore used to induce ultrasonic fnergy into a sail, One inieresting technigue is ilutated in Figure 5, where a rail impactor device was used to examine the natural fiterng effet ofthe rail on waves travelling in it. sample Fesult acquired approximately 200) ft away trom an impactor source is lustated in Figure 6. An accelerometer was used to ‘quite the data, The group Yelcity in the rail was approximately 3 mnmts. Noice the frequency contest, Significant content exists between 25 kHz up to 58 kHz, showing the ultrasonic ranges of wave propagation that might be considered inthe structure ‘Some interesting conclusions could be drawn from this work which was performed at the Bay Area Rapid Transit system test track in Hayward, California. The attenwation of guided ‘wave ultresonic energy in ral is Fequency-dependent and higher frequency waves in tis case actualy travelled further and faster than the lower frequency ones. Three ranges of frequency can be defined: an audio, mid-range, and so-called ultrasonic level. Although abitrarily defined, the ranges fit the data very well, fs illuszated in Figure 7. The ultrasonic waveform presented in Figure 6 does not show the audio portion since the wave velocity is so slow that is off the range depicted in the Figure, Notice the amplinide decay profiles in Figure 7 for the ultrasonic level ‘he audi, ad the mis-range evel i examining all ofthe signal, ‘Considering a reasonable aise level vale, propagation distances ‘upto 7000 feet were easily achieved Air-coupled transducers ‘A number of non-contact sensor sypes were evaluated: EMATS. laser vibrometers, microphones, and so forth, Only air-coupled transducers had the receiving capability and frequency response 10 detect signals inthe desired frequency range and atthe required Tivol distances. Transducers were obtained from Second Wave Corporation in Boal, PA. Infact, such transducers were shown to detect ultrasonic al signals ive feet vertically above a ahead Insight Vol 44 No 6 June 2002 “The techniques used to acquire data are illustated in Figure 8, long with sample waveform, “To further understand the reception response ofthe air-coupled transducers, set of experiments was conducted placing. the 2° Sa 2 Deanna) Vee ‘af, 20 xe san] : “| \ H cons] eS ¢ ‘unas 0, = 000s vee) . zoo ton icance te) Figure 7. Model of frequency-dependent attenuation and Comparigon of attenuation characteristics for three frequency ‘anges, Modes at spect requoncies have specific atenuation ‘haraciorities| ‘00 vowne TT Te a Figure 8 Alreaupled sensing and sample received wavetorm 385teansducer at selected distances and angles relative the hea op centre, See Figure 9 Five angles ad sx distances , were used. The angles were O° (ooking athe side ofthe rail ead), 30". 45°, 60°, and 90" (looking ‘down onthe head). The distances range from two inches to twelve inches in two-inch incremeats. ecauseof limitations to or mechanical configuration foranale measurements, only O° and 9X” measurements could be made for the to-inch distance, Figure 10 and 11 show the results ofthese experiments. Figure 8. Configuration for receiving angle experiments Rail cut experiments Several rl-cuting experiments were conducted (0 evaluate the possible relationships between signals acquited and depth of cut (OF primary concer was the possibility of Seeking out a monotonic change in amplitide with cut depth. For certain modes and frequency values this shouldbe possible, as indicated in iterate. [An overview of the exercise is indicated in Figure 12, Results for 1.60 kitzpiezostack normal bean tansduce loading in pulse echo fare presented in Figure [3 with data acquired from both sides of the cut approximately 15 feet away. Reasonable resol are obtained bout better result might be obiained with 2 differeat mode and frequency. Throggh-transmission results are shown io Figure 14 Anin, reasonable results ae obaine. ' gas 3 g 08 oe Eos a 3 a Bo2 0 = no Distance Gn) Figure 10. Variation of transducer response versus distance for all evaluated angles, L 3 4in ae Foe Ae Bas # Ba of o 8 0 4 © 7 ‘ng (deere) Figure 11. Variation of transducer response versus angle for the ‘istances shown ail oto ta Ril tengo ou stata) 4 a gees, ko crtrom 7 3 a a oie, 1 a 40 a Figure 12, Overview of the manner in which the cutting was performed ie can fs ca deshiam) cuseiiom Figure 13. Scatte amplitude asa function of cut depth including tren lines from ath sides of the cut i He & 2) cmeeyo on) et tte) Figure 18. Scatter plt of through transmission amplitude as 3 {ution of cut depth including trend lines for the 60 KHz piezo Stack sender and receiver fom the sender at both sides ofthe cut Aiecoupia ransducar Figure 15, Photographs of air-coupled transducers as used ‘uring the second ral cutting experiment “Another set of experiments was conducted with ans transducer os a receiver, soe Figure 15. Results, shown in Figure 16, are feasonable. Looking from the other side, similar sesuls are ‘obtained, see Figure 17 Two sample RF waveforms fr the ai wansdicer receiver case are shown in Figure 18. Another set of experiments wat conducted with 60. kH2 SH EMAT transducers, Excellent waveforms are obiineds sce Figure 19 for the waveform at various distances. Atenuation profiles ofthe data ae ilutated in Figure 20, Raileut experiments ‘were limite in this cas, but results are depicted in Figure 21. The result was as expected, Insight Vol 44 No 6 June 2002Fol pee | a Ee op Ege ees Fe sett Figure 18. Graphical representation of the results from the 60 kite plezo stack senderair transducer receiver experiment tvith trend lines shoven. Dashed lines incicae the total through ralldapth om dom) gh tnenscen Fd cmsen om i mr a ie a Seen a deans Figure 17. Graphical representation of the results from the 60 Kitz stack senderiair transducer receiver experiment with {tend lines shown from the opposite side ofthe cut 10 & Peakep-peak 52 Amplinde 2) © 20m 4000 eo 00010000 Time (s000) 3 E a amo 4000» ean 00010000 Time(s) Figure 18. Sample pulse echo response for the 60 kHz piezo ‘Mack sender and air transducer recelver (echo amplitude Fecorded in gate shown) Conclusions ‘We believe that we have demonstrated that urasound guided ‘wave propagation can be initiated in ral by tansducers. impact evives, and trains. The frequency ranges that suppor puided wave propagation in ril have been identified. We have shown that air-coupled, EMATs and piezoelectric wansducers can be used to ‘monitor propagation in rail, The acquisition of the data to support these conjectures was facilitated by the use ofthe PC lunchbox computer system that we developed Several approaches for rail defect detection systems we possible: Insight Vol 44 No 6 sane 2002 = Detestion of broken ral with pesanentyri- mounted semors with ether piezoelesticorimpactdevices factng a sound sources. wena 38 1 Detection of broken eal with signal processing schemes and tt Permaneatly mounted sensors - ting trins as sound source. : 1 Devection of defects in ral with guided waves generated and received by various senor ndreceiver elements mounted fm a railcar Ourexpeeatal evidenchas son tt Tere wes exhibit argc of ensonteiecton Shancerster and fen ab ww Mo te npingng trapnicenry Tus snort Bearer Writ onthe oer had, wre Some and exhib” pond Showing. anos oo" hs a Sonon of uranic ney Ange teria vaso telecon charac also Sac conte. tat mst ‘aporant tte par el Sie sais they efeced a Sauifian amon of wacoie hogy A socal “no break Grebe ca often Took “Sting eae 7 towfeqiey comtnuons wave EW) power utara Min usd, ere pied of the Sin ss eee 4 rh roma va sanding weve re casa noha fare ng) Sr ee tS epenica om probe posion and Soredoree theapstipeomeny ofall we Figure 19, Normalised graphs ‘Ve ogee os os os 02 Relative Aerolitude yo ATR -0.148 a2 R= 0.9276 ° 20 0 0 eo Pulse-echo distance (fet) Figure 20. Graphical representations, with trend analysis, of EMAT pulse-echo results, ostCut Dept oot oo ot 100 Respense(Sbit) om Figure 21, Pulge echo results 19° 2° away trom defect: logaritymic rena ine alse shown type, angle bar joint, switch, location, ete, Probe pesition was cic, I 8 our feeling that a sensor mounted on a tain could provide better discrimination results thana sensor aa fixed position ‘on the rail unless using five rsx fixed sensors o obain optimum signals Lastly, mach more work on quantification of defects in rail is seeded including tht wo determine defet classification and sizing. Acknowledgment An are given to the Federal Railway Administration, Bay Rapid Transit, TWUPlan Integrity, Nitany and Bald E Railroad for their support ofthis project. References Rose, L, “Standing on the Shoulders of Giants - An Example of Guided Wave Inspection’, Mehl Honor Lecture, Material Evaluation, Vol 60, No I, January 2002 Rose, J L, "Ukrasonics Waves in Solid Media, Cambridge University Press, 1999, Rose, 1, Avioli, MJ, “Elastic Wave Analysis for Broken Rail Detection’, TONE: Nondestructive Testing and Evaluation (NDT&E) forthe Railroad Industy, ASNT, December 2001 Rose. L, Avioli, MJ, “Elastic Wave Analysis for Broken Rel Detection,’ 15" Workd Conference on Nondstrotive Testing Rome, Hal, October 15-21, 2000, Rove. L, Zhao, X, “Anomaly roughwsll depth measurement Materials potential with “shear horizon guided wave Evaluation, Vol 59, No 10, pp 1234-1238, Oct. 2001 Ez ' Step on it & get the speed of service you deserve! Ether we get your equipment seniced within 7 working days, orwell foot the choge forthe work! ‘This new pois supports Hockngs continuing dedication o powcing| you with the best repai and atbratio services onthe market For moreinfrmation on ths offer, plese contact us on the numbers below, vit wa hocking convenes tm HOCKING SEN pay ooesane fan away ISouees svete com andtonsane __ an 4 DINSEARCH 1-00 [ Carbon Steel Tubes in Heat Exchangers | Electromagnetic System | | | Rapid Inspection | Dinsiey Devices Ltd, Wy House } ‘Streatlam Park, Bamard Castie, « ‘County Durham DL12 BTZ “TovFax 01833 637 671 Lp -y No 206-21 358 hr Enquiry No 206-22 Tneight Vol A No 6 tune 2002