Ultrasonic Phased Array Testing of Welded Components at NASA

ABSTRACTComprehensive and accurate tests of welded components have become

increasingly important as NASA develops new hardware, such as Aresrocket segments, for future exploration missions. Simulation and mod-eling will play an increased role in the future of nondestructive testingin order to better understand the physics of the testing process and helpexplain the experimental results. It will also help to prove or disprovethe feasibility of a test technique or scenario, help optimize testing, andallow, to a first approximation, determination of limits of detectability.This study presents simulation and experimental results for ultrasonicphased array testing of a critical welded structure important for NASAfuture exploration vehicles.Keywords: nondestructive testing, computational simulation, ultrason-ics, weld, modeling, phased array, probability of detection.

INTRODUCTIONAs NASA’s Constellation Program proceeds to develop new

crew and launch vehicles, many nondestructive testing challengeswill be encountered. Ultrasonic techniques will play a role in thetesting of space flight hardware such as beams, welds, lugs andother critical structures. The NASA Glenn Research Center is in-volved in building the upper-stage simulator for the Ares 1-X testrocket, scheduled to launch in 2009. The upper-stage is composedof multiple upper-stage simulator segments that will be stackedand attached to each other. The upper-stage simulator segmentshave critical welds that attach the skin half-cylinders, flange por-tions and the skin to the flange. In this study, the skin-to-flangeweld is of concern.

The skin-to-flange weld is required to pass a handheld ultrason-ic A-scan test for certification. Here, the modeling and use ofphased array ultrasonics to perform such a test is addressed as analternative to A-scan testing. Phased array testing is gaining wideacceptance and has many advantages over conventional A-scantesting, including the ability to perform scanning with no mechani-cal movement, the ability to perform angular (sectorial) scans anddynamic focusing, greater ability to test complex shapes and diffi-cult to test areas, and two-dimensional visualization of results. Sim-ulation and modeling will play an increased role in the future fornondestructive testing in order to better understand the physics ofthe test process and help explain the experimental results. It willalso help to prove or disprove the feasibility for a testing techniqueor scenario, help optimize tests, and allow, to a first approximation,determination of the limit of detectability (Aldrin and Knopp,2008).

TEST SAMPLEAskin-to-flange test sample of the variety used in pressure vessels

was fabricated. The skin and flange portions were approximately 12and 25 mm thick, respectively. The skin portion was double-beveled at one end, 45° on each side, and butted to the thickerflange. Flux core arc welding was performed to attach the skin tothe flange — the interior and exterior areas flanking the bevels ofthe skin were filled with weld metal of a similar composition to thatof the base steel. The external weld was shaved to make the transi-tion from skin-to-flange flat. Figure 1 shows the interior of an Aresupper-stage simulator segment and a solid model of the test part. A

60 Materials Evaluation/January 2009

Ultrasonic Phased Array Testing Simulations ofWelded Components at NASA

by D.J. Roth,* R.P. Tokars,† R.E. Martin,‡ R.W. Rauser,§ J.C. Aldrin** and E.J. Schumacher††

Submitted October 2008

* NASA Glenn Research Center, Optical Instrumentation and NDEBranch, 21000 Brookpark Rd., Cleveland, OH 44135; (216) 433-6017; fax(216) 977-7150; e-mail <[email protected]>.

† NASA Glenn Research Center, Optical Instrumentation and NDEBranch, 21000 Brookpar Rd., Cleveland, OH 44135.

‡ Cleveland State University, 2121 Euclid Ave., Cleveland, OH 44115.§ University of Toledo, 2801 W. Bancroft, Toledo, OH 43606.**Computational Tools, Inc., 4275 Chatham Ave., Gurnee, IL 60031.††Magsoft Corporation, 20 Prospect St., Ballston Spa, NY 12020.

Figure 1— Interior of an Ares upper-stage simulator segment: (a) theskin-to-flange specimen; (b) solid model of the skin-to-flange sample.

(a)

(b)

06_49_66_TPs_18pgs:08_1071_1090_TPs_20pgs copy 12/11/08 3:14 PM Page 60

Materials Evaluation/January 2009 61

side-drilled hole approximately 40 mm in length and 1.61 mm in di-ameter was located at one end of the part, simulating lack of fusionin the weld.

PHASED ARRAY ULTRASONIC TECHNIQUEA portable ultrasonic flaw detector was used for the actual test-

ing. Ultrasonic phased array transducers are made of multiple crys-tals that are electronically pulsed in a sequence using a specifieddelay between pulses to create linear scanning, beam steering or fo-cusing capability. Phased array probes generally consist of a trans-ducer and wedge. The transducer employed was a 32 crystal ele-ment linear phased array type with 5 MHz flat focus. The totalaperture was 16 by 10 mm. The element width was approximately0.45 mm and the gap between elements was approximately0.05 mm. Only 16 of the 32 elements are active at any one time as 16pulser/receivers are available in the instrument. The active aper-ture area was thus 8 by 10 mm. The transducer was attached to awedge made of Rexolite plastic with an incidence angle of approxi-mately 36° and subsequent shear wave refraction angle into steel ofapproximately 54°. Note that an incidence angle of 36° results in asetup beyond the critical angle for longitudinal waves; therefore,longitudinal waves are not of concern in this study. Figure 2 showsthe probe sitting on the external skin portion of the skin-to-flangetest sample. This is the single-sided test scenario for an actual test ofthe welds, whether using A-scan or phased array techniques. A sec-torial scan with a range of 40 to 75° was performed. The probe wasmoved manually along the exterior surface of the skin in order todetermine whether an indication related to the side-drilled hole inthe interior weld was observed.

ADDITIONAL ULTRASONIC TESTING5 MHz shear wave ultrasonic velocity and attenuation coeffi-

cient measurements were performed on the test sample skin (basesteel) and also on a flat sample of the weld steel using a precisioncontact ultrasonic technique (Roth et al., 1995). The shear wave ve-locity values for the base and weld steel were measured to be3.25 and 3.28 mm/µs, respectively, the difference of which is closeto the measurement uncertainty for the velocity measurement tech-nique. The velocity value for the base steel was used as a setup pa-rameter in the ultrasonic flaw detector, and the velocity values forboth base and weld steels were used in the ultrasonic model, as de-scribed below.

Ultrasonic ModelingIn this investigation, software developed by the French Atomic

Energy Commission was utilized to model the phased array ultra-sonic test. The software allows bulk wave beam field predictionsusing the elastodynamics pencil technique and discontinuity re-sponse predictions using Kirchhoff, geometric theory of diffractionor Born models for beam/discontinuity interaction (Achenbach,

1973; Achenbach, 1992; Cinquin et al., 2007; Darmon et al., 2003;Gengembre, 2003; Lonne et al., 2006; Raillon and Lecoeur-Taïbi,2000; Rose, 1999). The software has many options available, includ-ing the ability to compute delay laws for phased array ultrasonic se-tups (Chaffai-Gargouri et al., 2007; Mahaut et al., 2004) and the abil-ity to overlay resulting beam profiles and beam/discontinuityresponses onto the model of the part. Components requiring dis-continuity response predictions need to be either imported as two-dimensional DXF files or drawn in two-dimensional profile in thecomputer aided design (CAD) facility included in the software.After the two-dimensional profile is created or imported, it is sub-sequently extruded to create the three-dimensional solid model.Three-dimensional solid models can be imported into the software,but currently only for beam field predictions. A reasonable two-dimensional profile facsimile of the skin-to-flange test part wasdrawn using dimensions from the actual test part in the CAD facil-ity (Figure 3). For the CAD model, all sides were color coded asfront, back, side and interface, so that the ultrasonic probe attacheditself properly, and the beam field and discontinuity response pre-dictions were calculated correctly. The four zones shown in the two-dimensional profile consisted of two base steel zones and two weldzones. Each was assigned the appropriate ultrasonic velocity asmeasured for the base and weld steels. The software has interfacescreens for the user to describe the transducer crystal element char-acteristics and geometry, phased array crystal assembly, test/scanparameters, wedge material and geometry, test sample materialand geometry, discontinuity material, and geometry and computa-tion parameters. The model parameters selected matched closelywith actual experimental parameters. The side-drilled hole wasmodeled as a cylindrical discontinuity.

The “half-skip” sound path interaction with the discontinuitywas chosen in the computation parameters. When the half-skip op-tion is chosen, beam/discontinuity interactions that are in the directsound path from the probe to the discontinuity and vice versa areconsidered. Additionally, waves incident on discontinuities after re-flection off the bottom surface are also taken into account. Reflec-tions off the bottom surface from the probe toward discontinuitiesare in this case considered, as well as reflections off the bottom sur-face from the discontinuities toward the probe. This option enablesthe computation of corner echoes or, more generally, echoes result-ing from the inward or outward reflection off discontinuities via thebottom surface. Therefore, for the half-skip option, the echoes takeninto account can include the specular reflections, tip diffractionsand corner-like echoes.

Backwall echoes were included in the computation (surfaceechoes were not). At the 40 to 75° sector in steel, shear waves are theactive wave mode and thus were selected as the wave mode formodeling. A computation zone was selected that encompassed theweld area. Computation accounted for mode conversion, but didnot account for material noise, attenuation or shadowing from dis-continuity over geometry. These are all options that can be mod-eled, but increase computation time. The computed delay laws con-trolling the sector scan were not compared with the actual delaylaws for the instrument because they are not readily exportablefrom the instrument. The overall discontinuity response calculationcomputes the signal received by the probe as the summation of in-dividual contributions for the various resulting wave modes. Thecomputation mode was two-dimensional — the side-drilled holewas meshed along its profile vertically perpendicular to the soundbeam projection. Figure 3 shows the two-dimensional profile of theskin-to-flange test sample drawn in the CAD facility, the subse-quent three-dimensional solid model extrusion and the addition ofdiscontinuity, probe and ultrasonic setup parameters.

The Kirchhoff approximation (Schmerr, 1998) was used in thediscontinuity response calculations. The Kirchhoff approximationassumes that the elastic wave is entirely scattered by the disconti-nuity. The discontinuity is meshed and the wave scattered by thediscontinuity is the product of amplitude and time dependent func-tions of incident wave and complex scattering coefficient at eachmeshed location. The Kirchhoff approximation relies on the as-sumption that each point at the discontinuity surface contributes asif it were part of an infinite plane (no interaction with neighboring

Figure 2—Ultrasonic phased array probe positioned on the externalside of the skin of the skin-to-flange test part.


points). The Kirchhoff approximation is assumed to give accurateresults for discontinuities providing specular or near specular re-flection over planar or volumetric discontinuities. Quantitativeerror is expected to increase when the scattered direction movesaway from the specular direction.

The Kirchhoff approximation is a high frequency approxima-tion, valid when

(1)

(the discontinuity diameter is significantly greater than the wave-length). In this case, the weld steel has a shear wave velocity around3.28 mm/µs, and at a frequency of 5 MHz, the wavelength is ap-proximately 0.65 mm (using wavelength = velocity/frequency).Then,

(2)

(with a side-drilled hole diameter about three times larger than theultrasonic wavelength), which satisfies the Kirchhoff criterion.

EXPERIMENTAL RESULTSThe probe was ultimately positioned so that the front of the probe

wedge was about 5 mm away laterally from the discontinuity. Figure4b shows the experimental sector scan results obtained at the probeposition shown in Figure 4a. The large amplitude indication shownin Figure 4b at the lower right portion of the sector image is from thedirect path reflection off the side-drilled hole. The center of this indi-cation, where the highest amplitude occurs, is located at the 66°angle. Moving the probe laterally away from the side-drilled hole(which is 40 mm in length and extends about two-fifths of the wayacross the weld) results in the disappearance of the indication. Theultrasonic flaw detector shows a display that indicates that the side-drilled hole is located about 4 mm from the front of the wedge and is12.67 mm in depth, which agreed well with the actual location.

2 15 5πλ

a = .

2 1πλ

a?


Figure 3— Simulation steps: (a) two-dimensional profile of the skin-to-flange test sample is drawn in the CAD facility; (b) the profile isextruded to create a three-dimensional solid model; (c) the material, setupand discontinuity parameters are defined to create the test simulation;(d) test simulation. Beam computation can proceed after extrusion.Discontinuity response can proceed after the discontinuity is defined andinserted into the structure.

(c)

(d)

(b)

(a)

Figure 4—Ultimate probe positioning: (a) probe location; (b) 40 to 75°sector scan results.

(a)

(b)



Other major indications include one just to the right and oneto the lower right. These correspond to a creeping wave reflec-tion (Aldrin and Knopp, 2008) and indirect reflection(s) that trav-el to the fillet, then to the side-drilled hole and back to the probeor to the hole, then to the fillet and back to the probe. The directand indirect ray paths are indicated in Figure 5. Creeping wavesare not taken into account by the Kirchhoff approximation. Othersmaller amplitude indications are likely due to skip reflectionsoff the various surfaces.

ULTRASONIC MODELING RESULTSThe model probe was positioned so that the front of the probe

wedge was about 5 mm away laterally from the discontinuity in thetest model to match with the final experimental setup. Figure 6shows the beam profile ultrasonic modeling results for several

angles in the sectorial scan. These profiles indicate the ray paths,taking into account mode conversion and backwall reflection overthe angular sector scan. Examining the profiles, one can see the var-ious reflections in addition to the direct mode path.

Figure 7 shows the discontinuity response ultrasonic modelingresults. In the simulation, the direct and indirect reflections (the re-flections that travel to the fillet, then to the side-drilled hole andback to the probe or to the hole, then to the fillet and back to theprobe) are superimposed due to the positioning of the hole with re-spect to the fillet. By changing the positional relationship betweenthe side-drilled hole and the fillet in other model iterations, a sepa-ration between direct and indirect reflections could be observed.The actual experiment revealed greater separation between direct

Figure 5— Ray paths for direct reflection from side-drilled hole (black)and interactions with fillet and side-drilled hole (red).

Figure 6— Beam profile ultrasonic modeling results at several angles in the sectorial scan: (a) shot 11; (b) shot 21; (c) shot 31.

(a) (b)

(c)

Figure 7—Ultrasonic modeling results for phased array ultrasonictesting of skin-to-flange test sample.


and indirect reflections (Figure 4). If the simulation is done onlywith direct path reflections selected, the large indication will showless structure at its lower portion, which provides evidence for theexistence of the indirect reflection when the half-skip option is se-lected as a computation parameter. The 60° angle (shot 21) gave thehighest reflection response amplitude related to the direct reflectionfrom the side-drilled hole.

Figure 8 shows the ultrasonic results overlaid onto the model ofthe skin-to-flange test sample at a model probe position. Figure 9shows a direct comparison of the experimental and modeling re-sults for the phased array ultrasonic test of the skin-to-flange testsample with a side-drilled hole in the interior weld. The qualitativediscontinuity response for the experiment is in excellent agreementwith that predicted from the model, except that the angle where thestrongest reflection response occurs differs in location by 6°. It islikely that geometrical differences between the drawn model andactual test part were responsible for the difference in angle ofstrongest response. Additionally, the creeping wave indication can-not be modeled in the software. Looking carefully, other smalleramplitude indications are also present in the simulation andexperiment that correspond to each other. These indications appearprior in time to the large indication and are probably related to skip

reflections off the various surfaces and back to the transducer. Theseless major indications are also demarcated in Figures 4 and 7, whichshow magnified views of the experimental and simulation results,respectively.

DETECTABILITY LIMIT ESTIMATIONAlthough modeling tools can provide the measurement re-

sponse for varying the discontinuity size, several factors must beconsidered in evaluating the true detection capability of a nonde-structive testing technique. In this simulation, the direct reflectionwaveform (A-scan) off the side-drilled hole was obtained at 60°(shot 21), which was the angle predicted to provide a maximum re-sponse in the sector scan for the phased array testing simulations.Figure 10 shows the simulation results as the side-drilled hole di-ameter is reduced from 1.5 to 0.19 mm using identical simulationand computation parameters. The reflected amplitude of theA-scan is predicted to be reduced; this trend is shown clearly in Fig-ure 10b. The finite element mesh size used in the discontinuity re-sponse calculation will affect the predicted response.

To quantitatively evaluate the detection limits of a nondestruc-tive technique for a given test, a signal to noise analysis is required.Noise in nondestructive testing measurements can be attributed tomany sources: the measurement hardware, surface conditions andcouplant, material related noise, and superimposed signals reflect-ed from the part geometry. Often, experimental studies are per-formed to get a handle on the background noise in the measure-ment device for a known calibration sample. Calibration runs canalso be simulated in the software, and signal levels of the experi-mental and simulated results can be matched. Then, the noise mea-surement from the experimental data can be used to set a thresholdof detection in the simulation results for small discontinuities. Inaddition to experimental data, noise models, including grain noise,can be simulated in the software. Additionally, attention must bepaid to all of the model assumptions like the Kirchhoff high fre-quency limit when evaluating the nondestructive testing measure-ment response (Achenbach, 1973; Achenbach, 1992; Rose 1999). Fu-ture work is proposed to explore the accuracy of the modelpredictions with experimental data over the range of simulatedside-drilled hole sizes and test the detection capability using bothexperimental noise data and noise models run in the software.


Figure 9—Direct comparison of model results versus experimental results.

Figure 8—Ultrasonic modeling results overlaid onto model of skin-to-flange test sample.


CONCLUSIONThis paper describes simulation/modeling and its comparison

to experimental results for phased array ultrasonic testing of a steelskin-to-flange test sample with a side-drilled hole in the interiorweld. Good qualitative agreement was observed between experi-mental results and modeling predictions with regards to ultrasonicreflection response from the side-drilled hole in the weld. To quan-titatively test the detection limits of a nondestructive techniquefor a given test, a signal to noise analysis and attention to model

assumptions are required. Future applications of the simulationtechniques will be applied for important structures and testing sce-narios at NASA.

REFERENCESAchenbach, J.D., Wave Propagation in Elastic Solids, Amsterdam, North Hol-

land Publishing, 1973.Achenbach, J.D., “Measurement Models for Quantitative Ultrasonics,” Jour-

nal of Sound and Vibration, Vol. 159, 1992, pp. 385–401.Aldrin, J.C. and J.S. Knopp, “Modeling and Simulation for Nondestructive

Testing with Applications to Aerospace Structures,” Materials Evaluation,Vol. 66, 2008, pp. 53–59.

Chaffai-Gargouri, S., S. Chatillon, S. Mahaut and L. Le Ber, “Simulation andData Processing for Ultrasonic Phased-Arrays Applications,” Review ofProgress in Quantitative Nondestructive Evaluation, Vol. 26A, D.O Thomp-son and D.E. Chimenti, eds., Melville, New York, AIP, 2007, pp. 799–805.

Cinquin, M., L. Le Ber, S. Lonne and S. Mahaut, “Results of 2006 UT Model-ing Benchmark Obtained with CIVA at CEA: Beam Modeling and FlawSignal Prediction,” Review of Progress in Quantitative Nondestructive Evalu-ation, Vol. 26B, D.O Thompson and D.E. Chimenti, eds., Melville, NewYork, AIP, 2007, pp. 1870–1877.

Darmon, M., P. Calmon and C. Bele, “Modeling of the Ultrasonic Responseof Inclusions in Steels,” Review of Progress in Quantitative NondestructiveEvaluation, Vol. 22A, D.O Thompson and D.E. Chimenti, eds., Melville,New York, AIP, 2003, pp. 101–108.

Gengembre, N., “Pencil Method for Ultrasonic Beam Computation,”Proceedings of the 5th World Congress on Ultrasonics, Paris, available at<www.sfa.asso.fr/wcu2003/procs/cd1/index.html>, 2003, pp. 1533–1536.

Lonne, S., L. de Roumilly, L. Le Ber, S. Mahaut and G. Cattiaux, “Experi-mental Validation of CIVA Ultrasonic Simulations,” International Con-ference on NDE in Nuclear Industry (ICNDE), San Diego, 10–12 May,2006.

Mahaut, S., S. Chatillon, E. Kerbrat, J. Porré, P. Calmon and O. Roy, “NewFeatures for Phased Array Techniques Inspections: Simulation and Exper-iments,” 16th World Conference on Nondestructive Testing, Montréal,Canada, 30 August – 3 September, 2004.

Raillon R. and I. Lecoeur-Taïbi, “Transient Elastodynamic Model for BeamDefect Interaction: Application to Nondestructive Testing,” Ultrasonics,Vol. 38, 2000, pp. 527–530.

Rose, J.L., Ultrasonic Waves in Solid Media, Cambridge, Cambridge Universi-ty Press, 1999.

Roth, D.J., J.D. Kiser, S.M. Swickard, S. Szatmary and D. Kerwin, “Quantita-tive Mapping of Pore Fraction Variations in Silicon Nitride Using an Ul-trasonic Contact Scan Technique,” Research in Nondestructive Evaluation,Vol. 6, 1995, pp. 125–168.

Schmerr, L.W., Fundamentals of Ultrasonic Nondestructive Evaluation — AModeling Approach, New York, Plenum Press, 1998.


Figure 10— Simulation results as the side-drilled hole is reduced from1.5 to 0.19 mm using identical simulation and computationparameters: (a) A-scan versus discontinuity diameter for directreflection off side-drilled hole in flange to skin weld (from simulation);(b) A-scan peak amplitude versus side-drilled hole diameter obtainedfrom A-scans in (a).

(a)

(b)


Documents

Ultrasonic Phased Array Testing of Welded Components at NASA