2002 ULTRASONIC BENCHMARK PROBLEM: OVERVIEW AND DISCUSSION OF RESULTS R. Bruce Thompson Center for Nondestructive Evaluation Departments of Materials Science and Engineering & Aerospace Engineering and Engineering Mechanics Iowa State University Ames, IA USA 50011 ABSTRACT. The predictions of four measurement models are compared for the 2002 Ultrasonic Benchmark Problem. The problem involves the pulse-echo responses of spherical and cylindrical cavities. The ultrasonic waves are considered to be generated and detected by either a planar or spherically focused probe, each of finite diameter and positioned to produce normally incident or refracted waves (longitudinal or shear) of the desired angle. Among the results are a new expression for the response of the cylindrical cavity and a quantitative comparison of the various models. Noteworthy is the differences in the predictions of the beam models, for refracted angles near critical angles and for focused probes. INTRODUCTION In the previous Review of Progress in Quantitative Nondestructive Evaluation, solutions to an initial ultrasonic benchmark problem were presented by four sets of investigators, one from Korea, one from Germany and two from the United States [1-4]. That problem involved the prediction of the pulse-echo response of spherical cavities and circular cracks placed below planar, liquid-solid interface. The measurements were assumed to be made with both a planar or a spherically focused probe, illuminating the surface at normal incidence. In solving this problem, it was necessary to use an ultrasonic beam model (to determine the fields illuminating the flaw) and a scattering model (to determine the response of the flaw) with a reciprocity relation used to predict the observed signal. In the course of investigation, four beam models and five scattering models were examined. A discussion of the philosophy of the benchmark comparison and an overview of these initial results was presented by the author [5]. Among the conclusions of that paper was the observation that the models were in good general agreement with one another with a few cases being observed where further work was needed to clarify differences that were observed. A second benchmark problem was considered by the same groups during the current Review of Progress in Quantitative Nondestructive Evaluation. This problem CP657, Review of Quantitative Nondestructive Evaluation Vol. 22, ed. by D. O. Thompson and D. E. Chimenti © 2003 American Institute of Physics 0-7354-0117-9/03/$20.00 1769 introduced two new features. The list of defects considered was extended to consider cylindrical cavities, or side-drilled holes. In addition, the case of oblique incidence was considered. An overview of the results obtained by the four groups [6-9] follows. PROBLEM DEFINITION The problem is to predict the signals observed, in a pulse-echo measurement, from spherical and cylindrical cavities in a solid. Both normally and obliquely incident waves are considered in an immersion geometry, as illustrated by Figure 1 in the papers of both Schmerr and Sedov [8] and Gray [9]. The waves were excited and detected by both planar and spherically focused probes. The material is an aluminum block with a planar surface: (density = 2.71 gm/cc, longitudinal wave speed = 6374 m/sec, shear wave speed = 3111 m/sec). The block is placed in a water bath (density = 1 gm/cc, wave speed = 1470 m/sec). Two sets of six scattering objects are considered. Each member of the first set is a spherical cavity, centered at a distance of 25.4 mm (1 inch) into the material. They should have respective diameters of 0.125 mm, 0.25 mm 0.5 mm, 1 mm, 2 mm and 4mm. Each member of the second set is a cylindrical cavity. They should have the same depths and diameters as the pores. The diameters have been selected to range from the regime in which the defect is small with respect to the wavelength to the case in which it is large with respect to the spot size of the focused beam. Two transducers are considered. The first is a circular, unfocussed transducer of 12.7 mm (1A inch) diameter and of 5 MHz center frequency: The transducer is assumed to act as a piston source, with all points on its surface moving with equal amplitude. The second is a circular, focused transducer of 25.4 mm (1 inch) diameter, of 5 MHz center frequency: The transducer is assumed to have a curved element with a radius of curvature of 152.4 mm (6 inches) and also act as a piston source. (The radius of curvature is equal to the geometrical focal length, i.e. the point of maximum amplitude in the absence of diffraction.) The ability of the driving electronics and transducer to generate and detect ultrasonic waves is defined by a reference signal, which is taken to be the signal observed during the normal incidence reflection from the front surface of the aluminum block. This contains information about the bandwidth of the ultrasonic signal and its absolute level, as influenced by transducer efficiency and driving/receiving electronics. The proper interpretation of the reference waveform requires a specification of the distance of the transducer to the part surface. For the unfocused probe, the transducer to frontsurface distance is taken to be 50.8 mm (2 inches). For the focused probe, the transducer to front-surface distance is taken to be 152.4 mm (6 inches). This is equivalent to placing the geometrical focal plane on the part surface. In either case, the reference waveform is taken to have the following, timedomain waveform: 1-cos fl/f), forQ<t< —— otherwise 1770 Here t is in microseconds and f is the center frequency of the transducer, in MHz. Given the specification of this reference waveform in a particular reference experiment, models are used to predict signals from the various scattering objects and experiments. The specific goals are twofold. • Predict the amplitude of the insonifying velocity field at the location of the flaw center, normalized by that of the transducer face, for the 5 MHz center frequency of the transducer. • Predict the time domain response (i.e. the radio frequency waveform as a function of time) of the signal scattered from each combination of scattering objects, transducers and angles. The transducers should be positioned such that they excite longitudinal waves propagating normal to the surface, and refracted waves, both longitudinal and shear, propagating at angles of 30°, 45°, 60°, 75° with respect to the surface normal. For all calculations, it is to be assumed that the transducer is positioned such that the water path along the central ray is 50.8 mm (2 inches) before that ray enters the solid. At normal incidence, this should place the flaw in the far field of the radiation of the unfocused probe and near the focal point for the focused probe. OVERVIEW OF RESULTS The four groups adopted approaches that differed in the way that both the radiation from the transducer to into the material and the scattering from the flaw were treated. Table 1 summarizes the approaches. References to the individual beam radiation and flaw scattering techniques may be found in the citations in the individual papers. In each of the contributions, the electromagnetic reciprocity relationship of Auld [10] was used to predict the flaw response based on the beam radiation and crack scattering models. Detailed results may be found by consulting the individual manuscripts. Here, some comparisons will be made. Note that there are multiple problems being considered, pulse-echo scattering from spherical and cylindrical cavities as observed with focused (planar) and unfocused probes, with both normal and refracted incidence (longitudinal and shear). For each of these cases, insonifying field and flaw response predictions have been made for various combinations of beam and scattering models. This benchmark problem allows these models to be compared to one another. In addition, it contains several important insights into the differences in the scattering from spherical and cylindrical cavities and of the effects of incident angle on the predictions of various beam models. Response of a Cylindrical Cavity The cylindrical cavity, a model for the side-drilled hole commonly used as a calibration reflector, was not considered in the previous benchmark problem and introduces new scattering issues. Because of the axially symmetric geometry of this reflector, the scattering issues are the same for all angles of incidence. Schmerr and Sedov [8] make a number of assumptions that greatly facilitate the modeling of the response of this reflector and provide insight into the underlying wave physics. Starting with Auld's electromechanical reciprocity relation [10] and describing the incident wave 1771 TABLE 1. Comparison of the modeling approaches of the four groups. Cylindrical Cavity Model Beam Pore Radiation Scattering K PSS Spies 7 Spherical Cavity Model Pore Beam Scattering Radiation K PSS Song et al 6 MG Kl MG K3 RS Kl RS K3 8 MG Kl MG K2,K3 9 GH sov GH K3 Author Schmerr and Sedov Gray Ref Beam Model Legend: PSS-Point Source Superposition RS-Rayleigh-Summerfeld Integral (high frequency limit) MG-Multi-Gaussian (paraxial approximation) GH-Gaussian-Hermite (paraxial approximation) Scattering Model Legend: K-Kirchhoff Approximation Kl-Kirchhoff Approximation for front surface response of sphere in quasi-plane wave limit K2-Kirchoff Approximation for front surface response of cylinder in quasi-plane wave limit K3-Kirchoff Approximation for front surface response of cylinder when variations in velocity over curved surface neglected. SOV-Separation of Variables fields in a quasi-plane wave approximation, they develop a series of approximate measurement models in a systematic way. The first, appropriate to the spherical cavity, reduces to the Thompson-Gray measurement model [11]. The second and third develop the analogous results for the case of the cylinder, with the essential issue being how to treat the fact that the amplitudes of the insonifying fields vary along the length of the cylinder. Two forms of the Kirchhoff approximation are derived and shown to give very similar results for the cases studied. In applying these measurement models, the authors show that the response of the cylindrical cavity is always greater than that of a spherical cavity of the same radius, a result that is not surprising. They also note that, when compared to the response of a crack of the same radius, the cylindrical cavity produces a stronger pulse echo response for smaller radii, but a weaker response for greater radii. Effects of Oblique Incidence on Ultrasonic Radiation Pattern Four different forms of beam models were used in the solution of the benchmark problems. The paraxial approximation greatly speeds the calculation of the beam pattern, and various forms of this model were used in the papers of Song et al [6], Schmerr and Sedov [8], and Gray [9]. One way of describing this approximation is to note that the radiation of a transducer can be described by an angular spectrum of plane waves. In a 1772 rigorous description of the propagation of a beam through an interface, each plane wave component must be multiplied by the appropriate interface transmission coefficient. The beam propagating away from the interface is still described by an angular spectrum of plane waves, by the modification of their relative amplitudes and phases by the interface transmission coefficients leads to new structures in the beam pattern. Within the paraxial approximation, one essentially makes two assumptions. The transmission coefficient is taken to be the same for all plane wave components. In addition, small angle (with respect to the central ray) approximations are used to relate the angle of refraction of the plane wave components to the angle of the incident plane wave component. These approximations are considered to be good when the range of angles of propagation within the incident beam is not too large and when one is not too close to a critical angle, where the interface transmission coefficient can change rapidly with angle. A consequence of the paraxial approximation is that, within the plane of refraction, the beam profile will be symmetric about the central ray. Song et al [6] and Schmerr and Sedov [8] use what is know as the Multi-Gaussian (MG) beam model whereas Gray [9] uses the Gauss-Hermite (GH) beam model. Each makes use of the paraxial approximation. They differ in that they decompose the beam into different sets of elementary solutions to the wave equation. More accurate, but computationally slower, beam models were also considered. Song et al [6] utilize a generalized Rayleigh-Sommerfeld (RS) model with high frequency approximations to treat the propagation through the interface. Spies uses a closely related point-source-synthesis method (PSS). Song et al [6] present a systematic study of the differences between the MG and RS models. Table 1 of their paper presents a comparison of the insonifying velocities at the center of the flaw, as predicted by these two models. The predictions of the two models agree very well for most of the cases studied. The two exceptions are for shear waves when the angle of shear-wave refraction is 30°, close to the longitudinal wave critical angle, for both planar and focused probe. Figure 1 of that paper shows a comparison of the beam patterns predicted for the two beam models for the case of the planar probe and the 30° refracted shear-wave angle. They also noted differences in the beam pattern for the planar transducer at the 75° refracted shear-wave angle. Although the velocity insonifying the flaw is similar for the two cases, there are noticeable differences in the radiation fields off of the axis. The authors note, however, that neither model may be accurate in this limit because of the rapid variation of the refracted angle with the angle of incidence. They also note that, in predicting the time domain waveforms after introducing a scattering model and using Auld's electromechanical reciprocity relation, they obtained very good agreement on the shape of the waveforms. Based on the point-source-synthesis technique, Spies makes another interesting observation [7]. He examines plots of the radiation pattern for various incident angles. For the case of planar probes, the plots shown for the case of 45° refracted longitudinal and shear waves indicate that the beam is relatively symmetric in the refracted plane and that the center of the beam is close to the direction of the central refracted ray. However, this result is not achieved for the case of focused probes. Considering again the case of 45° refracted longitudinal and shear waves, he observes significant asymmetry. For example, he finds that, when the intended refracted wave is a longitudinal wave at 45°, the actual fields are centered as an angle closer to 35°. Thus, to achieve the desired refracted angle, it would be necessary to make a correction in the nominal refracted angle. 1773 Prediction of Ultrasonic Signals The predicted behavior of the ultrasonic signals follows closely the expectations based on the behavior of the illuminating fields discussed above. Regarding the angular dependence of the signal, three of the investigators [6,8,9] predicted that the longitudinal wave signals will decrease monotonically as the refracted angle increases for all flaw sizes. Likewise, they predict that, as the refracted angle increases, the shear wave signals will increase from a value of zero at 0° to a peak at 45°, followed by a decrease at higher angles. This result is quantified in Figure 1 of [8] and Figures 3-6 of [9], as well as in Tables 2 and 3 of [6] and Tables 3-6 of [9]. The strength of the peak at 30° depends on which beam model is used, consistent with the discussion above. Since Spies [7] compares his signals to those of a reference reflector at the same angle, his results are not in a form that can be compared to the above predictions. He does, however, note an interesting point. Application of the Kirchhoff theory in the point source superposition technique leads to deviations from the expected amplitude behavior when ka is equal to 5 or less, where a is the radius of the scatter. He notes that this behavior is consistent with results reported by Gray and Thompson in the study of the previous benchmark problem [4]. This is because, as the flaw becomes small with respect to the wavelength, the Kirchhoff approximation, that each point on the flaw scatters as if it were on a planar interface, breaks down. Both Song et al [6] and Gray [9] provide detailed tables of the numerical values of the predicted signals, as was noted in the first paragraph of this sub-section. Based on a casual comparison of these numbers, the general trends follow one another quite closely. Differences between the outputs of the two models are generally on the order of 20% or less (1.6 dB). The reasons for these differences have not been explored. Relationship to More Approximate Engineering Models Under additional assumptions (transducer and reflector in the mutual far-field and reflector smaller than the beam width in at least one dimension), approximate expressions for the amplitude dynamic curves measured in a contact, pulse-echo mode have been derived by Werneyer et al. [12]. Spies [7] compares the predictions of his more exact calculations to those engineering approximations for a number of cases. CONCLUSIONS A number of interesting points were revealed by the study of the 2002 Benchmark Ultrasonic Problem. These can be summarized as follows. • New approximations were derived for the response of the cylindrical cavity (sidedrilled hole). • Several beam models were compared. A direct numerical comparison showed the predictions of the paraxial beam models to be in good agreement with those of the generalized Rayleigh-Sommerfeld model in the high frequency limit. Differences were observed when the refracted wave was near a critical angle. The Point Source Superposition method predicted, for the case of focused probes, that the actual angle of the beam could deviate from the intended angle (as would be computed by applying SnelFs Law to the central ray of the beam). Further work 1774 would appear to be desirable to explore the various cases where the predictions of the models do not agree. • Numerical comparisons were made for a number of models. For the case of the models of Gray [8] and Song et al [6], systematic differences on the order of 20% (1.6 dB) were observed. Since this is on the order or less than the error in typical industrial NDE measurements, this difference is of little practical significance. However, it should be explored to improve our fundamental understanding of the models and their behavior. Future benchmark studies, which will involve experimental validation, are planned. ACKNOWLEDGEMENTS This work was supported by the National Science Foundation, Industry/University Cooperative Research Program at the Center for Nondestructive Evaluation, Iowa State University. REFERENCES 1. S.-J. Song, H.-J. Kim, and C.-H. Kim, "Prediction of Flaw Signals of the Ultrasonic Benchmark Problem by Sungkyunkwan University", in Review of Progress in Quantitative Nondestructive Evaluation 21B, D. O. Thompson and D. E. Chimenti, Eds. (AIP, New York, 2002), pp. 1941-1948. 2. M. Spies, "Simulating a Standard Type Problem of Ultrasonic Testing-A Contribution to the Comparison of Models", in Review of Progress in Quantitative Nondestructive Evaluation 21B, D. O. Thompson and D. E. Chimenti, Eds. (AIP, New York, 2002), pp. 1949-1955. 3. L. Schmerr, "Ultrasonic Modeling of Benchmark Problems", in Review of Progress in Quantitative Nondestructive Evaluation 21B, D. O. Thompson and D. E. Chimenti, Eds. (AIP, New York, 2002), pp. 1933-1940. 4. T. A. Gray and R. B. Thompson, "Solution of an Ultrasonic Benchmark Problem Within the Paraxial Approximation", in Review of Progress in Quantitative Nondestructive Evaluation 21B, D. O. Thompson and D. E. Chimenti, Eds. (AIP, New York, 2002), pp. 1925-1932. 5. R. B. Thompson. "An Ultrasonic Benchmark Problem: Overview and Discussion of Results", in Review of Progress in Quantitative Nondestructive Evaluation 21B, D. O. Thompson and D. E. Chimenti, Eds. (AIP, New York, 2002), pp. 1917-1924. 6. S.-J. Song, J. S. Park, and H.-J. Kim, "Prediction of the Insonifying Fields and Flaw Signals of the 2002 Ultrasonic Benchmark Problem", these proceedings 7. M. Spies, "Prediction of the Transient Flaw Signals of the Ultrasonic Benchmark Problem", these proceedings 8. L. W. Schmerr and A. Sedov, "Modeling Ultrasonic Problems for the 2002 Benchmark Session", these proceedings 9. T. A. Gray, "Ultrasonic Benchmark Problem: Application of a Paraxial Model to Side-Drilled Holes and Oblique Incidence", these proceedings. 10. B. A. Auld, "General Electromechanical Reciprocity Relations Applied to the Calculations of Elastic Wave Scattering Coefficients," Wave Motion 1, 3 (1979). 11. R. B. Thompson and T. A. Gray, "A Model Relating Ultrasonic Scattering Measurements Through Liquid-Solid Interfaces to Unbounded Medium Scattering Amplitudes", /. Acoust. Soc. Am, 74, 1279-1290 (1983). 12. R. Werneyer, F. Walte, and W. Muller., Technical Report No. 78052-TW (in German), IZFP Saarbrucken, Germany (1978). 1775
© Copyright 2025 Paperzz