qnde2004-benchmark

Bourdeau

Informations

• Publié par: Shuel
• Nombre de lectures: 90
• Langue: English

Extrait

1

RESULTS OF THE 2004 UT MODELING BENCHMARK OBTAINED WITH THE CIVA SOFTWARE DEVELOPED AT THE CEA O. Diligent, S. Chatillon, M. Darmon, A. Lhémery, and S. Mahaut Commissariat à l’Énergie Atomique, LIST, CEA-Saclay, bât. 611, 91191 Gif-sur-Yvette cedex, France ABSTRACT. The CIVA software developed at the French Atomic Energy Commission for processing and simulating NDT data (ultrasonics, eddy-current) includes tools for simulating the whole inspection of a component in which virtual defects are positioned. In this paper, the Kirchhoff’s approximation is applied and the reciprocity principle allows to predict the reception by transducers of waves scattered by defects. Simulated results are compared to experimental data for the various problems of the 2004 UT benchmark session. INTRODUCTION

This paper describes an approach and results of predicting a wave field interaction with certain types of flaws of a set of ultrasonic benchmark problems proposed by the World Federation of Non Destructive Evaluation Centers [1]. The Center of NDE at Iowa State university has conducted a series of experimental studies to allow direct comparison and validation with ultrasonic models in some simple configurations. The results obtained from the simulation may be compared in a later time with calculations performed in different universities. Benchmark problems have been running for a few years at the QNDE conference (see for example [2] and [3]). This year’s problem is a natural evolution of the past conferences. Here, the response of a side-drilled hole, a flat-bottomed hole or a spherical pore when insonified by normal or oblique incidence, compressional or shear waves probes is studied. All ultrasonic computations presented in this paper were performed using the CIVA software [4], which has been developed at the French Atomic Commission (CEA) for the last 10 years and has been presented at several QNDE conferences since 1994 (see for example [4-6]). Two main aspects are developed in the CIVA software. The first aspect is the wave field prediction. CIVA calculates wave fields generated by arbitrary transducers (standard or phased arrays [7]) in immersion or in contact. It also takes into account the mode conversion at the interface and with or without after a rebound on the back wall of the tested specimen. The specimen can be either a homogeneous or a multi-layered medium, isotropic or anisotropic and of arbitrary geometry [8].

Transducer

Field radiated at M ?

M

Pencil

Arbitrary transducer: discretized as source points

2

FIGURE 1. Field prediction sketch. In order to calculate the wave field radiated at point M (see Fig. 1), the transducer is discretized as a series of source points over the transducer’s surface. For each point source, elementary contributions are obtained through a pencil method (high frequency approximation). These contributions are calculated such as: follow: firstly, a pencil-matrix formulation in order to predict wave-front radii of curvature along the wave path as well as the amplitude and the time of flight, and secondly, a plane wave transmission/reflection coefficient at all interfaces is associated to each pencil. The impulse responses are then synthesized from the contributions and convoluted with the input signal. The second main aspect developed in CIVA is the simulation of the interaction of a wave field with flaws. Different configurations can be modeled, starting with either pulse echo or tandem methods, different types of flaws can also be used (side drilled holes, flat bottom holes, planar defect or spherical inclusions). Fundamentally the model is based ona priorifield computation for radiation (and reception) whenever it is possible. This field is simplified to the form: Φ(M,t)=A(M)ϕ(t−t of flight (M)) (1) where�the scalar potential (for P-waves) and is A(M)the parametric distributions. In CIVA, the model of the wave field interaction with the defect can be either based on the Kirchhoff approximation (pulse-echo) or on the Geometrical Theory of Diffraction (TOFD). In this paper, the Kirchhoff model will be used and the use of reprocity principles leads to the only integration over the defect surface.  This paper will be in four parts. After defining the problem, comparison between experimental data and calculations performed by CIVA will be discussed for different types of flaws, such as side-drilled holes, flat bottom holes and a spherical pore. Finally a conclusion will be drawn. PROBLEM DEFINITION

The different transducers and flaws modeled in CIVA for the UT benchmark are presented as well as the properties of the different blocks in which the flaws are situated. Transducers Definitions

Two types of transducers are used for the predictions. The first is a circular, unfocussed transducer of 12.7 mm in diameter and of 5 MHz center frequency. The second is a circular spherically focused transducer of 12.7 mm diameter, 172.9 mm focal length and of 5 MHz center frequency. An example of a wave field calculated in CIVA for a focused transducer generating a P-wave at 30 degrees is shown in Fig. 2a. In both cases, the waveform used is (Fig. 2b.):   2πf 3 costcos 2f t,for0t  1−    ( π ) ≤ ≤  (2) v(t)= 3      0otherwise 

(a)

(b)

3

Position of defect FIGURE 2. (a) incident wave field calculated for a focused probe, P-wave 30 degrees and (b) incident waveform In all testing configurations, the transducer is immersed in a water medium (density 3 -3 2 = 1 g/cm , wave speed = 1484 mm/sec, att. = 0.248x10f); keeping a path of 50.8 mm (2 inches). Also, the wave fronts are assumed to be spherical for planar probes and planar for focused probes when calculating the wave-fields interactions with flaws. Flaws Definitions

In term of defects, the benchmark problem is separated into three different parts which will be presented and discussed in the next section. The first part deals with the modeling of the interaction of a wave field generated by each of the transducers described above with two side-drilled holes (SDH), which diameters are 1mm and 4mm and which are placed in an aluminum block. The material properties for aluminum are presented in Tab.1. Both of the centers of the SDHs have the same depth, 25.4mm (1 inch), into the material. Normal incidence testing is performed on both SDHs, and then the transducers are rotated in order to generate refraction angles of 30, 45, 60, 70, and 75 degrees for both P- and SV-waves and for each type of transducer (planar of focused). The second part presents the study of the interaction of the wave field with either a flat-bottomed hole of diameter n3, n5 or n8 (where n is 1/64 inch) and placed at 25.4 mm into the material or a spherical pore of diameter 692 µm, and placed at 19.6 mm into the material. The flat-bottomed holes are placed in a steel block and the spherical pore is placed in a fused quartz block. Their properties are shown in Tab. 1. For these two studies, only normal incidence testing was performed using the planar and the focused transducer. It is important to note that for all calculations run in this paper, only the attenuation of water was taken into account. TABLE 1.Material properties for the materials used. P- wave speed of block (mm/µs): 6.416 Aluminum SV- wave speed of block (mm/µs): 3.163 3 2.75 Density of block (g/cm ): P- wave speed of block (mm/µs): 5.94 Steel 1018 SV- wave speed of block (mm/µs): 3.23 3 7.86 Density of block (g/cm ): P- wave speed of block (mm/µs): 5.9694 Fused quartz SV- wave speed of block (mm/µs): 3.7741 3 2.2 Density of block (g/cm ):

(b)

(a)

SDH –Ø1 mm

SDH –Ø4 mm

(c)

4

75.2 time [µs] 80.1 75.2 time [µs] 80.1 FIGURE 3.Example of a calculation using CIVA. Incident wave characterization: P-wave; Focused probe; refracted angle of 30 degrees. (a) B-scan of the 2 SDHs (white low amplitude, black high); (b) A-scan corresponding to the 1 mm in diameter SDH shown on (a); (c) A-scan corresponding to the 4 mm in diameter SDH shown on (a). Note that the amplitude scales of the two a-scans are identical. OVERVIEW OF THE RESULTS AND DISCUSSION

In this section, comparisons between the results calculated using the CIVA software and the experimental data given by the Iowa State University is presented and discussed. Response from a Side-Drilled Hole, Comparison between CIVA Calculations and Experimental Data.

The first part of the benchmark is to model the reflection from a series of side- drilled holes when two types of transducers (planar or focused) excite a wave field at different refraction angles. Hence, the interaction between 1 mm and 4 mm in diameter SDH and the incident field is calculated at normal incidence, 30, 45, 60 and 70 degrees for the P-wave and at 30, 45, 60, 70 and 75 degrees for the Sv-wave. Fig. 2 shows an example of a reflection from the two side drilled holes. Fig. 2a shows the B-scan (horizontal axis displacement, vertical axis time) when the incident wave field is generated by a focused probe and rotated so that the refraction angle is equal to 30 degrees. Fig 2b and Fig. 2c are the A-scans measured on Fig. 2a where there is a maximum in reflection. It has been found that for some refracted angles, the maximum does not exactly occur where the theorical beam hits the side-drilled hole; nevertheless, this maximum peak-to-peak response was measured and used in the following comparisons. Peak-to-peak response values were calculated from all A-scans and processed as follows. All amplitudes were normalized to the corresponding normal incidence response. For example, for the case of the 1mm in diameter SDH when the incident wave field is generated by a planar probe, all other amplitudes (P30, …, P70; Sv30, …, Sv75) were then normalized to the one from P0. The results are displayed in Fig. 3. Fig. 3a shows the results for the 1mm SDH/planar probe; Fig. 3b for the 1mm SDH/focused probe; Fig. 3c for the 4mm SDH/Focused probe; and Fig. 3d for the 4mm SDH/focused probe. It can be seen from the figure that there is excellent agreement for the P-waves cases between the

5

0

5

10

15

20 0

5

0

5

10

15

a

10

c

20 30 40 50 60 refracted angle

70

80

5

0

5

10

15

20

25 0

5

0

5

10

15

20

b

10

d

20 30 40 50 60 refracted angle

70

80

5

20 25 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 refracted angle refracted angle FIGURE 4. Amplitudes for all SDH cases run normalized to the normal incidence reflection. (a) 1 mm SDH normalized reflections when incident field is generated by a planar probe; (b) 1 mm SDH normalized reflections when incident field is generated by a focused probe; (c) 4 mm SDH normalized reflections when incident field is generated by a planar probe; (d) 4 mm SDH normalized reflections when incident field is generated by a focused probe. Solid lines are P-waves and dashed lines are Sv-waves. predictions made by CIVA and the experimental data (average difference of 1 dB). On the other hand, the Sv-wave predictions, although showing good agreement in the reflection behavior (maximum at the same angle for example), show a discrepancy much larger in amplitudes (average of 3 dB). There are still some investigations running at the CEA at the moment to understand this phenomenon but the first results show that the software overestimates the Sv-wave in relation to the P-wave.. Response from a Flat Bottom Hole and a Spherical Pore, Comparison between CIVA Calculations and Experimental Data.

The second part of the benchmark is to predict the relative peak-to-peak voltage responses for all the other normal incidence FBH and SPH responses and normalize them to the reference value (normal incidence planar probe 4 mm SDH, focused or planar). These were then compared to the experimentally observed ratios. This should test the models across different flaw types. Fig. 4a shows the B-scan (white low amplitude, black high amplitude) of the interaction between a normal incidence wave field generated by a planar probe on three flat-bottomed holes. Their diameters are 3/64, 5/64, and 8/64 inch. Fig. 4b and 4c show the A-scan for the 3/64 in diameter FBH and 8/64 in diameter FBH, respectively. They are measured at their maximum amplitude on the B-scan. Fig. 5a shows the B-scan (white low amplitude, black high amplitude) of the interaction between a normal incidence wave field generated by a planar probe on a spherical pore. The diameter if the pore is 692 µm. Fig. 5b shows the corresponding A-scan measured at the maximum amplitude on the B-scan.

6

Fig. 6 shows the comparison between the predictions calculated using the CIVA software and the experimental data. The results are processed as follows: each peak to peak response from the A-scans (Fig. 4b for example) is normalized by the 4 mm in diameter SDH normal incidence peak to peak response (planar or focused probe). Fig 6a shows the results when the incident wave field is generated by a planar probe and Fig. 6b when a focused probe generates the incident wave field. Excellent agreement is found between the predictions and the experimental data as an average error of less than 1 dB is observed.

(b)

(a)

FBH – n3

FBH – n5

(c)

FBH – n8

72.9 time [µs] 82.9 72.9 time [µs] 82.9 FIGURE 5.Example of a calculation using CIVA. incident wave characterization: P-wave; planar probe; normal incidence. (a) B-scan of the 3 FBHs (white low amplitude, black high); (b) A scan corresponding to the n3 in diameter FBH shown on (a); (c) A scan corresponding to the n8 in diameter FBH shown on (a). Note that the scales of the two a-scans are identical and that n corresponds to 1/64 inch. (a) (b)

70.3 time [µs] 80.3 FIGURE 6.Example of a calculation using CIVA. Incident wave characterization: P-wave; planar probe; normal incidence. (a) B-scan of the spherical pore (white low amplitude, black high); (b) A scan

7

5 5 (a) (b) 0 0 5 5 10 10 15 15 20 20 25 30 25 n3 n5 n8 SPH n3 n5 n8 SPH type of flaw type of flaw . FIGURE 7. Comparison between CIVA predictions and experimental measurements, all peak-to-peak responses normalized to the reflection of the 4mm diameter SDH for (a) planar probe and (b) focused probe. Magnitude of the Incident Wave Field Velocity

Finally, It is asked the for each of the twelve 4 mm SDH cases and at a fixed frequency (the center frequency of the transducer) to plot the magnitude of the incident wave field velocity,V z , versus the x- and y-directions (perpendicular to the refracted ray which is along the z-axis - see Fig. 7 below) where the origin is located at the center of the flaw. It is assumed that the wave field is generated by a piston transducer whose velocity, V 0 , is unity on the transducer face, as shown in Fig. 7. The purpose of this study is to compare directly the model-based predictions (from different modeling approaches) of these incident wave fields at the flaw, which, of course, are a key aspect of modeling the flaw responses. This may help to understand where modeling differences, if any, may occur. Fig. 8 shows the predictions of the velocity fieldV z  at the center of a 4mm in diameter SDH for the P-waves (Fig. 8a) and the Sv-waves (Fig. 8b) at different refracted angles. The dashed lines represent the focused probe and the solid lines represent the planar probe.

V 0 =1

X

FIGURE 8.Definition of coordinates for incident beam model. (a)

75°

60°

Z

(b)

75°

60°

0° 30° 45° 0° 30° 45° FIGURE 9.amplitudes plotted as a polar graph. (a) P-waves; (b) Sv-waves. Solid lines: planar Velocity probe; dashed lines: focused probe.

CONCLUSIONS

8

Comparison between predictions undertaken at the CEA using the CIVA software and experimental data given by the Center of NDE, Iowa State University have been presented in this paper. This comparison was achieved within the QNDE UT benchmark with other universities. Several simple testing configurations were modeled, involving pulse-echo set-ups where the reflector (side-drilled holes, flat-bottomed holes, or spherical pore) is interrogated with either a planar or a spherically focused transducer. The comparison showed excellent agreement for all P-waves cases with an average of 1 dB difference between predictions and experimental data. Also a very good agreement was achieved for the effect of the defect geometry (FBH and SPH versus SDH). On the other hand, average agreement was found for Sv-waves predictions. The behaviors of the reflected waves are good but there are gaps in amplitude (averaging 3 dB). First investigations undertaken at the CIVA software overestimates the Sv-wave amplitudes compared to the P-wave amplitudes. REFERENCES

1. 2.

3.

4.

5.

6. 7.

8.

2004 Ultrasonic benchmarks, http://www.wfndec.org/benchmarkproblemscurrent.htm Thompson, R. B.,Review of Progress in Quantitative Nondestructive Evaluation, Vol. 21, eds.D.O. Thompson and D.E. Chimenti, AIP Conference Proceedings, Melville, 2002, p.1917. Thompson, R. B.,Review of Progress in Quantitative Nondestructive Evaluation, Vol. 22,op. cit.(2003), p. 1769. Benoist, P., Besnard, R., Bayon, G. and Boutaine, J. L.,Review of Progress in QNDE, Vol. 14, eds. D. O. Thompson and D. E. Chimenti, Plenum Press, New York, 1995, p. 2553. El Amrani, M., Calmon, P., Roy, O., Royer, D. and Casula O.,Review of Progress in QNDE, Vol. 14,op. cit.(1995), p. 1075. th Calmon,P.,inproceedings of the 5 World Congress on Ultrasonics, 2003, p. 443. Mahaut, S., Chatillon, S., Raillon, R. and Calmon, P.,Review of Progress in QNDE Vol. 23,op. cit.(2004), p. 777. Gengembre, N. and Lhémery, A.,Review of Progress in QNDE,Vol. 19,op. cit. (2000), p. 977.