ULTRASONIC PROPAGATION IN AUSTENITIC STAINLESS STEEL WELDS -

APPROXIMATE MODEL AND NUMERICAL METHODS RESULTS AND COMPARISON

WITH EXPERIMENTS

B . Chassignole ’ , D. Villard ’ and G. Nguyen Van Chi * ELECTRICITE DE FRANCE ’ R and D Division, Material Study Branch Les Renardieres, MORET SUR LOING 77818 France 2 R and D Division, Mechanics and Numerical Modeling Department CLAMART 92 14 1 France

N. Gengembre and A. Lhemery CEA / CEREM Bat.61 1, CEA SACLAY 91191 Gif-sur-Yvette cedex, France

INTRODUCTION

Ultrasonic testing of austenitic stainless steel welds is often difficult because the ultrasonic beam is subjected to many perturbations : deviation, partition, distortions and sometimes ghost echoes. These perturbations are due to the anisotropic and heterogeneous metallurgical structure of the welds. We are carrying out a research program in order to evaluate the influence of the main parameters of the structure on the perturbations of the beam.

This program largely leans on modeling which can bring a theoretical justification for ex- perimental results and allows to reduce the number of required tests. We present some modeling results obtained on various types of welds with two different computer codes : ULTSON, a 2D finite element code, and CHAMP-SONS, a 3D semi-analytical one. The modeling results are com- pared with experiments.

CHARACTERIZATION OF AUSTENITIC WELDS FOR MODELING INPUT DATA

Metallurgical Structure of an Austenitic Weld

This study was performed on mockups representative of 316L industrial austenitic stainless steel welds. They were manufactured using a manual welding process with coated electrodes (weld thickness 40 mm). We developed a method of characterization of the metallurgical structure of austenitic welds, which would allow to understand the perturbations of the beam and would be compatible with the needs of modeling codes. We deliberately choose a description at a macro- scopic scale in order to simplify the modeling process [ 11.

0 a @o Figure 1: Metallurgical structure of 316L austenitic weld (type 1). (a) Macrographic structure of type 1 weld in a plane perpendicular to the welding direction. (b) Delimitation of the homogeneous areas and mean direction of the columnar structure ($ angle values) in each area.

A cross section of type 1 weld (Fig.1) perpendicular to the welding direction, shows a co- lumnar structure, formed of long shape solidification dendrites. The size and the direction of the solidification dendrites fundamentally depend on the thermal conditions existing in the melted metal during solidification. These thermal conditions themselves mainly depend on the welding process and the weld’s geometry.

In the weld presented, we can observe that the solidification dendrites grew up through several welded layers keeping approximately the same direction (epitaxic structure). So they may keep nearly the same direction in rather large areas. A metallurgical examination in a section par- allel to the welding direction shows that the columnar structure is not tilted in the welding direc- tion.

We carried out a study by X-rays diffraction which showed that the columnar grain axis is parallel to a <lOO> cristallographic axis. It also showed that the areas in which the morphological texture keeps nearly the same direction, can be considered as homogeneous with an orthotropic symmetry.

These observations lead to the conclusion that austenitic welds can be considered as a set of anisotropic homogeneous areas. We can delimit these homogeneous areas by visual examination or, better, using a image processing software.

Ultrasonic Characterization of the Type 1 Weld

The computer codes need, as input of material characteristics, the stiffness tensor for each homogeneous area of the weld. We determined the elastic constants in one homogeneous area and for the others, we assumed that the tensor can be obtained from the first one only by a rotation representative of the variation of the columnar grain axis between the areas.

The elastic constants were evaluated by an optimization process using the values of com- pression and shear waves velocities measured for various incident angles in a transmission mode, in thin small plates taken off from homogeneous areas of the weld [ 11.

154

, Transducer

4h-m 32 mm

/ X

/f,

- Ultrasonic testing in pulse-echo mode

Y - Contact transducer, diam. 13 mm - Compression waves, 2.25 MHz - Normal incidence - Surface of the weld : machined

Side drill hole Diam. 1.5 mm - Acquisition step distance : 1 rnm

Figure 2 : Ultrasonic testing of a type A weld in pulse-echo mode with normal incidence

MODELING OF ULTRASONICS IN AN AUSTENITIC WELD WITH ULTSON (2D CODE)

Experimental Results

We present an example illustrating the complexity of ultrasonic testing of an austenitic weld. The test was performed on a type 1 weld mockup containing a side drilled hole, in a pulse- echo mode and normal incidence (Fig. 2).

On the Bscan views of the results (Fig. 3), we can see the discontinuities of the back-wall echoes produced by the complex shape of the opposite surface of the mockup (last machining, first pass penetration). The interaction of the ultrasonic beam with the side drill hole leads to two addi- tional echoes, obtained for two different locations of the transducer Pl and P2, whereas only one echo was expected. The ultrasonic times of flight of these echoes are rather similar and seem to correspond to two different reflections of the compression waves on the defect.

Modeling with the Code ULTSON 2D

We implemented the modeling of the above ultrasonic testing case with ULTSON 2D to understand the origin of these two surprising echoes. ULTSON is a 2D finite-element code, devel- oped by EDF, which can describe the propagation of ultrasonic waves, radiated by a contact or a focused transducer, in an anisotropic heterogeneous material, either in transmission or in reflection mode. It can simulate all the configurations of an ultrasonic examination (instantaneous wavefronts and full beam in solid areas, interaction with flaws, echoes received by the transducer) in a 2D geometry. The transducer is simulated by a thin piezoelectric plate located inside a Plexiglas block coupled with the testing surface with a thin water layer.

Side drilled hole echoes

1 I

0 a (b)

Figure 3 : Ultrasonic examination of the type 1 weld mockup. (a) Experimental B-scan. (b) Simu- lated B-scan.

155

Table I Elastic constants for type 1 austenitic weld (GPa)

~

We assume that there is a linear relation between the electric signal and the normal vibra- tion velocity on the surface of the transducer. The pressure at the interface between water and solid can then be calculated starting from this velocity and using a time-dependant Kirchhoff’s integral, which is valid for high frequencies. The pressure after return is calculated by the same way, knowing the calculated displacement field in the mockup [2].

The weld is described as a set of seven homogeneous anisotropic areas. Moreover, since we deal with the case of waves polarized and propagating in a symmetry plane of an orthotropic medium (the columnar structure is not tilted in the plane perpendicular to the incident plane), the resolution of the general equation of motion only requires four elastic constants : C,,, C??, C,, and - - C,, (table I).

We calculated the echoes received for forty different locations of the transducer along a di- rection perpendicular to the weld. The calculated B-scans views (Fig. 3) confirm there are two different echoes for only one defect. Moreover the echoe structure is very similar to the experi- mental one. The visualization of the beam paths inside the weld, corresponding to these echoes, shows clearly that the beam is progressively curved and finally strongly deviated by the heteroge- neous anisotropic structure (Fig. 4).

$+fi;: :. ;,Jr$ ‘3 f’: ,~ $22 ““’ 6 ” ” t

f: :;;:$ ‘: Q %.& / - ~..xpxr9rr~ Q$

” ypgd g ?$ “’ t

“L..,,, e .:: : ;,; g .t ,.... .‘. ..:,, : * (i p 8 ,. \ / ,..’ :..:. .:: t ” y :: / f i : ::. ::: $:. T$ 1 .:: ‘;: > f., y; .’ :.,.: .,.F .: f ‘,. f i. 2 :., ‘. ,. ., 4 : :: ;:. :: ‘:.

:; ,.$ 4 y ::y ‘$ f .: + : ‘, Z ::.:, “2 9 ( f A.’ , : ‘.. . . . . .:: ::. :.‘.:, “: 2 : f.’ .4. ‘5 ; . . ,,,,, ,.,.,...... ‘.V.’ ‘3 ..:. Z

B “, f . . ..:. ‘.. ::‘: :: .‘“< ” .,..’

:,,.; :,;.: j .:..:...:,:..: * * 8 +, ..$ ,................... ::‘I:.::?; .> .A. f ::,,,. .,:., f., 0: :...: fl$ .:.:.; .I ,.,: .y ‘..,.. “$ .; p: Q : p ,3 f.3 :;: :$ ii .‘X B ,,,) ,:, p< ,: 8 .:,y$ $,y: : ,.,::. ,” r., . . . ..I : . . ‘3 .,. ;s; j .‘Q, :*pj ; 2 $$,.g<.& ., ,. .,.y .,., .,< )‘$.:.. 8 I p ::/ .:,: $$ xy >y,: v, y, ;: .,., c’ .,. , ., ., n

0 C w Figure 4 : Modeling results - examination of type 1 weld mockup - image processing description. (a) Beam path inside the weld for position Pl. (b) Beam path inside the weld for position P2. (c) Echodynamic curves. (d) Instantaneous amplitude of the displacements.

156

Table II. Positions of the transducer source in y direction (hole position = 0) and time of flight values

Pl cm@ p2 b-w T1 w T2 w Experimental results -9 1 11.5 12.1

Modeling results -9 5 12.0 13.0

The visualization of snapshots of the ultrasonic field inside the weld shows the distortion of the wavefronts and allows to understand the origin of the various observed and predicted ech- oes.

We observe a very good agreement between calculated and experimental positions of the transducer source giving an optimal interaction with the hole, especially for the position Pl (table II). For the position P2 there is a little difference. It can be assumed that the metallurgical structure of the mockup used for the ultrasonic tests could be a little different from the mockup used for the structural characterization. On the other hand, we can note that the experimental time-of-flight is greater for position P2 than for Pl. The simulated results confirm and allow to understand this observation. Indeed, for position P2, the beam mainly propagates in a domain where the velocity is slow because of the grain orientation (propagation quasi-parallel to the grain orientation).

We note, on experimental results, that the amplitude of the echo for position P2 is higher than for position Pl but it is the opposite for simulated results (Fig. 4). The theoretical studies show that attenuation of the beam, due to back scattering at grain joints, strongly depends on the direction of the beam with respect to the columnar structure direction [3]. In its actual state of de- velopment, ULTSON 2D cannot take into account this attenuation.

We can see on Fig. 5 the results of another modeling implementation of the above ultra- sonic situation obtained with a simple and symmetrical description of the weld (five homogeneous domains with vertical grains in the middle).We can observe that this description predict only one echo generated by the side drilled hole and no beam skewing. This result outlines the necessity to provide a description of the weld as realistic as possible to get accurate modeling results.

Pressure

20 30 40 50 Transducer position (mm> (Side drilled hole position = 35 mm>

Figure 5: Modeling results - Examination of type 1 weld mockup - Symmetric description. a) Weld description. b) Bscan view. c) Echodynamic curve

157

MODELING OF ULTRASONICS IN AN AUSTENITIC WELD WITH CHAMP-SONS (3D CODE)

Modeling of a 2D Ultrasonic Testing Situation

The software Champ-Sons has been developed at the French Atomic Energy Comrnission (CEA), and recently extended to deal with heterogeneous and anisotropic media. The main princi- ple is to discretize the emitting surface into elementary point sources, and to evaluate approxi- mately each elementary contribution, by means of the pencil method, applied to elastodynamic fields computation. An overview of the principles of this method is given in [4].

The first application concerns the above described type 1 weld structure for which the crystallographic system is only disoriented in the incident plane. Here, only the field radiated into the weld is considered. The constant given in table I are used, and in addition for the third dimen- sion, the constants Cl1 = 245 GPa, Cl2 = 110 GPa, Cl3 = 145 GPa, C55 = 110 GPa, and C66 = 75 GPa.

The field predicted by Champ-Sons for the 13rnm - diameter LO transducer at the position Pl is plotted on Fig. 6. Only quasi-longitudinal waves, as transmitted waves, have been calculated. We assume that mode conversion in quasi-shear waves doesn’t lead to significant contributions.

We observe that the beam skewing in the incident plane, previously experimentally meas- ured and predicted by the numerical model Ultson, is also recovered by Champ-Sons.

Modeling of a 3D Ultrasonic Testing Situation

We performed other ultrasonic tests on another 316L specimen taken of a welding mold (type 2 weld). The main difference with type 1 weld is that the width of this welding mold is much larger. The solidification conditions of this mold results in a metallurgical structure greatly differ- ent (Fig. 7). The cross section, perpendicular to the welding direction, reveals a much more homo- geneous columnar structure, especially apart the chamfers (@ = 3”). In the plane Parallel to the welding direction we clearly observe a layback of the structure (<p = 9”).

Figure 6: Quasi-longitudinal wave field into type 1 Champ-Sons.

weld for transducer at position Pl, predicted bY

158

Columnar grain axis ’

,,@

i

:. ‘... \

. ‘.. ‘... Y ‘..

@=3”

Columnar grain axis ’

Fcp

ii

‘... X \

(()= 90 \

1 cm 1 cm

0 a @I Figure 7: Macrographic structure of 316L austenitic welding mold specimen (type 2). (a) plane transverse to the welding direction. (b) plane parallel to the welding direction.

One could verify, by a crystallographic study and image processing, that this weld could be considered as a quasi homogeneous and transversely isotropic medium. (Cl1 = 233 GPa, Cl2 = 100 @a, CM = 139 GPa, C?q - _ = 194GPaandCa= 106 GPa). In this case the incident plane is not a symmetry plane so we may expect that the ultrasonic beam will be deviated in the welding direc- tion. This situation must be simulated by a 3D code.

This configuration has been implemented in Champ-Sons and compared with an experi- ment performed in transmission mode with a LO0 transducer, 20 rnrn - diameter, 2.25 MHz. The position of the maximum amplitude on the simulated beam is in an excellent agreement with that of the experimental beam (fig. 8). We particularly observe beam skewing both in x and y directions. Moreover, a comparison between experimental and calculated beam shapes confirms that the model well predicts the distortion effect. However, the experimental beam is a bit larger than the simulated one. Two main reasons can explain these differences: first, the 6mm -diameter aperture of the receiver in the experimental set-up, and second, small local variations in the grain orienta- tions, observed in the metallurgic study.

max- -2.5dIh .): <*. ;: :: VW’

CONCLUSIONS

The presented results demonstrate that ultrasonic beam perturbations (beam skewing and dis- tortion effects) in industrial austenitic welds can be accurately simulated on condition that a conven- ient description of the weld is provided. We propose a description at a macroscopic scale which lies on the division of the weld in a set of homogeneous domains in which we are able to evaluate the texture orientation and the elastic constants.

The fact that this common description was used in two different modeling approaches and that in both cases results are in good agreement with experiment, leads us to conclude that we obtain a intrinsic description of the weld structure as far as ultrasonic wave propagation is concerned. We showed more particularly that, in the studied cases, unaccurate values of grain orientations give rise to a false prediction of experimental echoes.

This description can be implemented in a 2D finite element code such as Ultson 2D which is able to simulate pulse-echo ultrasonic testing situations when there is no skewing out of the incident plane.

The description can also be implemented in a 3D semi-analytical code (Champ-Sons) which is able to simulate the ultrasonic field in 3D situations with skewing effect out of the incident plane, in case of simplified description.

Future work will deal with determining the influences of the weld description (number and shape of domains) and the elastic constants values on beam propagation and validating the two codes on a large number of NDT cases (different welds and NDT configurations and different defects).

REFERENCES

1. B. Chassignole, D. Villard, M. Dubuget, J.-C. Baboux and R. El Guerjouma, in Review of Progress in Quantitative Nondestructive Evaluation, Vol. 19, eds. D. 0. Thompson and D. E. Chimenti (Plenum, New York, 2000), (present volume).

2. B.Nouailhas., G. Van Chi Nguyen, F. Pons and S. Vermersch, J. NDE, 9, 145 (1990).

3. S. Ahmed and R.B. Thompson, in Review of Progress in Quantitative Nondestructive Evaluation, Vol. 14, op. cit. (1995), p. 1617.

4. N. Gengembre and A. Lhemery, in Review of Progress in Quantitative Nondestructive Evaluation, Vol. 19, op. cit. (2000>, present volume.

160