外文原文-奥氏体不锈钢焊缝各向异性的研究

B. Chassignole et al. / NDT&E International 43 (2010) 273-2829Contents lists available at ScienceDirectELSEVIERNDT&E Internationaljournal homepage: /locate/ndteint Ultrasonic and structural characterization of anisotropic austenitic stainless steel welds: Towards a higher reliability in ultrasonic non-destructive testingB. Chassignolea, R. El Guerjoumab,*, M.-A. Ploixc, T. Fouquetda EDF R&D, Materials and Component Mechanics Department, Moret sur Loing 77818, France b Universite du Maine and CNRS, LAUM UMR CNRS 6613, Le Mans 72000, France c INSA de Lyon and CNRS, MATEIS UMR CNRS 5510, Villeurbanne 69621, France d EDF R&D, Sinetics Department, Clamart 92141, Franceabstractarticle infoArticle history:Received 22 September 2006 Received in revised form 28 November 2009 Accepted 7 December 2009 Available online 29 December 2009Keywords:WeldUltrasonicNon-destructive testing Nuclear engineering Anisotropy Modeling AttenuationThe non-destructive testing of austenitic stainless steel welds of the primary coolant piping system is a significant problem for the nuclear industry. Ultrasonic techniques would be very helpful to detect, locate and size potential defects. Unfortunately, austenitic welds are coarse-grained, heterogeneous and anisotropic. This leads to aberration and scattering of the ultrasonic waves. In this paper, we present several experimental results of ultrasonic testing of two austenitic welds exhibiting high anisotropy. In order to explain the observed display of wave propagation phenomena such as beam deviation, we use finite element modeling. The modeling is associated with a complete characterization of the inspected welds. Two essential characteristics of the welds are determined: the average elastic constants of the weld and the grain orientations. The capability of the model is illustrated in different testing configurations. This work associating structural characterization and modeling shows that a better understanding of the phenomena of ultrasonic propagation should allow the interpretation and reliability of the industrial inspections of heterogeneous anisotropic welds to be improved. 2009 Elsevier Ltd. All rights reserved.0963-8695/$-see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.ndteint.2009.12.0051. IntroductionTo assure the structural integrity of nuclear pressurized water reactors, the non-destructive testing (NDT) techniques must allow the detection of potential defects in the austenitic stainless steel pipes and welds of the primary coolant piping system. In particular, manufacturing flaws (inclusions, voids or lacks of fusion) or in-service flaws (cracks) could be present in the welded joints.In French nuclear power plants, the in-service inspections of austenitic welds are carried out mainly on the basis of radiographic techniques. Ultrasonic inspections are rarely used because they are confronted with all the problems of testing coarsegrained, heterogeneous, anisotropic material 1-3. However, ultrasonic techniques would help to complement the information provided by radiography which efficiently detects defects but has some limitations as to their location and size.A weld made of austenitic stainless steel in some cases presents strong anisotropy varying continuously from one area to another all over the weld giving rise to heterogeneous anisotropy 1-3. The effect of anisotropy on the propagation ofn Corresponding author.E-mail address: rachid.elguerjoumauniv-lemans.fr (R. El Guerjouma).ultrasound makes the velocity direction-dependent and makes the group velocity different from phase velocity in both amplitude and direction 4.Furthermore, deviation of the ultrasonic beam within the weld material may cause difficulties in locating defects. Scattering of the ultrasonic energy results in some directions of propagation having a poor signal-to-noise ratio. To understand the anisotropy of the wave amplitudes recorded in ultrasonic inspection experiments, the complexity of wave propagation in these media has to be considered.To study these phenomena, modeling is very helpful. Indeed, modeling can reveal the quantitative features and thus help in the optimization of conventional ultrasonic NDT techniques and in the development of new ones. The simulation of ultrasonic testing using appropriate models allows us to perform, for example, parametric studies and obtain quantitative simulated results. The theory of wave propagation into anisotropic and homogeneous media already allows the prediction of beam skewing and divergence effects 4-7. When considering the more complicated case of heterogeneous anisotropic structures, modeling studies require realistic descriptions from the mechanical and metallurgical point of view of the various kinds of weld structures, especially the ones where anisotropy is strong.Over the last few years many studies have been undertaken to evaluate the effect of metallurgical structures of austeniticstainless steel welds on wave propagation for the application of ultrasonic non-destructive testing 8-10. However, to be efficient in comparing experimental results and modeling, the weld characterization should take into account both the metallurgical and mechanical aspects and this is rarely done.In this paper, we present several experimental ultrasonic testing results of two austenitic welds. In order to explain the wave propagation phenomena observed in the case of strong anisotropy, we use finite element modeling. This modeling is associated with a complete characterization of the inspected welds.In the first part, we describe the characteristics of the mock- ups inspected. In the second part, we present results of weld structural characterization by metallographic and crystallographic analyses. In the third part, we determine the anisotropic elastic properties of the welds using an ultrasonic method. Then, in the last part, we compare experimental and modeling results in pulse-echo mode for different testing configurations (different probes and defect localization).2. Mock-up characteristics and material characterizationThe extensive columnar grain structure in multipass austenitic welds differs greatly from that in ferritic welds. In austenitic welds, the deposition of successive weld beads does not destroy the grain structures in the previous beads. Moreover, the graingrowth is parallel to the heat flow direction and is governed by an epitaxic process 11. The columnar grains then grow through the boundaries of the beads. Consequently, grains of substantial length and with specific crystallographic orientations are produced. The long dendrite axis is almost vertical along the center of the weld and nearly perpendicular to the fusion lines (and the upper boundary of the weld). The present study used two flat position welding molds in AISI 316L steel (austenitic stainless steel), and the welds were manufactured with an automatic arc welding process that used coated electrodes. The difference between those two welds comes from the electrode scanning. For weld B, the weaving rate is equal to 130 oscillations/min which lead to a run width of about 20 mm. For weld A, the width of the string beads is equal to 10 mm. For both welds the overlap is equal to 35%. The dimensions of the mock-ups are indicated on Fig. 1.Metallographic observations in cross sections are presented in Figs. 2 and 3, respectively, for weld A and weld B. For each weld, the first observation was made in the plane transverse to the welding direction (yz) plane) and the second one in a plane parallel to the welding direction (xz) plane). Those pictures clearly reveal a dendritic type growth parallel to the heat flow direction. In the case of a sufficiently low welding energy, this leads to elongated and oriented grains which can grow by epitaxic process on several millimeters length.An X-ray diffraction analysis was performed on samples of 3 mm thickness taken in a relatively homogeneous area of each weld (central area). The pole figure of 2 0 0 planes for weld A is given in Fig. 4. It clearly reveals the strong anisotropy of the material. In the first approach, for the sake of simplicity, the material could be considered as transversely isotropic with a symmetry axis parallel to a 1 0 0 axis. Such material symmetry was chosen by several authors to describe an austenitic stainless steel weld 10,12. However, the analysis of the orientation distribution functions (ODF) shows that in a strict way, the material must be considered as orthotropic.Fig. 2. Macrographic structure of weld A. (a) Plane transverse to the welding direction. (b) Plane parallel to the welding direction.bThe orientations of the symmetry axis, determined from the X-ray diffraction analysis and defined by the a and g angles, are given in Figs. 2 and 3. We confirm that the symmetry axis is very close to the elongation axis of the grain 11.Fig. 3. Macrographic structure of weld B. (a) Plane transverse to the welding direction. (b) Plane parallel to the welding direction.For weld A, the columnar structure in the middle zone is rather homogeneous with a tilt of the elongation axis of the grain (a angle) of about 20 in the transverse plane. Near the chamfers of the weld, the structure becomes more heterogeneous. In the (xz) plane, the grains are almost vertical (g=3). The tilt in this plane is directly linked to the welding velocity. For up vertical welds, which are manufactured with a lower welding velocity, a tilt of about 20 was found 1.As far as weld B is concerned, the tilt is weaker in the transverse plane (about 3) because of the large electrode scanning which leads to flatter weld runs (weaving rate is equalto 130 oscillations/min which lead to a run width of about 20 mm). This scanning also leads to a more developed grain growth and a more homogeneous structure than for weld A. In the (xz) plane we clearly observe a layback of the structure (g = 9). This layback is once again due to the flat shape of the runs.A multi-scale analysis (neutron diffraction, electron back- scattering diffraction and transmission electronic microscopy) on a weld metal deposit of316L stainless steel is proposed by Bouche et al. 13. The authors also found a fiber-like texture with reinforcements (orthotropic symmetry). Their weld is an intermediate case with both grain tilts in the (yz) plane (20) and in the (xz) plane (10).3. Determination of the elastic properties by ultrasonic methodsThe velocities and the polarizations of plane waves propagating in a specific direction in a perfectly elastic anisotropic solid can be determined from the Christoffel equations 14:|Gil-rV2dil 卜 0(1)with ru = Cijklnjnk (i, j, k, l=1, 2, 3) and where p is the density, n is the unit vector in the wave propagation direction, V is the phase velocity of ultrasonic waves in the medium, Cijkl are the elastic constants of the anisotropic medium and d,l is the Kronecker symbol. With appropriate index changes, the elastic constants can be expressed with the more simple matrix form Cmn.Christoffel equations admit three eigenvalues corresponding to the phase velocities V of the three wave modes which can propagate. The associated eigenvectors are the polarizations of each wave mode.Knowing the elastic constants of the material, it is possible to calculate any velocity for a given direction of propagation by solving Eq. (1). The inverse problem consists in recovering elastic constants from suitable experimental values of measured velocities.Such measurements are performed on a set-up in transmission mode 15: a homogeneous sample is immersed in water and theTable 1Calculated elastic constants (GPa) and a and g angles (0) from ultrasonic measurements.WeldCnC22C33C44C55C66C23C13C12agA2342402209911095146148118154B237247210122125701341328427s/UOAloolsos/UOAMoolaFig. 6. Variations of velocity versus the angle of propagation in the (23) plane (a) longitudinal waves (b) transverse waves.uo o o o o o o o c uoooooooonVJ86420864C- r33333222cvelocities are obtained in all the accessible planes of propagation by varying the angle of incidence.In practice, the unknown material properties are determined simultaneously by minimizing the sum of the squares of deviations between the experimental and calculated velocities 15. For an orthotropic medium, when the material symmetry coordinate system coincides with the geometrical axes of the sample, the unknowns are the nine independent elastic constants. In our study, the material symmetry axes are unknown. Then the grain disorientation with respect to the geometrical sample coordinate system results in three additional unknowns, the Euler angles (j1,f, j2) which are also determined by the optimization process. The sample geometrical axes and the material symmetry axes are presented in Fig. 5. We can also see in this figure some of the planes where ultrasonic velocities are measured.The values of the nine elastic constants determined by this method for our welds are presented in Table 1. These elastic constants are determined with a very good accuracy when the ultrasonic velocities are themselves measured with a good accuracy as it is the case for these materials 15. The 1, 2 and 3 axes are linked to the material symmetry, the 3 axis corresponding to the fiber axis. If we compare the phase velocities calculated in the (23) plane (Fig. 6), we can note that weld B presents a more pronounced anisotropy (higher variations of velocities). This is coherent with the conclusions of Section 1. Intermediate values were found for a sample taken in a weld representative of a butt weld of the surge line of nuclear power plants.In the table the values of a and g angles (orientation of the fiber axis) linked to the values of the Euler angles are also indicated. The results are close to those determined by the X-ray diffraction analysis.4. Ultrasonic inspections in pulse-echo mode4.1. Configurations of inspectionTwo side-drilled holes with a diameter of 2 mm were machined in the mock-ups (Fig. 7). Their depths are 20 and 40 mm. The same defects were also machined in a reference mock-up in wrought austenitic stainless steel which presents an isotropic structure.The mock-ups were then inspected in pulse-echo mode with four contact testing probes whose characteristics are given in Table 2. They radiate by refraction through a Plexiglas wedge. Three of them produce longitudinal waves and the last one produces transverse waves with vertical polarization. The characteristics of these probes (probe size, center frequency, andbandwidth) are very close to those used for in-service inspections. In particular, the center frequency of the probes is close to 2 MHz, which is the value recommended for the inspection of this type of weld. Moreover, the transmission pulse indication is highly damped to get good performances in terms of echo resolution. The transverse waves were only used to inspect weld B.Automatic and parallel immersion scans were performed with signal registrations every 0.5 mm in the y direction (Fig. 6). This allows perfect coupling conditions between the contact probes and the mock-up to be.Table 2Probes characteristics.Wave typeAngle of refraction in austenitic steel (0)Transducer size (mm)Frequency (MHz)Longitudinal00132.25Longitudinal45|202.25Longitudinal600202.25Transverse4515n152In this paper, we are going to study more particularly the beam skewings and attenuations in the plane of incidence (yz) plane). We will compare the experimental results to the modeling ones to demonstrate the relevance of the modeling approach to reproduce the ultrasonic phenomena.Heterogeneity of most actual welds is known to be the origin of poor signal-to-noise ratios that greatly complicates the NDT. We have considered this crucial problem in a detailed study 17. The two welds considered here for experiments are not representative of all real welds as their large volume leads to a more homogeneous anisotropic structure. For such welds the signal-to-noise ratio is rather high in several directions of propagation and the problems of ultrasonic testing are mostly related to the anisotropy of the material, resulting in beam splitting and skewing and which makes the attenuation direction- dependent. In this study we focus on such problems and we consider specific welds which exhibit strong anisotropy. Those phenomena are not clearly illustrated on welds that exhibit very strong heterogeneity with poor signal-to-noise ratios 18.4.2. Modeling codeFor this study, we used a two-dimension code called ATHENA developed by EDF R&D and the French National Institute for Research in Computer Science and Control. This code solves the equations of elastodynamics expressed with the stresses and the velocities of the displacements by a finite element method. The convergence of the code was demonstrated in previous work. The elements are squares with a size equal to a fifteenth of the wave length which is known to be sufficient for an accurate solution.ATHENA simulates the propagation of the ultrasounds in an anisotropic and heterogeneous complex medium, with the assumption that such a medium can be described by a finite number of anisotropic and homogeneous areas.In this study, 2D computations are helpful because we limit the analysis to the ultrasonic propagation in the plane transverse to the welding direction. X-ray diffraction analysis has shown that this plane can be assumed as a plane of symmetry of the orthotropic material. So the beam skewing out of this plane will be weak for the majority of the configurations. For a weak grain la