# Simulation of a TIG Weld using Finite Element Method: Part 2 – Structural Analysis

1

1. Introduction Welding is very frequently used as a manufacturing process in engineering. The fact that welding leads to the generation of residual stresses is well known. Residual stresses are defined by Cullity [1] as auto-equilibrating stresses that remain in the material in absence of external loads (which
(Recebido em 12/12/2005; Texto Final em 07/04/2006). Artigo baseado em versão apresentada no COBEM 2005, Ouro Preto, 06 a 11 de Novembro 2005. Publicado com permissão da ABCM (Associação Brasileira de Engenharia e Ciências Mecânicas).
63

include thermal gradients). In the specific case of welding, these stresses are the result of non-uniform plastic strains caused by the elevated temperature gradients and phase transformations that occur during the process. Generally, these stresses are undesirable due to the possibility of prejudicing the quality of welded components [2]. Nevertheless, the minimization or elimination of these stresses is a difficult task and a good practice is to consider them inherently in the design. This way, the characterization of welding residual stresses has been the object of several studies, which use either experimental or numerical methods. It is widely known that experimental methods present critiSoldagem Insp., Vol. 11, No. 2, Abr/Jun 2006

plastic strains will be developed when the following condition is satisfied: σv M −σ y = 0 (5) It should be noted that the equivalent stress can never exceed the material yield limit since. 2. plastic strains can be defined as time-independent irreversible strains. typical drawbacks are that. Thus. Scotti. there is still a lack of information on the capability of these models to characterize the real distributions of welding residual stresses. many simplifications were adopted. which has attracted the attention of the scientific community to a greater extent. with 3D complete models. It is also aimed at characterizing the deformed shape of the plate and the transverse displacements to be compared to experimental results extracted from the literature. it is possible to solve a 3D complete problem. εtherm is the thermal strain and εp is the plastic strain. Nonetheless. The primary mechanism responsible for transformation plasticity is the volume change. 64 . Still in 2D models. A. plastic strains would develop instantaneously. No. the stress measurements must be performed at one point at a time. it is necessary to define the elastic domain. In this context. C.. temperature dependence of the material properties. within the range limited by the yield strength of the material (recalling that these properties depend on temperature). Initially. Moreover. is the solid phase transformation. This is because. More detailed simulations. according to Lemaitre and Chaboche [11]. as well as to use 2D simplified models. and so on) and difficulties related to the numerical computations themselves (computational cost and convergence problems). The modeling of plasticity is very complex.9. In the scope of numerical methods. assuming temperature-independent material properties (except the yield stress) and the determination of the thermal distribution through the simplified analytical formulation of Rosenthal [4]. increase the strain. 2. The total strain can be expressed as: ε Total = ε e + ε therm + ε p (1) where εe is the elastic strain. This strain is irreversible because transformation back to the original phase does not reverse the strain. in many cases. cal limitations. this work seeks to contribute to the resolution of welding structural problems by evaluating the application of a Finite Element Method (FEM) and 3D solid modeling in order to obtain the residual stresses generated by a TIG weld without filler metal on a stainless steel AISI 316L plate. Depradeux [10] points out another feature of the structural analysis. This is due to different factors. the number of degrees of freedom. several researchers have studied the resolution of structural welding problem. included the effect of solid phase transformations that occur during the process. the elements have different characteristics. when de-coupled from the thermal problem. which occurs during the phase transSoldagem Insp. On the other hand. which was based on the finite difference method. the resolution of the structural problem. consists of a static equilibrium analysis for each instant of time considered. despite it not being a requirement. in particular.10]. The elastic component of strain can be determined directly by the Hooke’s law. Vol. D. using a 2D finite element model. Like in the thermal analysis. However.. σ2 and σ3 are the principal stresses. multi-pass welding was simulated by Free and Goff [6] and material hardening was included in the modeling by Hong et al. Abr/Jun 2006 formation. Moreover. This is. in this case. which is defined as [11]: σv M = 1 (σ 1 − σ 2 )2 + (σ 2 − σ 3 )2 + (σ 3 − σ 1 )2 2 [ ] (4) where σ1. depending on the particular technique employed. there is no need to perform time integration in the structural analysis. such as a larger number of degrees of freedom and a more complex non-linear behavior. thereby reducing the stress to the material yield [12]. using the Young modulus. such as those used in the study carried out by Kamtekar [3]. which is normally based on the von Mises criterion: σv M −σ y < 0 (3) where σy is the yield strength and σvM is the von Mises equivalent stress. The plastic strain caused by this type of transformation is known as transformation plasticity. T is the current temperature and Tref is the reference temperature at which the expansion is null. despite the difficulties in achieving accurate modeling of complex phenomena (plasticity behavior. A. A. there may be numerical instability setbacks due to the fact that the material presents low stiffness at high temperatures. were recently performed [8. In the earliest simulations. According to Francis [9]. (5) is known as the yield surface. numerical methods for this purpose have been more and more developed. According to the von Mises criterion. the mesh used for the structural analysis is generally the same as that used for the thermal analysis. It may. However. Structural analisys The welding structural analysis constitutes an even more difficult problem than the thermal analysis. such as being costly and time consuming. Rade. Papazoglou and Masubuchi [5]. the cause for non-convergence. The thermal strain is obtained by using the thermal expansion relation: ε therm = α T − Tref ( ) (2) where α is the thermal expansion coefficient. can be restricted to specific types of materials and can even provide incorrect results when plastic strain occur..Bezerra. 11. [7]. The surface represented by Eq. another very important aspect. in fact. mainly due to the evolution of computational technology.

Numerical Modeling Numerical modeling was performed using the FE code ANSYS®. Kinematic hardening. Perfect-plasticity is considered for strains above 1. 65 Soldagem Insp.30. The Poisson ratio was considered constant and equal to 0. a very small Young’s modulus was adopted to simulate the low stiffness of the weld pool. which is analogous to a one-pass welding. In isotropic hardening. Table 1 shows the thermal expansion coefficient and the Young’s modulus as a function of temperature [10]. decreasing restraint will increase the distortion. 3. is used to represent the case of multi-pass welding [13]. This type of hardening is adequate to represent a monotonic loading. Figure 2. (a) (b) Figure 1. on the other hand. while increasing restraint will increase the residual stress. Above the melting temperature.5 %. assumes that the yield surface remains constant in size. The stress-strain curve used as function of temperature. It is important to point out that residual stress and displacement are sensitive to the type of restraints applied at the supports.e. In particular. since it has an austenitic matrix which is stable from room temperature to a value near to the fusion temperature. (a) Expansion of yield surface in isotropic hardening and (b) translation of yield surface in kinematic hardening. 11. so that the conditions (i.Simulation of a TIG Weld using Finite Element Method: Part 2 – Structural Analysis The hardening rule describes the changing of the yield surface with progressive yielding. Two hardening rules are available: isotropic hardening and kinematic hardening [11]. Abr/Jun 2006 . in association with isotropic hardening. is illustrated in Figure 2.. It was considered that the material does not suffer any phase transformation in the solid state. which was obtained experimentally by Depradeux [10]. 2. the yield surface remains centered about its initial centerline and its size increases as the plastic strains evolve (Figure 1a). Stress-strain curve as a function of temperature [10]. No. This type of hardening. but it translates in stress space with progressive yielding (Figure 1b). stress states) for subsequent yielding can be established. Vol. The modeled piece is a 250 mm × 160 mm × 10 mm stainless steel AISI 316L plate. Multi-linear isotropic hardening based on the von Mises plasticity criterion was assumed and large strain effects were also taken into account.

2. Young modulus and thermal expansion coefficient [10]. The time step employed in the calculation was 1. Results The numerical results obtained were compared to experimental results presented by Depradeux [10]. The meshed model (17. C. Room temperature (301 K) was imposed as reference. which inputs the results of thermal analysis previously obtained.. A. Therefore. Rade. all of them obtained by Depradeux [10].25 s and it was used a convergence tolerance criteria of 0. The experimental results of residual stress were obtained via a X-ray diffraction technique. The welding process simulated was the TIG process. 4. In this manner. which is a 3D solid element with eight nodes having three degrees of freedom per node (translations in the nodal x. A. The points P1 to P6 were used to evaluate the transverse displacements on the bottom face as a function of time during welding and cooling. The element used to mesh the model is SOLID45. as described by Bezerra et al.. [14].0 on the force. The welding parameters used were: current of 150 A. while the sections S1 and S2 were used to evaluate the final deformed shape on the bottom and upper faces. 11. Scotti.574 nodes) is shown in Figure 3 and it is the same used for the thermal analysis presented in a companion paper [14]. the corresponding temperature field was applied and the new displacement field was determined considering the state of the previous load-step. A routine in MATLAB® was implemented to write an ANSYS® input file. for each instant of time considered. The welding residual stresses were compared at section x=150 mm on the bottom face. D.Bezerra. Vol. a displacement restriction in z direction was imposed at these points on the bottom face of the plate. The plate is supported on three supports situated according to Figure 4. the transverse displacements via inductive sensors and the final deformed shape by using a feeler gage. A. Soldagem Insp. y and z directions).1 on displacement and 1. Abr/Jun 2006 66 . Table 1.. (a) (b) Figure 3. in which the time evolution is considered by applying sequentially the transient temperature field by means of load-steps. voltage of 10 V and travel speed of 1 mm/s. according to Figure 4. Finite element mesh of the studied plate (a) and its cross-section view (b). No. A series of static analysis were carried out. The transverse displacements were compared using three sections (S1 to S3).

The evolution of the transverse displacement as a function of time at points P1 to P6. analyzed at section x=150 mm. which makes the plate deform into a “V” shape is clearly verified. It is possible to verify a good agreement between the results both during welding (until 230 s) and during cooling. a satisfactory agreement between numerical and experimental results. No. Final deformed shape of sections S1 and S2 on bottom (a) and upper (b) faces. disagreement is noticed only at y=0. However. As expected. Figure 4. Cross-sections and points used for comparing experimental and numerical results. it is possible to see that bending occurs not only about the x axis. while the experimental one suffers a slight drop (about 100µm). the numerical results for point P6 remains constant and equal to zero. Again. Vol. 67 . those that remain after complete cooling of the welded piece. Figure 5 shows the numerical and experimental results describing the evolution of the transverse displacements as a function of time for the points indicated in Figure 4. can be observed. but also about the y axis. the occurrence of high tensile stresses in the longitudinal direction near the weld bead is confirmed. The residual stress field is illustrated in Figure 8 on x (longitudinal) and y (transversal) directions as a contour plot. in sections S1 and S2 for bottom and upper faces. high compressive stresses near the start and end points of the bead are observed. is shown in Figure Soldagem Insp. For the upper face. Analyzing these results. that is to say. Abr/Jun 2006 (b) Figure 5. 2. (a) (a) (b) Figure 6. during cooling.Simulation of a TIG Weld using Finite Element Method: Part 2 – Structural Analysis Figure 6 shows a comparison between experimental and numerical results for residual transverse displacements.. mainly for the bottom face. 11. The occurrence of bending about the x-axis. A comparison between the numerical and experimental stresses. In the transverse direction. Figure 7a illustrates the residual transverse displacement field (after cooling) and Figure 7b depicts the final deformed shape with an amplification of 20 times.

A. It can be seen that. (a) Residual transverse displacement field (in µm) and (b) final deformed shape (amplified 20×). which is the case near the fusion line. (a) (b) Figure 7. D. for stresses in the longitudinal direction (Figure 9a).. 2. which. A.. the experimental results present a compressive stress of about 100 MPa on the plate edge. Vol. C. Consequently. Scotti. 9. an inconsistency between the results for points far away from y=0 is verified. Soldagem Insp.. this can be explained by the fact that the experimental technique used is X-ray diffraction. does not indicate a true stress in points where plastic strain has occurred. A. the numerical results obtained (distortions and stresses) can be considered satisfactory. which is confirmed by Depradeux [10].Bezerra. 11. according to Cullity [1]. However. For stresses in the transverse direction (Figure 9b). Nevertheless. Thus. Residual stress field on longitudinal (a) and transversal (b) directions (in MPa). there is an uncertainty about the experimental measurements of residual stresses. there is a good agreement between both sets of results only for the points farther than 20 mm from y=0. Rade. where is known that the transversal stress should be zero (because it is perpendicular to a free edge). Abr/Jun 2006 68 . Figure 8. No.

. [2] Parlane. pp. [4] Rosenthal. 1997. [8] Fricke. 365378.. Inc. 2006. Journal of Pressure Vessel Technology. [12] ANSYS. Numerical Weld Modeling – a Method for Calculating Weld-Induced Residual Stresses. J. N. [11] Lemaitre. 11. S. [6] Free. Simulation of a TIG Weld using Finite Element Method: Part 1 – Thermal Analysis (Simulação de uma Soldagem TIG usando o Método dos Elementos Finitos: Parte 1 – Análise Térmica). The Welding Institute. B.. Vol. n. U.. MS Thesis. Leggatt. Keim. Nº 1. 7. vol. Computers and Structures. 1981. 104. E. and Schmidt.C. Rade. Thus. n.K. vol. [13] Murthy. L. Residual stresses on the bottom face in the longitudinal (a) and transverse (b) directions at x=150 mm.F. [10] Depradeux. A. Predicting Residual Stresses in Multi-Pass Weldments with the Finite Element Method. Elements of X-Ray Diffraction. 1998. J. Simulation Numerique du Soudage – Acier 316L – Validation sur Cas Tests de Complexite Croissante. K.G. pp.K. R. 1982. Harrison.G. the results of numerical simulations have shown to be in satisfactory agreement with experimental ones provided in the literature. A. Theory Reference Manual. and Goff. 69 Soldagem Insp. Y.-L.J.D. pp.P. L’Institut National des Sciences Appliquees (INSA) – Lyon. Dwight. J. and Dong. pp. of the Brazilian Ministry of Science and Technology. 1996..A.. In general. and Chaboche. for research scholarships and the grant of doctorate. 60. vol. Residual Stresses and their Effect. P. vol. FEM was used to get the transient thermal stress and residual stress fields and distortions. J. J. Virginia Polytechnic Institute and State University. 2004. [7] Hong. based on the temperature time-history obtained through a previous thermal analysis. 2002. 131-154. The Calculation of Welding Residual Stresses in Thin Steel Plates. pp 2-9. Welding Journal. Assessment of Numerical Procedures for Residual Stress Analysis of Multipass Welds. D. Dunod. 2nd edition. 1988. J. 32. which is based in a series of simplifying assumptions and a set of parameter values used as inputs in both the thermal and structural analyses. No. Abr/Jun 2006 .B.. 220-234. 1978.. although more specialized software could be more adequate for this purpose.V. and CAPES.4. 6. Procter. R. International Journal of Mechanical Sciences. pp. L.K. The support of agency FAPEMIG through Project TEC 317/03 is greatly appreciated and acknowledged.D. The limitations of the experimental procedure must be taken into account and one can conclude about the feasibility and accuracy of the modeling procedure adopted.J.A. and Iyer.H. Ph. Welding Journal. G. 11. [5] Papazoglou. Ecole Doctorale des Sciences de L’Ingenieur de Lyon. G. Numerical Simulation of Welding and Quenching Processes using Transient Thermal and Thermo-Elastic-Plastic Formulations.Simulation of a TIG Weld using Finite Element Method: Part 2 – Structural Analysis of the Brazilian Ministry of Education. and Masubuchi. pp. N. [9] Francis. ISSN 0104-9224. C. References [1] Cullity. 2001. J. P.S. 1989.D Thesis. 1941. Mécanique des Matériaux Solides. D. Welding Simulations of Aluminum Alloy Joints by Finite Element Analysis. A. 5. 2nd edition. (a) (b) Figure 9. 77.. J.. Allen. 2. Nuclear Engineering and Design. Conclusions This study presented the structural analysis of a TIG weld without filler metal on a stainless steel AISI 316L plate. 207-227. 1. 1978. “Mathematical theory of heat distribution during welding and cutting”. J. release 5. Acknowledgements The authors wish to thank the agencies CNPq. and Scotti. Vol. [3] Kamtekar.V. Numerical Analysis of Thermal Stresses during Welding Including Phase Transformation Effects. It is believed that the modeling procedure described herein can be successfully extended to simulate other welding processes and the welding of parts with more complex geometries. 139-150. 2. Addison-Wesley Publishing Company. vol. pp.A. Computers and Structures. 206. 372382. 198-203. 20. [14] Bezerra. Revista Soldagem e Inspeção. No.. A. Vol. the two companion papers fully describe the modeling procedures necessary to carry out the complete simulation of welding-induced residual stresses and distortions. 9. Bailey. E.D. Tsai. Rao. Jan/Mar 2006.D. V. 20. and Saunders.