Shinsuke Itoh

Development of idealized explicit FEM using GPU parallelization and its application to large-scale analysis of residual stress of multi-pass welded pipe joint

In this research, the authors developed the idealized explicit finite element method (IEFEM) to achieve shorter computing time and lower memory consumption in analyses of welding deformation and residual stress. IEFEM was parallelized by a graphics processing unit (GPU) to achieve even faster computation. To show its applicability to large-scale problems, the proposed method was applied to the analysis of the multi-pass welding of V-groove pipe joint that has 1 million elements, 13 layers, and 33 passes. In the analysis, isotropic hardening, kinematic hardening, and combined hardening were considered to investigate the influence of hardening rule on residual stress distribution. As a result, it is found that residual stress distributions were larger in the order of isotropic hardening, combined hardening, and kinematic hardening. In addition, the analyzed residual stress and experimental measurements showed good agreement. The computing time was approximately 70xa0h. From these results, it was shown that IEFEM can analyze a large-scale welding residual stress problem in realistic time with high accuracy.

Development of in situ measurement system for welding deformation using digital cameras

Abstract A three-dimensional (3D) deformation (in plane and out of plane deformations) measurement method is developed using digital cameras, which require no special equipment. This method is a non-contact method, and it can sequentially measure over the entire photographed image. Furthermore, since image analysis is based on the technique of image matching, the method is applicable even when the deformation to be measured is large. In addition, since it is possible to use all pixels as measuring points, the number of available measuring points at one time is the same as the number of effective pixels of the camera. In this study, the proposed method is applied to the sequential measurement of displacement under strong lighting levels in arc welding. Through the comparison of the results measured by a 3D shape measurement system (LAT-3D) using a laser displacement gauge and digital caliper, the quantitative validity of the proposed method is also verified.

Full-field time-series measurement for three-dimensional deformation under welding

Displacement during welding provides important information to understand the mechanisms of welding deformation and residual stress. In particular, if welding deformation can be measured sequentially and the displacement distribution over full field can be measured such as the results obtained by finite element analysis, they can be valuable information. Therefore, in this study, a 3-dimensional (3D) deformation (in-plane and out-of-plane deformation) measurement method is developed using a digital camera, which requires no special equipment. This method is a non-contact method and it can sequentially measure over the entire photographed image. Furthermore, since image analysis is based on the technique of image matching, the method is applicable even when measuring deformation is large. In addition, since it is possible to use all pixels as measuring points, the number of available measuring points at one time is the same as the number of effective pixels of the camera. This is currently more than 15 million points, and the measuring precision is expected to increase as the camera pixel resolution continues to increase. Therefore, this method is expected to have future potential. In this study, the proposed method is applied to the sequential measurement of displacement under the strong lighting levels in arc welding. By comparing the time history of transverse shrinkage, longitudinal shrinkage and angular distortion with the results of FEM thermal elastic-plastic analysis, the qualitative validity of the proposed method is verified. To investigate the measurement precision and usefulness of the method, a 3D shape measurement system (LAT-3D) using a laser displacement gauge and digital caliper are used. The distributions of residual transverse shrinkage and residual angular distortion are measured by the proposed method, LAT-3D and digital caliper. Through the comparison of the results measured by these methods, quantitative validity of the proposed method is also verified.

Computational simulation of weld microstructure and distortion by considering process mechanics

Highly precise fabrication of welded materials is in great demand, and so microstructure and distortion controls are essential. Furthermore, consideration of process mechanics is important for intelligent fabrication. In this study, the microstructure and hardness distribution in multi-pass weld metal are evaluated by computational simulations under the conditions of multiple heat cycles and phase transformation. Because conventional CCT diagrams of weld metal are not available even for single-pass weld metal, new diagrams for multi-pass weld metals are created. The weld microstructure and hardness distribution are precisely predicted when using the created CCT diagram for multi-pass weld metal and calculating the weld thermal cycle. Weld distortion is also investigated by using numerical simulation with a thermal elastic-plastic analysis. In conventional evaluations of weld distortion, the average heat input has been used as the dominant parameter; however, it is difficult to consider the effect of molten pool configurations on weld distortion based only on the heat input. Thus, the effect of welding process conditions on weld distortion is studied by considering molten pool configurations, determined by temperature distribution and history.

Development of Ultra Large Scale Computation for Transient Welding Deformation and Stress Using Idealized Explicit FEM Accelerated by GPU

Numerical simulations such as Finite Element Method (FEM) are widely used as tool of structural analyses in both design and production. However, in the application of FEM to welding problems, the simulation scale is usually limited to the welding joint level. Only a few large-scale welding analyses are performed on existing research because welding problems are transient and show strong nonlinearity. In such cases, it is necessary to use static implicit FEM to achieve an accurate analysis, but the larger analysis scale requires larger memory consumption and computing time. Thus, we previously proposed idealized explicit FEM (IEFEM) to achieve shorter computing time and lower memory consumption.Since IEFEM is based on dynamic explicit FEM, it is not needed to solve the stiffness matrix of the whole system and it is possible to analyze by only performing the calculation for each degree of freedom (DOF) and element. Such characteristic indicates that IEFEM is suitable for parallelization. Then, in this study, we developed parallelized IEFEM using a graphics processing unit (GPU). The usefulness and validity of the developed method are considered by analyzing a 3-dimensional multi-pass moving heat source problem, which is very difficult to analyze with commercial FEM software because of its analytical scale. As a result, it is found that parallelized IEFEM accelerated by a GPU can analyze a large-scale problem having over 1,000,000 DOFs on a single PC.Copyright

Prediction of Residual Stress in Multi-Pass Welded Joint Using Idealized Explicit FEM

Heavy thick steel plate is used for pipes and also ship structures, and multi-pass welding is usually adopted for the welding. Because of the heavy thickness, residual stress plays an important role, particularly in crack propagation. Implicit Finite Element Method (FEM) is often used as a welding analysis method to examine the residual stress of the welded plate, but it is not easily applied to multi-pass welding problems with tens of thousands of degrees of freedom, because of the huge computational time and memory consumption. Alternatively, it is possible to simulate the residual stress in shorter time with lower memory consumption by using Idealized Explicit Finite Element Method developed by the authors. Moreover, the computational time can be shortened by using Idealized Explicit FEM using a Graphics Processing Unit (GPU). In this research, Idealized Explicit FEM parallelized using a GPU is applied to the analysis of the residual stresses of the multi-pass welding joint of a pipe structure made of heavy thick steel plate.As the result, the residual stress simulated by the Idealized Explicit FEM corresponds to the measured residual stress. Furthermore, it is found that the grouping method may affect to the residual stress distribution.Copyright

Study on residual stress in multi-pass welded joint using idealized explicit FEM

Heavy thick steel plate is used for pipes and also ship structures, and multi-pass welding is usually adopted for the welding. Because of the heavy thickness, residual stress plays an important role, particularly in crack propagation. Implicit Finite Element Method (FEM) is often used as a welding analysis method to examine the residual stress of the welded plate, but it is not easily applied to multi-pass welding problems with tens of thousands of degrees of freedom, because of the huge computational time and memory consumption. Alternatively, it is possible to simulate the residual stress in shorter time with lower memory consumption by using Idealized Explicit Finite Element Method developed by the authors. Moreover, the computational time can be shortened by using Idealized Explicit FEM using a Graphics Processing Unit (GPU). In this research, Idealized Explicit FEM parallelized using a GPU is applied to the analysis of the residual stresses of the multi-pass welding joint of a pipe structure made of heavy thick steel plate. As the result, the residual stress simulated by the Idealized Explicit FEM corresponds to the measured residual stress. Furthermore, it is found that the grouping method may affect to the residual stress distribution.

Three-dimensional in situ measurement system for welding deformation using digital cameras

A 3-dimensional deformation (in-plane and out-of-plane deformation) measurement method is developed using digital cameras, which require no special equipment. This method is a non-contact method and it can sequentially measure over the entire photographed image. Furthermore, since image analysis is based on the technique of image matching, the method is applicable even when measuring deformation is large. In addition, since it is possible to use all pixels as measuring points, the number of available measuring points at one time is the same as the number of effective pixels of the camera. n nIn this study, proposed method is applied to the sequential measurement of displacement under the strong lighting levels in arc welding. Through the comparison of the results measured by a 3-dimensional shape measurement system (LAT-3D) using a laser displacement gauge and digital caliper, quantitative validity of the proposed method is also verified.

Investigation of factors influencing welding deformation of ship block by inherent strain analysis using idealized explicit FEM

It is impossible to avoid welding deformations when steel structures such as ships are assembled. Therefore, quantitative prediction and effective control of residual welding deformation are necessary. However, the current finite element analysis requires very long computational time and memory. To solve this problem, we have developed an inherent strain analysis method using idealized explicit FEM and have applied this proposed method to a ship block model. The simulated results agree well with the measured deformation. Furthermore, the influence of various factors on the welding deformation of targeted ship block is also investigated.