An accelerated explicit method with GPU parallel computing for thermal stress and welding deformation of large structure models

作者：Ma Ninshu

来源：International Journal of Advanced Manufacturing Technology, 2016, 87(5-8): 2195-2211.

DOI：10.1007/s00170-016-8542-3

• Accelerated explicit method
• GPU parallel computing
• Thermal stress
• Welding deformation
• Large scale models
• Welded structure

To simulate welding induced transient thermal stress and deformation of large scale FE models, an accelerated explicit method (ACEXP) and graphical processing units (GPU) parallel computing program of the finite element method (FEM) were developed. In the accelerated explicit method, a two-stage computation scheme is employed. The first computation stage is based on a dynamic explicit method considering the characteristics of the welding mechanical process by controlling both the temperature increment and time scaling parameter. In the second computation stage, a static equilibrium computation scheme is implemented after dynamic thermal loading to obtain a static solution of transient thermal stress and welding deformation. It has been demonstrated that the developed GPU parallel computing program has a good scalability for large-scale models of more than 20 million degrees of freedom. The validity of the accelerated explicit method is verified by comparing the transient thermal stress and deformation with those computed by an implicit FEM. Finally, welding deformation and residual stress in a structure model assembled from nine high-strength steel plates and 26 weld lines were efficiently analyzed by ACEXP and GPU parallel computing within 45 h. The computed welding deformation agreed well with measured results, and a good accuracy was obtained.

• 出版日期2016-11