Numerical modeling of welding processes

Welding processes induce microstructural changes and residual stresses and strains that are as difficult as they are important to control.

High residual stresses combined with the presence of hard phases can lead to joint embrittlement, or distortions induced by welding can lead to part misalignment, making it impossible to carry out a welding sequence.

Numerical simulation of welding processes often appears to be a privileged means of access to the quantities that characterize them. For example, numerical simulation of welding is of great interest when it comes to studying the mechanical strength of a welded joint or verifying the feasibility of a welding sequence. The simulation of such processes requires the modeling of complex interactions between thermal, metallurgical and mechanical phenomena, and the implementation of specific numerical methods. Spurred on by the nuclear industry in the 1980s and 1990s, a great deal of research was carried out into predicting residual stresses in welds on mechanical components. The methods and models developed formed the basis of numerical simulation software used today by a large number of industrial companies in the aeronautics, automotive and metallurgy sectors.

The aim of this article is to take stock of numerical modeling methods for welding processes.

The first section describes the benefits of welding simulations.

In the second part, the physical phenomena involved and their modeling are presented. The various metallurgy models that are key to this type of simulation are described, as well as how to take into account couplings with thermal aspects (phase-dependent thermophysical properties, latent heats of transformation) and mechanical aspects (volume changes, transformation plasticity in particular).

The third part is devoted to the implementation of these models in a numerical finite element approach. Various modeling algorithms and methods are described, such as step-by-step transient analysis with adaptive meshing, steady-state resolution or a two-scale method (local/global). Finally, applications based on the models and methods presented are described and commented on. In each case, we demonstrate the validity and efficiency of the numerical simulation. It concludes with an assessment of what can be expected from current simulation resources, and what developments can be expected in the coming years.