Multiobjective Genetic Algorithm applied to inverse finite element analysis: T-joint welding case study

Abstract

ABSTRACT Welding processes play as an important role in the manufacturing of a lot of products in most of industrial sectors. Despite its wide applicability, the welding process is subjected to some inconsistencies in quality due to controllable and uncontrollable variables. This work is aimed to use inverse Finite Element Analysis (FEA) to simulate the movement of the heat flow in the base metal submitted to the Gas Metal Arc Welding (GMAW) process. To this end, control limits were established, such as voltage and electric current (in the form of the heat source) and welding speed in order to obtain optimized values of deformations and stresses along the plates subjected to the GMAW welding process. For the numerical simulation process modelling, empirical data from weld beads applied on both sides of two ASTM A36 steel plates (thickness of 9.5 mm each) were used in a T-joint configuration. In this case, a multiobjective algorithm to find the optimized solution was used. Non-dominated Sorting Genetic Algorithm II (NSGA II) in conjunction with FEA was applied. The results presented are consistent with literature data within the pre-established limits, which are deformations and stresses less than 2 mm and 600 MPa, respectively. This demonstrates the potential of using FEA in conjunction with the NSGA II genetic algorithm to predict input variables in welding processes. This can be considered an important contribution to the industrial technological advancement for predicting welding process’s variables.