A Taguchi–SVR Methodology for Friction Stir Welding of Dissimilar Aluminium Alloys for EV Battery Tray Applications

The need for Lightweight EV battery system parts is driving an increasing demand for welding technologies capable of joining distinct Aluminium alloys while ensuring adequate mechanical integrity. This work demonstrates a systematic experimental approach for friction stir welding (FSW) of AA6082-AA5083 and AA6061-AA7075 alloys couples. Emphasis was placed on the reduction of the experimental efforts with the investigation focused on the relationship between welding parameters and joint performance. The four welding parameters – tool rotational speed, traverse speed, axial force and tool pin profile were considered at 3 levels and planned using a Taguchi L27 orthogonal array design.  Tensile and yield strength, hardness and impact properties of the welded samples were evaluated. Taguchi's Signal-to-Noise ratio was used to identify optimal levels and combinations of the welding process parameters and Analysis of Variance (ANOVA) used to determine the relative significance of the selected welding parameters. Support Vector Regression (SVR) with radial basis function kernel was used as a data driven modelling technique to predict tensile and yield strengths of welds as a function of welding parameters. This method was used with 5-fold cross validation and hyperparameter optimization in order to improve predictability of models against the limitations of available experimental data.  This combined strategy involves design of experiments, experimentation, analysis and machine learning into a consistent framework for investigation of various Aluminium alloy FSW. This strategy could support evaluation of weld parameters and prediction of performance of welded Aluminium alloys in light weight EV battery trays.