Prediction of Weld Geometry in Laser Welding by Numerical Simulation & Artificial Neural Networking

Abstract

ABSTRACT Advanced manufacturing processes have increased industrial productivity and enhanced marketing prospects. Optimization techniques coupled with simulative methods in laser welding can immensely contribute to the development of the industries of the future. This work reports the research results emanating from the application of simulative techniques for the prediction of welding parameters of the laser welding process of aluminium alloy 2024. Numerical simulation was conducted in a proprietary software environment and subsequently, artificial neural network (ANN)-based simulation was conducted. The simulatively obtained data were compared with experimentally obtained data and were found to be in good accord. The ANN was trained, validated, and tested with the experimental data. The validity of the ANN-based model was ascertained with linear regression analysis and the correlation coefficients were computed. A comparison was made between the numerical simulation-based model and the ANN-based model and it was observed that the artificial neural network-based model predicts the responses faster hence it could be viewed as a potent tool for quicker and more efficient calculation of the weld dimensions and strength in both laboratory and workshop environments.