Optimization of Laser Beam Welded Novel Dissimilar Material UNS S32304 and 304 L Steel Process Parameters Using Deep Learning
Abstract
In recent years, laser beam welding of dissimilar materials has become crucial for diverse industrial applications. Our research targets the optimization of this process for joining UNS S32304 and 304 L steel, necessitating a delicate balance of input parameters like peak power, weld speed, pre-heat temperature, undercut, deformation and tensile strength. The challenge lies in the intricate relationship between these parameters and weld quality, demanding a robust prediction model. To tackle this, we propose a deep learning strategy incorporating feature scaling, SMOTE for class imbalance and Multi-Agent Salp Swarm Optimization (MASSO). Employing a Multi-Layer Perceptron (MLP) neural network ensures precision, bridging traditional welding with advanced deep learning for more efficient and reliable industrial applications.
