An Optimal Design of an MLFNN Coupled with Genetic Algorithm for Prediction of MIG-CO2 Welding Process

In this paper, an optimal design of multilayer feed forward neural network coupled with real coded genetic algorithm has been demonstrated for predictive modeling of MIG-CO2 welding process parameters for EN-3A grade mild steel. The predictive modeling of this procedure has been established in the forward direction by designing a multilayer neural network model through updating its connection weights by back propagation algorithm and by real-coded genetic algorithm on the data set which was collected experimentally. Finally, a comparison study on the most efficient neural network design using Python programming has been carried out, and it was discovered that the multiple regression model and the back propagation neural network (BPNN) are both outperformed by the welding geometry predicted by the genetic algorithm tuned neural network (GANN) model.