MODELING OF TIG WELDING PROCESS BY REGRESSION ANALYSIS AND NEURAL NETWORK TECHNIQUE

In the present paper, neural network-based expert systems have been developed for process parameter to weld bead geometry for tungsten inert gas (TIG) welding process welding. However linear regression analysis is used for the process modeling and analysis of numerical data consisting of the values of dependent variables (responses) and independent variables (input parameters). The numerical data are utilized to obtain an approximation model correlating the outputs and inputs by showing the influences of the parameters on responses. Once trained, the neural network-based expert systems could make the predictions in a fraction of a second. The analysis of variance for all factor a pareto chart of effect of the responses on parameter and their interaction, which effect maximum on the welding process responses on weld bead geometry. Here, a performance analysis has been attempted to check the viability and performance of regression analysis and back propagation neural network (BPNN) based tool for predicting modeling of TIG welding process.