Prediction and Control of Residual Stress Distribution in Welded Joint Using Artificial Neural Network

The welding process induces Residual tensile stress that is detrimental to Fatigue life. Tensile stress act to stretch or pull apart the surface of the material. With enough loads cycle at a high enough tensile stress, a metal surface initiate a crack. Significant improvement in Fatigue life can be obtained by modifying the Residual stress level in the material. The intent of this Project is an artificial neural network approach was used to predict the residual stresses in welded joint. MATLAB was used to train the neural network using the Levenberg-Marquardt technique. The correlation results between experimental data and ANN outputs confirmed the feasibility of using artificial neural networks for modelling and predicting the residual streses.