Artificial neural network modeling studies to predict the friction welding process parameters of Incoloy 800H joints

The present study focuses on friction welding process parameter optimization using a hybrid technique of ANN and different optimization algorithms. This optimization techniques are not only for the effective process modelling, but also to illustrate the correlation between the input and output responses of the friction welding of Incoloy 800H. In addition the focus is also to obtain optimal strength and hardness of joints with minimum burn off length. ANN based approaches could model this welding process of INCOLOY 800H in both forward and reverse directions efficiently, which are required for the automation of the same. Five different training algorithms were used to train ANN for both forward and reverse mapping and ANN tuned force approach was used for optimization. The paper makes a robust comparison of the performances of the five algorithms employing standard statistical indices. The results showed that GANN with 4-9-3 for forward and 4-7-3 for reverse mapping arrangement could outperform the other four approaches in most of the cases but not in all. Experiments on tensile strength (TS), microhardness (H) and burn off length (BOL) of the joints were performed with optimised parameter. It is concluded that this ANN model with genetic algorithm may provide good ability to predict the friction welding process parameters to weld Incoloy 800H.