Prediction of material removal rate in wire electrical discharge turning using artificial neural networks and adaptive neuro-fuzzy models

This work intended to assess the prediction and simulation effectiveness of the artificial neural network (ANN) with adaptive neuro-fuzzy inference system (ANFIS) approaches for modeling the material removal rate (MRR) in wire electrical discharge turning for fabrication of micro-pin made by Ti6Al4V. 16 experiments have been conducted according to full factorial design by varying four different WEDT input attributes namely pulse intensity, voltage open, wire tension and spindle speed. This dataset is aimed to be used for training and then, five more trials with random selection of input attributes is conducted to be established as the validation data. In developing the ANN model, Levenberg–Marquardt backpropagation training algorithm with ten neurons of hidden layer is employed and the Gaussian curve built-in membership function is used for developing the ANFIS model. The ANN and ANFIS model have been compared with experimental results. Both models indicated good predictions, however, the comparison revealed that the ANFIS model produced the closest result with the experiment compare than ANN.