Artificial neural networks and adaptive neuro-fuzzy models for predicting WEDM machining responses of Nitinol alloy: comparative study

Abstract

This article reports a comparative study of artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) models for better prediction of wire electro-discharge machining (WEDM) responses like material removal rate and surface roughness of a Nitinol alloy. Pulse on time (Ton), pulse off time (Toff), peak current (Ipeak) and gap voltage (V) were selected as input attributes. Experimental results were performed to verify the results from ANN and ANFIS models. ANN model, back-propagation with three different algorithms Levenberg–Marquardt (LM), Elman regression neural network and generalized regression neural network and ANFIS model, were developed using the same input variables. The most suitable algorithm and neuron number in the hidden layer were found as LM with 10 neurons for ANN models whereas the most suitable membership functions and number of membership functions are found to be gauss and two, respectively. The statistical validation measures such as root mean square error, mean square error and mean absolute percentage error are obtained through ANN and ANFIS models. The statistical values are given in the tables. As per the statistical measures perspective, the ANFIS model will have better accuracy for anticipation of WEDM attributes of a Nitinol alloy.