Advanced Taguchi-Neural Network Prediction Model for Wire Electrical Discharge Machining Process

The present case study throws light on developing a new computational prediction model to predict the response, namely, material removal rate (MRR) more precisely with only few experimental runs using lower-level Taguchi’s orthogonal array (OA) in a wire electrical discharge machining (WEDM) process. The input parameters are pulse-on-time (TON), pulse-off-time (TOFF), peak current (IP), and servo sensitivity (SS). The developed model is based on artificial neural network (ANN) and the Taguchi method (TM), thus called as advanced Taguchi-ANN model. This method overcomes the limitations of both traditional ANN which requires a large number of experimental data for predicting the responses accurately and TM which can predict the response at only chosen parametric levels. However, the response can be predicted more precisely at any point in between and at the chosen parametric levels by this method. The optimal machining condition for maximizing MRR was found to be TON of 6 μs, TOFF of 7/9 μs, IP of 6 Amp, and SS of 8 mm/min. The maximum error % between the experimental and predicted values for MRR by advanced Taguchi-ANN at phase 2 was found to be within 5%, which is quite acceptable, whereas the same by traditional ANN method at phase 1 was higher.