Prediction of wire electrical discharge machining process based on GRNN

The general regression neural network (GRNN) was used to predict the process of WEDM multiple cutting, in order to reduce the blindness of parameter selection. The experiment on cutting speed and surface roughness was studied by orthogonal test method, factors including discharge pulse width, pulse interval, peak current, wire speed, the working fluid and each cutting offset. Mean square deviation of error sequence was generalized evaluation index of GRNN. The experiment shows that GRNN prediction cutting speed error is less than 4% and the surface roughness error is less 2%. The prediction of WEDM multiple cutting has a high forecast precision and can effectively select the processing parameters.