Comparison of Artificial Neural Network Methods for Modeling of Turning of Polymer Matrix Composite

In this research, polymer matrix composite filled with aluminum particles was synthesized and turned with different machining condition namely: cutting speed, weight fraction of particle, depth of cut and feed. Then, surface roughness was measured and two artificial neural networks models Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) were developed to estimate effects of four turning parameters on surface roughness. Correlation between training data and experimental data were shown that MLP network was better than RBF as a compatible network (correlations were 0.835 for MLP network and 0.542 for RBF network). Because of higher correlation for MLP network, this network was proposed as a model for investigation the effects of turning parameters on surface roughness.