Current Proceedings on Technology
Yazarlar: İlhan Asiltürk, Bekir Karlik
Konular:-
Anahtar Kelimeler:Surface roughness,Prediction,Neural networks,Multi layered perceptron
Özet: This proposed work deals with the development of surface roughness prediction model for turning of AISI4140 steel, using artificial neural network (ANN). Experiments have been carried out on conventional turning machine with carbide cutting tool with seven inputs, namely, three axes vibrations of the tool holder, as well as cutting speed, feed rate, depth of cut and profile angle. The data obtained by experimentation is used to construct predictive models. In this study a Multilayer perceptron (MLP) network using Feed Forward and Error Back propagation was chosen as the neural network structure to predict surface roughness. The process model roughnesses of the machined surfaces corresponding to these conditions are the outputs of the neural network. The Learning rate and momentum coefficient were found as 0.9 and 0.1 respectively for the smallest mean squared error (MSE) obtained as 0,000797 after 3000 epochs, and these values served as a measure of prediction precision.