Optimization of heat treatment process parameters using neural networks and Nelder-Mead algorithm

Abstract

Metallurgical processes consist of different and complex production operations. One of them is heat treatment. Hardness value is an important response variable for heat treatment process. Heat treatment parameters and interactions between each other are not known clearly. Hence it is hard to define convenient parameters for requested hardness value. In this study, effects of heat treatment parameters on hardness are modelled using back propagation artificial neural network (BPANN) model. BPANN is used to formulate a fitness function for predicting the value of the response based on the parameter settings and then Nelder-Mead algorithm takes the fitness function from the trained network to search for the optimal heat treatment parameters (furnace heat and heat treatment time) combination.