Application of Neural Networks to Optimal Selection of Cutting Length of Bars

Abstract

A new method of optimal selection of cutting length of steel bars etc. using mutually connected Hopfield type neural networks is proposed. When each raw bar of various length is cut to several finished bars of required length, the combination of cutting length in a bar should be determined to keep the amount of scraps produced minimum as well as to maintain the balance among the products in accordance with the customer's order. The proposed neural networks can determine the optimal combination of length by searching the networks using energy function. The configuration of networks, the definition of network energy function and the derivation of the network constants, such as weight and threshold, are explained, along with the results of simulation studies.