Optimal design of alloy steels using evolutionary computing

Abstract

Over the last four years efforts have been devoted towards the development and validation of mechanical test result models relating to a range of alloy steels. Several neural-network based models have been developed, two of which are related to the mechanical test results of Ultimate Tensile Strength (UTS) and Reduction of Area (ROA). The ultimate aim of developing these models is to pave the way to process optimisation through better predictions of mechanical properties. In this research we propose to exploit such neural network models in order to determine the optimal alloy composition and heat treatment temperatures required, given certain predefined mechanical properties such as the UTS and ROA. Generic Algorithms are used for this purpose. The results obtained are very encouraging.