Development of an ANN for the Prediction of Heat Treatment Temperatures for Martensitic Stainless Steels

This paper presents the current status of the research project aimed at predicting the optimum tempering temperature during the quenching and tempering heat treatment process of martensitic stainless steels, based on the chemical composition and previous metallurgical history. At present, in the iron and steel industry, time-consuming and cost-intensive laboratory tests often relies on to minimize the tolerance ranges of the mechanical properties. This paper contains a proposal as to how an artificial neural network (ANN) can be used in future to predict the optimum tempering temperature without considering complex metallurgical and mathematical models describing the evolution of the microstructure during the quenching and tempering process. Prior to developing an ANN for the prediction, the entire quenching and tempering process must be clearly understood. This stage will be followed by a comprehensive literature survey to determine the current state of the art. The usefulness of ANN for process control is demonstrated through a model which was developed by Siemens GmbH to predict the mechanical properties of hot rolled steels. This prediction is based on the outcome of a combination of a calculation and an ANN. Although the hot rolling operation affects the quenching and tempering process, we will show why a purely mathematical approach is insufficient. Thus, this paper will outline the input and output parameters and the design of the structure required for a new ANN.