Application of neural networks for selection of steel with the assumed hardness after cooling from the austenitising temperature

Purpose: The aim of the study is to establish a system that supports the choice of steel grade for quenching and tempering at a required hardness curve as function of cooling rate from the austenitising temperature. Design/methodology/approach: It has been assumed that the steel will meet the criterion provided that the hardness curve, defined by the user, is included within the range of hardness change that is characteristic of a certain steel grade. In order to determine the steel hardness ranges it has been necessary to work out a suitable calculation model. Therefore, a neural network has been designed and verified numerically to calculate the steel hardness on the basis of chemical content for the predetermined cooling rate. To develop the relationship between the chemical composition, austenitising temperature, cooling rate and hardness of the steels for quenching and tempering the artificial neural network was used. The obtained results were used for determination of neural classifier. The classifiers based on the neural networks carries out the task of selection of the steel grade. Findings: Artificial neural networks can be applied for selection of steel with the assumed hardness after cooling from the austenitising temperature. Practical implications:The system presented can be applied to selection of steel grade intended for machine parts of predetermined hardness in the section of a hardened or normalized element. Originality/value: The research presented in this paper offers a new strategy useful in selection of steel grade.