Prediction of mechanical properties of cast irons

Abstract

The adaptive hierarchical model reflecting behavior of different types of cast irons was designed and verified. It enables to evaluate an industrial metallurgical process with respect to its parameters, such as many possible combinations of chemical composition or various approaches to heat treatment and, consequently to estimate mechanical properties of the final product. Because of the multivariate nature of the presented problem, it is impossible to make technological conclusions based only on formulae or diagrams. That is why an intelligent computational technique can help, for example, in the development of new metal materials or act as a powerful tool in total quality management systems. In the proposed method, inductive learning and classification principle was selected and justified as suitable prediction tool. Thanks to the built-in adaptivity the suggested technique is more flexible than commonly used statistical methods and, in contrast to the numerical simulation of material characteristics, it can easily incorporate also historical or technology-specific values. After the series of computational experiments the presented intelligent approach to prediction was found as an applicable alternative to the traditional ways of laboratory investigations and processing of experimental data. Its responses were verified on real samples and compared with experts’ opinion.