Realization of quality design for steel products by using locally weighted regression modeling

Abstract

This paper presents a quality design method using locally weighted regression which is a type of Just-in-time modeling. The quality design system using the proposed method have been developed by utilizing a data mining tool, and applied practically to control of the various quality properties of steel products. Design accuracy was improved and design time was shortened in comparison with the conventional technique. Reducing product quality deviations and model maintenance load have been achieved in the commercial plants at JFE Steel's West Japan Works. The systems have made significant contributions to product quality improvement, production cost reduction and employee's satisfaction.