Hybrid Associative Classification Model for Mild Steel Defect Analysis

Quality of the steel coil manufactured in a steel plant is influenced by several parameters during the manufacturing process. Coiling temperature deviation defect is one of the major issues. This defect causes steels metallurgical properties to diverge in the final product. In order to find the cause of this defect, various parameter values sensed by sensors are stored in database. Many approaches exist to analyze these data in order to find the cause of the defect. This paper presents a novel model HACDC (Hybrid Associative Classification with Distance Correlation) to analyze causality for coiling temperature deviation. Due to the combination of association rule, distance correlation and ensemble techniques we achieve an accuracy of 95 % which is quite better than other approaches. Moreover, to the best of our knowledge, this is the first implementation of random forest algorithm in analyzing steel coil defects.