Hot-rolled strip thickness diagnosis and abnormal transmission path identification based on sub stand strategy and KPLS-MIC-TE

Abstract

Multi-stand combined rolling is the key process to determine the quality of strip. The rolled piece presents different states in each stand unit, and there is information transmission and interaction between stands. To reduce the impact of quality-related faults on subsequent production systems, it is particularly important to effectively identify the transmission path of quality-related information for locating the root cause of abnormalities. However, the causes of quality defects are misidentified due to the neglect of genetic characteristics of quality in the overall quality diagnosis. Focus on hot rolled strip thickness diagnosis, an abnormal transmission path identification method based on sub stand strategy and Kernel Principal Least Square-Maximal Information Coefficient-Transfer Entropy (KPLS-MIC-TE) was proposed to improve the accurate identification of defect roots, the sub-stand was diagnosed, and the abnormal variable diagnosis of a sub stand was conducted with the KPLS algorithm. Secondly, based on the asymmetric characteristics of the transfer entropy matrix, a directed transfer topology with the causality of variables was constructed. Finally, the production data of the 1580 hot rolling line is used for industrial validation. The results show that the method can accurately diagnose process abnormal variables and identify the path of abnormal variables.