On line detection of faults in a rolling mill process a method for detecting faults in a rolling mill process using multivariate SPC and system identification

A method for detecting on line a fault and his nature in a rolling mill process controlled by SPC (Statistical Process Control) is presented. The method uses the integration of multivariate SPC, traditional APC (Automatic Process Control) and a System Identification technique based on the shape factor of the system's step response. The normal behaviour of the automation system is controlled on line computing the shape factor of several step responses registered during the process cycle. The faults due to a change of the automation system and the faults due to a change in the process are then separated. The results of an implementation in a real plant are presented.