A New Multiphase Process Monitoring Method with Application to a Hot Rolling Mill Process

A joint knowledge and data-driven process monitoring approach is proposed for batch courses with multipahse characteristics. In a hot rolling mill process, there is a transition area between two batches of steel plates, which has strong time-varying characteristics. At present, the monitoring methods for the hot strip mill process usually only consider the rolling phase, and do not take the intermediate transition area into account. Firstly, a method based on summation of first-order differences is developed for multiphase partitioning. Secondly, a combined method of slow feature analysis and cointegration analysis enable to be introduced to monitor the faults in both transition phase and stable phase. The newly proposed method is applied to a real HRM process.The practical performance of the application demonstrates that the proposed methodology can accurately detect failure in the HRM process..