Monitoring the transition of industrial process with set-point control using k-mean clustering

In order to achieve set-point control for a highly inertial production process, it is important to distinguish the start and end of the transitional state due to set-point changes. In this paper, we propose a new method to identify transition states using clustering method for monitoring production processes with set-point control. First, the historical data of the set-point and measurement error data are collected and clustered to extract the historical data of steady-state and transition states. Next, we design the statistics and threshold of the monitoring model for identifying the transition and the steady state of the production process. Finally, based on the identification of the transition state, a monitoring scheme of the production process with set-point control is proposed. The effectiveness of our methodology is validated through applying it to the industrial processes. Both results demonstrate that the proposed method has high ability of identifying the transition due to the set-point changes and exhibits the good performance for monitoring the variation of the process behavior.