Improved generalized predictive control in polymerization process

Abstract

The polymerization process is the basis stage of PAN based carbon fiber production, and its temperature control affects directly the quality and yield of the last products. However, the polymerization process releases a lot of heat rapidly and it has the serious time delay character. These make it is very complex to control the polymerization process. The paper firstly analyzes the polymerization process model, gives its CARIMA parametric equation, provides the original data excited by the generalized binary noise (GBN) signal, and identify the model using the recursive least square algorithm with fading memory. Secondly, the paper introduces an improved generalized predictive control (GPC) method, which has stronger fault tolerance and robustness with little process overshooting. At last, a system is realized in the cascade frame based on the control layer and the monitoring layer. The former realizes the model identification and the recursive computation in the improved generalized predictive control, and sends the results to the latter, and the latter realizes PID with dead-zone control of the mixed water temperature control using the results of the main regulator. The practice shows that the polymerize temperature cascade system runs well and has evident effect with effective control.