Research On Monitoring Method For Batch Process Based On Phase Division

Batch process is an important way of modern industrial production. Due to the characteristics of product diversification, flexible operation, small equipment space etc, batch process is widely used in food, biopharmaceutical, medicine, polymer and other fields. At the same time, in order to guarantee the safe in operation and quality of product, it is necessary to monitor the production process real-time. In the existing monitoring methods, batch process is modeled as a whole in which all the time have the same characteristics, which leads to alarm delay, as well as false alarm and miss alarm. Multi-phase is the inherent characteristic of batch process, and the process property is similar in one phase, therefore, it has important theoretical meaning and application value to research on phase division and monitoring method based on phase for batch process.In this paper, phase is divided using MPCA and a transition range identification method using the first principal component contribution rate is proposed based on the analysis of the batch process characteristics, which successfully realize the identification of transition range effective. On this basis, according to the different property of phase and transition, a monitoring strategy that steady phase modeled as a whole and transition modeled by the moving window is proposed, and the strategy is applied to the monitoring for batch process based on KICA. Finally, the monitoring method based on phase and transition KICA is confirmationed by applying to the monitoring for penicillin fermentation process.The experimental results show that the first principal component contribution rate can reveal the different characteristics of adjacent time effective, and reflect the different property of phase and transition, which identify the transition more accurate. The monitoring strategy based on phase and transition can improve the accuracy of the stable phase model, and compromise the stability and accuracy of the transition model. Meanwhile, the KICA monitoring method based on phase and transition can significantly reduce the false alarm, detect the fault more timely and accurate.