基于多向PLS方法的间歇过程质量预测

Quality Prediction of Batch Processes Based on Multiway PLS Method

间歇生产过程中,很多质量指标不能在线测量,通常是在反应结束后通过离线分析得到。本文应用多向部分最小二乘(MPLS)方法建立软测量模型,通过在线测量的过程变量对产品质量进行预测。针对异丁烯酸甲酯(MMA)聚合反应过程中的数据进行仿真,结果表明,MPLS方法能够在反应初期就有效的预测出最后的产品质量指标。

In batch processes, many quality variables cant be measured on-line and are usually obtained by laboratory analysis off-line after the completion of a batch. Based on multiway partial least squares (MPLS) method, a soft sensing model is built to predict the final quality variables using on-line measurements of the process variables. A MMA polymerization batch process is studied. The simulation results demonstrate that MPLS method can effectively predict the final quality variables from the process variables at the early stage of a batch.