Multi-PCA模型过程监测方法

为了研究具有大量高度相关的过程变量的非线性系统的故障诊断问题,提高用于故障检测和诊断的PCA模型的精度,提出一种基于多PCA模型的方法.设计的基于超椭球面的分类规则用来对过程数据分类,建立的多PCA模型用于过程监测,SOFM网络用于故障诊断.发酵过程中的仿真结果表明,多PCA模型方法能确定合理的受控限,提高了过程监测的精度,验证了方法的可行性和有效性.

A Process Monitoring Method Based on Multi-PCA Models

In order to solve the problem of fault diagnosis for nonlinear systems with correlative process variables and improve the precision of PCA models for fault detection and fault diagnosis, a fault diagnosis method based on multi-PCA models is presented. Hyper-ellipsoid bound clustering rules are adopted to classify the process data, multi-PCA models are then built up for process monitoring. SOFM network is used in fault diagnosis. Simulation results in fermentation process show that the method can give reasonable control limits and improve the precision in process monitoring, which illustrates the feasibility and effectiveness of the proposed method.