基于滑动窗的CVA故障诊断算法

CVA algorithm for fault diagnosis based on moving window

由于过程数据通常具有时变性,规范变量分析(CVA)在动态过程系统的故障诊断中不能得到较好的故障诊断准确率,因此提出一种基于滑动窗的规范变量分析(MWCVA)算法.该算法首先建立初始的CVA模型和计算监控统计量,通过滑动窗更新过程变量数据,计算更新建模所需数据,不断实时地更新出新样本的CVA模型和监控统计量.通过对Tennessee-Eastman过程的仿真,对比CVA、MWPCA和MWCVA的故障诊断效果,验证所提出算法的有效性.

Process data usually have time-varying characteristics,the fault diagnosis based on canonical variate analysis（CVA）cannot offer a better accuracy of fault diagnosis in dynamic process system.Therefore an algorithm of moving window-based canonical variate analysis（MWCVA）is proposed.First,the initial CVA model is set up for this algorithm and the monitoring statistical data are computed,then the data of process variable are updated with the moving window and the necessary data for model updating are figured out to update the CVA model of new sample and monitoring statistical data continuously and in real-time manner.The results of fault diagnosis of the CVA,MWPCA,and MWCVA are mutually compared among them by means of Tennessee-Eastman process simulation,and the effectiveness of the proposed method is verified.