基于PCA多变量统计的故障检测与诊断

主元分析(PCA)是一种能够对过程生产进行监测和质量控制的有效方法,在保证数据信息丢失最少的情况下,大大降低了原始数据空间的维数。为了更好地进行故障检测与诊断,介绍了基于PCA多变量统计的故障检测与诊断,给出了广泛应用在多变量统计过程上的T2和Q(或SPE)统计量。利用PCA分析建模可以消除变量间的非线性关联,降低噪声影响。用田纳西-伊斯曼过程TEP(Tennessee-Eastman Process)平台产生仿真数据,并利用Matlab软件建立故障检测与诊断模型。通过T2和Q(或SPE)统计量与其阈值的判断,进行对系统的故障检测与诊断。实验表明,基于PCA的故障诊断方法能够对过程的非正常变化做出反应,也能较正确地找出发生故障的原因以及相应环节。

Fault Detection and Diagnosis Based on PCA for Multivariable Statistics

Principal component analysis(PCA) is an effective method to detect and diagnose fault.PCA offers T2 and Q(or SPE) statistical quantity.PCA is an effective method to monitor and control quality in process produce.In the case of the least lost data information,it can reduce the dimension of the data space largely.Analyzing and building model based on PCA can eliminate the non-linear in variable and reduce impact of noise.TEP(Tennessee-Eastman Process) is used to produce the emulational data.And the model of fault detection and diagnosis is built by Matlab software,then fault detection and fault diagnosis is implemented by comparing T~2and Q(or SPE) statistical quantity with it's threshold.The test result shows that using the model based PCA can find the fault's reason and corresponding tache accurately.