一种基于可预测元分析的故障诊断方法

A Fault Diagnosis Method Based on Forecastable Component Analysis

将可预测元分析(Fore CA)引入到过程监控中,通过选取合适的可预测元并构造能够反映系统运行状况的统计量对在线数据进行统计监控,克服了主元分析(PCA)方法假设数据服从高斯分布且无法反映系统动态时序特性的缺陷,能很好地描述工业过程的动态特性并进行故障检测。TE模型上的仿真结果证明了Fore CA在工业过程监控中的可行性与有效性。

The forecastable component analysis （ForeCA） was introduced into the industrial process monito- ring. Through selecting proper ForeCA and constructing the statistics which reflecting system＇ s operation state, the on-line data can be monitored to overcome drawbacks that principal component analysis （PCA） as- sumes the data to be in Gaussian distribution and incapable of displaying the system＇ s dynamic timing charac- teristics. It can better reflect the dynamic nature of industrial process and can be used for fault detection. Sim- ulation in Tennessee Eastman （TE） model proves ForeCA＇ s feasibility and effectiveness in the industrial process monitoring.