基于独立元分析算法的过程故障检测方法

文章针对目前实际工业生产中变量不能严格服从高斯分布,且大量变量之间存有严重相关性的特点,运用ICA方法提取高维数据中独立的信号,在保留数据信息的前提下对噪声加以抑制。信号提取后分别构造监控统计量,实施过程监控和故障诊断,并利用独立元模型对CSTR仿真实时数据进行故障检测研究,仿真结果表明该方法能快速准确的检测到运行中发生的异常。

Process Fault Detection Method Based on Independent Component Analysis Algorithm

The actual industrial processing variables could not be strictly Gaussian distribution.There was a serious correlation between the large number of variables.Independent signals could be extracted from high-dimensional datas in the ICA method.Noise to be suppressed.Subjected to the provisions of retain the data messages.The extracted sign-als could be constructed the monitoring statistics,therefore Implement process monitoring and fault diagnosis.A simul-ation of CSTR Breal-time data was analysed.Simulation results showed that ICA method couldrapidly and accurately dete-cted the exception occurred in the production processes.