基于鲁棒ICA-PCA的TE故障诊断

Robust ICA-PCA based TE process monitoring and fault diagnosis

针对复杂工业过程混合分布的问题,提出了基于鲁棒ICA-PCA(independent component analysis-principal component analysis)的故障诊断新方法。由于实际工业过程数据不可避免地带有大量干扰,为降低数据粗糙的影响,首先采用小波去噪算法提高建模数据质量;然后利用鲁棒ICA-PCA算法提取过程的非高斯和高斯信息,并构建了三个统计量进行故障的监控;最后将上述方法应用到田纳西—伊斯曼(Tennessee Eastman,TE)化工过程。仿真结果表明,相比于传统PCA算法、ICA-PCA等算法,鲁棒ICA-PCA方法能够有效地检测故障的发生,具有较好的鲁棒性和灵敏性。

This paper developed a robust new method of fault diagnosis based on independent component analysis-principalcomponent analysis (ICA-PCA) in chemical process, for complex industrial process hybrid distribution problems. In view ofthe practical industrial process data was inevitable with a large number of interference, first of all, it used wavelet denoising todeal with the real data for reducing the influence of outliers in the data. Then it established a robust ICA-PCA algorithm monitoringmodel. It applied the above method to the Tennessee Eastman (TE) chemical process and compared with the traditionalPCA algorithm, the algorithm of ICA-PCA, etc. The simulation results show that the proposed method has strong robustnessand sensitivity, can effectively detect the fault occurs.