基于独立分量分析(ICA)与小波变换的过程监测方法

Process monitoring method based on independent componint analysis and wavelet transform

提出了一种ICA与小波变换技术相结合的过程监测方法。通过ICA方法分析出独立分量,经过小波分解后构造平均能量作为过程特征量。然后以相似度为监测指标实现过程监测。应用ICA方法比应用主分量(PCA)方法能更准确地提取非高斯分布信号信息,可以更加有效地实现对过程的监测。ICA能从原始的输入特征提取出更紧致、更适合后端处理的二次特征。由于二次特征能体现出数据中的本质信息,所以ICA方法相对于那些只考虑方差信息的特征提取方法有更好的性能。

A process monitoring method based on independent component analysis(ICA) and wavelet transform was presented. which used ICA to calculate independent component and wavelet decomposition to construct average energy as the process feature respectively. Then the process monitoring can be conducted by comparing similarity degree that was considered as a monitoring perofrmance index. ICA is more accurate than principle component analysis(PCA) in extraction of non-Gaussian distribution signal, and it can get second power of signal features that are more compact and suitable for post-end treatment from original input. Since these features can represent essential information in the input data, ICA method is better than the feature extraction mehods only by considering variance information.