基于小波变换的多尺度独立元分析方法及应用

Multi-scale Independent Component Analysis Based on Wavelet Transform and Its Application

实际过程数据大多不满足正态分布的条件,且大多数过程监控方法对数据进行分析的尺度较为单一.为此,本文提出了一种基于小波变换的多尺度独立元分析的过程监控方法.该方法对初始数据进行多尺度细化分析,并根据信息最大化准则提取独立元信号,在数据的低维子空间上对过程进行实时监控.通过对TE过程的仿真研究,表明了该方法的有效性.*

According to the condition that most practical process data can not meet the needs of normal distribution and many process monitoring methods analyze data at a single scale, this paper presents an improved process monitoring method named as multi-scale independent component analysis （MSICA） based on wavelet transform. The method analyzes initial data at different scales carefully, extracts independent signals according to the information maximization criterion, and monitors the process in real time in a low-dimensional subspace of data. Results of the simulation in Tennessee-Eastman （TE） process verify the efficiency of the method.