自动化学报2025，Vol.51Issue(11)：2520-2533,14.DOI:10.16383/j.aas.c250134
独立慢特征分析建模方法及其在动态故障检测中的应用
Independent-Slow Feature Analysis Modelling Method and Its Application in Dynamic Fault Detection
摘要
Abstract
Fault detection and diagnosis technologies serve as critical technical supports and effective means to ensure the normal operation of complex equipment or industrial processes.As a typical multivariate statistical process monitoring method,independent component analysis(ICA)can fully exploit high-order statistical informa-tion from data.Conventional ICA methods employ principal component analysis(PCA)for whitening and dimen-sionality reduction during the pre-processing stage.However,the static nature of PCA compromises ICA's effective-ness in dynamic process monitoring.To address this issue,an independent-slow feature analysis modelling method is approached.Specifically,ISFA constructs a dual-objective optimization function using the original observation mat-rix and whitening matrix as independent variables,solves the objective function via Newton's iteration method,op-timizes weight coefficients through grid search,modifies statistical metrics using exponentially weighted moving av-erage,and establishes a comprehensive detection index.Finally,numerical simulations and electric servo mechan-ism experiments are conducted to validate the effectiveness of the proposed method.关键词
独立慢特征/动态过程/故障检测/网格搜索Key words
Independent-slow feature/dynamic process/fault detection/grid search引用本文复制引用
张晨,孔祥玉,胡昌华..独立慢特征分析建模方法及其在动态故障检测中的应用[J].自动化学报,2025,51(11):2520-2533,14.基金项目
国家自然科学基金(62273354,62227814)资助Supported by National Natural Science Foundation of China(62273354,62227814) （62273354,62227814）
