基于MSPM的故障诊断技术研究现状与展望  

Research Status and Prospect of Fault Diagnosis Technology Based on MSPM

作　　者：马洁[1] 党爱民 李刚[3] 周东华[3] 

机构地区：[1]北京信息科技大学自动化学院,北京100192 [2]洛阳双瑞精铸钛业有限公司,河南洛阳471003 [3]清华大学自动化系,北京100084

出　　处：《华侨大学学报（自然科学版）》2012年第6期601-607,共7页Journal of Huaqiao University(Natural Science)

基　　金：国家自然科学基金资助项目(61273173;61028010;61021063);北京市自然科学基金资助项目(4122029)

摘　　要：首先,阐述基于主元分析(PCA)模型、偏最小二乘法(PLS)模型和独立分量分析(ICA)模型的统计过程监控方法的基本思想及应用情况,并综述各种方法的研究现状及发展趋势.其次,介绍将传统统计过程监控技术与故障预测技术相结合,并实现基于多元统计过程监控(MSPM)的故障预测的方法及其研究成果.最后,分别就多元故障预测技术中出现的非高斯、非线性、多模态、概率分布、间歇过程的故障预测和应用验证等6个难点问题进行讨论.First of all,this paper introduces the basic ideas and applications of statistical process monitoring method based on principal component analysis（PCA） model,partial least squares（PLS） model and independent component analysis（ICA） model.The present research situation and development trend about various methods are reviewed.Secondly,by combining fault prediction technology with the traditional statistical process monitoring technology,fault prediction method based on multivariate statistical process monitoring（MSPM） can be realized.And some research results are also introduced.Finally,six difficult problems in multivariate failure prediction technology such as non-Gaussian,non-linear,multi-modal.probability distribution,intermittent process fault prediction and application verification are discussed respectively.

关 键 词：多元统计过程监控 故障诊断 故障预测 主元分析 偏最小二乘法 独立分量分析 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]