摘要
针对重构贡献(RBC)方法仅适合单变量故障的定位及贡献图中出现拖尾效应(SE)的问题,本文提出一种基于改进重构贡献图(MRBCP)的故障定位方法。采用概率主元分析(PPCA)建立监视模型和统一量度的监视统计量,克服PCA方法中不同量度的监视统计量造成的诊断结果不一致的缺点。对于故障样本,以变量的重构监视统计量为贡献统计量,通过组合最大化思想对故障变量进行逐次定位。在历史故障信息未知的情况下,能够进行多变量故障的定位,然后在定位出的故障变量中进行贡献图分析,进一步对故障变量实现准确定位,从而避免了拖尾效应。通过数值案例和TE过程——实际化工过程的真实模拟过程进行实验,并与基本RBC方法、基于PCA的MRBCP方法进行比较,结果表明了所提方法的有效性。
Reconstruction-based contribution（ RBC） method is only suitable for univariate fault with the shortage of smearing effect in contribution plots. A modified reconstruction-based contribution plots（ MRBCP） method for isolation is proposed. The process monitoring model is established using probabilistic principal component analysis（ PPCA） and a unified monitoring statistic is adopted,avoiding the shortage of different diagnosis result as two kinds of statistics in PCA method. Variable＇s monitoring statistic as the contribution statistic is reconstructed,and the fault variable with the thought of combinatorial maximization is isolated. Firstly the multivariate fault is isolated in condition that the historical faulty information is unavailable. Then a contribution plots analysis within the selected faulty variables is carried out for further isolation,and the smearing effect is eliminated. A numerical example and the Tennessee Eastman Process（ TEP）is provided for simulation. Basic RBC and MRBCP based on PCA methods are compared with the presented work,which proves the effectiveness of the proposed method.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2015年第5期1193-1200,共8页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61034006)
国家自然科学基金(60774070
61174119)
辽宁省教育厅科学研究一般项目(L2013155)
辽宁省博士启动基金(20131089)项目资助
关键词
故障定位
概率主元分析
拖尾效应
改进重构贡献图
fault isolation probabilistic principal component analysis（PPCA） smearing effect modified reconstruction-based contribution plots（MRBCP）
