基于多元统计投影方法的过程监控技术研究  

Process monitoring based on multivariate statistical projection analysis

作　　者：陈国金[1] 梁军[1] 刘育明[1] 钱积新[1] 

机构地区：[1]浙江大学工业控制技术国家重点实验室,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2004年第12期1561-1565,共5页Journal of Zhejiang University：Engineering Science

基　　金：国家"863"高技术研究发展计划资助项目(2001AA411230).

摘　　要：为克服传统过程监控方法需假设过程特征信号服从多元正态分布的缺陷,提出了一种新的基于独立成分分析(ICA)和主元分析(PCA)的过程监控方法,该方法由两步组成:第一步:利用独立成分分析方法从过程信息中提取非正态分布特征信号,然后用Parzen窗法估计其概率密度确定控制限进行过程监控;第二步:利用主元分析方法对剩余过程信息提取正态分布特征信号,采用Q和HotellingT2统计量对此正态特征信号进行过程监控.通过对双效蒸发器进料浓度和加热蒸气温度发生异常的两种故障模式仿真研究表明,该方法比传统多元统计过程控制具有更少的漏报率.To overcome the shortcoming of the conventional process monitoring methods' assumption that the extracted features must be subject to multivariate normal distribution, a novel method based on independent component analysis (ICA) and principal component analysis (PCA) was presented for process performance monitoring by using a two-step procedure. Step I: Process operating information with non-normal distribution was extracted by means of ICA, and the density function of this part of the information was estimated by means of Parzen density estimator for calculating the confidence limits; Step H: The information with normal distribution was extracted from the underlying residual datasets by PCA, and the confidence limits were determined for Q and Retelling T2 statistic. With the advantage that no assumption of normal distribution on process datasets is needed, application of the proposed method to a double-effect evaporator yielded has less missing alarms than the conventional process monitoring methods.

关 键 词：独立成分分析 主元分析 性能监控 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]