摘要
针对传统PCA故障检测算法的结果有定论不明确的缺陷,提出一种基于Q统计量分离的故障检测新方法,把Q统计量分为PVR和CVR,前者代表显著与主元有关的变量信息,后者代表与主元无明显关系的变量信息,再配合T^2统计量共同用于监测过程,检测效果更细致。将此方法结合基于累积方差贡献率(CPV)和复相关系数(MCC)确定过程监测模型主元数的新方法,监测β-甘露聚糖酶发酵工业的过程,与传统的PCA故障检测方法比较,仿真研究结果表明该算法能够确保主元空间(PCS)中的信息存量,充分刻画过程变化,有效识别正常工况变化与故障,正确检测微弱故障,提高过程监控的准确性。
Because results were indefinite in the performance monitoring of industry process, a new faulty detection approach based on Q statistics separation was proposed. Q statistic was separated into two parts representing Principal component related variable (PVR) and common variable (CVR) respectively, and it can detect adequately the change of process, when it was used with T^2 statistic in monitoring process. A new determination of principal component in PCA model by using cumulative percent variance (CPV) and multi- correlation coefficients (MCC) jointly was proposed to ensure enough information in principal component subspace because of disadvantages of biggish subjectivity and no considering faulty information using only cumulative percent variance. This new faulty detection approach was applied to a β-mannanase fermentation process monitoring. The simulation result showed that it can identify effectively the normal work status change and fault, detecting exactly feeble fault and enhancing the veracity of process monitor comparing with traditional faulty detection approach.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2008年第12期1537-1542,共6页
Computers and Applied Chemistry
基金
国家科技部"863"项目(2006AA10A301)