基于鲁棒尺度的统计建模数据中异常点去除算法的研究及应用

针对基于主元分析 (PCA)的统计监控模型受到历史数据中异常点强烈影响的不足，鉴于建模历史数据中存在的异常点会影响过程监控效果，分析目前常用的鲁棒异常值检测算法原理及其缺陷，提出将中心最短距离(CDC)法与椭球多变量整理(MVT)法相结合，构成一种基于鲁棒尺度的CDC-MVT异常值综合检测算法，更加准确地检测异常点。将该算法应用于工业发酵过程，与CDC法和MVT法相比较，该算法能够有效去除建模数据中的异常点。

Study and application of method of extracting outliers in statistical monitoring model based on robust scaling

Because statistical monitoring model based on PCA is strongly affected by outlying observations. The outlier in historical data acquired from industry process can decrease ability of process performance monitoring. This paper proposed a new outlier detection combined method based on robust scaling closest distance to center (CDC) and ellipsoidal multivariate trimming (MVT) after a summary on principle and limitation of robust outlier detection method. Applied the algorithm to extract outliers from a fermentation process and compared with the CDC and MVT outlier detection algorithms. The application results show that the proposed algorithm can effectively extract the outliers from the modeling database.