Industrial processes are often operated under different modes. However, most of the multivariate statistical process monitoring (MSPM) methods, such as principal component analysis (PCA) which are effective in single mode process, do not perform well in multimode process. A novel multimode fault detection approach named distance space statistics analysis (DSSA) was proposed. First, every sample was represented by the deviations of its κ-nearest distances between itself and its neighbor in the training data. All the samples were mapped from the original space into the distance space by this way. Then, different order statistics of the distance samples in a moving window were calculated in the distance space. Finally, principal component analysis (PCA) was used to analyze the new statistics samples. The proposed method, PCA method and a multimode fault detection method using κ-nearest neighbor rule (FD-kNN) were applied to the Tennessee Eastman (TE) benchmark process. The comparison of monitoring results showed that the proposed method was superior to the PCA and FD-kNN for fault detection of the multimode process.%工业过程往往运行于多个生产模态,针对多模态过程数据的空间分布特点,提出了一种新的基于样本距离空间统计量分析的故障检测方法(DSSA).首先用每一个样本与其训练集样本中的邻居之间的k个最近邻距离之差来表示该样本,将样本从原始变量空间映射到对应的距离空间中.然后在距离空间中通过移动窗口的方式计算各阶统计量,最后对由各阶统计量组成的统计量样本进行主元分析(PCA).将DSSA方法、PCA方法以及另一种基于k近邻规则的多模态故障检测方法(FD-kNN)应用于TE过程中,仿真结果表明DSSA方法对多模态故障检测更为有效.
展开▼
