基于非负矩阵分解的多模态过程故障监测方法

针对传统的多元统计故障监测方法往往需要假设测量数据服从单一高斯分布的不足,提出了一种基于非负矩阵分解(NMF)的多模态故障监测方法。首先使用标准的NMF算法对训练集数据进行聚类,将多模态数据划分到各个模态中;然后使用稀疏性正交非负矩阵分解(SONMF)算法对各模态分别建模,同时构造监控统计量进行故障监测。将提出的基于非负矩阵分解的多模态故障监测方法应用于数值例子和TE过程的仿真结果表明,该方法能够及时有效地检测出多模态过程中的故障。

Fault detection method based on non-negative matrix factorization for multimode processes

The traditional multivariate statistical fault detection methods are designed for single operating conditions and may produce erroneous conclusions if they are used for the multi-mode process monitoring. A novel multi-mode process monitoring approach based on non-negative matrix factorization (NMF) is proposed in this paper. First, the training set of data is clustered by the standard NMF algorithm and the multi-mode data are divided into each mode. Then, the sparseness orthogonal NMF (SONMF) algorithm is used to model every mode and the monitoring statistics are constructed to perform fault detection. The proposed method is applied to a numerical example and the TE process. The simulation results show that this method can effectively detect multi-mode process failure.