Industrial process monitoring method and application based on improved principal component tracking

Abstract

The invention discloses an industrial process monitoring method based on improved principal component tracking and its application, belonging to the technical field of industrial process monitoring and diagnosis. First, the principal component tracking method based on the low-rank matrix representation is used to decompose the industrial collected data to obtain a low-rank coefficient matrix that contains all the variable relationships of the process. Second, use the low-rank coefficient matrix and the correlation coefficient weights of the variables in the training matrix to construct L 2 Statistics for fault detection and identification. According to the principles of low-rank matrix representation and principal component tracking method, the present invention integrates low-rank matrix representation algorithm into principal component tracking, constructs a model of principal component tracking algorithm based on low-rank matrix representation, and uses this model for online monitoring , making full use of the correlation between variables in the training matrix and the effective information contained in the variables of the training matrix. The method of the invention has higher accuracy for the detection and identification of industrial faults with abnormal values.

Description

technical field [0001] The invention belongs to the field of industrial process monitoring and fault diagnosis, in particular to an online industrial fault detection and identification based on principal component tracking represented by a low-rank matrix, using L 2 Statistics. Background technique [0002] Industrial process production is the pillar industry of national economic development, so it is very important to ensure the efficiency and stability of the production process. Process monitoring is mainly divided into four steps: model building, fault detection, fault identification, and process reconstruction. Fault detection and identification is a critical step in process monitoring. [0003] Industrial process monitoring methods are divided into three categories: methods based on quantitative mathematical models, methods based on knowledge and methods based on data-driven. Compared with mechanism-based models and knowledge-based methods, data-driven process monito...

AI Technical Summary

Problems solved by technology

This method applies the principal component tracking method and uses the PCA statistics in the low-rank matrix for fault detection, but as mentioned above, the statistics of the PCA method are not suitable for direct application in the principal component tracking method

Images