Multimodal industrial process modality identification and fault classification method

Abstract

The present invention discloses a multimodal industrial process modality identification and fault classification method. The method comprises: collecting industrial process historical normal data andfault data from different modalities; performing offline classification on the different modalities; collecting to-be-detected industrial process data; calculating two feature quantities of the to-be-detected sample according to the standard Euclidean distance: the minimum distance from the sum of the local density of the points to the high local density point; and determining the modality to which the to-be-detected sample belongs online according to the distribution of the two feature quantities. According to the method disclosed by the present invention, fault history data from different modalities is collected; offline classification is performed on the different faults; a to-be-detected fault sample is collected; and the fault type of the to-be-detected fault sample is determined online, so that without the prior knowledge, the modality and the fault type of the data can be identified, and without specifying the cluster center and the number of clusters when clustering, the amountof calculation is greatly reduced.

Description

technical field [0001] The invention belongs to the field of multi-modal industrial process mode identification and fault classification, and more specifically relates to a multi-modal industrial process mode identification and fault classification method. Background technique [0002] For a large-scale industrial system, due to the change of production strategy and production environment, it often presents the characteristics of multi-mode and multi-fault. Different models need to be established for different modes, so it is of great significance to conduct mode identification and fault classification for multi-mode industrial processes before modeling. [0003] The most widely used mode identification and classification methods are based on data-driven methods. There are two main methods, one is a clustering algorithm based on pure mathematics, and the other is modal identification based on the similarity between PCA (Principal Component Analysis) or PLS (Partial Least Sq...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the above defects or improvement needs of the prior art, the present invention provides a multimodal industrial process modal identification and fault classification method, thereby solving the problem between the current clustering algorithm based on pure mathematics and the PCA or PLS model. There are technical problems such as artificially setting the number of clusters for modal identification based on similarity, complex calculations, and inaccurate clustering results

Images