Multimode process monitoring method based on local and structural consistency manifold GMM

Abstract

The invention discloses a multimode process monitoring method based on local and structural consistency manifold GMM. The GMM is utilized to divide multi-mode data into multiple local data blocks, local tangent spaces of local data blocks are analyzed, a main role among the local tangent spaces is calculated, the similarity relationship among the local data blocks is studied, local and structuralconsistency manifold maps are constructed, according to the structural relationship of sub manifolds corresponding different modes, process mode changes are obtained, the consistency manifold smoothing monitoring statistical magnitude is designed, and fault detection is implemented. Compared with a general manifold GMM monitoring method, according to the data blocks corresponding to different Gaussian components, the similarity of the local tangent spaces is obtained, the local and structural consistency information is merged to construct a manifold learning map, the relationship among the multiple sub manifolds is analyzed, so incorrect division of the cross-overlapping data blocks is avoided, the geometry of multi-mode process data in the low-dimensional embedded space is accurately described, accurate fault detection is implemented, and the better monitoring result is achieved.

Description

technical field [0001] The invention belongs to the field of industrial process monitoring, in particular to a multi-modal process monitoring method based on local and structural consistency manifold GMM. Background technique [0002] Process monitoring in modern industry plays a pivotal role in ensuring production safety and increasing production. With the development of distributed control systems, the scale of production and the complexity of operations have increased dramatically, and a large amount of high-dimensional data has been collected during the process. Moreover, because the product grade and output will be constantly adjusted with market demand and seasonal effects, and process parameters such as product composition, process settings, and feed ratios will also fluctuate, modern industrial processes will operate in multiple different operating modes. switch between states. These random changes in the production process make the process data present characteris...

AI Technical Summary

Problems solved by technology

However, the graph constructed based on the Euclidean distance is easy to mistake samples close to each other in different sub-manifolds or overlapping overlapping manifolds as neighbors, which will build an inaccurate monitoring model

Images