Intermittent process mode recognition method based on Bayesian statistical analysis

Abstract

The invention discloses an intermittent process mode recognition method based on Bayesian statistical analysis, and belongs to the technical field of intermittent process monitoring. The method comprises the steps of firstly, expanding three-dimensional historical process data of an intermittent process into two-dimensional data along a batch method, and carrying out data standardization on the expanded data; secondly, performing clustering analysis on the standardized process data by utilizing a fuzzy C-mean clustering algorithm, setting a mode coarse division subordination rule, and obtaining a mode coarse division result; and finally, analyzing the mode coarse division result by utilizing a Bayesian network classifier, introducing a mode inference coefficient of a time sequence constraint, inferring a minimum risk criterion according to a mode, and judging final attribution of the mode, thereby realizing mode recognition of the intermittent process. According to the method, the timesequence constraint of intermittent process data is fully considered; effective division of a stable mode and a transition mode of the intermittent process is achieved through the Bayesian statistical analysis; and the method has relatively high mode recognition accuracy.

Description

technical field [0001] The invention relates to a batch process mode recognition method, which belongs to the technical field of batch process monitoring, in particular to a batch process mode recognition method based on Bayesian statistical analysis. Background technique [0002] As an important production mode in industrial production, batch process has multiple operating states and multi-modal characteristics, which makes the process characteristics of batch process different in different modes, and the correlation of variables is also significantly different. If the process data of different modes are modeled with the same model, it will lead to large modeling errors, which limits the application of the built model in the batch process. Therefore, it is necessary to accurately identify multiple modes with obvious differences in the batch process, so as to provide a basis for the monitoring and control optimization of the batch process. [0003] The existing batch proces...

AI Technical Summary

Problems solved by technology

Data-driven clustering analysis methods are widely used in batch process modal identification, such as K-means clustering method, fuzzy C-means clustering method, affine propagation clustering method, etc. During modal identification, the modal identification results are greatly affected by outliers in the process data, the iterative process has a large time complexity, and the timing constraints of intermittent process data are ignored, and parameter selection is difficult

Images