Layered non-Gaussian process monitoring method based on common and specific feature extraction

Abstract

The invention discloses a layered non-Gaussian process monitoring method based on public and specific feature extraction. Dividing the training data into a plurality of modes, and obtaining high-ordercommon features of the training data by using a plurality of feature quantities such as weight vectors and component vectors in each mode; obtaining low-order common features of the training data byusing a plurality of feature quantities such as weight vectors, component vectors and the like in the remaining low-order modals; constructing a statistical limit and a statistical magnitude in the common subspace according to the common features, and performing fault detection; and constructing a statistical limit and a statistical quantity in the remaining specific subspace, and performing faultdetection. Compared with other traditional methods, the multi-mode non-Gaussian process fault detection method is superior to other traditional methods, multi-mode specific features can be extracted.Meanwhile, common features can be extracted, and the mutual relation of the features and universality between different modes is considered, so that multi-mode process monitoring is more effective.

Description

technical field [0001] The invention belongs to the field of industrial process system engineering and relates to a multi-mode fault monitoring method applied to complex industrial processes such as chemical production. Background technique [0002] The complexity of complex industrial processes is reflected in the fact that the process of modern industrial production will generate a large amount of data. These process data usually have different process characteristics, and the distribution characteristics are unknown; the industrial process has multiple units, and the data has a mechanism relationship; and the traditional method It is mostly assumed that the chemical process operates under a single and stable operating condition, but in fact the plant-level data is multi-modal, so how to better monitor multi-modal issues is worth studying. [0003] In a large number of studies in the past, researchers have conducted a lot of research on advanced data classification and inf...

AI Technical Summary

Problems solved by technology

[0004]In order to overcome the traditional multi-modal process fault monitoring in the process of multi-modal monitoring, only the unique characteristics of each mode data are developed, ignoring the different The problem of the interrelationship of characteristics and generality between patterns, the purpose of the invention is to propose a layered non-Gaussian process monitoring method based on public and unique feature extraction, and use the data collected by sensors in industrial production to carry out real-time analysis and processing Obtain real-time detection and monitoring of fault status, which solves the technical problem of poor monitoring effect on multi-modal processes in the prior art

Images