Hierarchical Fault Monitoring Method Based on Hybrid Characteristic Evaluation and Subspace Decomposition

Abstract

The invention relates to a hierarchical fault monitoring method based on mixed characteristic evaluation and subspace decomposition, which comprises the use of mixed characteristic evaluation to obtain mixed characteristic subspaces of complex industrial processes, including Gaussian linear subspaces, Gaussian nonlinear subspaces, and non-Gaussian linear subspaces and non-Gaussian nonlinear subspace, based on PCA‑ICA‑KPCA‑KICA hierarchical subspace decomposition method to establish a fault monitoring model, using comprehensive statistics and hierarchical monitoring strategy for fault monitoring. The present invention combines the Omnibus test, weighted nonlinear measurement and hierarchical subspace decomposition based on PCA-ICA-KPCA-KICA, considers the coexistence of mixed characteristics of complex industrial processes, and overcomes the existing failure monitoring methods that rely on prior process knowledge or do not Considering the limitations of Gaussianity, non-Gaussianity, linear correlation and nonlinear correlation, etc., has theoretical and practical significance for monitoring abnormal working conditions and improving product quality.

Description

technical field [0001] The invention belongs to the technical field of fault monitoring and diagnosis, and specifically relates to a hierarchical fault monitoring method based on mixed characteristic evaluation and subspace decomposition. Background technique [0002] The wide application of intelligent manufacturing technology has significantly improved the quality and production efficiency of industrial products, but also increased the scale and complexity of industrial processes. Actual industrial processes often have complex and mixed characteristics. Fault monitoring is an effective means to discover abnormal working conditions of industrial processes in time, ensure production safety and improve product quality. However, the existing fault monitoring methods usually rely on prior process knowledge or do not consider the coexistence of Gaussian, non-Gaussian, linear correlation and nonlinear correlation. Therefore, in order to ensure production safety and improve produc...

AI Technical Summary

Problems solved by technology

The proposed method utilizes existing complex industrial process data resources, considers the coexistence of mixed characteristics of complex industrial processes, and overcomes the existing fault monitoring methods that rely on prior process knowledge or do not consider Gaussian, non-Gaussian, linear correlation and non-gaussian. There are limitations such as the coexistence of linear correlation, which has theoretical and practical significance for monitoring abnormal working conditions and improving product quality

Images