Failure prediction method based on nonlinear failure reconstruction

Abstract

The invention relates to a failure prediction method based on the nonlinear failure reconstruction. The method comprises the following steps that (1), a KPCA model is utilized for carrying out off-line nonlinear modeling on process data monitored when a rotary mechanical system operates, carrying out abnormity detection and extracting failure information, wherein the KPCA model is the kernel principal component analysis model; (2), the failure degree is quantificationally described through the KPCA model and the abnormity detection in the step (1), and failure estimation is achieved under the failure reconstruction through the optimization method; (3), the development tendency of the estimated failure amplitude f is predicted through the multi-layer hierarchical method. The failure prediction method based on the nonlinear failure reconstruction can effectively improve the failure prediction efficiency, reduce false alarm rate and omitted alarm rate and lay the foundation for further researching the prediction maintenance technology of a complex mechanical system. The failure prediction method based on the nonlinear failure reconstruction can also be widely applied to online monitoring systems of large electromechanical devices of petrochemical enterprises, metallurgy enterprises, coal enterprises and other many enterprises.

Description

technical field [0001] The invention relates to a fault prediction method, in particular to a fault prediction method based on nonlinear fault reconstruction in the technical field of mechanical fault detection. Background technique [0002] With the rapid development of science and technology and modern industry, the machinery, energy, petrochemical, transportation and national defense industries of the national economy are increasingly large-scale, high-speed, integrated and automated. Once an accident occurs in such a complex system, it will cause Huge property loss and casualties. Therefore, preventing safety issues caused by component performance degradation has become an important problem in the daily operation of complex systems. As an important part of predictive maintenance, fault prediction has been widely concerned by experts at home and abroad. [0003] In the complex industrial process environment, the traditional fault diagnosis technology based on the proces...

AI Technical Summary

Problems solved by technology

(2) Nonlinear problems

The result of traditionally ignoring nonlinear factors is often unsatisfactory. With the increasing requirements for fault detection and prediction, nonlinear problems that could be ignored are becoming more and more prominent.

Images