Fault predicting method based on fuzzy nearness and particle filter

Abstract

The invention aims at the characteristics that equipment system often has uncertainty and nonlinearity, and provides a fault predicting method which combines fuzzy nearness and a particle filter algorithm. The method provided by the invention comprises the following steps: describing a fault evolution process of a system by adopting a state-space model, and tracking and predicting the operating state of the system by utilizing the particle filter algorithm; designing a normal fuzzy subset which describes normal operation of the system and an abnormal fuzzy subset which describes abnormal operation of the system by utilizing a membership function; calculating a normal membership set of a predicting sequence solved by the particle filter algorithm; calculating the nearness of the normal membership set and the normal fuzzy subset, and calculating the nearness of the normal membership set and the abnormal fuzzy subset. When a real-time monitoring system is operated, if the nearness of the normal membership set and the abnormal fuzzy subset is greater than that of the normal membership set and the normal fuzzy subset, a fault is predicted; the method can predict the fault early, and is an effect fault predicting method.

Description

technical field [0001] The invention relates to a fault prediction method based on fuzzy proximity and particle filtering. Background technique [0002] Early failure prediction of equipment and systems in operation is an important way to take countermeasures in advance and avoid losses caused by failures. The particle filter method provides an effective technique to solve the state estimation problem in continuous signal processing, and its main advantage is applicable when the system is nonlinear or has non-Gaussian noise. For the problem of predicting when equipment will fail, there is often an uncertain relationship between potential failures and failure symptoms, and judging whether a failure occurs is ambiguous, so fuzzy mathematical methods can be used to study. At present, researchers at home and abroad have used particle filter and fuzzy closeness in fuzzy mathematics independently in fault detection and prediction, such as particle filter method using randomly per...

AI Technical Summary

Problems solved by technology

It can be seen that the above fault prediction / diagnosis methods either use particle filter or fuzzy closeness, and there is no method of combining fuzzy closeness and particle filter for fault prediction.

Images