Distributed intelligent fault diagnosis method and device based on particle filtering

Abstract

The invention discloses a distributed intelligent fault diagnosis method and device based on particle filtering. The technical scheme comprises the steps of building a normal model and various fault models according to system dynamics characteristics; obtaining various data parameters through distributed data acquisition; accurately estimating system state parameters by adopting a particle filtering algorithm; and comparing the filtering processed data parameters with the normal model and the various fault models by applying a mode recognition algorithm to obtain an operating mode of the system at present. The advantages of the invention lie in that the acquisition and centralized processing of multi-variable data are realized through a sensor network, and accurate estimation for the stateof a monitored object and intelligent diagnosis for a fault are achieved based on the particle filtering algorithm. The system can realize remote real-time monitoring and fault prediction, and effectively improves the application range and diagnosis level of the existing fault diagnosis system.

Description

technical field [0001] The invention belongs to the field of fault diagnosis, in particular to a particle filter-based distributed intelligent fault diagnosis method and a device thereof. Background technique [0002] In modern industry, most monitoring and control mechanisms are built on the assumption that the state of the system is precisely measurable. However, it is often difficult to accurately obtain the internal state of the system through sensors. The actual measurement system has random errors, so the measurement vector also contains random quantities, and the true value of the state cannot be directly obtained through the ideal measurement equation. [0003] As the most important state estimation tool, the filter has experienced the development process from non-recursive to recursive, frequency domain to time domain, non-stationary random process to state space model. Nowadays, there are many filtering algorithms for state estimation, the most typical ones are: K...

AI Technical Summary

Problems solved by technology

[0007] In terms of pattern recognition algorithms for fault diagnosis, there are currently methods such as neural networks and support vector machines. Yang Yu and others proposed a fault diagnosis method based on EMD (Empirical Mode Decomposition) and VPMCD (Variable Predictive Modelbased Class Discriminate) in rolling bearings. Application, avoiding the structure and type of neural network and the selection of support vector machine kernel function and its parameters, but it is more complicated and has a large amount of calculation, so it cannot be well applied in systems with high real-time requirements

[0008] In addition, the Chinese patent application number CN201510031512.9 discloses a rolling bearing fault diagnosis method using particle filter and spectral kurtosis, which also has the problem of complex diagnosis method and large amount of calculation

Images