Prediction method for machine fault

Abstract

The invention discloses a prediction method for a machine fault. The method comprises the following steps of using plurality of kernel functions to create a multi-core combined relevance vector machine model cluster; respectively inputting the signal characteristics of a field signal, the signal characteristics of a normal signal and the signal characteristics of the fault signal of a machine into a multi-core combined relevance vector machine model cluster; generating a model particle set corresponding to the signal; performing initialization weight operation, iterative prediction operation,weight updating operation and resampling operation on each model particle in the various model particle sets; screening out model particles with the largest number of particles to serve as a multi-core combined relevance vector machine with optimal corresponding signals; and finally, generating a corresponding map according to the optimal multi-core combined correlation vector machine, and selecting the map with the highest similarity with the field signal map as a prediction result of the machine fault by judging the similarity between the field signal map and the normal signal map and the similarity between the field signal map and the fault signal map, thereby realizing fault early warning of the machine.

Description

technical field [0001] The invention relates to the technical field of machine equipment fault prediction, in particular to a machine fault prediction method. Background technique [0002] Power generation equipment is an important basis for generating power resources. The normal operation of power generation equipment can generate a steady stream of power resources and ensure the normal use of electricity for people. [0003] In the industry, the fault detection of power generation equipment generally adopts planned maintenance or after-event maintenance. There are problems such as maintenance redundancy and waste of resources, and it is difficult to achieve optimal benefits. As for the health of power generation equipment, there is currently no complete evaluation and prediction method, when a failure occurs, data comparison, troubleshooting and cause analysis are performed manually, which is time-consuming and labor-intensive. It is difficult to predict possible failures ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem that existing power generation equipment generally uses manual methods for fault detection, but cannot give early warning of possible faults in power generation equipment, the purpose of the present invention is to provide a multi-core group correlation vector machine model cluster, and The signal generated by the actual operation of the power generation equipment is input into this model cluster for calculation, and the output map is compared with the fault signal map and the normal signal map to realize the prediction method of equipment fault early warning

Images