A Feature Reduction Method for Rotating Machinery Faults Based on Feature Evidence Discretization

Abstract

The invention provides a rotary machine fault feature reduction method based on feature evidence discretization. The method comprises the steps of firstly combining original measurement data by means of K-means for obtaining T data clusters, modeling each data cluster by means of a random fuzzy variable model, performing matching for obtaining the evidence of each measured value, deciding a discrete value after measured value conversion according to a decision rule and obtaining a decision table, and hereon a fault characteristic is a condition attribute, and the fault mode of a sample is a decision attribute; calculating a compression binary matrix for acquiring a core attribute, respectively calculating authenticity reliability of a union set of residual condition attributes and the core attribute, adding the condition attribute which corresponds with the union set with least authenticity reliability into the core attribute set until the least authenticity reliability is 1, and furthermore obtaining a final reduction result. According to the method of the invention, in discretization processing, boundary fuzziness between the data clusters is considered. Compared with a method in which only K-means is utilized, the method according to the invention has advantages of realizing higher accuracy in discretization processing, and reducing calculation amount and storage space through compressing the binary matrix.

Description

Technical field [0001] The invention relates to a rotating machinery fault feature reduction method based on feature evidence discretization, which is applied to the fault diagnosis of rotating machinery equipment, and belongs to the technical field of rotating machinery equipment fault monitoring and diagnosis. Background technique [0002] Fault diagnosis is an important method to improve the reliability of the system and reduce the risk of accidents. However, the monitoring and diagnosis technology based on single sensor has been unable to obtain enough reliable information for the diagnosis requirements of large-scale equipment with increasingly complex structures. With the development of sensor technology and the improvement of computer storage and computing capabilities, a large amount of monitoring data can be obtained through the multi-sensor system installed on the device, but if all the monitoring data is used for fault diagnosis, it will cause excessive calculation. It...

AI Technical Summary

Problems solved by technology

[0002] Fault diagnosis is an important method to improve system reliability and reduce the risk of accidents. However, monitoring and diagnosis technology based on single sensor cannot obtain enough reliable information for the diagnosis needs of large-scale equipment with increasingly complex structures.

With the development of sensor technology and the improvement of computer storage and computing power, a large amount of monitoring data can be obtained through the multi-sensor system installed on the equipment, but if all the monitoring data are used for fault diagnosis, it will cause excessive calculation. Large, seriously affecting the real-time performance of diagnosis

[0003] Due to the uncertainty of the occurrence of faults and the complex causes of faults, usually the same fault can exhibit multiple features, and the same fault feature may be caused by different faults; and these fault features are imprecise, irregular, redundant Sex and other characteristics

Images