Novel machine learning method for diagnosing gearbox fault based on multi-scale permutation entropy

Abstract

The invention relates to the technical field of motor gearbox predictive maintenance, discloses a novel machine learning method for diagnosing gearbox faults based on multi-scale permutation entropy, and solves the problems existing in a gearbox fault detection process in the current market by adopting a classical spectrum analysis method, a deep learning method and a machine learning method. The machine learning method for diagnosing the gearbox fault based on the multi-scale permutation entropy has the advantages that the diagnosis precision is improved based on the Adaboost classification model, a large amount of fault data does not need to be calibrated, the problem that permutation entropy multi-scale parameters are difficult to select is solved, finally, multiple pieces of data are fused through a majority voting method, the accuracy can be further improved, the method is simple and practical, and has strong innovativeness and novelty.

Description

technical field [0001] The present invention relates to the technical field of gearbox predictive maintenance, more specifically a classification method for accurately diagnosing gearbox faults, and in particular to a new machine learning method for diagnosing gearbox faults based on multi-scale permutation entropy. Background technique [0002] The gearbox of a large unit, such as a wind power gearbox, is a rotating mechanical system. Faults such as gear breakage, pitting, and wear are prone to occur during long-term operation. The fault diagnosis mainly includes classical spectrum analysis methods, fault mode recognition, etc. The input of wind turbine gearboxes is time-varying, causing their vibration signals to be unstable, and accompanied by strong background noise, these vibration signals are easily overwhelmed by noise. The structure of most gearboxes is relatively complex, and it is easy to have multiple faults at the same time. These fault signals may be coupled wit...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies in the prior art, the purpose of the present invention is to provide a new machine learning method for diagnosing gearbox faults based on multi-scale permutation entropy. Its advantage is that it improves the diagnostic accuracy based on the Adaboost classification model, without the need for a large number of calibration fault data, and Solved the problem of difficult selection of multi-scale parameters of permutation entropy. Finally, the accuracy rate can be further improved by fusing multiple data through the majority voting method

Images