Method for diagnosing failure of airplane generator bearing based on GentleBoost

Abstract

The invention discloses a method for diagnosing the fault of an airplane generator bearing based on GentleBoost. The method comprises the following steps of: creating an airplane generator bearing failure test platform, collecting vibration acceleration signals of a bearing under four different working conditions as training samples, forming a sample set, extracting 10 statistical characteristics of 256 time domain sampling points of each sample in the sample set, designing a binary weak classifier, enhancing the weak classifier by the adoption of a GentleBoost classification method, forming 6 pairs of pairwise binary classifiers aiming at the four working conditions of the rolling bearing, and forming a multi-type classifier according to a majority voting principle. The method for diagnosing the fault of the airplane generator bearing based on the GentleBoost, disclosed by the invention, has the advantages that the covering surface of failure characteristics is increased, the problem of a multi-classification generalized error of a characteristic overlaying area is avoided, the classification accuracy of a difficult-classification sample can be increased, and the higher failure identification rate of the airplane generator rolling bearing is obtained.

Description

technical field [0001] The invention relates to a fault diagnosis method for aircraft generator bearings. Background technique [0002] Rolling bearings are important components of aircraft generators. Due to the repeated long-term contact stress on the working surface, the inner ring, outer ring, rolling elements and cage of the bearing are prone to failures such as fatigue, pitting, peeling, and abrasions, resulting in The bearing broke, causing an accident. Therefore, the fault diagnosis of rolling bearings, especially the early fault detection and diagnosis of rolling bearings is very important. The so-called early failure refers to the fact that the failure has just started and before it causes harm to the performance and working state of the machinery, it is of great significance to identify the damage of the rolling bearing, judge its damage state and damage type, and ensure the normal operation of the aircraft generator. [0003] The patent "Rolling Bearing Fault A...

AI Technical Summary

Problems solved by technology

However, the characteristics of the neural network determine that it often has the disadvantages of too long training time and difficult selection of relevant parameters during the learning process, so it is not easy to realize in engineering

[0008] (3) "Bearing diagnostics" use acoustic signals for detection. This detection method is widely used today, but its disadvantage is that the sound signal is easily affected by external noise. This noise is superimposed on the original signal, so it is not easy to determine The eigenmode of the signal, therefore, it is not easy to accurately identify the fault

First of all, the Bayesian method requires a prior distribution. If the prior distribution is unknown, this method cannot be realized. Secondly, in the KNN method, the most critical parameter is the order of the classification function. If the order is selected small, the classification error will be biased. Large, if the order is too large, it will not only increase the complexity of calculation, but also cause overfitting

SVM is a kind of kernel method used by some diagnostic systems. The advantage of this method is that it can process small sample data, but the disadvantage is that the processing speed of data is slow, the selection of kernel function is also complicated, and there is no theoretical basis.

Images