基于小波时频图和卷积神经网络的行星齿轮箱故障诊断方法

Fault Diagnosis Method of Planetary Gearbox based on Wavelet Time-frequency Diagram and Convolutional Neural Network

针对行星齿轮箱故障诊断中故障类型难以区分的问题,提出了一种基于小波时频图和卷积神经网络相结合的行星齿轮箱故障诊断方法。首先,对原始信号进行连续小波变换,获取小波时频图;然后,对小波时频图进行统一处理和压缩,将处理好的小波时频图输入到卷积神经网络中进行分类识别,通过调整小波基函数和卷积神经网络参数,最终得到一个较为理想的诊断模型。试验证明,在训练集数据和测试集数据转速不同的情况下,该方法与BP神经网络相比,在诊断准确率和鲁棒性方面都有提升。该方法的研究为行星齿轮箱的故障诊断提供了参考。

Difficulties are always encountered when distinguish the fault types in the planetary gearbox diagnosis.A new fault diagnosis method with implementation of wavelet time-frequency diagram and convolutional neural network is proposed.At first,the continuous wavelet transform is used on the original signal to obtain the wavelet time-frequency diagrams.Secondly,the wavelet time-frequency diagrams are processed and compressed,the processed wavelet time-frequency diagrams are input into the convolutional neural network to classify and identify.Finally,the wavelet basis function and convolution neural network parameters are adjusted in order to get an ideal diagnosis model.Experimental results show that the proposed method has better diagnostic accuracy and robustness than the BP neural network when the speed of training set data and test set data is different.This approach provides a reference for planetary gearbox fault diagnosis.