International Open Access Journal Platform

基于进化神经网络的滚动轴承故障预测研究

Research on Rolling Bearing Fault Prediction Based on Evolutionary Neural Network

Mechanical Research / 2019,1(1): 1-19 / 2019-09-02 6579 9688

View Download PDF

• Authors: 李云飞      葛江华*      郑智杰      郭海乐      牛阳阳      徐海锋     
• Information:

  哈尔滨理工大学，哈尔滨

• Keywords: 滚动轴承；故障预测；遗传算法；均方根；进化神经网络
• Rolling bearing; Fault prediction; Genetic algorithm; Root mean square; Evolutionary neural network
• Abstract: 在针对传统的神经网络在故障预测过程中，存在学习效率低，网络权值难以确定且容易陷入局部极值的问题，本文提出基于进化神经网络的滚动轴承故障预测方法。首先，提出Otsu-EWT信号降噪方法，最大类间方差法（Otsu）是通过计算目标与背景的类间方差并将最大值作为图像划分阈值的标准，EWT能对频谱图频谱区间进行自适应划分，在各划分的区间频率带上构建相应的带通滤波器进行降噪；其次，根据均方根指标对早期故障的敏感性，对均方根指标进行故障特征提取；最后应用遗传算法的具有良好的全局搜索能力和BP神经网络的局部搜索能力相结合，提出基于进化神经网络的滚动轴承故障预测方法，实验结果表明，本方法能够有效地对轴承的故障进行预测。
• In the process of fault prediction for traditional neural networks, there is a problem that the learning efficiency is low, the network weight is difficult to determine and it is easy to fall into the local extremum. This paper proposes a rolling bearing fault prediction method based on evolutionary neural network. The Otsu-EWT signal denoising method is proposed. The maximum inter-class variance method (Otsu) is to calculate the inter-class variance of the target and the background and use the maximum value as the threshold for image partitioning. EWT can adaptively divide the spectral range of the spectrogram. The corresponding bandpass filter is constructed on each divided frequency band to reduce noise; the root mean square index is sensitive to early faults, and the feature extraction is performed by the root mean square index; the genetic algorithm has good global search ability and BP. Based on the local search ability of neural network, a prediction method of rolling bearing fault based on evolutionary neural network is proposed. Experimental results show that this method can effectively predict bearing faults.
• DOI: https://doi.org/10.35534/mr.0101001
• Cite: 李云飞，葛江华，郑智杰，等．基于进化神经网络的滚动轴承故障预测研究［J］. 机械研究，2019，1（1）：1-19.
  https://doi.org/10.35534/mr.0101001