摘要
针对滚动轴承不同运行状态振动信号具有不同复杂性的特点,提出一种新的基于多尺度熵(multiscale entropy,MSE)和概率神经网络(probabilistic neural networks,PNN)的滚动轴承故障诊断方法。该方法首先利用MSE方法对滚动轴承振动信号进行特征提取,并将其作为PNN神经网络的输入,再利用PNN自动识别轴承故障类型及故障程度。实验数据包括不同故障类型和不同故障程度样本,结果表明,相比于小波包分解和PNN结合的诊断方法,提出的方法具有更高的诊断精度,能有效实现滚动轴承故障类型及程度的诊断。
Considering different levels of complexity of vibration signals of rolling bearings in different operatingconditions, a novel fault diagnosis method has been proposed based on the multiscale entropy （MSE） and probabilisticneural networks （PNN）. Fault feature vector is firstly extracted from the vibration signals using MSE and then provided toPNN neural network as the input. The PNN network will identify the bearing fault type and severity level simultaneously.The experimental data are collected from an induction motor bearing involving various fault types and severity levels. Theresults demonstrate that the proposed method has a higher accuracy in rolling bearing fault diagnosis than the method of thecombination of wavelet packet decomposition with PNN.
出处
《噪声与振动控制》
CSCD
2014年第6期169-173,共5页
Noise and Vibration Control
基金
国家自然科学基金资助项目(51205130
51265010)
江西省教育厅科技项目(GJJ12318)
江西省自然科学基金项目(20132BAB216029)
作者简介
张磊（1989-）,男,湖北孝感人,硕士研究生,主要研究方向：工程信号处理与机械故障诊断.
熊国良,男,博士生导师.E-mail：lzhang0712@126.com
