基于独立成分分析与概率神经网络的滚动轴承故障识别方法的研究  

Research on Fault Identification for Rolling Bearing Based on Independent Component Analysis and Probabilistic Neural Network

作　　者：王宏[1] 徐长英[1] 邓芳明[2,3] WANG Hong XU Chang-ying DENG Fang-ming(Engineering Training Center, Nanchang Aviation University, Nanchang 330063, China Electrical and Electronic Engineering Institute, East China JiaoTong University, Nanchang 330013, China Institute of Electrical and Automation Engineering, Hefei University of Technology, Hefei 230009, China)

机构地区：[1]南昌航空大学工程训练中心,江西南昌330063 [2]华东交通大学电子与电气工程学院,江西南昌330063 [3]合肥工业大学电气与自动化工程学院,安徽合肥230009

出　　处：《仪表技术与传感器》2016年第12期161-164,共4页Instrument Technique and Sensor

摘　　要：为了提高滚动轴承故障诊断的准确性和适应性,文中提出快速独立成分分析(fast independent component analysis,FICA)和概率神经网络(probabilistic neural network,PNN)相结合的滚动轴承故障识别方法。首先,针对滚动轴承的故障振动信号非高斯特点,利用FICA算法提取出滚动轴承振动信号特征;其次,为了提高概率神经网络分类的适应性,采用正交最小二乘算法训练概率神经网络结构,基因算法优化概率神经网络参数。实验表明,该集合型FICA-OPNN故障诊断方法较传统概率神经网络(FICA-PNN)有更高的分类准确性和适应性。An ensemble approach based on fast independent component analysis （FICA） and probabilistic neural network was proposed to improve the accuracy and adaptability of rolling bearing fault detection.Firstly, the feature of the vibration signals of the rolling bearing, usually non-Gaussian, was extracted by the FICA algorithm.Then Orthogonal Least Squares method was adopted to train the probabilistic neural network structure and genetic algorithms optimize probabilistic neural network parameters to improve the classification adaptability of probabilistic neural network.The experimental results show that the accuracy and adaptability of classification by FICA-OPNN are better than that of traditional probabilistic neural network （FICA-PNN）.

关 键 词：滚动轴承 故障识别 振动信号 独立成分分析 概率神经网络 

分 类 号：TH17[机械工程—机械制造及自动化]