基于SOM-BP复合神经网络的齿轮故障诊断研究  

Study on Gear Fault Diagnosis Based on SOM-BP Composite Neural Network

作　　者：陈立爱 陈松 张舒 龚雪 Chen Li'ai;Chen Song;Zhang Shu;Gong Xue(College of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230601,China)

机构地区：[1]安徽建筑大学机械与电气工程学院,合肥230601

出　　处：《煤矿机械》2021年第7期187-190,共4页Coal Mine Machinery

基　　金：国家自然科学基金(51605169);安徽省重点研发计划项目(1804a0812091);安徽省自然科学基金(1908085QE243)。

摘　　要：采集齿轮4种常见状态的振动信号,采用MATLAB建立SOM神经网络、BP神经网络和SOM-BP复合神经网络模型并对模型进行训练,对齿轮故障进行分析。通过对比测试样本的实验结果,BP神经网络的识别准确率可达到94%,SOM神经网络的识别准确率可达到91%,SOM-BP复合神经网络的识别准确率可达到99%,证明了采用SOM-BP复合神经网络进行齿轮故障诊断的正确性和准确性。Acquired the gear vibration signals in 4 common states,established SOM neural network, BP neural network and SOM-BP composite neural network models and trained the models by MATLAB,analysed the gear fault diagnosis. Though comparing the experiment results of test samples, the recognition accuracy of BP neural network can reach 94%, that of SOM neural network can reach 91%,and that of SOM-BP composite neural network can reach 99%. It was proved to be correct and precise that use SOM-BP composite neural network in gear fault diagnosis.

关 键 词：齿轮故障 SOM网络 BP网络 SOM-BP复合网络 

分 类 号：TH132.41[机械工程—机械制造及自动化]