摘要
往复压缩机振动信号特性复杂,传统特征提取方法难以有效提取故障特征,从而影响故障诊断效果。提出了基于原始振动信号卷积神经网络(RVCNN)的方法,将采集的一维原始振动信号作为输入,充分利用卷积神经网络(CNN)自动提取信号特征的特性,对往复压缩机故障进行特征提取及诊断。使用从试验台获得的压缩机气阀故障数据样本进行测试,结果表明,与传统方法相比,RVCNN方法具有更高的故障识别率和更好的抗噪性能。
The features of the reciprocating compressor vibration signal are complex, and the traditional feature extraction methods are difficult to extract the fault features effectively, affect the compressor fault diagnosis rate. It proposes a method based on the raw vibration signal convolutional neural network （RVCNN）. The acquired one-dimensional raw vibration signal is directly input, and RVCNN is used to automatically extract the feature of reciprocating compressor signal and diagnose the fault. Using the compressor valve fault data samples obtained from the test bench, the results show that the RVCNN method has higher fault recognition rate and better anti-noise performance than the traditional methods.
作者
杨洪柏
聂昂
张江安
张宏利
刘树林
Yang Hongbai;Nie Ang;Zhang Jiang＇an;Zhang Hongli;Liu Shulin(School of Science and Technology,Shanghai Open University,Shanghai,200433,China）（2.School of Mechatronic Engineering and Automation,Shanghai University,Shanghai,200072,China）（3.Advanced Vocational Technical College,Shanghai University ofEngineering Science,Shanghai,200437,China)
出处
《机械设计与制造工程》
2018年第9期67-70,共4页
Machine Design and Manufacturing Engineering
基金
上海开放大学2018年度学科研究课题(KX1805)
上海市智能制造及机器人重点实验室开放课题(ZK1801)
国家自然科学基金资助项目(51575331)
关键词
原始振动信号
往复压缩机
故障诊断
卷积神经网络
raw vibration signal
reciprocating compressor
fault diagnosis
convolutional neural network
