摘要
为了解决齿根疲劳裂纹故障难以识别的问题,对齿轮箱正常和裂纹故障状态的声发射信号进行时间序列分析,利用AR模型的自回归系数作为齿轮箱不同状态时的特征向量,形成支持向量机的训练样本对支持向量机进行网络训练,实现对齿轮箱正常、轻微裂纹和严重裂纹故障状态的识别与诊断。实验结果表明:基于支持向量机和声发射技术的齿轮箱故障诊断系统能够准确地识别与诊断齿轮箱的裂纹故障状态,它对于齿轮裂纹故障检测是一种有效的诊断手段。
In order to correctly identify fault classes of gear crack,a gear crack fault diagnosis method was established based on time series analysis and support vector machine.Firstly,the AE signals from the normal and crack gears were analyzed through the time series analysis respectively.Then the AR model′s autoregressive coefficients were obtained which were inputs of support vector machine for neural networks training.Finally,the identification and diagnosis of gears in normal state,slight crack fault and severe crack fault states were accomplished.The experimental results indicate the methods based on time series analysis and support vector machine are effective for monitoring the gear crack fault.
出处
《海军工程大学学报》
CAS
北大核心
2010年第2期70-73,112,共5页
Journal of Naval University of Engineering
基金
湖北省自然科学基金资助项目(2006ABA011)
关键词
故障诊断
齿轮裂纹
声发射
时序分析
支持向量机
fault diagnosis
gear crack
acoustic emission
time series analysis
support vector machine
