基于云加端的电机轴承故障诊断应用研究

Application of motor bearing fault diagnosis based on cloud and terminal

针对以往的故障诊断系统实时性差、学习能力有限而且工程实现难的问题,提出了一种改进的云加端支持向量机(Cloud and Terminal Support Vector Machines,CaTSVM),并将其运用在电机轴承故障诊断中。CaTSVM方法把传统的故障诊断中的特征提取和特征分类两部分分别运行在终端设备和云端设备中,并且将"流水线"(Pipeline)数据处理结构引入到CaTSVM方法中,有效提升了该方法的实时性。在云端建立故障特征模型库(Cloud Feature Mode Library,CFML),将故障特征选择性的加入模型库,在传统的离线SVM训练中辅以在线SVM训练,选择性的使用更新的故障特征训练SVM模型,进一步提高其分类能力,使诊断系统拥有了"终生学习"的能力。经过大量的实验验证,云加端方法的使用显著提高了诊断的准确率,并且推进了故障诊断的实际工程应用。

Aiming at problems of previous fault diagnosis systems’ being poor to respond in real-time, limited to learn new knowledge and difficult to implement in engineering, an improved cloud and terminal support vector machine (CaTSVM) was proposed here and applied in motor bearing fault diagnosis. The CTSVM method was used to put operations of traditional fault diagnoses’ feature extraction and feature classification into running of terminal devices and cloud devices, respectively, introduce "pipeline" data processing structure into itself, and effectively improve its own real-time performance. The cloud feature model library (CFML) was built in cloud and terminal, and fault features were selectively added in this model library. In traditional SVM offline training, SVM online training was added and updated fault features were selectively used to train SVM model, and further improve its classification ability so that a diagnosis system can have the ability to learn throughout life. A great number of tests showed that the cloud and terminal method can significantly improve the correct rate of fault diagnosis and promote its engineering application.