基于数字孪生模型的故障特征生成与诊断

Fault Feature Generation and Fault Diagnosis Based on Digital Twin Model

现有的故障诊断方法需要对实体设备进行大量损伤实验来获取故障特征数据,针对上述方法在实施过程中存在的成本高、周期长、安全性低等问题,提出了一种基于数字孪生模型的故障特征生成与诊断方法。首先,构建物理实体的数字孪生体并进行虚拟故障注入,利用模型仿真获取故障特征数据,为故障诊断算法提供虚拟数据源;其次,利用深度学习技术,对“虚拟数据”进行迭代分析,并将训练好的诊断模型迁移到物理空间,通过接入物理感知数据实时诊断设备运行状况;最后,以轴承这一机电装备中应用广泛的易损件为例对提出的方法进行验证。实验结果表明,数字孪生模型生成的故障特征数据是可信的;利用“虚拟数据”得到的故障诊断模型可以对设备的健康状况做出较为准确的诊断,证明了所提方法的有效性。

Existing fault diagnosis methods need to conduct lots of destructive experiments on physical equipment to obtain the data of fault feature,which has high cost,takes a long time,a nd has safety issues in the implementation.In order to solve the problem above,a method of fault feature generation and fault diagnosis based on digital twin model is proposed.Firstly,digital twin model of physical entity is conducted and implement virtual fault injection.Then the data of fault feature is obtained by model simulation,which provides virtual data for fault diagnosis algorithm.Next,the"virtual data"is analyzed iteratively by deep learning technology.After the training,the model of fault diagnosis is transferred to physical space to evaluate the health status of equipment.Finally,the effectiveness of this method is demonstrated by bearing,which is an widely used wearing part in mechatronics equipment.Experimental results show that the data of fault feature generated by the digital twin model is credible,the fault diagnosis algorithm based on"virtual data"can accurately diagnose the health status of equipment,which proves the effectiveness of the proposed method.