基于数字孪生的加速器机组三维可视化与故障诊断方法

Fault diagnosis method and 3D visualization of accelerator device based on digital twin

基于数字孪生技术,提出一种机器学习故障诊断与3D可视化系统融合的DC 500 kV加速器故障诊断方法。依据设备通信原理搭建可视化系统整体框架,完成Unity与加速器机组设备的实时通信交互,实现对设备的数据管理和实验控制功能。融合可视化系统与加速器数字孪生模型,实现加速器机组设备的3D可视化和系统新增模型处理功能。针对正常运行状态、轴承故障、漏气故障、底座松动、泵体振动等5种状态下设备放电实验产生的振动信号,利用机器学习算法完成加速器机组设备故障分类预测和验证。采用决策树算法、随机森林算法、k-近邻(k-nearest neighbor,k-NN)算法等3种算法模型对振动信号进行模拟训练,预测精度达到0.96,实现加速器机组可视化系统的故障诊断功能。

A fault diagnosis method for DC 500 kV accelerator is proposed based on digital twin technology,which integrates machine learning fault diagnosis and 3D visualization system.The visual system framework is built based on device communication principles,and the real-time communication and interaction between Unity and accelerator unit equipment is completed,and the data management and experimental control functions for the equipment are achieved.The visualization systems with accelerator digital twin models are integrated,and the 3D visualization of accelerator unit equipment and new model processing functions in the system is achieved.Machine learning algorithms are used to classify,predict,and verify equipment faults in accelerator units based on vibration signals generated during equipment discharge experiments in five different states:normal operation,bearing failure,air leakage,loose base,and pump body vibration.Using decision tree algorithm,random forest algorithm and k-nearest neighbor(k-NN)algorithm,the vibration signals are simulated and trained on,and the prediction accuracy is 0.96,which means the visualized fault diagnosis of accelerator unit system is achieved.