基于数字孪生的牵引变流冷却系统PHM构建
Construction of PHM for Traction Inverter Cooling System Based on Digital Twin Technology

通过智慧化运维降本增效是提高中国城市轨道交通可持续发展能力的重要途径之一，将数字孪生技术应用在轨道交通车辆故障预测与健康管理(PHM)中，可以有效解决传统PHM的不足。本文提出了一种基于数字孪生技术的牵引变流器冷却系统PHM的构建方法。首先构建了牵引变流冷却系统关键子系统物理实体的数字孪生体，包括机理模型和数字模型；然后通过虚实之间的数字映射和健康域的标定，并采用岭回归、循环神经网络、随机森林等方式进行模型训练，得到最优的模型组合；最后以变流器柜体温度预测进行了模型测试。结果表明，该模型的预测效果与实际系统的运行结果基本吻合，可以进一步作为变流器冷却系统故障预测和故障定位的开发基础。
Reducing costs and increasing efficiency through Intelligent Operation and Maintenance (IOM) is one of the most important ways to improve the sustainable development of China’s urban rail transit. The application of digital twin technology in the Prognostics and Health Management (PHM) of rail transit vehicles can effectively address the shortcomings of traditional PHM methods. This paper proposes a construction method for a traction inverter cooling system PHM based on digital twin technology. Firstly, a digital twin of the physical entities of the key subsystems of the traction inverter cooling system was constructed, including the mechanism model and the digital model; then, the optimal model combination was obtained through the numerical mapping between the real and the imaginary and the calibration of the health domains, and the model training was carried out by using ridge regression, recurrent neural network, and random forests, etc.; finally, the model was tested for predicting the temperature of the inverter cabinet. The results showed that the prediction performance of the model was basically consistent with the operation results of the actual system, indicating that the model can be further used as a development foundation for the fault prediction and location of the inverter cooling system.