列车转向架数字孪生建模仿真关键技术研究

Key technologies of digital twin modeling and simulation for train bogie

为全面推进数字列车的工程化应用,以数字转向架建设为突破点开展数字孪生建模仿真关键技术研究,实现物理实体与数字孪生体之间的虚实交融,进一步提升转向架故障诊断及健康管理的有效率和准确率。转向架数字孪生体是一个数据驱动的多学科、多物理、多尺度、高保真度的虚拟模型,模型的输入为自身结构参数、车载在线监测系统数据、车库定期检修数据等多源数据,输出为实时动态展示的动力学性能评价指标(如脱轨系数和轮重减载率)、结构强度指标(如应力和累积损伤),能反映并预测转向架全寿命周期内的动力学及强度性能,从而革新现有的转向架运营维护模式。提出一种转向架数字孪生体建设的框架结构,可实现动力学、结构强度关键技术指标的实时动态展示与反馈设计功能。论述转向架孪生体建模的四大关键技术及技术路线,给出转向架动力学孪生子模型、结构强度孪生子模型的仿真方法及步骤,对比分析多种仿真试验设计方案并给出一种最优建议方案。提出一种基于深度学习神经网络的代理模型技术路线,并通过轮对的轮轨力预测、轮对应力预测案例验证模型的有效性及准确性。研究成果可用于指导构建列车转向架数字孪生工程应用的核心模块,并将其部署在地面中心。

In order to comprehensively promote the engineering application of digital train,the construction of digital bogie was taken as the breakthrough point to carry out the research on the key technologies of digital twin modeling and simulation.The blending of virtual and real between physical entity and digital twin was realized,and the effectiveness and accuracy of bogie’s fault diagnosis and health management were improved.The digital twin of bogie was a data-driven virtual model with multi-disciplinary,multi physics,multi-scale and high fidelity.The input of the model was multi-source data such as its own structural parameters,on-board online monitoring data and regular maintenance data.The output was performance evaluation indexes(such as derailment coefficient,wheel load reduction rate)and structural strength indexes(such as stress and cumulative damage)displayed in real time.It can reflect and predict the dynamics and strength performance of the bogie in the whole life cycle,so as to innovate the existing bogie operation and maintenance mode.A frame structure of bogie digital twin construction was proposed,which can realize the real-time dynamic display and feedback design of key technical indexes of dynamics and structural strength.The four key technologies and technical routes of bogie twin modeling were described.The simulation methods and detailed steps of bogie dynamics twin model and structural strength twin model were given.Some simulation test design schemes were compared and analyzed and an optimal proposal scheme was proposed.A technical route of agent model based on deep learning neural network was proposed,and the effectiveness and accuracy of the model were verified by the cases of wheel rail force prediction and wheel-set stress prediction.The results of this study can be used to guide the development of the core module of digital twin engineering application of bogie and deploy it on the ground center.