基于神经网络的自动空气制动系统仿真研究  

Simulation research of automatic railway air brake system basedon neural network

作　　者：成庶[1] 周昕怡 于天剑[1] 林磊 王佳 CHENG Shu;ZHOU Xinyi;YU Tianjian;LIN Lei;WANG Jia(School of Transportation Engineering,Central South University,Changsha 410075,China;Zhuzhou CRRC Times Electric Co.Ltd.,Zhuzhou 412001,China)

机构地区：[1]中南大学交通运输工程学院,湖南长沙410075 [2]株洲中车时代电气股份有限公司,湖南株洲412001

出　　处：《中南大学学报（自然科学版）》2024年第4期1591-1601,共11页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(52072413)。

摘　　要：通过分析空气制动试验数据研究10 000 t重载列车空气制动系统的空气传递特性,提取影响空气制动系统关键部件(列车管、副风缸和制动缸)的神经网络特征,建立一种基于神经网络的自动空气制动系统仿真模型。将模型作为制动激励输入纵向动力学模型,并将纵向动力学模型的预测结果与浩吉西峡东站—襄州北站铁路10 000 t重载列车的实际运行结果进行对比验证。研究结果表明:10 000 t重载列车的制动和缓解信号从机车向远端车辆传递,随着传递距离增加,传递速度几乎不变,但传递强度有所衰减;基于神经网络的制动系统仿真模型能预测10 000 t重载列车常用制动减压50 kPa工况的列车管、副风缸和制动缸风压变化,预测精度高达99.9%,在相同计算步长下,计算效率较传统的流体力学仿真模型提升了2 938倍,具有广阔的工程应用前景。The air transmission characteristics of the railway air brake system of 10000 t heavy haul train were studied by analyzing the railway air brake test results,the characteristics that affect the key components of the railway air brake system(train pipes,auxiliary reservoirs,and brake cylinders)were extracted,and a simulation model of automatic railway air brake system based on neural network was established.A simulation model of an automatic air braking system based on neural networks was established.The model was input into the longitudinal dynamics model as a braking excitation,and the obtained results were compared with the actual operation results of a 10000 t heavy haul train on the Hao−Ji Railway.The results show that the braking and release signals of a 10 000 t heavy haul train are transmitted from the locomotive to the remote vehicle. With the increase of thetransmission distance, the transmission speed is almost constant, but the transmission intensity is attenuated. Thebrake system simulation model based on the neural network can predict the changes of the air pressure of the trainpipes, auxiliary reservoirs, and brake cylinders under the condition of 50 kPa common brake pressure reduction for10 000 t heavy haul trains with a prediction accuracy of 99.9%. Compared with the traditional fluid dynamicssimulation models, the calculation efficiency is improved about 2 938 times and has wide engineering applicationprospect.

关 键 词：重载列车 制动系统 机器学习 神经网络 数值仿真 

分 类 号：U270.2[机械工程—车辆工程]