Article Details

Abstract

To address the problems of oversimplification and low accuracy in traditional traction characteristic modeling methods, a traction characteristic modeling method applicable to multiple locomotive types and featuring both high accuracy and practicality is developed. First, based on the main data tables for locomotive traction calculation, traction curves are fitted by means of piecewise polynomial fitting at continuous speeds and discrete notch levels with an interval of 0.1, thereby obtaining the traction characteristic functions. Then, a Multi-TrainMLP fusion estimator is designed. Built upon a lightweight backbone network, the estimator extends multiple output branches equipped with channel attention mechanisms to fit the traction forces of different locomotive types, demonstrating independence, scalability, and flexibility along the vehicle-type dimension. Finally, the data from the traction characteristic functions are used to train the Multi-TrainMLP fusion estimator. Experiments are conducted on data from five locomotive types. The average MAE between the fitted traction force and the lookup-table-based traction force is less than 0.3 kN, while the method can also obtain traction force at any speed.

Keywords

locomotive traction characteristics; modeling; polynomial fitting; MLP; fusion estimator; Attention Mechanism

References

1. [1] Train traction calculation regulations: TB/T 1407-1998 [S]. [Place of publication unknown]: Industry standard - Railway, 1998
2. [2] Test method for traction characteristics of electric locomotives: TB/T 2510-1995 [S]. [Place of publication unknown]: Industry standard - Railway, 1995
3. [3] Peng Qiyuan Traction Calculation of Urban Rail Transit Trains [M]. Southwest Jiaotong University Press, 2005
4. [4] Kang Xiong Mathematical Modeling and Curve Fitting of Electric Locomotive Characteristic Curve [J]. Journal of Railway, 1993 (1): 4-8
5. [5] Zhao Xiaolong, Xu Zhongwei, Zhu Long Fitting of Train Traction Characteristic Curve Based on Mobile Least Squares Method [J]. Science and Technology Association Forum (Second Half), 2013 (1): 109-111
6. [6] Cybenko G. Approximation By Superpositions of a Sigmoidal Function [J].Mathematics of Control, Signals and Systems，1989，2（4）：303-314.
7. [7] Hornik K，Stinchcombe M，White H. Multilayer Feedforward Networks are Universal Approximators [J].Neural Networks，1989，2（5）：359-366.
8. [8] Weigend A S，Huberman B A，Rumelhart D E. Predicting the Future: A Connectionist Approach [J].International Journal of Neural Systems，1990，1（3）：193-290.
9. [9] Rossi F，Conan-Guez B. Functional Multi-layer perceptron: A non-linear tool for Functional data Analysis [J].Neural Netw，2005，18（1）：45-60.
10. [10] Zhang Xiquan Electric Traction Transmission and Control [M]. China Railway Press, 2012
11. [11] Yang Zhiwei Analysis and improvement measures of the influence of traction characteristics on idle running of HXD3CA electric locomotive [J]. Harbin Railway Technology, 2023 (2): 24-27, 31

About the authors

Yang Yuanyuan (1990-), female, Han nationality, from Zhengzhou, Henan, assistant engineer, master's degree, research direction: research on object detection and segmentation algorithms based on deep neural networks, as well as the application and optimization of neural network technology in locomotive assisted driving systems; Chen Guotao (1987-), male, Han ethnicity, born in Xinyang, Henan Province, is an engineer with a master's degree. His research interests include research on locomotive assisted driving systems, train model modeling, and ground simulation systems for locomotive assisted driving systems; Deng Ruirui (1991-), male, Han ethnicity, from Qingyang, Gansu Province, engineer, bachelor's degree, research direction: transportation management; Jia Qunxi (2000-), male, Han ethnicity, from Zhumadian, Henan Province, assistant engineer, bachelor's degree, research direction: locomotive assisted driving control, locomotive assisted driving simulation modeling, embedded systems; Wu Han (1991-), male, Han ethnicity, from Zhumadian, Henan Province, assistant engineer, bachelor's degree, research direction: signal and information processing, communication and network technology, embedded systems and Internet of Things; Guo Jinwei (2001-), male, Han nationality, from Zhoukou, Henan, engineer, bachelor's degree, research direction: accurate modeling of traction characteristics of multiple locomotive types and neural network optimization methods.