Computational Technology Resources - CCC

W. Zhang^1,2, H. Wu^1,2, S. Fu^3,4, X. Liang^3,4 and X. Zeng^1,2

¹School of Future Technology, University of Chinese Academy of Sciences, Beijing, China
²Key Laboratory for Mechanics in Fluid Solid Coupling Systems, Institute of Mechanics, Chinese Academy of Sciences, Beijing, China
³Qingdao Sifang Co., Ltd., CRRC, China
⁴State Key Laboratory, High-speed Maglev Transportation Technology, Qingdao, China

Keywords: Maglev vehicle, neural network, nonlinear dynamics, model predictive control, state constraints, recursive radial basis function.

Abstract

The levitation control system plays a pivotal role in governing the intricate aerodynamic load-vehicle-rail coupling vibration process, a crucial determinant of vehicle stability. As speed escalates, the impact of aerodynamic load and vehicle-rail coupling on levitation stability becomes increasingly undeniable. Conventional proportional-integral-derivative (PID) controllers, while effective in simpler environments, exhibit diminished performance within the complexities of high-speed mechanical systems. To address the pressing need for accurate prediction of non-stationary aerodynamic performance in high-speed maglev vehicles, we propose a load prediction model based on the Recursive Radial Basis Function Neural Network (RRBF). This model, leveraging recursive history information in loop neurons, offers dynamic memory capabilities, thus enhancing its learning of temporal patterns. The RRBF network predicts real-time aerodynamic load based on time series states. Simultaneously, we introduce a robust Nonlinear Model Predictive Controller (NMPC), designed to consider the physical constraints of the chopper and the influence of aerodynamic loads on the model's future dynamic behaviours. The algorithm ensures optimal roll optimization within finite time, accounting for constraints. Stability assessments of an electromagnet under non-stationary aerodynamic loads and random track irregularities are conducted through a minimum levitation unit dynamics-control co-simulation model. Results highlight the RRBF's precise identification of aerodynamic loads. Moreover, the Recursive Radial Basis Function Neural Network-based Model Predictive Controller (RMPC) demonstrates superiority over the PID method, showcasing exceptional disturbance resistance and making it more suitable for high-speed maglev levitation control.

download the full-text of this paper (PDF, 12 pages, 1276 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCC CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Conferences ISSN 2753-3239 CCC: 7 PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON RAILWAY TECHNOLOGY: RESEARCH, DEVELOPMENT AND MAINTENANCE Edited by: J. Pombo Paper 16.5 Recursive Radial Basis Neural Network-Based Model for Predictive Control of Maglev Vehicles W. Zhang^1,2, H. Wu^1,2, S. Fu^3,4, X. Liang^3,4 and X. Zeng^1,2 ¹School of Future Technology, University of Chinese Academy of Sciences, Beijing, China ²Key Laboratory for Mechanics in Fluid Solid Coupling Systems, Institute of Mechanics, Chinese Academy of Sciences, Beijing, China ³Qingdao Sifang Co., Ltd., CRRC, China ⁴State Key Laboratory, High-speed Maglev Transportation Technology, Qingdao, China doi:10.4203/ccc.7.16.5 Full Bibliographic Reference for this paper W. Zhang, H. Wu, S. Fu, X. Liang, X. Zeng, "Recursive Radial Basis Neural Network-Based Model for Predictive Control of Maglev Vehicles", in J. Pombo, (Editor), "Proceedings of the Sixth International Conference on Railway Technology: Research, Development and Maintenance", Civil-Comp Press, Edinburgh, UK, Online volume: CCC 7, Paper 16.5, 2024, doi:10.4203/ccc.7.16.5 Keywords: Maglev vehicle, neural network, nonlinear dynamics, model predictive control, state constraints, recursive radial basis function. Abstract The levitation control system plays a pivotal role in governing the intricate aerodynamic load-vehicle-rail coupling vibration process, a crucial determinant of vehicle stability. As speed escalates, the impact of aerodynamic load and vehicle-rail coupling on levitation stability becomes increasingly undeniable. Conventional proportional-integral-derivative (PID) controllers, while effective in simpler environments, exhibit diminished performance within the complexities of high-speed mechanical systems. To address the pressing need for accurate prediction of non-stationary aerodynamic performance in high-speed maglev vehicles, we propose a load prediction model based on the Recursive Radial Basis Function Neural Network (RRBF). This model, leveraging recursive history information in loop neurons, offers dynamic memory capabilities, thus enhancing its learning of temporal patterns. The RRBF network predicts real-time aerodynamic load based on time series states. Simultaneously, we introduce a robust Nonlinear Model Predictive Controller (NMPC), designed to consider the physical constraints of the chopper and the influence of aerodynamic loads on the model's future dynamic behaviours. The algorithm ensures optimal roll optimization within finite time, accounting for constraints. Stability assessments of an electromagnet under non-stationary aerodynamic loads and random track irregularities are conducted through a minimum levitation unit dynamics-control co-simulation model. Results highlight the RRBF's precise identification of aerodynamic loads. Moreover, the Recursive Radial Basis Function Neural Network-based Model Predictive Controller (RMPC) demonstrates superiority over the PID method, showcasing exceptional disturbance resistance and making it more suitable for high-speed maglev levitation control. download the full-text of this paper (PDF, 12 pages, 1276 Kb) go to the previous paper go to the next paper return to the table of contents return to the volume description
Back to top	©Civil-Comp Limited 2023 - terms & conditions