Real-Time Hybrid Simulation Research On Vertical Semi-Active Suspension Of Vehicle Running On Suspension Bridges Based On LSTM

In recent years,high-speed railways have been developing rapidly in China,with the total mileage of high-speed railways in China exceeding 40,000 km by the end of 2021.The number of long span railway bridges acrossing mountains and valleys is also increasing,where longspan suspension bridges as flexible structures generate large static deformation and dynamic response under complex loading conditions,posing a great challenge to the stability of high-speed trains(HSTs)running on the bridge.In this paper,the coupled vibration system of longspan train-suspension bridge is studied,and the effect of semi-active suspension on the vibration reduction of HSTs and the real-time hybrid simulation(RTHS)method of semi-active suspension based on long shortterm memory(LSTM)are investigated.The main research results include:(1)Establishing a finite element model for long-span suspension bridges and a multi-body dynamics model for high-speed train-suspension bridge coupling vibration,and verifying the accuracy of the model through modal and time history analysis;based on the established train-bridge coupling vibration model,considering different train speeds,different levels of temperature changes and different damping ratio of the secondary suspension,analyzing and evaluating the train running stability on longspan suspension bridges under complex operating conditions,and comparing the damping ratio of different suspension systems.The results show that the dynamic response of the car body varies with the speed of the vehicle,the temperature effect,and the damping ratio of the suspension.The results show that the dynamic response of the car body increases with the increase of the vehicle speed and the change of the system temperature,and is more sensitive to the overall cooling of the system,and the peak vertical acceleration of the car body exceeds the limit of the train running stability index with overall temperature decrease of 27°C.(2)A modified Dahl magnetorheological damper model was established and validated,and a numerical simulation study of the magnetorheological damper was carried out;subsequently,based on the Simulink numerical simulation platform,the vibration reduction effects of different control algorithms on 1/4 trains were compared,the effects of control algorithms such as Skyhook control,ADD control and Mix-SH-ADD on train running stability at different vehicle speeds were evaluated.Finally,based on the Simulink-Simpack co-simulation model of train-suspension bridgedamper,the vibration reduction effects of magnetorheological dampers with semi-active control method on the vertical vibration of trains under different speeds and temperature loading conditions is investigated,and the effectiveness of semi-active suspension on the vibration control of trains passing through the suspension bridges is verified.(3)A RTHS framework for the evaluation of vibration reduction effect of the semi-active suspension system on bridges is built.The coupled trainsuspension bridge system is simulated in a computer as a numerical substructure,and the second suspension damper of the train with strong nonlinearity is dynamically loaded using an actuator as the test substructure.The feasibility of LSTM to replace complex numerical substructures in RTHS is analyzed,two reasonable recursive network architectures are selected,and numerical simulation data is used as the data set for network training and prediction.The prediction performance of the two trained network models is evaluated by time history comparison,regression analysis and normalized error distribution;Finally,the network models are applied to the RTHS of the semi-active damping devices,and their stability,accuracy and real-time performance are verified,which preliminarily proves the neural network is an efficient alternative method to the traditional numerical integration algorithm.Including 102 figures,14 tables and 151 references.