Pure longitudinal slip tire model identification based on vehicle dynamic response and BP neural network
The current identification of tire model mainly relies on nonlinear fitting of known data such as slip ratio-longitudinal force or slip angle-lateral force.However,these data need to be measured using specialized test benches or tire six-component force sensors,and the high cost limits the widespread application of such methods.In this paper,an off-line identification method for pure longitudinal slip tire model based on on-board sensors and vehicle dynamic responses was proposed to obtain accurate tire models at a low cost.In Carsim,a virtual vehicle dynamic model that matches the vehicle equipped with the tire to be identified was constructed in CarSim(with no matching requirements for the tire model).The simulation calculated the vehicle's dynamic responses with different tire model parameters,providing training samples for the BP neural network and then formed a mapping relationship between"different tire model parameters and the vehicle dynamic response".The state responses of the vehicle equipped with the tire to be identified were collected under braking conditions,and the tire model parameters were then off-line identified using the trained BP neural network model.Finally,the identification results of the Gim and Unitire tire models were validated in the Simulink-Carsim co-simulation environment,confirming the accuracy of the proposed identification method.
tire model identificationBP neural networkvehicle braking testGim tire modelUniTire model
NETL
NSTL
万方数据
