Research On Road Preview Model Predictive Control Strategy For Magneto-rheological Semi-active Suspension

With the development of technology and the progress of society,people have put forward higher and higher requirements for the comfort of cars.The stiffness of the elastic element of the passive suspension and the damping of the shock absorber are designed for specific road conditions.It is inevitable that the damping performance will deteriorate under certain road conditions,and it is difficult to make breakthrough progress in improving ride comfort.Magnetorheological damper semi-active suspension is a suspension with broad application prospects.It can meet the requirements for riding comfort under a variety of road surface excitations,and at the same time has the advantages of good controllability,low energy consumption,continuously adjustable damping force,and rapid response.However,the magnetorheological damper as a suspension actuator has strong hysteresis characteristics,and its modeling and controller design are difficult.At present,most control algorithms for semi-active suspensions calculate control inputs based on the current vehicle state,and do not consider the road surface information ahead.In addition,semi-active control algorithms rarely consider system constraints when optimizing control input.In this paper,the study of magnetorheological semi-active suspension system adopts a hierarchical control strategy.The lower controller designs a feedforward-feedback structure controller to solve the nonlinear control problem of magnetorheological damper,and the upper controller is a a semi-active suspension preview control algorithm that considers road surface information and is easy to implement.The main research content includes the following aspects:1.Magneto-rheological damper modeling and controller designFirst,the external characteristics of the key component magnetorheological damper are tested,and the experimental data of the output damping force of the magnetorheological damper and the displacement and velocity of the piston movement under different control currents are obtained.The forward model of the magnetorheological damper was obtained by using double hidden layer BP neural network identification,and the model was verified.The transfer function between the output damping force of forward neural network model and the actual damping force is obtained by the identification method.The Hammerstein model is formed by the transfer function in series with the forward neural network.The simulation experiment verifies that the accuracy of the Hammerstein model is higher than the pure neural network forward model.Then,another double hidden layer BP neural network was used to identify the inverse model of the magnetorheological damper,which was used as a feedforward controller.The LQR controller is designed according to the transfer function model in the Hammerstein model of the magnetorheological damper,and it forms a feedforward-feedback structure with the neural network inversion feedforward controller.Because the state of the linear part in the Hammerstein model is identified and does not exist in the actual magnetorheological damper,a full-dimensional observer is built to observe the state of the linear part.The simulation analysis shows that compared with only the feedforward inversion controller and the introduction of the PID controller,the LQR controller with a full-dimensional observer can reduce the tracking error in most cases.2.Design of Road Preview Model Predictive ControllerAiming at the upper control problem of magneto-rheological semi-active suspension,the road information is integrated to realize predictive control based on road preview.The vertical acceleration of the vehicle body,which replaces the ride comfort,is the control output,and the dynamic load of the tire,which characterizes safe handling and stability,and the suspension stroke,which characterizes the mechanical limitation,are used as the constraint output.At the same time,the output of the actuator―magneto-rheological dampers’ upper and lower limits is considered.The road preview model predictive controller is designed by taking the road surface information―the rate of change of road height as a measurable time-varying interference based on the idea of road preview.Simulation experiments show that,compared with passive suspension,the designed road preview model predictive controller can improve ride comfort while ensuring safety and meeting damping force and dynamic stroke constraints;compared with PID combined with neural network inversion controller,LQR Combining neural network inversion can further improve suspension performance.