Vibration control of vehicle semi-active suspension combined with the modified adaptive neuro fuzzy inference system
Suspension is one of the key components for reducing vehicle vibration and improving driving safety and ride comfort.Magnetorheological dampers are characterized by low energy consumption,strong controllability,short response time,and high failure safety.Thus,they are considered one of the most suitable vibration reduction devices for semi-active suspensions.However,due to the complex nonlinear characteristics of MR dampers,their damping force cannot be directly controlled.Therefore,it is urgent to build an accurate inverse model of MR dampers to predict the control signal of MR dampers and indirectly control the magnitude of damping force.Adaptive Neural Network Fuzzy Inference System (ANFIS) is a product of the combination of fuzzy control and neural network control.In recent years,the intelligent non parametric modeling technology dominated by ANFIS has become one of the mainstream technologies for building reverse modeling of MR dampers.However,the hybrid training algorithm used in standard ANFIS has certain limitations.Namely,the gradient descent method is prone to cause the search to fall into local optima,which undermines the accuracy and the generalization ability of ANFIS model.In addition,in the semi-active control using MR dampers,it is necessary to use effective control algorithms to calculate the ideal control force.In this study,a semi-active control method based on an improved adaptive neuro fuzzy inference system (ANFIS) is proposed to improve the damping effect of MR damper semi-active suspension.First,to address the difficulty in accurately determining the reverse dynamics model of MR dampers,an improved crow search algorithm (MCSA) is adopted to optimize ANFIS.MCSA and least squares method are employed to optimize the antecedent and consequent parameters of ANFIS respectively to overcome the defect of standard ANFIS being prone to get stuck in local optimal solutions.Among them,to improve the search accuracy of the standard crow search algorithm (CSA),a triangular probability distribution strategy is adopted to select the target crow,and the updated solution is subjected to reverse mutation operation.Then,based on the suspension response,an LQR controller is designed to calculate the ideal control force,which is combined with an improved inverse model to achieve the conversion between the ideal control force and the input control signal of the MR damper,thereby adjusting the magnitude of the damping force and achieving vibration control of the semi-active suspension system of the vehicle.Our simulation results show compared to GA-ANFIS and PSO-ANFIS,the proposed MCSA-ANFIS reverse modeling method achieves a higher prediction accuracy,improving the prediction accuracy of the input signal and damping force of the MR damper by 17.49％ and 30.62％ respectively.Random road signals are employed as for semi-active suspension.Our simulation results show compared to passive control and LQR-COC semi-active control,the proposed LQR-MCSA-ANFIS control strategy reduces the root mean square values of spring-loaded mass acceleration,suspension dynamic travel,and tire dynamic load by 12.37％,37.63％,30.70％,and 6.64％,14.89％,and 17.27％ respectively.The semi-active control strategy may serve as a feasible approach to the vibration reduction research of MR damper suspension systems.
vehicle suspensionsemi-active controladaptive neuro fuzzy inference systemcrow search algorithmLQR control
国家科技期刊平台
NSTL
万方数据
