Research On Semi-Active Control Of Quasi-Zero Stiffness Air Suspension System For Commercial Vehicles Based On Data-Driven Approach

Heavy-duty and specialized commercial vehicle products will always be transformed into green,low-carbon,energy-saving and efficient direction.Ride comfort,cargo damage rate and road friendliness are important indicators of high-quality operation of commercial vehicles.The direct component to meet these needs is the suspension assembly system.Because of its strong load-bearing capacity,low inherent frequency and adjustable body height,air suspension is widely used in commercial vehicles at this stage.However,as people’s requirements for vehicle performance continue to improve,performance of single air suspension can no longer well meet the requirements of vehicles driving under various complex driving conditions.The structure of air suspension has been continuously improved,and the control of its electrification and intelligent technology have been rapidly developed.However,most control of suspension system is based on model-driven approach,which cannot realize the self-adaptation of suspension system to different driving conditions.The suspension model based on the control cannot well reflect the nonlinearity of the actual suspension system.Considering above problems,my research team proposed a new quasi-zero stiffness air suspension system configuration.Based on this,this paper proposes a control strategy for the stiffness and damping of the suspension system through a datadriven method to reduce the vibration impact of commercial vehicles during driving and optimize and improve the adaptability of the quasi-zero stiffness air suspension system to different driving conditions by interacting with environmental data.Firstly,in order to obtain the training data required for subsequent control and ensure the reliability of the training data source,the Matlab / Simulink simulation model of the quasi-zero stiffness air suspension system is built according to the nonlinear mathematical model of the single-degree-of-freedom quasi-zero stiffness air suspension system.At the same time,a single-degree-of-freedom quasi-zero stiffness air suspension system test bench and an AMESim physical model considering pipeline and pneumatic system are built.The key parameters in the nonlinear quasi-zero-stiffness air suspension model are identified by data obtained from the bench test and the data of the AMESim model.Based on this,the vehicle quasi-zero stiffness air suspension system model is established,and the corresponding training data required for control process are obtained from the established model.At the same time,based on the support vector machine(SVM)algorithm,a classification and identification method of pavement grade is designed to be the basis for subsequent stiffness and damping control.Secondly,aiming at the problem of variable driving conditions of commercial vehicles,an adaptive switching strategy of quasi-zero stiffness air suspension stiffness based on data-driven method is proposed.The adaptive network-based fuzzy inference system(ANFIS)offline model of negative stiffness cylinder pressure optimization and the parameter values of the active disturbance rejection controller(ADRC)under different driving conditions are obtained by data training.The road grade,vehicle load and vehicle speed are used as input variables.The optimal negative stiffness cylinder pressure of each suspension system is obtained by ANFIS model,and the cooperative game theory is introduced to optimize the target pressure for the uncertainty of suspension system and vehicle performance.The ADRC controller completes the tracking control of the target pressure and realizes the stiffness control of the commercial vehicle quasi-zero stiffness air suspension system under different driving conditions.The simulation results show that the optimal control of the stiffness of the quasi-zero stiffness air suspension system is feasible and effective under different driving conditions.Then,in order to further improve the ride comfort and driving stability of commercial vehicles,the damping of each suspension system is controlled in real time based on the deep reinforcement learning method.Considering nonlinearity of the quasi-zero stiffness air suspension system and uncertainty of road conditions,the Deep Deterministic Policy Gradient(DDPG)algorithm is chosen.Through designing the network parameters and reward functions,the corresponding agent is then trained by interactive optimization with environmental data.The active damping force of each suspension system at each moment is obtained by iterative update,and the real-time control of damping is realized.The simulation results show that the damping control strategy can further improve the comprehensive vibration isolation performance of commercial vehicles under different driving conditions.Finally,hardware-in-the-loop(HiL)test platform is built to verify the feasibility and effectiveness of the stiffness and damping control strategy for quasi-zero stiffness air suspension system.Through test analysis,it can be seen that when the vehicle driving conditions change,the stiffness control strategy of the suspension system can accurately and quickly complete the switching control of the stiffness mode,so that the suspension system stiffness is adaptive to different driving conditions.During the vehicle dynamic test,the damping control based on DDPG algorithm reduces the centroid acceleration of the vehicle by 1.54 % compared with that before damping control,the body roll angle by 7.93 %,and the body pitch angle by 0.99 %.The test results show that the proposed stiffness and damping control strategy based on datadriven method can effectively improve the adaptability of the quasi-zero stiffness air suspension system to different driving conditions and further improve the comprehensive vibration isolation performance of commercial vehicles.