Actuator Fault Estimation For Vehicle Active Suspension Systems

Vehicle suspension system plays a key role in improving vehicle ride comfort,operation stability and ride comfort.Compared with passive and semi-active suspensions,active suspensions are more adaptable to different vehicle operating conditions,and can make a balance between driving safety and riding comfort,which is an important direction for the development of vehicle suspension industry.At present,many advanced control strategies have been successfully applied to the active suspension system.However most of them are under a full-functional and fault-free assumption.During the operation of the active suspension,actuators,sensors,and components may fail at any time.In this case,the desired control effect may not be achieved by the control strategy for the fault-free situation,or even the normal operation of the system cannot be guaranteed.Therefore,it is of great theoretical and practical significance to carry out research on fault estimation（FE）and FE-based fault-tolerant control for active suspension systems.The main research work is as follows:（1）A quarter-car active suspension and commonly used pavement excitation models are established in this thesis.Furthermore,according to the characteristics of the three typical kinds of actuator faults,which are gain change,drift and stuck faults,a unified mathematical expression of the actuator fault is obtained.Then,a multi-objective hybrid H₂/H_∞ controller for active suspension is designed by using the H₂/H_∞ control method.Simulation results show that the active suspension has better vibration control performance than the passive suspension under the fault-free situation,and the control performance of the active suspension will be significantly affected by the actuator fault.（2）A FE method is proposed based on fast adaptive observer and genetic algorithm.Firstly,a fast adaptive FE algorithm is introduced,which includes both the system output error and its derivative term.On this basis,considering three typical kinds of actuator faults,genetic algorithm is applied to optimize the learning rate and integral weight parameters of fault estimation algorithm.Further,according to the fault information obtained by the proposed FE method,the least square linear regression fitting method is used to identify the fault type online.Simulation results demonstrate that the proposed method can track three typical kinds of actuator faults more accurately and quickly than other FE methods.In addition,Fault-tolerant control results based on FE and fault type identification further verify the effectiveness and practicability of the proposed FE method.（3）Considering the possible sprung mass variation in active suspension system,a parameter-dependent FE method is proposed based on radial basis function neural network（RBFNN）.Firstly,the active suspension system is modeled as a parameter-dependent system with the actuator fault and external disturbance input.Then,RBFNN is introduced to approximate the additive actuator fault due to its good approximation ability and fast convergence.The corresponding adaptive fault observer is designed based on RBFNN.Finally,simulation results verify the accuracy,rapidity and applicability of the proposed method,as well as its good adaptability to the sprung mass variation.