Fault Diagnosis And Prognosis Of Electric Power Steering System Based On Bond Graph Theory

With the development of modern automobile technology,the driver’s requirements for vehicle steering performance are also increasing,in the low speed and parking requirements of light steering,but also hope that the car can maintain steering stability at high speed.Electric Power Steering(EPS)system has been widely used to solve the contradiction between "light" and "smart" in steering by virtue of the advantage that the size of assist torque can be designed according to the assist characteristic curve.EPS system is prone to failure in the driving process,which poses a threat to the safety of the vehicle.Therefore,this thesis studies the fault diagnosis and prediction scheme of EPS system based on the bond graph theory,which is of great significance for improving the reliability of the vehicle and ensuring the safety of the vehicle.Firstly,a bond graph model of EPS system with causal Relationship is established,and Analytical Redundancy Relationship(ARR)is deduced.Fault detection is realized by checking whether the residual value of ARR exceeds the threshold value.When the system parameters are uncertain,if the constant threshold value is used,there may be false positives and missed detection.Therefore,the linear fractional transformation is introduced into the bond graph model to describe the uncertainty of system parameters,and the adaptive threshold is derived from the EPS system uncertainty bond graph model to improve the fault detection ability.In addition,the concept of sensitivity signature is introduced to generalize the original two fault features(0 and 1)into three fault features(0,+1 and-1),which refines the combination of fault features and improves the fault isolation ability.Secondly,candidate fault sets are obtained through fault isolation.In order to locate the fault source,an improved bat algorithm is proposed for fault identification,which solves the problem of fault isolation in the case of the same fault signatures,and the fault amplitude and type can be determined by fault identification results.Then,the Gamma process is used to describe the evolution trend of progressive faults,and the probability distribution function of Remaining Useful Life(RUL)is obtained by combining the failure threshold and the time of fault detection.Finally,the EPS system test platform was built to carry out real-time fault detection and isolation experiments.In the process of the experiment,the sensor signal of the EPS system experimental platform is collected by the data acquisition card first,and the signal is analyzed and processed,and then the output of the experimental acquisition signal and the simulation model is compared to verify the correctness of the simulation model.Then the sensor and actuator faults are injected in a nondestructive way,and the feasibility of the proposed fault detection and isolation scheme is verified by experiments.