Adaptive Observer Based Fault Diagnosis And Fault-Tolerant Control With High Sensitivity For Active Vehicle Suspension Systems

Vehicle suspension system mechanically connects the chassis to wheels,which plays an important role in reducing vehicle vibrations,providing ride comfort,and ensuring maneuvering stability.With the constantly developing automotive industry and increasingly high performance demands,active suspension is the wave of the future for its superior control performance.However,modern active suspension systems are equipped with various actuators and sensors,and these components are prone to malfunctions due to long use and aging,which degrades suspension and even vehicle performance,reliability and safety.The reliability concerns hinder universal application of active suspensions.To investigate fault diagnosis and fault-tolerant control problems of active suspension is of great theoretical and practical significance in guaranteeing reliability and breaking through the bottleneck of broad use of active suspension.Therefore,this dissertation focuses on adaptive observer based fault diagnosis and fault-tolerant control with high sensitivity of active suspension systems,which is carried out from quarter-vehicle to fullvehicle suspension model.The brief summary of the dissertation is given as follows.For quarter-vehicle suspension systems with parametric and nonlinear uncertainties,adaptive fault detection and isolation scheme has been proposed to handle two major categories of suspension faults including actuator and system configuration faults.In most of the relevant research work,only one category of suspension fault is considered,such as actuator or sensor failure,and there lacks a more universal fault diagnosis method for active suspension systems.It is of much difficulty to address multiple categories of suspension faults,which requires a deep understanding into features of various fault types.In order to make the fault diagnosis scheme applicable to more suspension faults,an adaptive fault detection and isolation method is developed for two main categories of suspension faults.An adaptive fault detection estimator is designed to report occurrence of faults and a set of fault isolation estimators are specifically established for each possible suspension fault.If the residual of the fault isolation estimator exceeds its corresponding threshold,occurrence of the related fault is isolated.When determining adaptive fault detection and isolation thresholds,the uncertain system parameter appearing in the thresholds is periodically updated in real time so that sensitivity of fault diagnosis can be promoted.With the consideration of nonlinear uncertainties,the proposed method also shows good robustness to model uncertainties and external disturbances.In some cases,vehicle displacement is adopted as the output signal measured by a sensor to reduce complex wiring and lower costs.An adaptive sensor fault accommodation scheme is proposed for uncertain vehicle active suspensions via output feedback control where vehicle body displacement is the only measurable output signal corrupted by sensor bias.To solve the above problem,an adaptive observer with variable gains is developed to generate state estimates with the use of measured output and control input signal where parameter adaption of the unknown system parameters and sensor bias is involved.An output-feedback controller is then designed based on the measured output and estimates of the states,uncertain parameters,and sensor bias obtained from the observer to stabilize the vertical motion of the suspension system and achieve sensor fault accommodation at the same time.Furthermore,it should be mentioned that adaptive compensation for sensor failure is included both in the design process of the adaptive observer as well as the output-feedback controller to weaken the influence brought by the sensor measurement error.Variable observer gains are determined in real time using a switching strategy where values of the gains can be altered in finite times by monitoring the state estimates generated by the observer itself.Theoretical analysis and simulation results both show that the proposed sensor fault accommodation method can effectively weaken the influence of sensor fault and contribute to better vibration attenuation performance,compared with the output feedback controller without any compensation for sensor bias.Full-vehicle suspension model with high dimensions captures dynamics of suspension systems more thoroughly but brings a challenge to the fault diagnosis design.A joint fault-tolerant and fault diagnosis strategy is proposed for multiple actuator faults of full-vehicle active suspension systems.Different from traditional methods where fault detection and isolation is prior to fault-tolerant process to provide certain fault information,in our proposed scheme,fault diagnosis is carried out under the framework of fault-tolerant control to guarantee satisfactory suspension performance over the whole operation process.Since there will be not much obvious performance degradation to arise under the guarantee of the fault-tolerant framework,it brings much difficulty to actuator fault diagnosis by observing certain indexes dependent on the system states.Also,it is difficult to conduct the fault-tolerant controller design directly based on the control inputs,due to the fact that they are integrated together in the dynamic equation of each motion of the vehicle body.For the fault-tolerant technique,high gain filters are developed to establish a fast timescale system to approximate the variations of the control parts caused by actuator faults,which will be compensated for in the fault-tolerant controller.Under the framework of the fault-tolerant algorithm,a set of fault diagnosis observers are designed to online identify the force constants of the actuators.Actuator faults can be detected and both the location and severity of the actuator faults can be identified via monitoring the estimates of the force constants.Based on the fault diagnosis result,control commands are reassigned to alleviate the use of faulty actuators and thus protect them from further damage.An adaptive fault diagnosis scheme is proposed for multiple sensor faults of fullvehicle active suspension systems with model uncertainties including unmodelled dynamics,parametric uncertainty and external disturbance.Unlike actuator faults,sensor fault diagnosis can be greatly affected by model uncertainties.To solve the problem,the so-called integrated model uncertainty is introduced by combining unmodelled dynamics,parametric uncertainty and external disturbance of linearized full-vehicle suspension model into a unified term.A group of disturbance observers have been designed to online estimate the integrated model uncertainties of the full-vehicle suspension system,and the obtained estimates from the disturbance observers will be employed to compensate for influence of inaccurate suspension model on sensor fault diagnosis.For each sensor at risk of failure,a set of sensor fault diagnosis observers are constructed,where only the measurement of the specific sensor is taken as the input of the observers,making the observers sensitive to the possible failure of the single sensor.Online estimates of integrated model uncertainties obtained from the disturbance observers are injected to the fault diagnosis observers to emulate the unmodelled dynamics,parametric uncertainty,and external disturbance of the suspension system,which contributes to promoting accuracy,sensitivity,and robustness of the sensor fault diagnosis scheme.Each set of fault diagnosis observers operate independently and do not interfere with each other such that fault diagnosis for multiple sensor faults can be achieved.