Research On Fault-Tolerant Control Methods For Sensors In Intelligent Ships

With the development of automatic and intelligent technology,the research on ship system become more in-depth.In order to cope with growing demand for sea freight while improving the reliability of the ship’s navigation,the development strategies of countries around the world are gradually targeting the field of intelligent ships.Because of the special conditions and performance of ship,higher safety and reliability requirements will be imposed on the system.Ship control issues,especially in the event of equipment failure,will not only need to ensure the stability of the ship system,but also expect to quickly restore the normal performance of the ship within a certain range.As an essential part of the control system,the accuracy of the signal measured by the sensor device will have an impact on the state of ship motion.Hardware redundancy is one of the solutions to the problem of equipment fault,but it needs to provide additional sensor equipment,which increases the cost of ship construction.Therefore,designing fault diagnosis system and corresponding fault-tolerant control method for sensor equipment is an important link of the research and development of intelligent ships.This dissertation has carried out the following main research work on the above issues.Aiming at the problem of the uncertainty of the system mathematical model in the analytic model-based method and the small number of sensor fault data sets in the data-driven method,a sensor fault diagnosis method based on passive observer and wavelet packet neural network was designed.The algorithm analyzes the fault signal characteristics of the sensor,and the model-based and data-driven ideas are combined.To start with,a non-linear passive observer is used to perform state estimation and generate residual signals.Then the multi-band energy characteristics in the residual signal could be obtained by wavelet packet decomposition.By inputting it into a BP neural network for training,a classification network model with appropriate parameters is completed to solve sensor fault detection and classification problems.Simulation results verify that the method can effectively implement sensor fault diagnosis and improve the accuracy of detection and classification.In view of the fault-tolerant control of intelligent ships when sensor faults have been detected,this dissertation designs a reconfiguration method of control loop based on virtual sensor.And the adaptive variable universe theory is introduced to improve the nominal fuzzy PID controller.After the fault sensor is successfully detected by the fault diagnosis system,it forms a virtual sensor through the observer to track the actual ship motion status.Then add it to the control loop to replace the faulty sensor,which is equivalent to the output of normal sensor to achieve the reconstruction goal.On this basis,an adaptive variable universe method is designed according to the error term,which improves the accuracy of the parameter tuning of the nominal fuzzy PID controller,accelerates the response speed of the fault-tolerant control method,and reduces the overshoot generated during the control process.The simulation results verify that the method can rapidly and effectively resolve the fault impact after accurate detecting,and complete the fault-tolerant control of intelligent ship sensors.Due to the impact of false negatives on sensor fault diagnosis system,this dissertation discusses the sensor fault-tolerant control of intelligent ship without fault diagnosis system,and devises a robust fault-tolerant control strategy based on LMI and integrity conditions.It is assumed that part of the impact of the fault is manifested as a perturbation of the model parameters.At the same time,considering the possible uncertainty of the system itself.These two aspects are attributed to the bounded change of the system parameters.And the fault switch matrix is added to indicate the part of fault sensor.According to the system parameter matrix which is modified,the stability condition of the faulty system and the robustness condition to the uncertain parameters are transformed into the problem of solving linear matrix inequality.Simulation results verify that the method can effectively complete fault-tolerant control of sensors on intelligent ships without the fault diagnosis system.