Research On Sensor Fault Diagnosis And Fault-tolerant Control Of Liquid Level System

The control system plays an increasingly important role in the rapid development of industrial production.The failures of sensors cause the instability of the system.So it is necessary to perform the fault diagnosis on the sensors timely and accurately,such that to ensure the system can be maintained.Besides,when the sensors fail to work yet not be repaired,it is very necessary to guarantee that the system run normally with fault states,which is controlled by corresponding fault-tolerant control methods.In this paper,the liquid level system is considered as the plant,the problem of sensor fault diagnosis and fault-tolerant control is studied,the main works include:(1)Based on the analysis of the current situation of sensor fault diagnosis,fault-tolerant control technology at home and abroad and the characteristics of sensor failure,an experimental platform and a PID control system of two tank liquid level are designed,then set up the Simulink structure diagram.In the control system,three different faults are set for,namely,constant gain fault,constant deviation fault and constant gain with constant deviation fault.By collecting the data of normal state and three fault states,it is feasible to verify the fault diagnosis and fault-tolerant control of liquid level system sensor.(2)Based on the comparative study of common methods of fault diagnosis,a fault diagnosis method based on improved particle swarm optimization is proposed.Firstly,wavelet combines with neural network is used to establish the wavelet neural network model.Then,the improved particle swarm optimization(PSO)algorithm is used to optimize the network.After the pretreatment of the collected data,the model is trained and tested to achieve the purpose of sensor fault classification.A total of 330 sets of data samples were collected,250 groups were as training samples,80 groups were as test samples.Compared with traditional BP neural network and wavelet neural network,after the training,The wavelet neural network with particle swarm optimization is proved to be very good in terms of convergence speed,training error and accuracy,which provides strong support for sensor fault maintenance.(3)For the two tank liquid level control system,the design of linear time-varying parameter state observer and state controller to achieve fault-tolerant control.The sensor fault value and the system state are estimated.When the sensor is detected that it failed,the nominal controller does not work.At this moment,the state controller control the system.The liquid level system can still run normally although it is fault so that can achieve fault tolerance.When the sensor is subjected to constant gain fault,constant deviation fault and constant gain and constant deviation fault,the linear time-varying parameter state observer and the state controller are used to verify.The simulation results show that state observer can enable fault-tolerant control function when the liquid sensor is fault.