Fault Diagnosis Of Pole Classification And Reliable Control Based On Trapezoidal Area

Once a malfunction occurs in a chemical or aircraft system,the system may collapse,causing catastrophic consequences.Therefore,designing a reliable controller is necessary.But after adding a reliable controller,the system performance will decrease and the cost will increase.Reliable control with fault diagnosis can solve this contradiction.In linear system,once the system fails,the pole position changes,and when the system has different faults,the pole change trend is different.Simulate the pole database of different actuators in the system when faults occur.Once a fault occurs,diagnose the system based on the changes in the poles.The pole positions of the system matrix reflect the performance of the system.Due to the difficulty in achieving accurate pole allocation,regional pole allocation can meet certain practical needs.The pole configuration in the strip region limits the attenuation speed of the system,while the pole configuration in the sector region limits the oscillation frequency of the system.The pole configuration in the trapezoidal region is obtained by the simultaneous action of open sector region pole configuration and strip region pole configuration,which not only limits the stable speed of the system but also limits the oscillation frequency of the system.Therefore,this article conducts research based on the trapezoidal region.The work and achievements of this article include the following two aspects:firstly,based on the pole configuration in the trapezoidal region,taking the single channel fault of the actuator as an example,the pole observer is used to observe the changes in the system poles in real time.According to the different range of pole changes when different channels fail,the Gridsearch-SVM(Support vector machines,abbreviated as SVM)classification model is used to classify pole data,thereby achieving fault diagnosis and switching to the reliable controller of the corresponding channel,implementing precise and reliable control.The effectiveness of fault diagnosis was demonstrated through specific numerical simulations;Secondly,in order to improve the accuracy of fault diagnosis,Ant Colony Optimization(ACO)and Bat Algorithm(BA)are used to optimize the hyperparameter combination in SVM classification model.The positive feedback feature of ACO algorithm improves the speed of parameter optimization in SVM classification model.The BA algorithm improves the speed of optimizing SVM classification model parameters by introducing selection probability.The effectiveness of ACO algorithm and BA algorithm in parameter optimization was verified through simulation using the same numerical examples.The results indicate that although the Gridsearch-SVM classification model has high accuracy,its diagnostic speed is the slowest,with ACO-SVM classification model in the middle and BA-SVM classification model in the fastest.