Fault Adjustment Research Based On LSSVM

Modern control engineering systems are becoming more and more complex and large-scale.Improving the security and reliability of the system is a key problem in fault detection and fault-tolerant control.In fault diagnosis,least squares support vector machine(LSSVM)has the advantage of learning faster and more generalizing.It has become an effective way to deal with fault system.In addition,reliable control has the advantage of reducing the sensitivity of the system to faults and the difficulty of system design.However,there are still many problems to be improved in dealing with specific problems.Therefore,research on it is significant both in theoretical popularization and in practical application.In this topic,based on machine learning theory,parameter optimization technology and fault handling method,the problem of fault adjustment for dynamic systems are studied.In the fault regulation system designed in this paper can diagnose and deal with the faults effectively at any time regardless of whether the components are faulty or not.The main contents and results of this research mentioned in this paper include the following aspects.1 The problem of arbitrary single channel faults of dynamic system actuator based on LSSVM is studied.A fault-tolerant control method using closed-loop system poles as fault characteristic information is presented.Finally,according to the fault separation information,the corresponding single channel fault reliable controller is switched to make the system running under precise fault-tolerant control.Through the simulation of Unmanned Aerial Vehicle's longitudinal flight,the accuracy of fault diagnosis in the designed fault-tolerant control system is verified.2 The fault diagnosis method of LSSVM optimized by improved cuckoo optimization algorithm(MCS)is studied.In order to enhance the efficiency of LSSVM's pole learning,the CS is improved from the aspect of discovery probability.In order to obtain the closed-loop system poles constantly,the method of pole observer design is given.A benchmark example of adaptive reconfiguration control shows,the MCS-LSSVM fault diagnosis method with the pole learning has better modeling effect and higher classification accuracy.3 An modified particle swarm optimization algorithm(MPSO-GA)is proposed to optimize LSSVM and add Deep learning model to fault diagnosis.By adding depth learning model,the system can not only locate the fault of a part,but also more accurately fit the estimated value of fault gain.The simulation results show that the method is effective and advanced.