基于MPSO-SVM的扇形区域故障诊断研究

针对一类线性定常系统,基于扇形区域,研究了执行器单一部件故障诊断与可靠控制的问题。首先,对于文中极点信息难于获取的问题,给出全维状态观测器的设计方案,实现对极点信息的实时观测。同时为解决支持向量机在故障诊断中选取参数易受主观先验知识影响的缺陷,提出用MPSO-SVM(ModifyParticleswarm optimization algorithm optimize the SVM)建立优化模型,设计惯性权重自适应调整公式进行算法优化,既能获取核参数及惩罚因子最优参量,又能克服PSO-SVM算法的传统不足。该方法与SVM(SupportVectorMachine,SVM)、Gridsearch-SVM、PSO-SVM相比,诊断准确率明显得到改善,从而验证MPSO-SVM模型对执行器故障诊断是可靠的。

RESEARCH ON FAUIT DIAGNOSIS OF SECTOR REGION BASED ON MPSO-SVM

Be aimed at a class of linear time-invariant systems, in terms of the sector region, the problem of reliable control of actuator single fault based on particle swarm optimization algorithm was studied. Firstly, for the problem that pole information was difficult to obtain, the design scheme of the full-dimensional state observer was given to realize the real-time observation of the pole information. At the same time, in order to solve the defect that the selection of parameters in the fault diagnosis of support vector machine was easily affected by subjective prior knowledge, an optimization model was established by using MPSO-SVM, and the self-adaptive adjustment formula of inertia weight was designed to optimize the algorithm, which not only could obtain the optimal parameters of kernel parameters and penalty factors, but also could overcome the traditional shortcomings of PSO-SVM algorithm. Compared with SVM, Grid search SVM and PSO-SVM, the accuracy of the proposed method was significantly improved, which verified that the model was reliable for actuator fault diagnosis.