基于Gridsearch-SVM梯形区域极点分类的故障诊断

针对一类线性定常系统，基于梯形区域极点配置，给出了执行器部件故障诊断的一种方法。首先，利用极点观测器，通过测量系统的状态，得到极点的动态信息；其次，根据模拟各通道执行器故障，实时采集闭环系统的极点信息，形成极点分类数据库；最后，利用支持向量机算法（Support Vector Machine,SVM）根据不同通道发生故障时极点所处位置不同，设计极点分类器，对极点进行分类，实现对系统的故障诊断。针对SVM中惩罚因子和核宽度系数需要依靠先验知识的缺陷，采用Grid search优化其参数，缩小寻优范围。仿真结果表明设计方案的可行性以及故障诊断的有效性。

FAULT DIAGNOSIS BASED ON GRIDSEARCH-SVM TRAPEZOIDAL REGION POLE CLASSIFICATION

For a class of linear time-invariant systems, the fault diagnosis of actuator components was studied based on trapezoidal region. Firstly, the pole observer was used to obtain the dynamic information of the pole by measuring the state or output of the system. Then, the fault of each channel actuator was simulated, and the pole information was collected in real time to form the pole classification database. Secondly, support vector machine (SVM) algorithm was used for the state classification, and the pole classifier was designed. The penalty factors and kernel width coefficient parameters of SVM algorithm were optimized by Gridsearch. The poles were classified according to the poles'' positions when faults occured in different channels to carry out fault diagnosis; Finally, the simulation results showed the feasibility of the design scheme and the effectiveness of fault diagnosis.