基于遗传优化最小二乘支持向量机的导引头故障诊断  

Intelligent fault diagnosis for seeker based on least squares support vector machine with genetic algorithm

作　　者：李轩[1] 李建宏[2] 何玉珠[1] 许金玖 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100191 [2]北京航空航天大学可靠性与系统工程学院,北京100191 [3]空军驻西南地区军事代表室

出　　处：《计算机应用研究》2011年第3期899-901,904,共4页Application Research of Computers

摘　　要：为提高导引头故障诊断准确率,提出了一种采用改进遗传算法优化的最小二乘支持向量机(LSSVM)构造导引头多故障分类模型的方法。该方法基于一对一策略及改进的投票法建立两层LSSVM多故障分类器,并利用一种自适应变步长搜索策略改进的遗传算法对LSSVM的核参数和正则化参数进行自动优选。通过对某型导引头实测数据的仿真并与标准SVM和BP神经网络诊断方法相比较,结果表明,该方法具有更高的诊断准确率和计算效率。In order to improve the correct rate of seeker fault diagnosis,this paper proposed a multi-class classification model for seeker based on LSSVM optimized by the improved genetic algorithm.The model constructed the multi-class LSSVM classifiers of two layers according to the one-against-one strategy and the improved vote method.And employed the genetic algorithm which was improved by an adaptive search strategy with a variable step length to select the kernel function parameter and the regularization parameter of LSSVM.Then brought the data of diagnosing missile seeker into the model to test its functions.Result shows that the method has higher diagnosis accuracy and computational efficiency by comparing with the standard SVM and BP neural network diagnosing method.

关 键 词：故障诊断 最小二乘支持向量机 遗传算法 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]