基于LS-SVM的多步故障预测模型  

LS-SVM-based Multi-step Failure Prediction Model

作　　者：杨珍贵[1] 郝永生[1] 赵双龙[1] 

机构地区：[1]军械工程学院导弹工程系,河北石家庄050003

出　　处：《现代电子技术》2011年第3期25-27,共3页Modern Electronics Technique

摘　　要：针对控制箱等复杂系统故障难以判断的不足,将控制箱作为一黑箱,分析输入输出信号特点,利用osu_svm 3.0和LS-SVMlab 1.5工具箱,在Matlab 7.1环境下分别建立了控制箱的故障映射模型和俯仰通道预测模型。通过试验仿真,验证了该故障诊断和预测模型的有效性,并且通过比较BP神经网络与支持向量机的分类和预测效果,得出选择支持向量机作为故障诊断和预测工具的优越性。On account of the control box and other complex systemts failure is hard to determine, the characteristics of input and output signals of the control box are analyzed as a black box. The fault mapping model and the pitch channel prediction model of the control box were established in Matlab 7.1 environment with osu_svm 3.0 and LS-SVMlab 1.5 toolbox. The experimental simulation shows effectiveness of the fault diagnosis and prediction of the model. On the basis of comparing the classification and prediction results between BP neural network and support vector machine, superiority of the support vector machines as tools for diagnosis and prediction was obvious.

关 键 词：支持向量机 故障预测 LS-SVM 控制箱 

分 类 号：TN911-34[电子电信—通信与信息系统]