GASA-SVM改进算法及其在柴油机供油系统故障诊断中的应用  

The Application of SVM Based on Improved GASA Algorithm to Fault Diagnosis in Diesel Engine Fuel System

作　　者：何俊强[1] 李建勇[2] 姜涛涛[3] 代勤芳[4] 唐超[1] 

机构地区：[1]重庆通信学院控制工程重点实验室 [2]重庆通信学院电力工程系 [3]重庆通信学院政治部 [4]中国人民解放军78088部队

出　　处：《西南科技大学学报》2011年第2期44-48,共5页Journal of Southwest University of Science and Technology

基　　金：国防科研项目(TZ-CQTY-Y-A-2010-002);重庆市高校优秀成果转化项目(Kjzh10219)

摘　　要：针对目前支持向量机(SVM)参数选择的盲目性,结合遗传算法GA的并行搜索和模拟退火算法SA的概率突跳特性,提出一种改进的基于遗传退火算法(GASA)混合策略优化支持向量机惩罚函数和核函数参数的GASA-SVM算法。利用柴油机供油系统油压波形的实测数据,归一化处理后作为诊断模型的特征值,建立了基于GASA-SVM的柴油机供油系统故障诊断模型。通过与BP神经网络、RBF神经网络、SVM和GA-SVM故障诊断模型比较表明:应用GASA-SVM建立的故障诊断模型在故障识别准确性上优于其它网络模型,能够有效进行柴油机供油系统的故障诊断。For the current choice blindness of parameters in support vector machine, combined with parallel search feature in genetic algorithm and probability sudden jump characteristic in simulated annealing structure, an improved GASA-SVM based on GASA mixed strategy to optimize the parameters of support vector machines penalty function and the kernel function is proposed in this paper. According to the pressure data measured from fuel supply system of diesel engine, which is taken as the characteristic values for diagnostic model after normalizing, the fault diagnosis model of diesel engine fuel supply system based on GASA-SVM is established finally. Compared with fault diagnosis models of BP network, RBF network, SVM and GA-SVM model, the results show that the fault identification accuracy of fault diagnosis model based on GASA-SVM for diesel engine fuel supply system is superior to the other network models, which is practical and effective in fault diagnosis for diesel engine fuel supply system.

关 键 词：遗传退火算法(GASA) 支持向量机(SVM) 供油系统 故障诊断 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TK428[自动化与计算机技术—控制科学与工程]