基于SVM的船用柴油机增压换气系统的故障诊断  

Failures Detection of Turbo-charging and Air-exchange System of Marine Diesel Engine Based on SVM

作　　者：汪猛 胡以怀[1] 曾存 方云虎 张陈 王东 Wang Meng;Hu Yihuai;Zeng Cun;Fang Yunhu;Zhang Chen;Wang Dong(Merchant Marine College,Shanghai Maritime University,Shanghai 201306,China;AVIC Dingheng Shipbuilding Co.,Ltd.,Yangzhou,Jiangsu 225217,China)

机构地区：[1]上海海事大学商船学院,上海201306 [2]招商局金陵鼎衡船舶(扬州)有限公司,江苏扬州225217

出　　处：《机电工程技术》2022年第3期67-73,共7页Mechanical & Electrical Engineering Technology

基　　金：上海市科技计划项目(编号:20DZ2252300)。

摘　　要：采用SVM对不同运行工况下的柴油机增压换气系统进行故障诊断,为了提高SVM对柴油机换气系统故障诊断效率和准确率,选取最佳的支持向量机模型具有重要意义。热工参数相对偏差分析可以解决柴油机不同运行工况下的标准建模问题,分别采用交叉验证、网格搜索算法、粒子群算法、遗传算法对支持向量机模型进行参数(c、g)寻优,且得出网格搜索算法最适合换气系统的故障诊断,寻优的SVM分类准确率也最高,基本上都在90%以上,然后以此为基础,提出改进的网格搜索算法,分类准确率与传统网格搜索算法相差不大,但是时间大幅度减少,达3倍左右。研究成果为柴油机换气系统的故障数据分类识别提供更有效的方法参考,也为柴油机系统的诊断提供训练样本和趋势分析。In order to improve the failure diagnosis efficiency and accuracy,it is of great significance to select the best SVM model for the failure diagnosis of turbocharged ventilation system of diesel engine under different operating conditions.Relative deviation analysis of thermal parameters can solve standard modeling problems under different operating conditions of diesel engine,cross validation,grid search algorithm,particle swarm optimization algorithm,and genetic algorithm were used to optimize the parameters(c and g)of the SVM model to find the SVM model parameter algorithm suitable for fault diagnosis of marine pressurized ventilation system,grid search algorithm was most suitable for fault diagnosis of ventilation system,and the optimized SVM classification accuracy was also the highest,basically above 90%.Based on this,an improved grid search algorithm was proposed,with the classification accuracy not much different from that of traditional grid search algorithm,but time has been greatly reduced to about three times.The research results provide a more effective method reference for the fault data classification and identification of diesel engine ventilation system,and also provide training samples and trend analysis for the diagnosis of diesel engine system.

关 键 词：SVM 相对偏差 改进网格搜索算法 

分 类 号：U664.1[交通运输工程—船舶及航道工程]