基于物理模型和支持向量机的柴油机冷却系统故障诊断算法  

Fault diagnosis algorithm of diesel engine cooling system based on physical model and support vector machine

作　　者：朱观宏 宋康[1] 谢辉[1] 陈韬[1] 钱振环 ZHU Guanhong;SONG Kang;XIE Hui;CHEN Tao;QIAN Zhenhuan(State Key Laboratory of Combustion of Internal Combustion Engines,Tianjin University,Tianjin 30072,China)

机构地区：[1]内燃机燃烧学国家重点实验室,天津大学,天津300072

出　　处：《汽车安全与节能学报》2020年第4期529-537,共9页Journal of Automotive Safety and Energy

基　　金：国家重点研发计划(207YFE0102800);国家自然科学基金资助项目(51906174)。

摘　　要：为了对可监测变量少、时间尺度大、耦合性强的柴油机冷却系统的故障进行有效监测和准确诊断,设计了一种结合同步运行物理模型和小样本数据驱动的智能诊断算法。算法中建立了一个基于冷却系统物理原理的简化模型。利用模型实时预测的水温和实际水温的残差作为故障诊断的信息依据,并将信息输入支持向量机(SVM)进行分类,辨识故障原因。利用GT-SUITE柴油机模型对算法进行仿真测试,在车辆故障工况下对算法进行了试验测试。结果表明:该算法对故障的识别准确度在97%以上,诊断用时在45 s以内,显示出该诊断算法对冷却系统故障有良好的监测能力和准确辨识的潜力。An intelligent fault diagnosis algorithm was developed by using synchronous operating physical model and small sample data-driven to effectively monitor and accurately diagnose the faults of the cooling system of diesel engine with strong coupling,large time scale and few variables to be monitored.A simplified physical model that based on the physical principle of cooling system was built in the algorithm.The support vector machine(SVM)was used to classify the fault information based on the residual of actual water temperature of the engine and the predicted water temperature of the synchronous operating model to identify the cause of the fault.The algorithm was tested on a precisely calibrated GT-Power diesel engine model and a real bus with fault.The results show that the identification accuracy of the algorithm is above 97%,and the diagnosis time is within 45 s after fault occurred;the algorithm has good monitoring ability and accurate identification potential for cooling system faults.

关 键 词：柴油机 冷却系统 故障诊断 物理模型 支持向量机(SVM) 

分 类 号：TK428[动力工程及工程热物理—动力机械及工程]