Research On Fault Diagnosis Method Of Automobile Engine Cooling System

The cooling system of automobile engine is an important system to ensure the normal operation of automobile engine.Its performance will directly affect the safety performance and emission performance of automobile.Due to the large number of engine cooling system components and long-term operation in high temperature and high pressure environments,the cooling system is prone to occur various faults.Therefore,it is necessary to diagnose and locate the fault of engine cooling system so as to ensure the stability of the system and reduce engine damage caused by the fault of cooling system.In view of the characteristics of engine cooling system with many variables and large amount of fault diagnosis calculation,data dimension reduction(PCA,KPCA)method is introduced into the fault diagnosis of engine cooling system,and the data processing method of cooling system based on PCA and KPCA is studied in detail.Different data processing methods are used to process the sample data.On this basis,the fault diagnosis models based on RBF neural network and support vector machine(SVM)are designed respectively to achieve efficient and accurate diagnosis and location of engine cooling system faults.The main contents of this thesis include:Firstly,considering that there are many signal variables in the engine cooling system,the engine cooling system model based on AMESim simulation platform is established,the causes of the cooling system failure is analysed and the faults are classified.Through model simulation analysis,the relationship between different fault types and system variables is obtained,and the variables needed for fault diagnosis of cooling system are determined.Secondly,PCA and KPCA methods are used to reduce the dimension of cooling system data in order to solve the problem of high correlation and data redundancy between the collected sample data variables.By solving the covariance matrix of the pretreated sample data,the number of principal components is determined,and the principal component values are calculated to form a new sample set.The new sample set not only retains the original data information to a high degree,but also reduces the sample dimension,which is conducive to improving the efficiency of fault diagnosis.Then,respectively based on the sample data processed by different data processing methods,the RBF neural network is designed to diagnose the cooling system faults.Particle swarm optimization is used to optimize the structure parameters of RBF neural network.With the processed sample data as the input of RBF neural network and each fault type as the output of the network,the network is trained and the fault diagnosis of cooling system is carried out by using the trained network model.Then it is compared with the fault diagnosis method based on BP neural network.Then a fault diagnosis model of cooling system based on SVM is designed,and the parameters of SVM are optimized by genetic algorithm.The trained SVM model is applied to the fault diagnosis of cooling system.Then the k-nearest neighbor algorithm is used for fault diagnosis,which verifies the reliability of the SVM model.By comparing the diagnostic performance of different fault diagnosis methods,it can be seen that the diagnostic accuracy of SVM fault diagnosis method using normalized data processing method is the highest,and its accuracy rate is 96.7%.The k-nearest neighbor fault diagnosis using PCA to process data has the shortest running time,which is 0.1550 s,but its fault diagnosis accuracy rate is low,which is only 81.7%.The fault diagnosis method based on KPCA-RBF neural network can effectively balance the running time and diagnostic accuracy.Finally,in order to verify the reliability of the fault diagnosis method,an engine cooling system test bench is built and a rapid prototype of the controller is developed.The SVM fault diagnosis algorithm using normalized sample data is downloaded to the rapid prototype.The running bench is used to diagnose the cooling system faults.The experimental results show that the fault diagnosis algorithm designed can effectively identify the fault types.