Research On Remote Predicting Diagnosis Of Engine Fault Based On DEBA-SVM Algorithm

With the rapid development of social economy,the demand for automobile has increased significantly.While the demand for the development of automobile intelligent operation and diversified functions is met by China's automobile industry,the types of automobile fault are also increasing.As the core of the automobile,engine is the high-frequency parts of the fault.Therefore,the condition monitoring and fault prediction of automobile engine become the top priority.This paper used support vector machine,4G network,CAN bus and other technologies to develop a set of engine remote fault predicting diagnosis system to achieve remote fault real-time prediction,condition monitoring,data storage and other functions for vehicle engines.Main tasks are as follows:(1)The overall structure and functional design of the engine remote fault predicting diagnosis system.The vehicle terminal hardware was designed based on the principle of "core board + baseboard".CAN bus was selected and CAN communication protocol was developed to realize the data transmission between the vehicle terminal and the engine controller.The 4G network was used to realize the wireless communication between the vehicle terminal and the remote fault prediction software.The support vector machine was selected to build a fault prediction model to realize the predicting diagnosis of the engine mechanical fault.(2)Vehicle terminal platform design.The S5P4418 control chip was selected as the vehicle intelligent terminal main control chip.EC20 was selected as the 4G communication module.MCP2515 was selected as the CAN communication module.Linux system was used for the control software program design.The CAN bus communication program and 4G network communication program were designed respectively and the vehicle terminal platform design was completed.The CAN communication module and 4G communication module of the vehicle terminal were tested separately to verify the communication functions.(3)Research on pre-diagnosis method of engine fault.The normalized processing method and the grey relational analysis method were used to process the engine operating parameter sample data respectively.An engine fault prediction control algorithm model based on SVM was built.The DEBA algorithm was used to optimize the support vector machine key parameters.The fault prediction model was tested by the sample data and the accuracy of fault prediction diagnosis reached 97.5%,which verified the feasibility of the engine fault prediction model.(4)Development of remote fault prediction software.Python was used to design and develop remote fault prediction software based on the Django framework.MySQL was used to establish a database to store engine operating parameter data.The interaction between the software and the database was realized.The web interface was designed to intuitively display real-time monitoring and fault pre-diagnosis functions of the system.Email was used to notify users of fault diagnosis information.System test platform was built to test the overall system functions.The test results showed that the system could predict engine faults in time,which verified the effectiveness of the overall system functions.