The Design Of On-Board Engine Diagnosis System Based On SVM

With the rapid development of car industry, traffic accidents cause veryserious personal injury or death, and now it has become a social problem, whichcannot be ignored. As people’s demand for car function becoming higher andhigher, the structure of the car becomes more and more complex, and anincreasing number of electronic equipment is used in car, so it becomes more andmore difficult for automotive fault diagnosis and elimination. Therefore aneffective method in automotive fault diagnosis is what we expect. Based on theincrease of difficulty on automotive diagnosis, a special online fault analysis andforecasting device is designed in this paper, so that the cause of car malfunctioncan be detected by the owners in advance, which can be self-repaired or promptlysent to the maintenance department to repair. This equipment shows the cause ofthe problem very intuitively, so the serviceman can remove the trouble quickly.A vehicle fault diagnosis system based on the ARM9processor is inventedin this article. The system is an embedded detection equipment. Based on thetypical data of engine running state on-time transferred from CAN bus, throughthe established fault diagnosis model, analyze and process these data. Thediagnostic system will judge the engine running state based on the correspondingfault code, and predict the probability of possible fault, eventually get thefunction of fault diagnosis and early warning.The main research content of this paper is following:On the embedded hardware side, the M3C2440is used as the core chip ofhardware design, and CAN communication interface, LCD, USB interface, audio,power supply circuit are extended on the bottom plate. At the same time, thecorresponding schematic diagram and the drawing and generation of PCB isfinished.On the embedded software side, transplantation and customization of Linux operating system and communication protocol of CAN and software compilerand debugging of each module in the cross-compile environment are completed,and use QT compiled graphical interface to complete the man-machine interfacedesign.On the intelligent learning side, this article further studies on the algorithmof support vector machine (SVM). From the study we know that the parameter ofpenalty factor and radial basis function parameter play a very important role inthe classification accuracy, so the genetic algorithm is used to optimize theparameters, and use this optimization algorithm into fault diagnosis. Take enginemisfire as an example, establish the mapping relations between gas volumefraction and the cause of the misfire; use principal component analysis (PCA)method to reduce the dimension of data, and reduce the amount of datacalculation.