Study On Calibration Technology For Electronic Controlled Gasoline Engine Based On Neural Network

With the exhaustion of petroleum resources all over the world, the constant deterioration of the global environment and the continual improvement of the requirements for the automobile technology, the automotive industry is facing severe challenges. As electronics is regarded as one of the most important method for reducing the exhaust emissions and improving the power performance and economical efficiency, people pay more and more attention to the development and study of engine electronic control technology. In order to follow the China Emission Standard and get better power performance and economical efficiency, a large number of matching experiments between electronic control unit and engine are carried out.This thesis takes domestic produced 479Q gasoline engine as the research subject. Through much literature reading, this thesis fully and accurately analyzes the development and present situation of domestic and foreign engine calibration technology. With the development of the engine calibration gradually from the off-line calibration to the online calibration, the automatic calibration and the transient state calibration, the degree of the automaticity is getting higher and higher, the control parameters are getting more and more, the precision of the control parameters is also greatly enhanced. According to analyzing the restriction and purpose of calibration, this thesis adopts the online automatic calibration method.The steady-state engine calibration test bench is built. AVL PUMA OPEN dynamic test bed is used as the calibration test bed. The integrated development environment of Electronic Engine Control is established by combining with the electronic control systems, emission performance test systems, calibration software and PC. It is easy to accomplish the on-line and automatic calibration of electronic control system by communicating and sharing the data between the ECU and the Up-PC.The calibration of steady state for 479Q gasoline engine is carried out. Aiming onλclose-loop control, the experiment is performed point by point. Knock sensor is used to ensure the ignition advance angle of the closed-loop control. Basing on this, this thesis obtains the map of fuel inject pulse width and ignition advance angle . After the study of artificial neural network theory, the thesis builds an electronic control system parameter model of the steady-state condition to forecast the electronic control parameters by using neural network highly nonlinear mapping ability. This thesis completes the BP neural network structure design and the sample training by choosing the number of network layer and neuron. The results show that the model can accurately forecast the electronic control parameters of the steady-state condition. This is a new method for fast and accurately forecasting the electronic control parameters of the engine steady-state condition.