Model-Based Calibration For Air Path Control System Of Gasoline Engines

With the introduction of new electronic control technology in modern engines, theengine control system is becoming more and more complex, which makes engineers ex-tremely anxious about its mechanism modeling. Meanwhile, the model-based controlalgorithm generally is very complicated, difcult to calculate, and it is hard to meet theneeds of real-time capability. So, an efective map is always used in the process of enginemodeling and control, on one side, map modeling instead of mechanism modeling is easyto implement, on the other side, map stored in ECU takes smaller storage units thanmodel and its control algorithm.The calibration of engine control system requires considerable engineering resources.Traditional calibration methods use a combination of engine dynamometer and vehicletesting, which is time-consuming. Full Factorial Experiment is a traditional calibrationmethod by testing all possible combinations of parameters on an engine bench to get theoptimal value. However, this approach is deemed too detrimental on the grounds of timeand costs.The optimized calibration technology is considered as one of the most importantissues in modern engine control.In this paper, in order to reduce the workload of calibration and development costs,the Model-based Calibration (MBC) technology is applied to develop the torque controlsystem of SIS4gasoline engines, which consists of the following three parts:(1)Various design of experiment methods are being explored in this paper. The coreof model-based calibration technology is the selection of operation test points in designof experiment (DoE). The operation points need to be tested and relevant data should bemeasured on test bench, it is important to make reasonable experimental arrangementsin order to get desirable results and scientifc conclusions with fewer number of tests andshorter development cycle. The method of Latin Hypercube Sampling (LHS) space-fllingdesign has been chosen in this paper by the perspective of reducing the number of testpoints without afecting the results of the testing.(2)For the control thought of the torque demand as the center that have already completed in our lab, the pedal model which reflects torque demands in the form of mapis added by MBC method. Meanwhile, in order to obtain the maximum output torque anddo not afect the emission and detonation, the desired air-fuel raito λ and spark advanceangle S are obtained by MBC method. Finally, the ofine simulation experiments arepresented to validate the efectiveness of the designed torque control system.(3)This paper designs an optimal calibration method based on artiftial neural net-work(ANN) model, through the study of MBC toolbox in Matlab. The DoE, frst, needsto be designed by this method to get the relavant response data. Then an ANN modelcan be trained with these data to meet the requirement of accuracy. Next, genetic algo-rithm(GA) is taken to search for the optimal parameters. Finally, the ofine simulationexperiments with the high fdelity engine model enDYDA are presented to validate theefectiveness of this method.As the development of engine control system, the model-based calibration thechnol-ogy is improved continuously. Some problems in this paper need to be solved gradually.For example, the response data is collected on virtual engine bench instead of real enginebench, which provides a great convenience to get the steady and accurate data. However,there is external noise appearing at any time and unsteady operating mode when realengine works, which will have influence on the accuration of calibrated map.