Study On Cyclic Variation Of Indicated Mean Effective Pressurein A Gasoline Engine

Cyclic variation is the special combustion phenomenon in spark ignition engines, which has a significant influence on the overall performance of engines. The minimization of cyclic variation could improve the power performance, fuel economy and emission. The studies on cyclic variation controlling are not deep enough and there is few prediction model to describe the non-linear relationship between the cyclic variation and its affect factor. So the experimental researches of the cyclic variation in a gasoline engine are accomplished at the external characteristic condition and NEDC condition in this study. Based on the statistical analyses of experimental data, the changing characteristics of cyclic variation are analyzed and the prediction models are established.Firstly, the comparison of relative deviation is employed and the experimental data at external characteristic condition and NEDC condition are analyzed to investigate the differences of cyclic variation with different sampling numbers. It is believed that the most rational sample number at overall test points is 4000 cycles in this study.Based on the correlation analyses of cyclic variation parameters, it is found that the IMEP has little correlation with peak pressure. So they can’t be replaced with each other when they are considered as the characteristic parmeters of cyclic variarion. The flame development duration and the main combustion duration are strongly associated with the IMEP at some conditions. It is indicated that both of them have a powerful influence on the cyclic variation of IMEP at this conditions.Meanwhile, the change trends of the COV(Coefficient of Variation) of IMEP with speed, load, ignition timing and excess air ratio are analyzed. Results show that the COV of higher speed conditions are much higher than that of lower speed conditions. There is a point at low or medium load and the amplitude of variation is relatively higher in this area. The trends of COV present in a form of parabola with the change of ignition timing and the COV could reach a minimum value with the ignition timing of MBT. There is a sharp rise in the COV when the excess air ratio is higher than the best value which could minimize the COV.Finally, the prediction methods of the cyclic variation are explored in this study, using the statistical approaches of regression analysis and neural network. In the predictive models, the COV of IMEP is set as the input variable and six parameters are identified as the output variables due to their importance for the fluctuation of IMEP, including speed, load, excess air ratio, flame development duration, CA50 and main combustion duration. The relative deviation between the predicted value and the experimental value of COV are calculated and the prediction accuracy of models are evaluated according to the relative deviation. Results show that the prediction accuracy of regression models is a bit higher. The percentage of validation points, where the prediction error are smaller than 10%, is up to 83.3%. Furthermore, the prediction accuracy of the model established by back propagation neural network is a bit higher than that of regression model. So the generalization ability of neural network is much better with the same experimental data.