Abstract
Along with the recent increase of the interest in lower exhaust emissions and better fuel economy of automobiles, it has been required to improve the accuracy of the engine air-fuel ratio control. In improving the accuracy of the air-fuel ratio control, how to improve the estimation accuracy of in cylinder intake air amount is the challenge. For preceding studies, the estimation accuracy has been steadily improved, but the estimation error caused by the transient state remains. In this study, we propose to use Gaussian Process (GP), a statistic model, to estimate the in cylinder intake air amount takes to the transient state into account. To improve the estimation accuracy using GP, it is required to obtain the exhaustive data of engine operation, but it is not easy because of cost constraint. Therefore, we have studied a method to extract data which contributes to the estimation accuracy out of the limited measurement data. We have verified the impact on the estimation accuracy where modeling is done by extracting the training data from engine operation data according to a rule and the results are reported in this paper．
