Statistical modelling of exhaust gas recirculation process in a turbocharged engine

This paper proposes a new methodology for constructing engine exhaust-gas-recirculation(EGR) rate model, which includes the procedure of determining the model structure, selecting model input items, and generating identification data. Simulator based on first principles models were utilized to investigate identification procedure. Then several candidate model structures are identified and estimated by Akaike Information Criterion(AIC) criterion. The redundant input terms are eliminated by using stepwise regression method to improve the model accuracy. Finally the design of experiment is introduced based on the D-optimal method.