Real-time implementation of multiple model based predictive control strategy to air/fuel ratio of a gasoline engine

Growing safety, pollution and comfort requirements influence automotive industry ever more. The use of three-way catalysts in exhaust aftertreatment systems of combustion engines is essential in reducing engine emissions to levels demanded by environmental legislation. However, the key to the optimal catalytic conversion level is to keep the engine air/fuel ratio (AFR) at a desired level. Thus, for this purposes more and more sophisticated AFR control algorithms are intensively investigated and tested in the literature. The goal of this paper is to present for a case of a gasoline engine the model predictive AFR controller based on the multiple-model approach to the engine modeling. The idea is to identify the engine in particular working points and then to create a global engine's model using Sugeno fuzzy logic. Opposite to traditional control approaches which lose their quality beside steady state, it enables to work with satisfactory quality mainly in transient regimes. Presented results of the multiple-model predictive air/fuel ratio control are acquired from the first experimental real-time implementation on the VWPolo 1390cm3 gasoline engine, at which the original electronic control unit (ECU) has been fully replaced by a dSpace prototyping system which execute the predictive controller. Required control performance has been proven and is presented in the paper.