The problem of adaptively controlling the mixture ratio can be reduced to the problem of identifying the respective non-linear system dynamics [ 19].
In the present paper, convenient models of the significant dynamic processes, i.e., intake manifold, wall-wetting and oxygen sensor dynamics, arc deduced. We will separate the analysis in terms of an air and a fuel path. Concerning the fuel path we restrict our attention exclusively to linear sensor models in order to keep the modelling overhead small. Nevertheless, considering the overall dynamics, we will have to deal with some inherent non-linearities.
Suitable parametrizations of these models with respect to the demands imposed by the filtering techniques are then introduced. In the case of linear dynamics we aim to achieve a linear regression form whereas in the case of non-linear dynamics, we will augment the system state and apply extended Kalman filter theory.
It turns out that the proposed Kalman filtering methods provide highly effective means to solve the present classes of identification problems.