The need for a stoichiometric air-to-fuel ratio in an SI engine with a catalytic converter makes the accurate knowledge of the air and fuel paths indispensable. This investigation is focused on the prediction of the air mass flow into the cylinder without the use of an air mass flow meter. A dynamical mean value engine model of the intake manifold has been derived. Combining a gain-scheduling and a self-tuning algorithm has been found to be a good strategy for the persistent adaptation of the intake manifold model to the changing ambient conditions and actuator parameters such as aging or malfunctions.
The adaptation algorithm is based on the direct identification of the air mass flows entering and leaving the intake manifold, thus the identified parameters can be interpreted as the throttle and the filling characteristics.
The recursive least squares algorithm has been used for parameter identification. Different modelling approaches and discretization methods have been applied for parametrization. The direct identification of the air mass flows by use of bilinear discretization techniques has brought the best results.
To reduce the noise level, a segment filter for identification has been developed. It calculates the mean value of the measure d signal from a collection of oversampled data. The use of the least squares algorithm has been shown to be a good approach to the prediction of the air mass in the cylinder without using an air mass flow meter. Yet the sensitivity of the least squares algorithm against coloured noise necessitates the investigation of other algorithms.