A System Identification Approach to the Modelling of Engine Transients 900237
Mathematical models of transient engine dynamics have traditionally been produced with analysis from first principles of the physical processes within the engine. However, system identification techniques allow a ‘black box’ model to be derived purely from experimental data. These discrete time models have the advantage of being both straightforward and in a form which is suitable for digital control algorithm design and implementation.
This paper describes work which is underway to identify the transient dynamics of a four cylinder 2.0ℓ DOHC multi-point injection engine. The experimental arrangement allows the fuel injection for each power stroke to be individually controlled while data acquisition is synchronised with the engine crankangle.
The requirements for the input signal to the engine are discussed and methods used for data analysis are described.
Results showing the response of IMEP to perturbations in fuelling are presented. These demonstrate that the calculated models can provide an accurate simulation of the IMEP produced by each power-stroke at operating points for which misfires are infrequent. However, the inherent non-linearity of misfires at lean operating points is shown to reduce the usefulness of the models. It is therefore concluded that a system identification approach is most suitable for aiding the design of stoichiometric controllers.