STOCHASTIC MODELLING AND ESTIMATION FOR CYCLIC PRESSURE VARIATIONS IN SPARK IGNITION ENGINES

A new method of fitting linearised, parametric stochastic models of cycle-by-cycle variations of pressure, during the combustion region of a spark ignition petrol engine, is described. The technique is based on stochastically fitting the combustion models to the covariance function of the measured pressure fluctuations, obtained by averaging over the entire ensemble of measured cycles. Comparisons, for two specific combustion models, with corresponding results obtained by deterministic fitting on a cycle-by-cycle basis, show that the new method gives a similar degree of fit, but with much improved computational efficiency. It is also demonstrated that the degree of fit to the data can be further improved by modelling the residual error between the data and the combustion models in terms of Chebyshev polynomials: the parameters in these polynomials may be determined by stochastic fitting. The technique has wider applications in the condition monitoring of rotating machinery.