Analysis of a tumbling motion using a clustering algorithm on dual-piv measurements: application to the in-cylinder flow of a miller cycle engine

Abstract

This paper aims to study the internal aerodynamics of a Miller cycle gasoline engine. This cycle uses an over-expanded stroke to recover more work, for the same amount of fuel, than the classical Otto cycle. To achieve this, the effective compression stroke is made smaller than the expansion one by closing the intake valves prematurely. The in-cylinder tumble flow then stops being driven by the valve jet before the piston reaches the bottom dead center, which disturbs the movement. The flow is observed here on a motored single-cylinder transparent engine coupled to a dual-PIV system. This system allows two velocity fields per engine cycle to be measured. The temporal evolution of the flow structures could thus be followed, while keeping the high number of instantaneous fields measured by the traditional PIV systems. The data are then well adapted to statistical analysis. A clustering analysis method, using the k-means algorithm and adapted to dual-PIV data, is described here. The first part presents a method defining a criterion on the instantaneous center of rotation of the tumble motion that can improve the rotation rates by 17%. A second analysis classifies the velocity fields into three groups of strong, medium and weak rotation intensity. The evolution of the characteristic quantities of each group shows that the movements with high rotation rates, created upstream of the compression phase, can give high turbulence rates close to the top dead center of the piston.Graphical abstract