Prechamber ignition technology receives increasing attention due to its considerable improvement on engine combustion efficiency and stability. However, fundamental knowledge concerning flame propagation inside the pre-chamber and jet formation in the main chamber is still quite scarce. In this study, a small (<0.5% VTDC) un-scavenged pre-chamber was tested in a medium size gas engine with pressure transducers installed in both pre- and main chamber. Three-dimensional computational reactive fluid dynamics Reynolds-averaged Navier–Stokes simulations were carried out using a level-set combustion model –G-equation – towards improved understanding of the combustion processes occurring inside the pre and main chamber. The characteristics of the turbulence and the flame at locations just ahead of the propagating turbulent flame front were recorded and analysed by means of the well-known Borghi–Peters diagram. The results revealed that the characteristics of the flame inside the pre-chamber differed greatly from those inside the main chamber due to considerably reduced turbulent length scales. In addition, a wide range of turbulence intensity and length scales are covered throughout the combustion event, presenting a significant challenge to modelling of flame–turbulence interaction. Various turbulent flame speed ( ST) closures widely used in internal combustion engine simulation were therefore assessed and the ranges of their respective model constants explored. A correlation for ST is subsequently proposed by blending two formulations of Gülder developed for small and large scale turbulence, respectively, and compared to the well-known Peters correlation. With appropriate model constants, both successfully reproduce the pre and main chamber combustion for the reference case in terms of evolutions of cylinder pressure, heat release rate and pressure difference between pre and main chamber. Following successful calibration of the reference operating condition, variations in engine speed, load, spark timing and lambda were calculated using both correlations, demonstrating encouraging predictive capabilities of the proposed modelling strategy.

No abstract

No abstract