Parametric comparison of well-mixed and flamelet n-dodecane spray combustion with engine experiments at well controlled boundary conditions

Extensive prior art within the Engine Combustion Network (ECN) using a Bosch single axial-hole injector called ‘Spray A’ in constant-volume vessels has provided a solid foundation from which to evaluate modeling tools relevant to spray combustion. In this paper, a new experiment using a Bosch three-hole nozzle called ‘Spray B’ mounted in a 2.34 L heavy-duty optical engine is compared to sector-mesh engine simulations. Two different approaches are employed to model combustion: the ‘well-mixed model’ considers every cell as a homogeneous reactor and employs multi-zone chemistry to reduce the computational time. The ‘flamelet’ approach represents combustion by an ensemble of laminar diffusion flames evolving in the mixture fraction space and can resolve the influence of mixing, or ‘turbulence-chemistry interactions,’ through the influence of the scalar dissipation rate on combustion. Both combustion methodologies are implemented in the Lib-ICE code which is an unsteady Reynolds-averaged Navier-Stokes solver with k-ɛ turbulence closure based on OpenFOAM® technology. Liquid length and vapor penetration predictions generally fall within the experimental measurement uncertainty at 7.5% O2, 900 K, and 15.2 kg/m3. Flame liftoff length, cylinder pressure, apparent heat release rate, and ignition delay time from the two computations are compared to experiments under single parametric variation of ambient density 15.2 kg/m3 and 22.8 kg/m3, temperatures of 800, 900, and 1000 K, at 13, 15 and 21% Oxygen and injection pressure of 500, 1000, and 1500 bar. Both models generally provide apparent heat release rate maxima to within the uncertainty. The flamelet model better predicts the sensitivity of lift-off length while the well-mixed model better predicts ignition delay.