Model Calibration for Spray Penetration and Mixture Formation in a High Pressure Fuel Spray Using a Micro-Genetic Algorithm and Optical Data

Correct prediction of mixture formation in fuel sprays is a prerequisite in the framework of 3D CFD engine simulations using a reliable combustion model. To understand the process of fuel evaporation and mixture formation in dense atomized sprays a simultaneous Mie/Shadow imaging technique and a 1D-linear Raman scattering technique are applied to investigate spray formation of high pressure direct injection. Mixture composition is one of the most important parameters in fuel sprays and is difficult to measure because liquid and vapor phases appear simultaneously. Ethanol is used as a model fuel since the phase-dependent spectral shift of the OH stretching vibration allows the Raman signal separation of liquid and vapor phase. The investigations are carried out in a high temperature high pressure injection chamber, where pressures and temperatures can be set up to 5MPa and 800 K. A passenger car common rail injection system equipped with a 5 hole nozzle (VCO) with double needle guidance is used. The results provide simultaneous information on the propagation of liquid/vapor phase and quantitative results of air/fuel ratio for comparison with 3D CFD simulations. The commonly wide used discrete droplet model (DDM) with its sub-models accounting for droplet breakup, collision, coalescence and evaporation is applied in the numerical simulations. Since none of these mentioned sub-models can be studied and validated independently, the appropriate model parameters and initial conditions are somehow unsettled.

In this study, 6 parameters are identified which have to be correctly adjusted in order to get a good agreement between the numerical and experimental results. Since the manual search for an optimal point in a 6 dimensional parameter space is nearly impossible, a Micro(μ)-Genetic Algorithm (μGA ) is applied. The outcome of this optimization strategy is a set of parameters that resolves the spray penetration and mixture formation correctly when compared to the experimental data.