Assessment of the performance of conventional spray models under high pressure and high temperature conditions using a “Design of Experiments” approach

An integrated Design of Experiments (DoE) with Reynolds Averaged Navier Stokes (RANS) approach is suggested and implemented to model turbulent spray combustion. In the automotive industry, DoE is often combined with an optimizer and is used to find an optimum set of internal combustion engine calibration parameters for set criteria at reduced experimental effort. The novelty of the numerical approach suggested here, is that the methodology is adjusted to provide an optimal set of model “tuning constants” for the 3D CFD simulations which best matched experimental data at three conditions taken from the Engine Combustion Network (ECN) database. Multi-variable DoE were run for each condition. The goal of this work is to use these DoE derived coefficient sensitivities and link the observed trends to real physical processes. The analysis is based on both microscopic (droplet statistics) and macroscopic (liquid & vapor penetration and heat release) spray characteristics. Results indicate that a single coefficient matrix exists that can model a wide range of injection pressures. This finding is important since it paves the way for using conventional spray models for high pressure injection conditions, if tuned appropriately. Moreover, a separation of the model coefficients between the ones that affect mostly nonreactive predictions and the ones that affect reactive cases is suggested. This reduces the computational cost of the suggested methodology since the reactive DoE can be restricted on a sub-set of coefficients. The physical meaning of these coefficient groups reveals the link between the various sub models when turbulence and evaporation are the only processes acting on the droplets as well as when these processes are coupled with combustion.