Numerical Simulation of Internal Sprays in a Constant Chamber Using Large Eddy Simulation Techniques

For conventional compression ignition (CI) engines, the phenomena of spray fuel injection are important processes. The challenge of spray simulation is to accurately capture complex physical phase changes and interactions between vaporized fuel and ambient air. The present work describes numerical simulation of spray structures in a constant vessel chamber. Large eddy simulation (LES) technique is used to simulate the complex spray structure. The prediction results from the standard Smagorinsky (LES-SML) and the dynamic structure (LES-DS) based LES turbulence model are mainly compared, while the commercial CFD software package ANSYS-Forte is employed for this purpose. The non-reacting spray case is simulated with diesel surrogate such as n-dodecane at ambient temperature of 900 K. The present research starts with an initial grid size of 2 mm, and uses a solution adaptive grid refinement (SAM) to obtain the minimum grid sizes of 0.125 mm. The existing experimental data from the database ECN are employed to validate, and the results from time averaging RANS simulations are included to compare the spray global trends. The results showed that the similar global spray characteristics results from LES-DS and LES-SML were captured. The over-predicted results were observed during the early stage when time after start of injection was less than 0.3 ms. However, the dynamic structure model (LES-DS) was able to capture gradients of vapor penetration well compared to existing experimental data, while the development was higher than those of LES-SML results during the last stage of injection. The liquid penetration length predictions from LES-DS models show good agreement with the experiments at a certain point, whereas the liquid penetration length predictions from LES-SML steadies at different levels. The global trends of mixture fraction, gas-phase temperature and gas phase velocity along the axial distance were also presented. The simulated mixture fraction performed the maximum value near nozzle exit and decreased along the axial locations. For temperature, a cooling in the central zone was observed, while the highest value of predicted gas velocity was near the tip of spray centerline. Moreover, corresponding with temperature and mixture fraction simulated contours, more large spray structures were observed with LES-DS model, and LES-DS model can provide a better local information when compared with those of LES-SML model. In conclusion, the LES-DS model was a better choice for spray simulation when compared with the standard Smarkorinsky model.