Modélisation aux grandes échelles de la combustion Dual-Fuel

Dual-fuel combustion (DFC) is a promising concept for low-emission, highly-efficient internal combustion engines. In DFC, the injection of a highly reactive pilot fuel ignites a premixed low-reactivity bulk charge, which supplies the majority of the released energy. The combustion process involves the transition from pilot auto-ignition to premixed flame propagation under highly turbulent conditions. Large-Eddy Simulation (LES) has become a widely used tool to predict the combustion physics and turbulent flow in internal combustion engines. However, the presence of multiple combustion regimes poses a challenge when simulating dual-fuel engines, since existing models usually rely on assumptions regarding a specific combustion regime. This thesis develops a new combustion model for dual-fuel engine simulations based on the Thickened Flame Model (TFM) approach. TFM directly resolves the chemistry of an artificially thickened flame front on the grid. This approach is well validated for premixed combustion but fails to properly predict auto-ignition, because thickening artificially delays ignition. This work proposes a new formulation, called TFM-DF, which modulates the thickening factor to the combustion regime. A burnt gas detection is introduced as a reliable criterion to indicate the transition to flame propagation. This regime criterion is combined with a temporal relaxation of the thickening factor based on the value of a transported ignition sensor. Thanks to this formulation, TFM-DF prevents thickening during the auto-ignition phase and progressively increases the thickening factor to revert to a classical TFM approach during flame propagation.Results obtained in one-dimensional dual-fuel auto-ignition cases confirm the validity of the proposed formulation. An academic case of dodecane pilot ignited methane dual-fuel jet ignition is set up to validate the model under 3D turbulent conditions. Fully flame resolved simulations provide reference data for detailed model validation. TFM-DF achieves good agreement to the reference in 3 operating points with varying pilot jet sizes and injection durations.To evaluate the model for industrial use, LES of n-heptane/methane combustion in a single cylinder Caterpillar dual-fuel engine are validated against experimental pressure traces. TFM-DF operates in a stable way and successfully avoids thickening the auto-ignition. These results demonstrate that TFM-DF is functional even at low grid resolutions required to carry out industrial engine simulations at achievable computational cost. The engine simulations reveal limitations of equilibrium-based efficiency models under transient and highly stratified conditions. To overcome the resulting errors, this work proposes a non-equilibrium efficiency correction that relaxes the efficiency factor using a transport equation. An algebraic initial wrinkling model to initialize the efficiency transport equation is also developed and validated in an academic HIT setup. In the industrial engine case, the non-equilibrium model improves the agreement to the experimental data during the transient conditions shortly after transition. Finally, a variation of the pilot injection timing is investigated in three operating points. Increased advance in injection timing leads to higher pressure rise rates, eventually resulting in rapid consumption of the bulk charge. TFM-DF reproduces this behavior qualitatively but fails to predict the rapid pressure rise indicating “aggressive” combustion, pointing to a potential shortcoming of the chemical mechanism.