Alternative fuels derived from renewable energy offer a promising approach to reduce greenhouse gas emissions while utilizing existing infrastructure. To fully exploit their potential, a detailed understanding of the complex pollutant formation process is required. However, the available models are not accurate enough, especially when predicting soot. A major challenge is modeling turbulence-chemistry interactions (TCIs). Since the gas-phase chemistry is significantly influenced by the local mixing field, TCIs are often considered using mixture fraction-based models, which have relatively low computational costs. These models solve for the mixture evolution in physical space, while reactive scalars are calculated based on mixture fraction. Conventional models, however, often disregard that scalar transport in mixture fraction space must be formulated consistently with the mixture fraction evolution in physical space. This thesis demonstrates that this inconsistency leads to significant mass conservation errors for certain scalars, such as soot mass. Another contribution of this work is the development of two analytically consistent mixture fraction- based models, which exhibit better predictive accuracy than conventional approaches, particularly regarding soot formation. These new models are applied in Large Eddy Simulations (LES) of auto-igniting spray flames. It is shown that using dissipative numerical methods in the LES solver also causes significant scalar mass conservation errors, which can be substantially reduced by employing low-dissipative methods or discretely consistent approaches. The revised combustion model is finally applied to simulate a compression ignition engine running on biofuels.weiterlesen