This work develops a computational framework for modeling turbulent combustion in multi-feed systems that can be applied to internal combustion engines with multiple injections. In the first part of this work, the laminar flamelet equations are extended to two dimensions to enable the representation of a three-feed system that can be characterized by two mixture fractions. A coupling between the resulting equations and the turbulent flow field that enables the use of this method in unsteady simulations is then introduced. Models are developed to describe the scalar dissipation rates of each mixture fraction, which are the parameters that determine the influence of turbulent mixing on the flame structure. Furthermore, a new understanding of the function of the joint dissipation rate of both mixture fractions is discussed. Next, the extended flamelet equations are validated using Direct Numerical Simulations (DNS) of multi-stream ignition that employ detailed finite-rate chemistry. The results demonstrate that the ignition of the overall mixture is influenced by heat and mass transfer between the fuel streams and that this interaction is manifested as a front propagation in two-dimensional mixture fraction space. The flamelet model is shown to capture this behavior well and is therefore able to accurately describe the ignition process of each mixture. To provide closure between the flamelet chemistry and the turbulent flow field, information about the joint statistics of the two mixture fractions is required. An investigation of the joint probability density function (PDF) was carried out using DNS of two scalars mixing in stationary isotropic turbulence. It was found that available models for the joint PDF lack the ability to conserve all second-order moments necessary for an adequate description of the mixing field. A new five parameter bivariate beta distribution was therefore developed and shown to describe the joint PDF more accurately throughout the entire mixing time and for a wide range of initial conditions. Finally, the proposed model framework is applied in the simulation of a split-injection diesel engine and compared with experimental results. A range of operating points and different injection strategies are investigated. Comparisons with the experimental pressure traces show that the model is able to predict the ignition delay of each injection and the overall combustion process with good accuracy. These results indicate that the model is applicable to the range of regimes found in diesel combustion.