Resolution of droplet-scale processes occurring within engine sprays in multi-dimensional Computational Fluid Dynamics (CFD) simulations is not possible because impractically refined numerical meshes or time steps would be required. As a result, simulations that use coarse meshes and large time steps suffer from inaccurate predictions of mass, momentum and energy transfer between the spray drops and the combustion chamber gas, or poor prediction of droplet breakup and collision and coalescence processes. Several new spray models have been proposed to address these deficiencies, including use of an unsteady gas jet model to improve momentum transfer predictions in under-resolved regions of the spray, a vapor particle model to minimize numerical diffusion effects, and a Radius of Influence drop collision model to ensure consistent collision computations on different meshes. The present work combines these models with improved KH-RT models to improve the consistency of drop breakup predictions. A modified mean collision time model is also proposed to reduce timestep dependency of droplet collision prediction. The models have been implemented into the KIVA CFD code and are demonstrated to achieve independency with respect to both mesh sizes and time steps. The code was validated for non-evaporating and evaporating sprays, and also for diesel engine simulations with variations of larger than one order of magnitude in mesh cell volume and around two orders of magnitude in time steps. The numerical results were found to match available experimental measurements very well, including spray tip penetration, local drop velocity and Sauter mean diameter (SMD) and averaged mean diameter (AMD) of non-evaporating diesel sprays. The new spray models were also applied to simulate evaporating sprays and good agreement was found with measured liquid and vapor penetration lengths. Finally, the engine simulations were also found to agree well with experimental engine data.