The use of natural gas in internal combustion engines (ICEs) improves thermal efficiency and reduces exhaust emissions at lean mixture operating conditions. However, as the mixture is leaned out beyond the Lean Misfire Limit (LML), several technical problems are more likely to occur. The flame propagation speed gradually decreases, leading to a slower heat release, thus increasing the occurrence of misfiring and incomplete cycles. This gives in turn a steep increase of CO and UHC emissions, and of cycle-by-cycle variations. In order to limit the above-mentioned problems, several solutions have been proposed so far. Among them, the stratification or the partial stratification of the charge has been demonstrated to successfully extend the lean limit if compared with traditional lean burn engines. This result has been accomplished through the formation of a richer mixture in the vicinity of the spark plug location, improving the stability of the combustion and ignition processes. Some works have demonstrated that this strategy allows for reducing the cycle-by-cycle variation. In the context of the development and optimization of such strategies, Computational Fluid Dynamics (CFD) techniques are key to thoroughly understand the phenomena occurring during the evolution of mixing and combustion processes. A detailed description of the turbulent parameters related to the PSC injection process is in fact crucial to reliably represent the mixture stratification. The LES approach has already been validated to provide high accuracy in capturing the scales of motion typical of this mixing process into a Constant Volume Combustion Chamber (CVCC) using an OpenFOAM based solver. The previous works done have highlighted the influence of the main turbulent parameters close to the spark location on the mixture formation and on the subsequent combustion process, showing that a high turbulence level provides an enhancement in the kernel development speed and stability. This work aims at extending the lessons learnt into the CVCC device toward the behavior of a real engine. The mixture formation process has been carefully studied with a LES approach to provide a comparison of the local flow conditions between the two case studies (CVCC and engine), to highlight their differences and similarities. The CONVERGE CFD Toolbox has been used to carry out the engine simulations due to its excellent capability of dealing with moving geometries and the possibility of using an Adaptive Mesh Refinement (AMR) to deliver high resolution where high temperature and velocity gradients are calculated without significantly increasing the total number of computational cells. Preliminary results with regard to the representation of the combustion process have been presented already, compared with experimental data collected at the University of British Columbia by E.Chan et al., showing a good agreement between such data and the numerical ones in terms of pressure trace over time. The solver allowed for capturing the performance enhancement due to the adoption of the PSC strategy if compared with the homogeneous counterparts, thus confirming the potential of CFD as a valid alternative to a purely experimental approach.