The main topic of this work's research is obtaining the methodological framework for large eddy simulation of turbulent premixed combustion. Obtaining this framework implies good knowledge of the state-of the-art in LES combustion, understanding the sub-grid scale (SGS) modelling of combustion and turbulent transport, proper definition of boundary conditions, quality mesh generation and proper numerical setup. In this work two different approaches have been used for SGS modelling of turbulent transport in Smagorinsky modelling framework. The first approach is the approach with constant value of Smagorinsky parameter. This approach is heavily relying on the chosen value of the parameter, which depends on flow configuration and simulation expertise and can introduce modelling errors if not chosen properly. Second approach is the Coherent Structure Method (CSM) approach, which is able to modify the Smagorinsky parameter according to the local coherence in the velocity field. Additionally, the CSM approach is in this work first-time used in the scope of the combustion modelling. The two approaches are mutually compared according to several criteria. The most significant comparison criteria is the comparison against available experimental data for two significantly different premixed combustion flames. The two approaches are also compared according to their level of numerical and discretization errors. Simulation results show that with the proposed methodological framework it is possible to get results which are comparable with experimental within approximately 20% of local discrepancy, with remark that CSM approach better reproduces fluctuations in the flow field. Furthermore, it is evaluated that probable cause of local discrepancy in validation comes from too excessive influence of combustion SGS modelling in regions of the flame front. This is directly related to insufficient resolution of the mesh resolution, even though the formal criteria of 80% of the resolved-to-total turbulent kinetic energy was satisfied. Conclusion is that the proposed methodological framework can be used for simulation of turbulent premixed combustion and that further improvement in result validation can be obtained by reducing the cell size in the flame region and further monitoring of numerical and discretization errors.