This thesis describes one of the first computational works to investigate the physical feedback mechanisms associated with self-excited, combustion-driven instabilities in gas turbines. For this purpose, a novel numerical method based on large eddy simulation is devised. The method (called BOFFIN) uses a fully compressible formulation to account for acoustic wave propagation and applies a transported probability density function approach for turbulencechemistry interactions. The latter is solved by the Eulerian stochastic fields method and is complemented by two different 15-step/19 species chemical reaction schemes. This approach is shown to be flame burning regime independent and therefore highly applicable in the context of partially premixed gas turbine combustion. Combustion instabilities are a phenomenon often encountered in the late design stages of modern gas turbine combustors. Under certain conditions, these types of instabilities can develop into sustained limit-cycle oscillations with potentially severe consequences on a combustor’s operating behaviour. In order to study the various physical feedback mechanisms driving such limit-cycle oscillations, two different test cases are simulated in the present work. Firstly, the combined effects of thermo-acoustic and hydrodynamic instabilities are examined in the lab-scale PRECCINSTA model combustor. Secondly, the superposition of a longitudinal and azimuthally spinning instability mode is investigated in the industrial SGT-100 combustor.

Amongst the different feedback mechanisms identified and studied in these cases are: mass flow rate and equivalence ratio oscillations, as well as hydrodynamic phenomena such as flame angle oscillations, periodic vortex shedding and a precessing vortex core. It is further demonstrated that in addition to reproducing longitudinal instability modes, the applied LES approach is capable of accounting for modes acting in the transverse direction. Overall, the findings of this research project strongly suggest that BOFFIN is a reliable and accurate method for the prediction of self-excited combustion instabilities in gas turbines.