This paper discusses the use of LES to predict the performance of an annular combustion chamber in stable operating conditions and in presence of self-exited dynamics. The availability of high-accuracy data taken in a full-scale combustion test facility allowed an extensive validation of the prediction capability. The analysis focuses on a small size heavy duty annular gas turbine whose size allows to test and compute the entire 360 degrees combustion chamber. The comparison with measurements confirms that, if the correct operating conditions are implemented, LES is capable to discern between stable and unstable operating conditions, as well as predict several other engineering relevant parameters, although the model is sometime affected by a limited shift in frequency. The post processing of LES results in presence of combustion dynamics is not a trivial task. Here the results of the simulations have been-post-processed by means of a triple decomposition method to determine a mean flow, a deterministic unsteady flow at the main instability frequency and a turbulent stochastic flow. Such decomposition indicated the instability triggering mechanism together with the cross-talk mechanism between different components. This approach is currently used for design phase, while further validation is on-going to include different geometries and operating conditions with the goal of reducing both risks and number of tests.
