Center for Turbulence Research Proceedings of the Summer Program 2012 1 Experimental and numerical study of the influence of small geometrical modifications on the dynamics of swirling flows By T. Poinsot†, J. Dombard‡, ,V. Moureau¶, N. Savary‖ G. Staffelbach†† and V. Bodoc‡‡ This report presents a joint experimental / numerical study of the non-reacting flow in a swirler where small geometrical variations on a row of holes are imposed. Simulation results are compared to experimental data in terms of mean and RMS velocity fields but also in terms of unsteady activity (hydrodynamic modes such as the PVC (Precessing Vortex Cored) which is characteristic of such flows). Results demonstrate that LES captures the mean and RMS fields very well and that it can also predict the fact that in one case, two hydrodynamic modes are identified experimentally and that one of them disappears when the row of holes is displaced. 1. Introduction Swirling flows are commonly used in gas turbines to stabilize combustion. The first advantage of swirl is that the rotation imposed to the flow creates a large Central Recirculation Zone (CRZ) which is essential for flame stabilization (Gupta et al. 1984; El-Asrag & Menon 2007; Poinsot & Veynante 2011). On the other hand, it has been recognized for a long time (Billant et al. 1998; Roux et al. 2005) that swirling flows are often submitted to absolute instability (Ho & Huerre 1984) leading to strongly unstable hydrodynamic modes which can compromise the overall stabil- ity of the combustion chamber and cause combustion instability when they couple with acoustic modes (Huang & Yang 2004; Staffelbach et al. 2009). The present design of swirlers used in com- bustion chambers relies on complex geometrical shapes, often based on multiple swirler passages and subject to permanent optimization because the swirler controls a large part of the chamber performances: flame stabilization, mixing between fuel and air, flame stability, ignition capacities, etc. In this context, being able to optimize swirlers numerically has become a major issue because this optimization would be too expensive experimentally. This optimization essentially requires to simulate three types of problems: (1) obtain the correct velocity fields, (2) predict the correct pressure losses and (3) capture hydrodynamic modes to limit their amplitudes. One of the specificities of swirling flows is that Types 1 to 3 problems are sensitive to small geometrical changes. Experimentalists know that a minor design variation in a swirler can cause a strong flow change so that hysteresis and bifurcations are common features in many swirling flows (Dellenback et al. 1988; Vanierschot & den Bulck 2007). Vanierschot et al show that four different mean flow patterns can be obtained by small variations of the swirl number and that the transition between these four states depends on the method used to vary swirl: decreasing or increasing swirl number leads to different transitions. These transitions are observed on the velocity field but also on the pressure losses which are another important characteristic of the flow. The study of Vanierschot et al focused on the mean flow patterns but the instability modes can † IMF Toulouse, INP de Toulouse and CNRS, 31400 Toulouse , France ‡ Center for Turbulence Research, 488 Escondido Mall, Stanford, CA 94305-3035, USA ¶ CORIA, Rouen, France ‖ TURBOMECA, Bordes, France †† CERFACS, 31057 Toulouse, France ‡‡ ONERA-CT, 2 avenue Edmond-Belin, 31055 Toulouse Cedex 04, France