Experimental and Numerical Calculation of Turbulent Timescales at the Exit of an Engine Representative Combustor Simulator

Abstract

To deepen the knowledge of the interaction between modern lean burn combustors and high pressure (HP) turbines, a nonreactive real scale annular trisector combustor simulator (CS) has been assembled at University of Florence (UNIFI), with the goal of investigating and characterizing the combustor aerothermal field as well as the hot streak transport toward the HP vanes. To generate hot streaks and simulate lean burn combustor behaviors, the rig is equipped with axial swirlers fed by a main air flow stream that is heated up to 531 K, while liners with effusion cooling holes are fed by air at ambient temperature. Detailed experimental investigations are then performed with the aim of characterizing the turbulence quantities at the exit of the combustion module, and specifically evaluating an integral scale of turbulence. To do so, an automatic traverse system is mounted at the exit of the CS and equipped to perform hot wire anemometry (HWA) measurements. In this paper, twopoint correlations are computed from the time signal of the axial velocity giving access to an evaluation of the turbulence timescales at each measurement point. For assessment of the advanced numerical method that is large Eddy simulation (LES), the same methodology is applied to a LES prediction of the CS. Although comparisons seem relevant and easily accessible, both approaches and contexts have fundamental differences: mostly in terms of duration of the signals acquired experimentally and numerically but also with potentially different acquisition frequencies. In the exercise that aims at comparing highorder statistics and diagnostics, the specificity of comparing experimental and numerical results is comprehensively discussed. Attention is given to the importance of the acquisition frequency, intrinsic bias of having a short duration signal and influence of the investigating windows. For an adequate evaluation of the turbulent time scales, it is found that comparing experiments and numerics for high Reynolds number flows inferring smallscale phenomena requires to obey a set of rules, otherwise important errors can be made. If adequately processed, LES and HWA are found to agree well indicating the potential of LES for such problems.