Validation of a large eddy simulation methodology for accelerated nozzle flows

Prediction of aeroengine exhaust plume near-field development requires knowledge of velocity and turbulence distributions at nozzle exit. The high Reynolds number nozzle inlet boundary layers of engineering practice are fully turbulent, but acceleration can induce re-laminarisation. Thus, to reproduce nozzle exit conditions accurately, large eddy simulation (LES) for plume prediction must be capable of capturing re-laminarisation and any subsequent boundary layer recovery. Validation is essential to establish a credible LES methodology, but previous studies have suffered from lack of nozzle inlet/exit measurements in the test cases selected. Validation data were here taken from an experiment on a convergent round nozzle with a parallel exit extension to allow boundary layer recovery. LES inlet condition generation applied a rescaling/recycling method (R²M), whose performance was validated against measurements of first and second moment statistics as well as the turbulence integral length scale. Simulations employed two sub-grid-scale (SGS) models; these produced similar predictions up to the end of the nozzle convergent section, but marked differences appeared for the nozzle exit turbulence field. The Smagorinsky model predicted much lower turbulence levels than measured, whereas the Piomelli and Geurts model revealed the presence of a small separation region at the convergence/parallel section corner, which led to higher exit turbulence and much better agreement with measured data.