Numerical Prediction of Mean Flow and Acoustic Field of a Supersonic Impinging Jet Konstantin A. Kurbatskii, Saravana Kumar and Hossam El-Asrag 1 © 2011 ANSYS, Inc. December 6, 2016 ANSYS, Inc. Outline 1. Objectives 2. Problem description 3. Numerical model – – – – Pressure-based coupled solver Physical models and boundary conditions Computational meshes Turbulence models 4. Numerical results and comparison with test data – Steady-state results – Transient results – impingement tone radiation – Evaluation of structural resonant effects 5. Summary 2 © 2011 ANSYS, Inc. December 6, 2016 Objectives • Steady-state simulation of hover lift loss due to impinging cold and hot jets • Unsteady simulation of generation and near-field propagation of acoustic tones of a hot impinging jet • Evaluation of a potential resonant response of the structure to impinging acoustic tones 3 © 2011 ANSYS, Inc. December 6, 2016 Problem Description • Convergent-divergent nozzle – • • • • design Mach = 1.5 D = 2.54 cm – nozzle throat diameter De = 2.75 cm – nozzle exit diameter 2.0 ≤ h/D ≤ 8.0 – steady-state h/D = 4.0 – transient • Nozzle pressure ratio NPR = 3.7 – perfectly expanded jet • Two temperature ratios TR = 1.0 – cold jet – TR = 1.4 – hot jet – Ref: Krothapalli, A., Rajkuperan, E., Alvi, F., and Lourenco, L., “Flow Field and noise characteristics of a supersonic impingement jet,” Journal of Fluid Mechanics, Vol. 392, 1999, pp. 155-181. Kumar, R., Lazic, S., and Alvi, F. S., “Control of High-Temperature Supersonic Impingement Jets Using Microjets,” AIAA Journal, Vol. 47, No. 12, 2009, pp. 2800-2811 4 © 2011 ANSYS, Inc. December 6, 2016 Pressure-Based Coupled Solver • • • • Unsteady numerical simulation using CFD code ANSYS Fluent 15.0 Pressure-based coupled double-precision solver Control-volume-based technique QUICK-type scheme for interpolation of face values from cell centers in the momentum and energy equations (steady-state) – a weighted average of 2nd order upwind and 2nd order central differencing – uses a variable, solution-dependent value of the weight factor to avoid introducing new solution extrema • Bounded 2nd-order central differencing (BCD) for momentum and energy equations (transient) – BCD is a composite normalized variable diagram (NVD)-scheme that consists of: pure central differencing, , = ( + )+ ( · + · ) blended scheme of the central differencing and the second-order upwind scheme, and 1st order upwind scheme (used only when the convection boundedness criterion violated) 5 © 2011 ANSYS, Inc. December 6, 2016 Pressure-Based Coupled Solver (continued) • 2nd-order scheme for reconstructing face values of pressure • Least-squares cell-based calculation of gradients – preserves 2nd-order spatial accuracy • Non-differentiable Minmod gradient limiters • Fully implicit coupling between momentum and continuity equations – implicit discretization of pressure gradient terms in momentum equations – implicit discretization of face mass flux and pressure dissipation terms • System of discretized algebraic equations solved using coupled algebraic multigrid (AMG) scheme • Incomplete Lower Upper (ILU) decomposition to smooth residuals between levels of AMG – ILU has better smoothing properties than Gauss-Seidel for block-coupled systems – allows for more aggressive coarsening of AMG levels 6 © 2011 ANSYS, Inc. December 6, 2016 Time-Marching Scheme • Second-order fully implicit time marching algorithm – implicit equations solved iteratively at each time level before moving to the next time step – unconditionally stable with respect to the time step size – time step size chosen to resolve impingement tone frequencies up to 20 kHz ∆t = 0.004 D / Vj 7 © 2011 ANSYS, Inc. December 6, 2016 Physical Models and Boundary Conditions • Air modeled as a single-species calorically perfect gas • Molecular viscosity as a function of temperature by Sutherland's law • Pressure inlet condition at the nozzle inlet – total pressure and total temperature • Far-field boundary – Stead-state: pressure outlet specifies static pressure, all other variable extrapolated from interior – Transient: acoustically non-reflective pressure boundary based on characteristic wave relations derived from the Euler equations • Adiabatic no-slip walls 8 © 2011 ANSYS, Inc. December 6, 2016 Computational Mesh • 3D conformal hexahedral meshes using ANSYS Meshing • h parameterized for automatic regeneration of the solid model and mesh Cell size at nozzle exit in radial direction Cell size at nozzle exit along jet axis Cell size far-field Total cell count h/D = 4.0, ¼ model Coarse 0.016D 0.06D 0.08D 1.4 million Medium 0.008D 0.03D 0.06D 5.5 million Fine 0.0047D 0.03D 0.03D 24.7 million DDES Medium mesh 9 © 2011 ANSYS, Inc. December 6, 2016 Turbulence Models • SST k-ω (steady-state) • Wall-modeled Large Eddy Simulation (WMLES) model (transient) – WMLES with the S-omega formulation – covers the inner portion of the boundary layer by RANS and the outer portion by LES – avoids the very high resolution requirements of LES in the inner wall layer along the impingement plate 10 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results • Numerical runs carried out in parallel on two 48-core nodes (total of 96 compute processors) • Converged steady-state jet flow RANS solution is used as an initial flowfield for transient simulations • The solution is then time-marched until it locks into a self-sustained impingement tone feedback loop • No synthetic turbulence is generated at the nozzle inlet, all the turbulence is produced by jet flow instabilities 11 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Steady-State Hover Lift Loss Normalized static pressure Velocity vectors 12 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Steady-State Hover Lift Loss 0.7 0.9 CFD - medium mesh CFD - coarse mesh Test -Lift / Thrust 0.7 0.6 0.5 0.4 0.3 TR = 1.4 0.2 0.5 0.4 0.3 0.2 0.1 0.1 0 0 0 1 2 3 4 5 6 h/D 7 8 9 CFD - medium mesh CFD - coarse mesh Test 0.6 -Lift / Thrust 0.8 10 TR = 1.0 0 1 2 3 4 5 6 7 8 h/D Comparison of calculated hover lift loss as a function of h/D with experimental data 13 © 2011 ANSYS, Inc. December 6, 2016 9 10 Numerical Results – Steady-State 1.2 1.2 1.0 Test z/D = 0.5 CFD 1.0 0.8 0.6 V / Vj V / Vj 0.8 TR = 1.0 0.4 0.0 0.0 -0.2 -0.2 0 1 z/D = 0.5 CFD 0.4 0.2 -1 Test 0.6 0.2 -2 TR = 1.4 -2 2 -1 1 1.2 1.2 1.0 0.8 TR = 1.0 Test z/D = 4.0 CFD 1.0 0.8 V / Vj 0.6 0.4 TR = 1.4 Test z/D = 4.0 CFD 0.6 0.4 0.2 0.2 0.0 0.0 -0.2 -0.2 -2 -1 0 x/D 1 2 -2 -1 0 1 x/D Normalized mean velocity distribution across the jet, h/D = 5.0 14 2 x/D x/D V / Vj 0 © 2011 ANSYS, Inc. December 6, 2016 2 Impingement Tone Generation Mechanism Impingement tone feedback loop 15 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Transient Instantaneous contours of Mach number showing development of large-scale structures in the shear layer (medium mesh). 16 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Transient Instantaneous turbulence structures by Q-criterion, colored by the ratio of eddyviscosity to molecular viscosity 17 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Transient Instantaneous pressure field plotted on a log scale (medium mesh) 18 © 2011 ANSYS, Inc. December 6, 2016 Numerical Results – Transient Comparison of calculated and experimentally measured noise spectrum. x/D = 2.0 location on the lift plate (left), and microphone location at x/D = 15 (right). TR = 1.4 19 © 2011 ANSYS, Inc. December 6, 2016 Analysis of Spectral Data • Favorable prediction of the dominant peak frequency at 7 kHz • SPL of the dominant peak is under-predicted • The fine mesh did not improve SPL prediction of the dominant peak – mesh resolution may not necessarily be the primary factor – under-prediction of the dominant peak SPL was observed even in a very large-scale simulation of the same problem* using higher-order numerics • Other factors not included in the CFD simulations may contribute to SPL levels – structural resonant response *Uzun, A., Kumar, R., Hussaini, M. Y., and Alvi, F. S., “Simulation of Tonal Noise Generation by Supersonic Impinging Jets,” AIAA Journal, Vol. 51, No. 7, 2013, pp. 1593-1611. 20 © 2011 ANSYS, Inc. December 6, 2016 Structural Modal Analysis • Modal analysis of 20 mm thick aluminum lift plate using ANSYS Mechanical to predict its natural frequencies • Fixed support at the inner ring surface where the plate is attached to the nozzle 21 © 2011 ANSYS, Inc. December 6, 2016 Structural Modal Analysis of the Lift Plate Natural frequencies of the lift plate f, Hz 923.01 923.19 1115.1 1242.9 1597.2 1597.2 3509.6 3509.7 4959 4959.1 Mode 11 12 13 14 15 16 17 18 19 20 f, Hz 5942.3 5942.3 6415.1 7051.6 7052.1 8760.6 8760.6 9396.2 9396.3 9563.8 Mode 21 22 23 24 25 26 27 28 29 30 f, Hz 9563.9 11879 11879 12855 12855 13328 13328 14034 14034 14167 Lift plate h/D = 4.0, x/D = 2.0 170 160 SPL, dB/17 Hz Mode 1 2 3 4 5 6 7 8 9 10 180 150 140 130 120 110 100 Test 90 80 1 f, kHz 10 • Natural frequencies of the plate at 7 kHz coincide with the dominant impingement tone frequency • Two other plate natural frequencies at 3.5 kHz and 14 kHz are also nearly coincidental with two other impingement tone peaks • Fluid-structure resonant response may amplify impingement tone SPLs – will be confirmed by a fluid-structure interaction simulation in a future investigation 22 © 2011 ANSYS, Inc. December 6, 2016 Summary • Pressure-based coupled solver (PBCS) proves to be a robust and effective method for predicting flow and acoustic field of impinging jets – PBCS is less memory and CPU intensive than a traditional density-based approach – PBCS is an attractive alternative to density-based algorithms for obtaining transient solutions to supersonic jet noise problems • Steady-state predictions of hover lift loss and mean jet velocity distributions are in favorable agreement with experimental data • Predicted frequency of the dominant impingement tone matches with the experimental data – this frequency coincides with predicted natural frequency of the lift plate structure – a potential for fluid-structure resonance which can significantly amplify SPL at this frequency 23 © 2011 ANSYS, Inc. December 6, 2016 Questions? 24 © 2011 ANSYS, Inc. December 6, 2016 Nozzle Parameters Static discharge coefficient, Cd & m & i Cd = m & Mass flow Rate from CFD at m nozzle exit, Isentropic Mass Flow Rate & i = PA t m γ 2 RT γ + 1 γ +1 γ −1 P Total Pressure at Nozzle Inlet T Total temperature at Nozzle Inlet p∞ Ambient pressure At Throat Area Nozzle thrust performance, Cfg Cfg = Fj / Fi Isentropic Thrust 25 © 2011 ANSYS, Inc. &i Fi = m December 6, 2016 2γ p RT 1 − ∞ P γ −1 Fj Thrust from CFD γ −1 γ & Ve + (p e − p ∞ )A e Fj = m pe Area-averaged pressure at Exit Ae Exit Area Ve Mass-Averaged velocity at Exit