Navier-Stokes High-Lift Airfoil Computations with Automatic Transition Prediction using the DLR TAU Code Andreas Krumbein German Aerospace Center Institute of Aerodynamics and Flow Technology, Numerical Methods Thank you very much! Ladies and Gentlemen, ‚eN Transition Prediction for 3D Wing Configurations using Database Methods and a local, linear Stability Code‘ is the title of my talk in which I would like to give you a short overview over the status of the transition capabilities of one of our CFD codes, in this case the FLOWer code. Additionally, I want to point out a problem which arises when validation work for eN transition prediction methods has to be carried out using experimental data from wind tunnel tests, especially when different eN methods are applied.

Outline Introduction Transition Prediction Coupling Structure Test Case: 2D A310 take-off configuration Computational Results Conclusion Outlook Let me first give you a brief overview over the contents of my talk. I am starting with an introduction, I will then outline the transition prediction coupling procedure which is the basis for the results I am going to show, I will present one of the 3d wing test cases I have worked on so far which is based on the ONERA M6 wing, and I will show some results which underline the problems which arise when validation work for eN transition prediction methods has to be done. I will end with a conclusion and with an outlook.

Introduction Aircraft industry and research requirements: RANS based CFD tool with transition handling Better numerical simulation results Capturing of otherwise unconsidered physical phenomena At first: impact on lift and drag Characteristics Transition prescription Transition prediction Modelling of transitional flow areas Automatic: no intervention of the user Autonomous: necessary user information as little as possible One of the tasks of DLR is to provide simulation software for the German aircraft industry, for example for Airbus. Thus, we have to face industry requirements on our CFD tools, and one of them is to provide a CFD tool based on the Reynolds-averaged Navier-Stokes equations with integrated transition handling especially transition prediction. The transition capability must be able to be applied in an automatic manner without intervention of the user and it shall reduce the modelling based uncertainties because it is known that often the accuracy of computational results from fully turbulent flow or from flow with estimated prescribed transition is not satisfactory. Especially with respect to high-lift flows an improved simulation of the interaction between the transition locations and separation is expected.

Introduction Reduction of modelling based uncertainties Accuracy of results from fully turbulent flow or flow with prescribed transition often not satisfactory Improved simulation of the interaction between transition locations and separation At first in FLOWer code 3d multi-element wing configurations Later in TAU code Fuselages and nacelles TAU transition prediction module developed by Institute of Fluid Mechanics, Technical University of Braunschweig in German research initiative MEGADESIGN One of the tasks of DLR is to provide simulation software for the German aircraft industry, for example for Airbus. Thus, we have to face industry requirements on our CFD tools, and one of them is to provide a CFD tool based on the Reynolds-averaged Navier-Stokes equations with integrated transition handling especially transition prediction. The transition capability must be able to be applied in an automatic manner without intervention of the user and it shall reduce the modelling based uncertainties because it is known that often the accuracy of computational results from fully turbulent flow or from flow with estimated prescribed transition is not satisfactory. Especially with respect to high-lift flows an improved simulation of the interaction between the transition locations and separation is expected.

Different approaches: Introduction Different approaches: RANS solver + stability code + eN method RANS solver + boundary layer code + stability code + eN method RANS solver + boundary layer code + eN database method(s) RANS solver + transition closure model or transition/turbulence model Apart from the application of transition criteria the following different approaches can be found in the literature which adress transition prediction in RANS solvers: # a RANS solver together with a stability code and the eN method # a RANS solver and a laminar boundary layer code together with a stability code and the eN method # a RANS solver coupled to a laminar boundary layer code which itself is coupled to eN database-methods # a RANS solver with integrated transition closure or transition/turbulence models

Different approaches: Introduction Different approaches: RANS solver + stability code + eN method RANS solver + boundary layer code + stability code + eN method RANS solver + boundary layer code + eN database method(s) RANS solver + transition closure model or transition/turbulence model At DLR we have gone and continue to go the third way.

Different approaches: Introduction Different approaches: RANS solver + stability code + eN method RANS solver + boundary layer code + stability code + eN method RANS solver + boundary layer code + fully automated stability code + eN method RANS solver + boundary layer code + eN database method(s) RANS solver + transition closure model or transition/turbulence model But also we go back one step, because we now have a fully automated linear stability code at hand, which can be used instead of the eN database methods in an automatic manner.

Transition Prediction Coupling Structure cycle = kcyc FLOWer The transition prediction coupling procedure looks as follows: Now I am saying, what I always say!

Transition Prediction Coupling Structure cycle = kcyc FLOWer & TAU The transition prediction coupling procedure looks as follows: Now I am saying, what I always say!

Transition Prediction Coupling Structure cycle = kcyc FLOWer & TAU cycle = kcyc TAU The transition prediction coupling procedure looks as follows: Now I am saying, what I always say!

Transition Prediction Module of TAU: Coupling Structure Transition Prediction Module of TAU: RANS infrastructure part: BL data from RANS grid (BL mode 2)  Transition inside separation bubble possible  High mesh density necessary External codes: Laminar boundary-layer method COCO (G. Schrauf) for swept, tapered wings (BL mode 1)  Transition inside separation bubble NOT possible  Laminar separation approximates transition if transition downstream of laminar separation point eN database-methods for TS and CF instabilities (PD mode 1) Local, linear stability code LILO (G. Schrauf) (PD mode 2) 2d, 2.5d (infinite swept) + 3d wings + 3d fuselages/nacelles (only BL mode 2) Single + multi-element configurations N factor integration along: Line-in-Flight cuts Inviscid streamlines Attachment line transition & by-pass transition not yet covered The transition prediction module has the following characteristics, additionally to the handling of the some geometry issues: Now, read the list!

Transition Prescription: Coupling Structure Transition Prescription: Automatic partitioning into laminar and turbulent zones individually for each element Laminar points: St,p  0 PTupp(sec = 2) PTupp(sec = 1) PTupp(sec = 3) PTupp(sec = 4) The generation of laminar regions is based on an automatic partitioning of the computational grid into laminar and turbulent or non-laminar zones, which is done individually for each element of the geometry, as you can see here (left) for a multi-element airfoil: this is the slat and this is the main airfoil nose with laminar zones defined by an upper and a lower transition point and a wall normal extension. Here, on the right you see the outcome for a single-element wing with a 4-point polygonial transition line on upper and lower side and the wall normal extent of the laminar zone in 4 selected grid planes.

Algorithm: no yes STOP set stru and strl far downstream Coupling Structure Algorithm: set stru and strl far downstream compute flowfield check for RANS laminar separation  set separation points as new stru,l cl  const. in cycles  call transition module  use outcome of prediction method (PD modes 1&2) or BL laminar separation point (BL mode 1) set new stru,l underrelaxed  stru,l = stru,l d, 1.0 < d < 1.5 convergence check  Dstru,l < e no yes STOP

ONERA M6 wing Transition lines for 11 wing sections h = 0.000, 0.110, FLOWer results Transition lines for 11 wing sections h = 0.000, 0.110, 0.220, 0.325, 0.420, 0.800, 0.860, 0.900, 0.930, 0.960, 0.975 Calibration of both critical N factors for lower side and a = 5°: NCFcr = 5.157 → h = 0.42 NTScr = 4.75 → h = 0.96 ONERA M6 wing a = 0°, 5°, 10°, 15° Re = 3.5106 M = 0.262 upper side lower side taken from *) TS upper side lower ls a = 0° a = 0° ls a = 5° a = 5° a = 15° a = 15° TS TS all ls *)Schmitt, V., Cousteix, J., “Étude de la couche limite tridimensionelle sur une aile en flèche,” ONERA Rapport Technique N° 14/1713 AN, Châtillon, France, July 1975 all ls CF all CF

FLOWer results TC 214 from EUROLIFT II Re = 1.35 mio., M = 0.174, SAE, eN database methods a = 14°, upper side predicted a = 14°, lower side predicted

FLOWer results TC 214 from EUROLIFT II Re = 1.35 mio., M = 0.174, SAE, eN database methods TS a = 14°, upper side predicted CF a = 14°, lower side predicted

Comparison of cp-distributions: h = 0.20, 0.38, 0.66, 0.88 FLOWer results Comparison of cp-distributions: h = 0.20, 0.38, 0.66, 0.88 a = 14.0°

Test Case 2d A310 take-off configuration M = 0.221, Re = 6.11 x 106, a = 21.4° grid 1: 22,000 points grid 2: 122,000 points, noses refined SAE turbulence model prediction on upper sides, lower sides fully laminar, NTS  8.85 (F1) exp. Transition locations  slat: 15% & flap: 34.5% kink on main upper side  19% different mode combinations: a) BL mode 1 & PD mode 1  BL code & TS database method b) BL mode 1 & PD mode 2  BL code & stability code c) BL mode 2 & PD mode 2  BL in TAU & stability code Test Case

Surface pressure grid 1 grid 2 TAU results Surface pressure grid 1 grid 2 a.) & b.) results identical  all lam. seps. a.) & b.) results identical  all lam. seps. c.) no convergence  grid too coarse c.) all from stability code

Skin friction grid 1 grid 2 TAU results Skin friction grid 1 grid 2 a.) & b.) no separation bubbles a.) & b.) very small sep. bubble on slat c.) no convergence c.) much larger slat bubble & flap improved

grid 2 slat flap Skin friction very small bubble transition locations: TAU results Skin friction grid 2 slat very small bubble transition locations: error reduced by 40% flap large bubble

Transition locations and separation TAU results Transition locations and separation grid 2 grid 2

Transition locations and separation TAU results Transition locations and separation grid 2 grid 2

Conclusion/Outlook TAU transition prediction module works fast and reliable for 2d multi-element configurations Transition inside laminar separation bubbles can be detected with high accuracy when appropriate prediction approach is used Therefor, high grid densities are required much more testing necessary: more test cases needed with TS transition (e.g. CAST 10, A310 landing) full aircraft WB+HTP+VTP (wing with full-span flap without slit) WB high-lift configuration with full-span slat and flap from EUROLIFT II transition criteria: - transition in lam. sep. bubbles - attachment line transition - by-pass transition development of a stream-line oriented bl code with transverse pressure gradient COCO-3d → replaces COCO in 2007 unsteady transition prediction method based on eN method alternative approaches based on transport equations in future DLR T&T-project RETTINA done by TU-BS