Proceedings of GT2006 ASME Turbo Expo 2006: Power for Land, of GT2006 Proceedings Sea and Air May

8-11, for Land, Sea and Air ASME Turbo Expo 2006: Power 2006, Barcelona, Spain

May 8-11, 2006, Barcelona, Spain GT2006-90766

GT2006-90766
CFD ANALYSIS OF HOT GAS INGESTION MECHANISMS FOR THE VERTICAL DESCENT PHASE OF A HARRIER AIRCRAFT
G.A. Richardson PDRA, CFD Lab Engineering Department Cambridge University, UK W.N. Dawes Professor, Fluid Mechanics Group Engineering Department Cambridge University, UK A.M. Savill Professor, Computational Aerodynamics Design Group School of Engineering Cranfield University, UK ABSTRACT Hot Gas Ingestion (HGI) can be a problematic feature of STOVL aircraft during the descent phase of landing, or while on the ground. The hot exhaust gases from the downwards pointing nozzles can be re-ingested into the engine intakes, causing power degradation or reduced engine surge margin. The flow-fields that characterise this phenomenon are complex, with supersonic impinging jets and cross-flows creating large ground vortices and fountain up-wash flows. As a partner in the PUMA DARP (Unsteady Methods Focus Group), the Cambridge University Engineering department CFD Lab are trying to model this flow-field in order to validate the RollsRoyce HYDRA CFD code against experimental data obtained from detailed Rolls-Royce HGI tests. The HYDRA code has been developed to include a suitable mesh deformation technique for the descending aircraft configuration. The code is applied to predict the occurrence of HGI, by simulating experimental results from a 1/15th scale model of a descending Harrier. Based on these computational results, this paper studies the aerodynamic mechanisms that govern HGI, in terms of the near-field and far-field effects and their impact on the magnitude of temperatures at the engine intake. The trends in experimental engine intake temperature profiles are explained by analysis of these mechanisms. Following a more thorough validation of these results, the HYDRA code will provide a valuable tool for predicting the occurrence of HGI. The CFD method can then be used for the analysis of other STOVL aircraft as well as configuration changes aimed at preventing HGI. NOMENCLATURE ADF CFD DARP Advanced Data Format Computational Fluid Dynamics Defence Aerospace Research Partnership DTI Department of Trade and Industry FOD Foreign Object Debris hex hexahedral HGI Hot Gas Ingestion HPCx High Performance Computer IGES International Graphics Exchange Standard LES Large Eddy Simulation OPLUS Oxford Parallel Library for Unstructured Solvers PDRA Post-Doctoral Research Associate PUMA Partnership in Unsteady Methods for Aerodynamics RANS Reynolds Averaged Navier-Stokes RR Rolls-Royce plc STOVL Short Take-Off / Vertical landing tet tetrahedral URANS Unsteady RANS VL Vertical Landing VR Virtual Reality VTO Vertical Take-Off VTOL Vertical Take-Off and Landing VRML Virtual Reality Mark-up Language

Copyright 2006 by ASME

INTRODUCTION (HGI) The flows below a Harrier are characterised by four high-speed jets impinging on a ground plane. These flows are further complicated by interaction with any cross-flows present which serve to drive the exhaust flows back towards the descending aircraft. The flow-field contains several components: - supersonic jets impingement on the ground plane - fountain up-wash flows between the jets - a low speed head-wind induced 'horse-shoe' vortex The combination of these effects can cause hot gas ingestion leading to engine power degradation and/or reduced engine surge margin. Other potential operational problems include: - variable aircraft pitching moment distributions - engine ingestion of ground debris (FOD) - adverse effects on ground crew working environment The dominant mechanisms that govern HGI are illustrated in Figures 1, 2 and 3. The influence of HGI can be divided into near and 'far-field' effects. Figure 1 shows how the hot gases from the nozzles impinge on the ground plane and spread out in all directions. A proportion of the ground plane flow from the front two nozzles extends upstream and encounters the low speed onset boundary layer flow. This causes it to separate and form a horse-shoe shaped ground vortex which rolls-up back towards the aircraft. Some of this far-field flow can be sucked into the engine intake causing a rise in intake temperature and temperature distortion levels. 'Near-field' effects are dominated by fountain upwash flows caused by the interaction of the exhaust jets with the ground plane (and themselves). The upwash from the impinging jets is forced towards the aircraft fuselage (see Figure 4) causing localised heating of the aircraft belly and enabling suction from the engine intake to draw these hot gases into the engine. The consequence of these near-field HGI effects can be potentially more severe as the hot gas takes a shorter path to the intake, compared with far-field flows and thus undergoes less mixing and dilution. In practice, one or other mechanism may dominate depending upon factors such as; onset flow direction and strength, cross-flow ratio (ratio of onset flow to jet speed) boundary layer height, aircraft height, fuselage geometry, nozzle configuration, and aircraft descent rate. Broadly speaking, near field effects are more significant closer to the ground (far-field effects being more influential as the aircraft first descends into the hot ground vortex region). The results presented in this paper will show how the fountain upwash flows drive two secondary (near-field) vortices, immediately below the engine intake, which are thought to be responsible for the most significant increase in engine intake temperatures.

Figure 1: multiple jet impingement in head wind

Figure 2: schematic diagram of hot gas ingestion

Figure 3: ground plane jet impingement pattern

Figure 4: schematic of HGI flows (section A-A in Fig.3)

Copyright 2006 by ASME

METHOD (HYDRA DESCRIPTION) The RR HYDRA code is a suite of non-linear, linear and adjoint CFD solvers developed collaboratively by RR and its University partners. HYDRA is a general purpose code for hybrid unstructured meshes which uses an efficient edge-based data structure by Moinier and Giles Ref. [4]. The flow equations are integrated around median-dual control volumes using a MUSCL (Monotonic Upwind Scheme for Conservation Laws) based flux-differencing algorithm. Turbulence is modelled using the either the Spalart-Allmaras Ref. [6] or kepsilon turbulence models. A Smagorinsky based LES capability is also available. The discrete flow equations are preconditioned using a block Jacobi pre-conditioner Ref. [4] and iterated towards steady state using the 5-stage Runge-Kutta scheme of Martinelli Ref. [3]. Convergence to steady state is further accelerated through the use of an element-collapsing multi-grid algorithm Ref. [5]. Time-accurate solutions are obtained using either implicit dual time-stepping, with preconditioning and multi-grid on the inner steps, or with explicit time-stepping. Mixing and sliding plane capabilities are available for multistage turbo-machinery. The HYDRA solver has been parallelised using the domain decomposition method and runs efficiently on both shared and distributed memory computers. The user guide and programmers guide for HYDRA are given Ref. [8] and [9]. There are additional references for the HYDRA code Ref. [7] and [14]. METHOD (HYDRA PROCESS) Mesh generation The aircraft/sting geometry is imported to the ICEMCFD grid generator from a CAD package, in the form of IGES files. The geometry is cleaned and a hybrid mesh is created within ICEMCFD, containing a mixture of tetrahedral, hexahedral and pyramid volume cells. The mesh is then exported in the form of a FLUENT mesh file. The ICEMCFD package has been installed on a 64-bit machine to enable the memory limitations of 32-bit machines (4GB limit) to be overcome and thus enable larger meshes to be generated. The current mesh size for the results show here is 8.5 million cells, which relates to just one half of the flow domain a symmetry boundary is assumed at the aircraft centre-line (x-y plane). It is estimated that roughly 20 million cells will be required to model the full flow domain for this problem. This will allow adequate resolution of the geometry and key areas of the flow-field, for a URANS solution. Considerable time has been spent in this study, arriving at a suitable hybrid meshing strategy. It was considered that a hexahedral (hex) mesh was required in a large proportion of the flow domain. The flow solution is considered too dissipative in regions of tet mesh. Hence a tet mesh is retained for less important regions around the aircraft geometry which, are too difficult to mesh using a structured approach. The hex mesh is applied for most of the flow domain, including a region along the aircraft belly, and immediately beneath the exit planes of the four nozzles. In particular it was considered important to use hex mesh to resolve the four jet plumes, the ground vortex,

the fountain upwash flows, and the flows along the underside of the aircraft. One draw-back of the hybrid meshing approach is that it can be very time consuming to generate meshes on complex geometries, as described in following sections. Pre-Processing Once the FLUENT mesh file is generated, it may then be read into FLUENT to check for any import errors and to view the mesh and boundary conditions. The FLUENT mesh file is then read into the RR HYDRA pre-processor JM52. JM52 is used to scale/translate the mesh, and specify boundary conditions. The initial solution is specified and the resultant grid and flow solution are exported in ADF format. The grid and flow ADF files are then imported into JM56, which is used to calculate the edge weights, generate any extra connectivity information required and create any additional levels of multigrid if needed. The resultant grid and initial flow solution are then exported from JM56 as ADF files. JM52/56 packages are both installed on 64-bit machines for handling large meshes. Solution generation The HYDRA solver is installed on the HPCx IBM 64-bit supercomputer allowing it to be run in parallel on a large number of processors. The HPCx system is located at the UK's CCLRC's Daresbury Laboratory and operated by the HPCx Consortium. The HPCx system uses IBM eServer 575 nodes for the compute and IBM eServer 575 nodes for login and disk I/O. Each eServer node contains 16 1.5 GHz POWER5 processors. There are 104 frames, i.e. 1664 processors, available to users. The total main memory of 32 GB per frame is shared between the 16 processors of the frame. The peak computational power of the HPCx system is 10.0 Tflops peak, or at least 6 Tflops sustained. So far, typical runs for this test-case have used 128 processors for a maximum period of 12 hours at a time. This means that an entire unsteady vertical descent calculation may be completed within 12 hours, assuming an initial hover solution is available. Jobs are submitted to the system via a batch queuing system, which means that jobs may be queuing for several days before execution. On completion of the solution, typically 100 unsteady solution files are created, each containing current (deformed) grid and flow-field data. Following this, the solution is copied back to the users local workstation for post-processing. The dual time-stepping implicit solver was used and the time-step size was 0.001sec. Unsteady solutions files were saved every 10 time-steps. Post-processing The xmgrace package was used to plot profiles, which show the temperature at the engine intake thermocouple rakes, throughout the duration of the unsteady vertical descent. The RR Visual3 based Spy package is used to analyse the flow solutions. For smaller meshes (<1million cells) spy can be used to read in all flow solutions and unsteady animations can be created. This is useful for looking at changing flow contours during the unsteady descent (temperature or Mach number)

Copyright 2006 by ASME

For larger meshes (>1million cells) the memory requirements of the PC (4GB RAM) become a limitation on the number of unsteady flow solutions which can be read into Spy. The solution then has to be loaded into Spy in a series of chunks (say 20 flow files at a time) for analysis of separate segments of the unsteady solution. For the larger mesh solutions used in this study, a single precision version of Spy was created (on a 32-bit machine) and the solution was converted to single precision files. This was done to allow visualization of just three unsteady flow files at a time (for a 8.5M cell mesh). Path-lines may also be generated, upstream or downstream using the Spy package. This is very useful for reverse tracking flows from the engine intake region, to see where the intake flow originated from. The dvConvert and dvMockup packages were also used to create 3D (VR) animations of path-lines within the ground vortex and temperature iso-surfaces below the nozzles, during the vertical descent phase. This is described in more detail below in the section on Virtual Reality methods; See Fig 11-14. METHOD HYBRID MESHING APPROACH Figure 5 shows the geometry which is imported into the ICEMCFD software in the form of IGES files. The geometry is then repaired to eliminate any gaps/holes between the surfaces. The surfaces are divided into the respective boundary conditions, so that areas of similar mesh density can be easily assigned a surface mesh size. Additional geometry is created where necessary, such as outer boundaries and symmetry plane. The blocking strategy is created first within HEXA to generate a structured hex mesh (see Fig.6). The block edges are projected onto the geometry curves and block faces are projected onto geometry surfaces, until a valid structured hex mesh is produced. The user checks the quality of the resultant mesh and addresses any areas of poor quality. It is important to minimise the aspect ratio of cells on the interface surface (between tet/hex region) to facilitate the later merge operation. The surface mesh (quad cells) on the interface surface are then converted to a tri mesh and saved as a surface mesh file. The user then imports this tri mesh to define the interface surface. The tri mesh is then used as a boundary condition for the tet mesher so that the tet mesh is created only between the aircraft geometry and the surface of the existing hex mesh. The hex and tet meshes are then merged, which then automatically creates pyramid cells at the interface surface. In the case where this is not successful, some of the pyramid cells will be missing, and the user has to improve the aspect ratio of the hex mesh to prevent this from occurring at the next attempted merge operation. The resultant hybrid mesh is then saved, and the user selects an output option for the mesh (ASCII FLUENT format in this case) so the ICEMCFD hybrid mesh can be converted into a format which may be imported into the JM52 pre-processor.

Figure 5: IGES geometry imported to mesh generator

Figure 6: blocking strategy for a hybrid mesh METHOD - MESH DEFORMATION The existing moving mesh capability within HYDRA is the same as that described in Ref. [2]. This project has further developed the existing mesh movement capability to allow 3D mesh deformation for a full aircraft in descent. A detailed description of the linear mesh movement algorithm, which has been added to the code, is described in Ref. [1]. Figures 7 and 8 shown below demonstrate the mesh deformation at the aircraft centerline plane. In this case the grid nodes move only in the vertical y-direction. It can be seen that there is a square region of unstructured mesh immediately surrounding the aircraft, outside which the mesh is fully structured. Figure 7 shows the starting mesh which represents the aircraft in hover. Figure 8 represents the mesh as the aircraft lands on the ground plane. This demonstrates how the structured mesh below the aircraft becomes compressed, as the model descends and approaches touch-down. For this reason the undercarriage is not included in the analysis, since there should be a clear gap between aircraft and ground, on landing. If the mesh did include any geometry in this region, it would become distorted during the descent phase of the simulation.

Copyright 2006 by ASME

The computational setup matches the experimental conditions. The test-piece (Fig.9) descends a distance of 0.446m at a rate of 1m/sec, in a low speed head wind of 6m/sec. The engine intake flow conditions are Ttotal = 295K, mass-flow = 0.445kg/sec. Both front and rear nozzles are chocked flow, and therefore flow conditions corresponding to Mach > 1.0, are specified at the exit plane of the nozzles.

Figure 7: mesh deformation at start of descent Figure 9: photo of Harrier test rig model The engine intake of the aircraft test-piece contained three annular thermocouple rakes to record engine face transient temperatures. Each rake has 16 fast response thermocouple probes to measure the intake temperatures during the aircraft descent (Fig.10). The experimental data is available as engine intake temperature against elapsed descent time as shown in Fig.16. Only the test results from the middle rake have been used here for comparison with the equivalent predicted results.

Figure 8: mesh deformation at end of descent

TEST-CASE DESCRIPTION A complete description of the experimental details can be found in Ref. [12] and [13]. The tests were carried out on a 1/15th scale geometrically representative model of the AV-8B aircraft fitted to a VTOL rig inside a low speed wind tunnel. The aircraft model was set at 7.5 nose-up incidence with the auxiliary doors open. The production standard deep strake and dam geometry (without gun-pods) was used and variations on the dam/strake geometry were also tested. The model was tested under VL and no-go VTO conditions, for a range of headwinds. The tests were all carried out at full-scale pressures and at temperatures less than full scale due to a rig limit of 750K. Repeat runs were performed for every configuration to allow statistical treatment of the results in view of the stochastic nature of the HGI process.

Figure 10: test-rig engine intake thermocouple locations

Copyright 2006 by ASME

METHOD - VR TECHNIQUES

Figure 11: front view of VR path-lines at start

Figure 12: front view of VR path-lines at middle

Virtual reality techniques have been used to visualise the ground vortex flows around the Harrier during the descent phase. This enabled a 3D path-line animation to be viewed. The user can get a better feel for the fluid dynamic mechanisms by viewing in 3D while manipulating the view to move around the aircraft and ground vortex during the animation. The method of post-processing was achieved by analysis of 100 unsteady solution files in ADF format. The ADF files were interrogated to produce VRML files to represent the pathlines for each flow solution. The VRML files are then imported into dvConvert software, which converts them into a format that can be read by dvMockup software. The Mockup package is CAD-like software which allows 3D VR viewing on a twin polaroid projector system. This VR system uses a metallic viewing screen to preserve the polaroid nature of the images, and the user wears polaroid lens glasses to view the images. Once the 100 path-line solution files have been imported into Mockup, the user can easily set up an animation to skip though all of the unsteady solutions at a selected speed in order to create an animation. Figures 11-13, opposite, show views of path-lines within the ground vortex region, at various stages of the aircraft descent. Figure 14 shows a side view of the ground vortex path-lines. Current work is underway to post-process the ADF files and then view the results directly using OpenGL commands which allow stereo viewing using a twin projector, or, red/blue mode which works locally on desktop PC. The red-blue mode is more convenient since it allows the user to check the VR post-processing locally, without needing a twin-projector. Once the post-processing is set-up, the user can then switch to the polaroid twin-projector system for more detailed analysis of results, for meetings or group discussion. Twin projector polaroid mode has the advantage that full colour analysis of the results can be done (e.g. for surface pressure contour plots) which is not practical for red/blue mode viewing. This method eliminates the need to use intermediate software such as Mockup. Fig.15 below shows an example of this, where an ADF file has been interrogated to show the surface mesh for a hybrid mesh case, for 3D flow over a cylinder.

Figure 13: front view of VR path-lines at end

Figure 14: side view of VR path-lines at middle Figure 15: example of red/blue stereo view

Copyright 2006 by ASME

RESULTS AND DISCUSSION A quantitative comparison of predicted engine intake temperature data with equivalent test data is shown in Fig.16. The test rig was not instrumented with a view to CFD validation, therefore only predicted engine face temperatures can be compared with test-data. These results show that the temperature increase at the engine intake associated with HGI is being predicted by the CFD analysis. It can be seen that the predicted (URANS) results do not simulate the unsteady nature of the experimental data, yielding instead a smooth variation with time. It is considered that LES is required to capture this unsteady behavior of the fluctuating experimental temperature profiles. It is planned that LES will be applied to this test-case, towards the end of the project. The experimental results (Fig 16) also show an overall larger range of temperatures compared with the predicted data. To explain this, the experimental data is divided into region 1 and region 2. In region 1 the predicted temperatures are on average, approximately the same as the experimental ones. During this time period, the aircraft descends into the large scale far-field ground vortex upstream of the aircraft, which is well predicted by the simulation. In region 2 the experiment shows a much larger temperature rise (compared to predicted data) for the probe locations that are situated towards the bottom of the engine intake. We speculate that during this time period the intake ingests hotter flows, which originate from the near-field fountain upwash flows, beneath the aircraft. The computational modelling does not adequately resolve these flows, so they are not evident in the predicted temperature profiles. However, the CFD results do show some evidence of HGI originating from the fountain upwash flows (Fig.21). Figure 17 and 18 show the Mach number and temperature contour plots at the symmetry plane of the flow domain. The aircraft is in its initial hover position, before the unsteady descent phase calculation begins. The size and location of the far-field ground vortex is clearly shown as a warm air region in Fig 18. It can be seen that the engine intake is outside of this region in the initial start position.

supporting sting

aircraft model

Figure 17: side view of mach contours at symmetry plane

far-field warm ground vortex

experimental results predicted results

Figure 18: side view temp contours at symmetry plane Figure 19 presents reverse flow pathlines from the engine intake, which shows the position of the ground vortex as the aircraft approaches the ground plane. Initially when the aircraft is in hover, the ground vortex is aligned with the ground plane as shown schematically in Figure 1. During vertical descent, the shape and orientation of the ground vortex changes considerably. Initially it is aligned with the (horizontal) x-z ground plane, but as the aircraft descends, this vortex structure becomes aligned with the (vertical) y-z plane. This was also observed in earlier virtual reality results (see Figures 11-13). It is thought that this change in shape of the original horse-shoe ground vortex is due to interaction with the near field effects as they become more dominant, when approaching touchdown.

Temp

start of descent

touch-down Region 1 Region 2

0.0

0.1

0.2

0.3

0.4

0.5

Time (sec)

Figure 16: engine intake temperatures (CFD vs test-data)

Copyright 2006 by ASME

far-field warm ground vortex

Figures 20 and 21 show pathlines released from the exit plane of the front nozzle. Fig 20 is a front view, which clearly shows the jet impingement and the resulting fountain upwash flows. This jet impingement and fountain upwash drives the near-field ground vortices - there are two of them, either side of the centre-line plane. In Figure 21 the ground vortex on the port side of the aircraft has been highlighted to show its location and close proximity to the engine intake. Pathlines from the ground vortex can be seen to enter the engine intake.
Y X

Figure 19: reverse flow engine intake path-lines

Only a few snapshots of the flow field are shown here, but the unsteady solutions for vertical descent phase have been analysed using animations of flow properties and pathlines, and have been viewed using both 2D and 3D/VR techniques. This has allowed a better understanding of the flow-field at various stages of the descent phase. The results shown above (Fig 1621) are explained more easily in Figure 22, which provides a simplified representation of the experimental engine intake mean temperature profiles. The descent phase is labeled as regions A-H, to mark the various stages of the aircraft model descent on the time axis.
edge of far-field ground vortex edge of near-field ground vortex

near-field hot fountain upwash

G F
front nozzle Impinging jet

Temp A B C

D
far-field effects dominate

near-field effects dominate

Figure 20: front nozzle pathlines (front view)

Time
hover descent phase touch-down

hot gas ingestion

Fig 22: simplified representation of experimental engine intake mean temperature profiles In the test-rig, before the aircraft begins to descend (region A), it is static, simulating the aircraft in hover, and is not showing any HGI effects at the engine intake thermocouple rakes. As the aircraft starts to descend (region B) the intake temperatures remain roughly constant since the aircraft engine intake remains outside of any hot gas regions. As the aircraft descends further, the engine intake becomes engulfed in the warm re-circulating air of the far-field ground vortex. This causes the initial rise in temperatures at the engine intake (region C). As the aircraft descends further into the far-field ground vortex, the intake temperatures increase more gradually as the intake temperatures approach the highest temperature within the far-field ground vortex (region D).

near-field hot ground vortex

Figure 21: front nozzle pathlines (isometric view)

Copyright 2006 by ASME

As the aircraft model descends further it begins to suck in the hotter flows from the near field fountain upwash driven ground vortex. This causes a sudden increase in the intake temperatures (region E) since the near-field flows are less well mixed (less cooling) compared to the far-field flows. The mean engine intake temperatures continue to rise quite rapidly as the aircraft approaches touchdown (region F). After this the aircraft is on the ground (region G) and CFD results have shown that the intake temperatures continue to increase after this time (region H), although the experimental data is not available after the model touches down. CONCLUSIONS The objective of the present work was to predict the vertical descent phase of a Harrier aircraft. This has been achieved by development of a linear mesh movement technique based on an existing sinusoidal mesh movement capability within the Roll-Royce HYDRA code. The method allows the user to prescribe a linear translation to a full 3D aircraft body to allow it to simulate a vertical descent. The method has been applied to simulate a hot gas ingestion test-rig experiment, in which a 1/15th scale model of the Harrier aircraft descends and lands, and the temperatures around the engine intake are monitored to assess the severity of HGI. Various meshing strategies have been explored and current results are based on a URANS approach with an 8.5 million-cell hybrid mesh, which represents one half of the flow domain. The HPCx supercomputer has been used to apply the RR parallel HYDRA code over a large number of processors, to obtain hundreds of unsteady flow-field solutions on large meshes, within a period of 12 hours. Issues of visualisation for large data-sets have been addressed in order to post-process the results. Virtual reality techniques have been used to study unsteady 3D animations of path-lines within the ground vortex, in order to better understand the way that the ground vortex changes shape and orientation as the aircraft descends. It is considered that VR techniques are the best way for visualising the complex flow field associated with this type of application. The CFD results were compared with experimental results and it was shown that the key mechanisms that govern HGI can be predicted. Further validation against test data is ongoing. Issues of structured verses unstructured mesh have been considered, and as a result, hybrid meshes have been used to capture the complex aircraft geometry while retaining structured mesh for key areas of the flow-field. Analysis of the flow-field predictions has helped to understand the combination of near-field and far-field mechanisms governing HGI, and to interpret the experimental intake temperature profiles.

For the configuration modelled, it has been found that the occurrence of HGI is mainly due to ingestion of flow from the front two nozzles. The weaker far-field effects are shown to be responsible for the initial increase in intake temperature profiles, whereas the stronger near-field effects are shown to cause the latter more rapid increase in engine intake temperatures. This improved understanding of the HGI mechanisms, could enable the design of potential HGI palliatives such as dams / strakes to be refined/optimised. FUTURE WORK The current work will proceed with the validation exercise for improved comparison of computational and experimental results using the URANS approach. This will include a sensitivity study for the effects of grid size, turbulence modelling and inlet boundary conditions. It is anticipated that grid sizes of at least 20 million cells will be solved using HPCx for implicit (and explicit, if possible) solution of the full descent phase. The turbulence modelling aspect will proceed with SpalartAllmaras, and progress to k-epsilon (wall function and low-Re versions) and then to LES. It is anticipated that the application of LES will be limited to the final part of the descent phase, where the aircraft is close to touchdown. It would be highly computationally expensive to apply LES for the full descent. It may also be necessary to limit the LES calculations to key areas of the flow, such as the jet and fountain upwash regions. Development of the virtual reality capability will proceed, so that sets of unsteady flow solutions from HYDRA (ADF format) may be directly viewed in 3D (twin polaroid projector mode or red-blue mode) using bespoke OpenGL code. The aim is to develop unsteady 3D animations which allow the user to manipulate/interrogate the solution, while the results are being viewed in real-time. ACKNOWLEDGMENTS The authors wish to acknowledge the following bodies and people: EPSRC/DTI/MoD (who funded the project number NMZB/013); Rolls-Royce (Peter Stow, Leigh Lapworth, Richard Bailey, Tony Ponton and Kevin Menzies); ANSYS (Mark Allen ICEMCFD support); Sussex University (Nick Hills, for his parallel version of HYDRA and support); Loughborough University (Gary Page, Jim McGuirk, Qinling Li); Daresbury Lab (Andy Sunderland HPCx support). The HPCx computer time was provided through the UK Applied Aerodynamics Consortium (UKAAC) under EPSRC grant GR/S91130/01.

Copyright 2006 by ASME

REFERENCES [1] G A Richardson, W N Dawes, A M Savill, Hot Gas Ingestion Modelling for the Vertical Descent Phase of a Harrier, Paper AIAA-5338-2005, 17th AIAA CFD Conference, Toronto, June 6-9, 2005. [2] Michael Giles, UNSFLO: A Numerical Method For The Calculation Of Unsteady Flow In Turbo Machinery, GTL Report No.205, May 1991. [3] L Martinelli, Calculations of Viscous Flows with a Multigrid Method, Ph.D Thesis, Dept. of Mech. And Aerospace Eng., Princeton University, USA, 1987. [4] Pierre Moinier and Mike B Giles, Preconditioned Euler and Navier-Stokes Calculations on Unstructured Grids, 6th ICFD Conference on Numerical Methods for Fluid Dynamics, Oxford, UK,1998 [5] J-D Muller, M B Giles, Edge-Based Multigrid Schemes for Hybrid Grids, 6th ICFD Conference on Numerical Methods for Fluid Dynamics, Oxford, UK, 1998 [6] P R Spalart and S R Allmaras, A One-Equation Turbulence Model for Aerodynamic Flows, La Recherche Aerospatiale, Vol.1, pp.5-21, 1994 [7] Pierre Moinier, Algorithm Developments for an Unstructured Flow Solver, PhD Thesis, Oxford University, 1999 [8] Leigh Lapworth, The HYDRA User's Guide, RollsRoyce, August 2004. [9] Leigh Lapworth, The HYDRA Programmer's Guide, Rolls-Royce report, August 2004. [10] Jens-Dominik Muller, A user's guide to Hip, Oxford University Computing Laboratory Report, Jan 2001. [11] John T Batina, Unsteady Euler Algorithm with Unstructured Dynamic Mesh for Complex Aircraft Aerodynamic Analysis, AIAA volume 29, pages 327333, March 1991 [12] C J Penrose, Harrier 2 hot gas reingestion model tests, INTERNAL Rolls-Royce (Bristol) Report, GN 30224, August, 1990 [13] L R Harper, Harrier 2 hot gas reingestion model tests, phase II, INTERNAL Rolls-Royce (Bristol) Report, GN 30345, September 1990. [14] Michael Giles, UNSFLO: A Numerical Method for the Calculation of Unsteady Flow In Turbomachinery, GTL Report, 205, May 1991.

10

Copyright 2006 by ASME