Analysis of the Stability and Control Characteristics of the F/A-18E Super Hornet using the Kestrel CFD Flow Solver

Transcription

1 50th AIAA Aerospace Sciences Meeting including the New Horizons Forum and Aerospace Exposition January 2012, Nashville, Tennessee AIAA NAVAIR Public Release SPR Distribution Statement A Approved for public release; distribution is unlimited. Analysis of the Stability and Control Characteristics of the F/A-18E Super Hornet using the Kestrel CFD Flow Solver Bradford E. Green * Naval Air Systems Command, Patuxent River, MD, The goal of this project is to evaluate the ability of the Kestrel flow solver to analyze the stability and control characteristics of the F/A-18E Super Hornet. The Kestrel flow solver is currently in development and real applications are being used to assess the accuracy of the code and find areas where improvements must be made. The F/A-18E is one of the most challenging Navy aircraft to model and significant amounts of data are available to assess the accuracy of the results. For this study, a variety of configurations and flow conditions were analyzed. Specifically, Kestrel was used to analyze: the longitudinal stability and control characteristics at low-speed and high angle of attack for configurations with full and neutral nose-down control; the longitudinal and lateral/directional stability and control characteristics at transonic speeds; transonic roll damping; the longitudinal stability and control characteristics at Mach 0.6 at 30,000 ft; and the stability and control characteristics of the aircraft doing a constant-g wind-up turn maneuver. To assess the accuracy of the calculations, most of the results were compared to wind-tunnel data, data from a flight database, or known trends of the data. In most cases, the correlation between Kestrel and the truth data is extremely good. C A C D C L C l C m C N C n C Y Re c s Nomenclature = aircraft axial-force coefficient = aircraft drag coefficient = aircraft lift coefficient = aircraft rolling-moment coefficient = aircraft pitching-moment coefficient = aircraft normal-force coefficient = aircraft yawing-moment coefficient = aircraft side-force coefficient = Reynolds number based on mean aerodynamic chord = angle of attack (degrees) = angle of sideslip (degrees) = horizontal tail deflection angle (degrees) I. Introduction HE Naval Air Systems Command (NAVAIR) is responsible for providing flight clearances for all types of T Navy air vehicles. These air vehicles include fighter/attack aircraft, reconnaissance aircraft, unmanned aerial systems (UASs), and helicopters. Data for decisions regarding flight clearances can be obtained from computational fluid dynamics (CFD), wind-tunnel testing, flight testing and/or similarity approaches. Each approach has its advantages and disadvantages. The cost, schedule, and accuracy of the approach is often a major driving factor in the decision as to which approach is best to use for a given flight clearance decision. In many cases, the decision regarding a flight clearance requires quick turnaround. While CFD can still be used in such a case, the accuracy and the efficiency of the CFD tools in place must already be known. When a flight clearance decision is required, there is no time at that point to assess the accuracy and efficiency of the CFD tools. * Aerospace Engineer, Applied Aerodynamics and Store Separation Branch, Bldg 2187, Unit 5, Suite 1320-B, Shaw Road, Senior Member. 1 This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

2 As a result, the accuracy and efficiency of the CFD tools must be evaluated before the flight clearance decision is required. Each different type of aircraft (fighter/attack, reconnaissance, UAS, and helicopter) must be evaluated with CFD to determine the accuracy and efficiency of the tools. Furthermore, each computational tool should be used on each of the aircraft separately and independently. Simply because one flow solver can accurately model a given aircraft at a given condition does not mean that a second flow solver can accurately do the same. Each tool must be evaluated separately on each different type of aircraft. A new CFD flow solver named Kestrel 1,2 is being developed as part of the Computational Research and Engineering Acquisition Tools and Environments (CREATE) Program. This program is funded by the High Performance Computing Modernization Program (HPCMP), which was recently transferred to the Department of the Army. The CREATE Program, which is a 12-year program, started in The development of Kestrel started in the same year. The CREATE Program is composed of several components. The component for air vehicles is called CREATE-AV. As part of the CREATE-AV Program, the Shadow-Ops program assesses the accuracy and functionality of the CREATE-AV tools. Kestrel is among the tools that are assessed by the Shadow-Ops program. During this process, projects are selected to Shadow Department of Defense acquisition programs. Within the last year, the F/A-18E Super Hornet was evaluated using the Kestrel flow solver. The F/A-18E Super Hornet is an extremely complex fighter aircraft that is difficult to accurately model with CFD. The pre-production F/A-18E experienced Abrupt Wing Stall (AWS) in ,4 While the problems with AWS were eliminated on the F/A-18E before aircraft production began, the AWS characteristics of the pre-production F/A-18E provide an opportunity to evaluate Kestrel with a combination of configurations and conditions that is difficult to accurately model. Partially due to the AWS issues with the pre-production aircraft, vast amounts of wind-tunnel and flight-test data is available for the aircraft. As a result, this aircraft is the perfect choice with which to assess the accuracy of the Kestrel flow solver. During this project, the pre-production F/A-18E Super Hornet was analyzed with the Kestrel flow solver. It is important to reiterate here that the configuration that was modeled in this study is the F/A-18E configuration that was susceptible to AWS, not the aircraft that is currently being used by the fleet. Several different configurations and conditions were considered. The goal of this project was to assess the ability of the Kestrel flow solver to accurately predict the stability and control (S&C) characteristics of the pre-production F/A-18E Super Hornet. Wind-tunnel data, data from the F/A-18E flight database, and known trends of the data were used to assess the accuracy of the CFD results. During this project, the following areas were the focus: (1) Longitudinal S&C characteristics at low-speed and high angle of attack for configurations with full and neutral nose-down control (2) Longitudinal S&C characteristics at transonic speeds (3) Lateral/directional S&C characteristics at transonic speeds (4) Transonic roll damping (5) Longitudinal S&C characteristics at Mach 0.6 at 30,000 ft (6) Constant-g wind-up turn maneuver Wind-tunnel data were used to assess the accuracy of the results for (1) through (4). Trends with the known flight characteristics of the aircraft were also used for (4) and (6). Results from the F/A-18E flight database were used for (5). This set of cases looks at a range of conditions and configurations for both wind-tunnel scale calculations and calculations of the aircraft in flight. The computations in (1) through (4) were investigated with another flow solver by Green between 2004 and ,6,7 These cases seek to answer the following questions: (1) Can Kestrel accurately predict the forces and moments of the F/A-18E at wind-tunnel scale for low-speed and transonic flow? (2) Can Kestrel accurately predict the forces and moments of the full-scale F/A-18E in flight at altitude? (3) Can Kestrel accurately predict the tendencies of the pre-production F/A-18E toward AWS? (4) Can Kestrel model a maneuvering F/A-18E? Both static and dynamic calculations were conducted with Kestrel to answer these questions. The results of these studies are presented below. Each area mentioned above is discussed within its own section. In each of these sections, the grid and geometry are discussed, the approach used to generate the results is presented, the results are shown and discussed, and then the computational efficiency of the Kestrel flow solver is presented. The paper ends with a summary. 2

3 II. Discussion of CFD Tools Used for this Study In this section, the CFD tools that were used during this study are discussed. These tools include the grid generation tools as well as the CFD flow solver. A. Grid Generator The Tetrahedral Unstructured Software System (TetrUSS) 8 was used to generate grids of the F/A-18E Super Hornet during this study. TetrUSS was developed at NASA Langley Research Center in Hampton, VA. TetrUSS uses GridTool, VGrid, and Postgrid to generate unstructured tetrahedral grids. While viscous grids were generated during this study, it is also possible to generate inviscid grids with TetrUSS. After a water-tight CAD geometry has been obtained, the grid generation process is begun using GridTool 9. In GridTool, a series of points and curves are used to form patches on the surface of the geometry. Next, sources that control the size and density of the cells in the grid are created. VGrid then uses these patches and sources to generate an unstructured tetrahedral grid on the geometry. VGrid is run three times during the grid generation process. During the first run of VGrid, a surface grid composed of triangles is generated. After obtaining an acceptable surface grid, VGrid is run again using an advancing-layers method 10 to generate the cells in the boundary layer. The normal spacing at the wall and the number of layers in the viscous part of the grid are controlled by three variables that are specified within GridTool. A separate tool called usgutil is used to determine the values for these three variables. The inputs to usgutil are the Reynolds number, the desired number of viscous layers, and the desired stretching of those layers. In the final pass through VGrid, the inviscid volume grid is generated using the advancing-front method 11. Postgrid is then used in the final step of the process to remove bad cells that were formed during the grid generation process. The final product is a full three-dimensional unstructured viscous tetrahedral grid. Each of the grids used during this project are symmetric full-span grids. The y-plus value of the first cell above the surface is approximately unity for each of the grids. B. Flow Solver During this study, the Kestrel flow solver 1,2 was used to analyze the grids that were generated with TetrUSS. Kestrel is being developed by the CREATE-AV team, which is funded by the HPCMP. Kestrel originated from the AVUS flow solver, which was developed at the Air Force Research Laboratory at Wright-Patterson Air Force Base in Dayton, OH. Kestrel is a second-order, cell-centered, finite-volume Navier-Stokes flow solver that is capable of analyzing grids with arbitrary cell topologies. Kestrel can be used to generate steady-state or time-accurate solutions. Kestrel is capable of using several different flux schemes, limiters and turbulence models. During this study, the inviscid flux scheme of Gottlieb and Groth was used, as well as the Original AVUS limiter and the Venkatakrishnan limiter. The Gauss-Seidel matrix scheme was used. The Spalart-Allmaras (SA) RANS turbulence model 12 was used to generate some of the results during this study. The remaining results were generated using the Delayed Detached Eddy Simulation (DDES) model with SA. Kestrel versions 1.1, 2.1.1b and were used to generate the results during this study. For documentation purposes, the specific Kestrel version will be noted when presenting results below. In addition, while the TetrUSS grid generation tool mentioned above generates grids with tetrahedral cells, it s important to mention that some of the cells in the boundary layer of these grids were converted to prisms prior to running Kestrel. This reduces the size of the grids, which reduces the memory and run-time requirements, and allows for more accurate computation of the flow in the boundary layer. III. Low-Speed, High Angle of Attack Longitudinal Stability and Control In this section, the results of the low-speed, high angle-of-attack calculations on the pre-production F/A-18E are presented and discussed. The longitudinal S&C characteristics were evaluated and the accuracy of the results was assessed with wind-tunnel data. Configurations with neutral and full nose-down control were evaluated. More details regarding the importance of calculations such as this and the results of previous work in this area were published by Green 5. A. Geometry and Grid Descriptions The geometry evaluated during this computational study was that of the 15%-scale pre-production F/A-18E wind-tunnel model. A picture of the computational geometry is shown in Fig. 1. The wing was modeled with 34/4/0 flaps, indicating that the leading-edge flap was deflected 34, the trailing-edge flap was deflected 4, and the aileron was neutral. The geometry was modeled with both horizontal and vertical tails present. Horizontal-tail deflections of 0 and 20 were evaluated and the rudders were toed-out 30. The model also included a leading-edge 3

4 extension (LEX) spoiler, which was deflected 60. The geometry with a horizontal tail deflection of 20 with the other control surfaces deflected as indicated is often referred to as the geometry with full nose-down control. The same geometry with a horizontal tail deflection of 0 is referred to as the geometry with neutral nose-down control. A Sidewinder missile and launcher were modeled at the wing tip and the inlet was modeled as flow-thru. A dualsting was also included in the geometry. The sting extends 3.5 fuselage lengths downstream of the aircraft. The sting was modeled as a viscous surface, but the forces and moments on the sting were not included in the total aircraft force and moment calculations. An afterburner nozzle was also included in the geometry. The 15%-scale wind-tunnel model had an internal cavity in the aft part of the model. This cavity was modeled in the CFD geometry and is shown in Fig. 2. The air enters the open cavity via the inlet duct and leaves the cavity through the nozzle. The sting, with a mounting plate attached to the end, extends through the nozzle and downstream of the aircraft. As mentioned above, two different geometries were modeled during this study. One of the geometries had a horizontal tail deflection of 0, while the other had a horizontal tail deflection of 20. The computational grids associated with these geometries had 15 and 16.3 million cells, respectively. B. Approach Kestrel was used to evaluate the low-speed, high angle-of-attack longitudinal S&C characteristics of the F/A-18E at Mach The DDES/SA turbulence model was used. For this case, the aircraft was pitched at a rate of degrees/second (equivalent to a pitch rate of 0.47 degrees/second for the full-scale aircraft) starting from a converged solution at -10 angle of attack. The pitch sweep was terminated at an angle of attack of 65. It is anticipated that the pitch rate is small enough to obtain quasi-static results that can be compared to the wind-tunnel data. Calculations were conducted in the exact same way for both geometries that were described above. The increments between the pitching-moment coefficients from these two geometries were used to calculate the tail effectiveness. C. Results While two geometries were evaluated during this study, only the forces and moments from the geometry with a tail deflection of 20 will be presented here. These results and the results for a tail deflection of 0 were used to calculate the tail effectiveness, which will also be presented. In Fig. 3, the forces and moments for the configuration with a horizontal tail deflection of 20 are shown. All of the results are plotted as a function of angle of attack, which varies between -5 and 65. The calculations were conducted at Mach at a Reynolds number based on mean aerodynamic chord (MAC) of 1.15 million. Fig. 3 shows results for the lift, drag, pitching-moment, axialforce, and normal-force coefficients. The results from CFD generally compare very well with the wind-tunnel data. CFD slightly under-predicts the wind-tunnel data for angles of attack under 30. This same under-prediction has been seen with other flow solvers 5, so this issue is not specific to Kestrel. The axial-force coefficient is overpredicted in the vicinity of 15 angle of attack, as is the pitching-moment coefficient at 0 angle of attack. Kestrel does a very good job at predicting the forces and moments at the higher angles of attack. Note the unsteadiness in the forces and moments. The pitching-moment and axial-force coefficients are particularly unsteady at the higher angles of attack. In Fig. 4, the increment in pitching-moment coefficient between the geometries with 0 and 20 tail deflections is shown. The results are compared to the wind-tunnel data. The unsteadiness in the pitching-moment coefficient that is seen in Fig. 3 is amplified in this plot because of the fact that there is unsteadiness in the results from both solutions. The plot in Fig. 4 indicates that Kestrel is capable of accurately predicting the pitching-moment increment due to tail deflection. D. Computational Efficiency The results presented for this case were obtained by a dynamic pitch sweep. Before pitching the aircraft in angle of attack, a converged solution was obtained at -10 angle of attack. This solution was obtained with 14,000 time steps using three Newton sub-iterations. This calculation required approximately 10,000 CPU hours on 512 processors of a Cray XE6. The dynamic pitch sweep itself required approximately 55,000 time steps using five Newton sub-iterations. This calculation required approximately 112,000 CPU hours on 1024 processors of a Cray XE6. The entire pitch sweep including the initial static run was completed in less than six days. 4

5 IV. Transonic Longitudinal Stability and Control In this section, the results of the transonic longitudinal S&C characteristics of the pre-production F/A-18E are presented and discussed. These calculations were conducted at Mach 0.8 and 0.9 and the results were compared to wind-tunnel data. In this regime, the pre-production F/A-18E is subject to AWS and this is reflected in the forces and moments. Thus, this is an opportunity to assess whether Kestrel is capable of predicting the presence of AWS on the aircraft. Previous calculations on the pre-production F/A-18E in this regime were done by Green 7. A. Geometry and Grid Descriptions The geometry evaluated during this computational study was that of the 8%-scale pre-production F/A-18E windtunnel model. A picture of the computational geometry is shown in Fig. 5. The wing was modeled with 6/8/4 flaps, indicating that the leading-edge flap was deflected 6, the trailing-edge flap was deflected 8 and the aileron was deflected 4. The geometry was modeled with both horizontal and vertical tails present. Horizontal-tail deflections of -6, 0 and 6 were evaluated. The rudders had a neutral deflection. A Sidewinder missile and launcher were modeled at the wing tip and the inlet was modeled as flow-thru. A sting was also included in the geometry as a viscous surface, but the forces and moments on the sting were not included in the total aircraft force and moment calculations. The model has aft-body distortion due to the presence of the sting. As mentioned above, geometries with tail deflections of -6, 0 and 6 were modeled during this study. The computational grids associated with each of these geometries had 16.7 million cells. B. Approach Kestrel 1.1 and 2.1.1b were used to evaluate the transonic longitudinal S&C characteristics of the F/A-18E at Mach 0.8 and Mach 0.9. The SA and DDES/SA turbulence models were used. Both static and dynamic pitch sweep calculations were evaluated. For the pitch sweep calculations, the aircraft was pitched at a rate of 2.5 degrees/second (equivalent to a pitch rate of 0.2 degrees/second for the full-scale aircraft) starting from a converged solution at -10 angle of attack. The pitch sweep was terminated at an angle of attack of 13. It is anticipated that the pitch rate is small enough to obtain quasi-static results that can be compared to the wind-tunnel data. Static calculations were conducted in the exact same way for all three geometries that were described above, in an effort to determine the ability of Kestrel to predict the impact of tail deflections on the lift and pitching-moment coefficients. C. Mach 0.8 Results The initial set of calculations being presented was conducted at Mach 0.8 at a Reynolds number of 3.9 million based on MAC. Kestrel 2.1.1b was used for these calculations, along with the DDES/SA turbulence model. The forces and moments for CFD and the wind-tunnel data are shown in Fig. 6 as a function of angle of attack. The lift, drag, pitching-moment, axial-force, and normal-force coefficients are shown in the figure. There are four different sets of wind-tunnel results provided. All four sets of data are plotted in an effort to show the consistency of the wind-tunnel data. Based on the wind-tunnel data, AWS occurs at an angle of attack of approximately 7. Although it is not apparent in the plots, there is a break in the wind-tunnel data at this angle of attack. Both static and dynamic pitch sweep results from CFD are presented in the figure. There is general agreement between the static and dynamic pitch sweep results, indicating that the pitch rate chosen is able to predict quasi-static results. Kestrel predicts AWS at approximately 5. Kestrel also predicts a shallow break rather than the sharp break that one would expect. The lift is well predicted above and below the AWS event. The drag coefficient is accurately predicted with Kestrel. The pitching-moment coefficient is fairly well predicted, although there is a slight disagreement at the higher angles of attack. The axial-force coefficient from Kestrel is accurate below 5, where AWS is predicted. Above the AWS event, Kestrel over-predicts the axial-force coefficient. The normal-force coefficient is also plotted, although its behavior is similar to the lift-coefficient. On the pre-production F/A-18E, there are several indicators in the forces and moments that AWS is present. These indicators are the sharp loss in lift and normal force, the sharp increase in axial force, and the sharp increase in pitching moment. Each of these characteristics is apparent in the results and data that are presented in Fig. 6. While Kestrel s prediction of these changes in the forces and moments are not perfect, Kestrel does predict the proper trends of these changes which indicate that Kestrel is capable of predicting the presence of AWS on this aircraft at these conditions. It is important to understand these indicators of AWS on the pre-production F/A-18E. For the constant-g wind-up turn maneuver this is presented later, these artifacts are also present in the Kestrel results, although there is no data available to corroborate the results. 5

6 D. Mach 0.9 Results Calculations with Kestrel were also conducted at Mach 0.9 at a Reynolds number of 4 million based on MAC. These calculations were conducted with Kestrel 2.1.1b using the DDES/SA turbulence model. Both static and dynamic pitch sweep results from CFD are compared to the wind-tunnel data in Fig. 7. Once again, four sets of wind-tunnel data are plotted in the figure. Based on the lift coefficient from the wind-tunnel data, AWS occurs at an angle of attack of 8. Once again, Kestrel under-predicts the AWS angle of attack at 5. Below that point, Kestrel predicts the lift coefficient fairly well. The results from Kestrel agree well with the data above 9. The agreement for the drag coefficient is very good. Kestrel does a good job of predicting the pitching-moment coefficient below AWS. The difference in the AWS angle of attack is also clearly evident in the plot of pitching-moment coefficient. At the higher angles of attack, there is a slight disagreement in the results between Kestrel and the wind-tunnel data. The axial-force coefficient from Kestrel agrees well with the wind-tunnel data, although Kestrel slightly overpredicts the wind-tunnel data at the higher angles of attack. This is likely a result of the difference in the AWS angle of attack between Kestrel and the wind-tunnel data. The normal-force coefficient is also plotted, although its behavior is similar to the lift-coefficient. E. Tail Effectiveness In an effort to show the ability of Kestrel to predict the tail effectiveness of the F/A-18E at transonic speeds, the lift and pitching-moment coefficients are plotted against the wind-tunnel data for tail deflections of -6, 0, and 6 in Fig. 8. These results were conducted at Mach 0.8 at a Reynolds number of 3.9 million based on MAC. Kestrel 1.1 with the SA turbulence model was used to generate these results. Static CFD runs were conducted for these calculations. These results clearly show that Kestrel can accurately predict the tail effectiveness of the F/A-18E at transonic speeds. F. Computational Efficiency The results presented for this study were obtained by static calculations and dynamic pitch sweeps. The typical static calculation required 14,000 time steps using three Newton sub-iterations. Each static calculation required approximately 11,000 CPU hours on 256 processors of a Cray XE6. This translates to less than two days per calculation. The dynamic pitch calculations were started from a converged solution at 2 angle of attack. As an example, one of the dynamic pitch sweeps required approximately 202,000 time steps using five Newton subiterations. This calculation required approximately 537,000 CPU hours on 1024 processors of a Cray XE6. The entire pitch-sweep calculation with the initial static run was completed in 24 days. V. Transonic Lateral/Directional Stability and Control In this section, the transonic lateral/directional S&C results for the pre-production F/A-18E Super Hornet are presented and discussed. Calculations were conducted at Mach 0.8 and 0.9 at several different angles of attack. The results were compared to wind-tunnel data. Previous work in this area was done by Green 7. A. Geometry and Grid Descriptions The geometry evaluated during this computational study was that of the 8%-scale pre-production F/A-18E windtunnel model. A picture of the computational geometry is shown in Fig. 5. The wing was modeled with 6/8/4 flaps, indicating that the leading-edge flap was deflected 6, the trailing-edge flap was deflected 8 and the aileron was deflected 4. The geometry was modeled with both horizontal and vertical tails present. The rudders and horizontal tails had neutral deflections. A Sidewinder missile and launcher were modeled at the wing tip and the inlet was modeled as flow-thru. A sting was also included in the geometry as a viscous surface, but the forces and moments on the sting were not included in the total aircraft force and moment calculations. The model has aft-body distortion due to the presence of the sting. The computational grid associated with this geometry had 16.7 million cells. This same geometry was used for the transonic longitudinal S&C results presented in the previous section. B. Approach The transonic lateral/directional S&C characteristics of the F/A-18E were evaluated with Kestrel at Mach 0.8 and Mach 0.9. Kestrel 1.1 with the SA turbulence model was used for these calculations. Static calculations were conducted at discrete angles of attack and angles of sideslip. Sideslip angles between -5 and 5 were evaluated. At each angle of attack, the initial calculations were conducted at a sideslip angle of 0. All other cases at each angle of attack were then restarted from the calculation at 0 sideslip. 6

7 C. Mach 0.8 Results At Mach 0.8 with a Reynolds number of 3.9 million based on MAC, calculations were conducted at 5 and 10 angle of attack. The rolling-moment, yawing-moment and side-force coefficients for Kestrel and the wind-tunnel data are plotted in Figs. 9 and 10. The agreement between Kestrel and the wind-tunnel data is very good for each of the cases. D. Mach 0.9 Results Calculations were also conducted at Mach 0.9 at a Reynolds number of 4 million based on MAC. These calculations were conducted at 5, 7 and 10 angle of attack. The results from these calculations are compared to the wind-tunnel data in Figs. 11 through 13. The results for 5 angle of attack are shown in Fig. 11. Once again, the agreement between Kestrel and the wind-tunnel data is very good. The results for 7 angle of attack are shown in Fig. 12. Two sets of wind-tunnel data are shown in the plots. The agreement between Kestrel and the wind-tunnel data for the yawing-moment and side-force coefficients is extremely good. There are differences between the two sets of wind-tunnel data for the rolling-moment coefficient. In general, the two sets of wind-tunnel data bound the Kestrel results. The results for 10 angle of attack are shown in Fig. 13. Once again, Kestrel does an excellent job of predicting the yawing-moment and side-force coefficients. For the rolling-moment coefficient, the wind-tunnel data is linear. The Kestrel results, however, are not linear. It is interesting to note that the Kestrel results for the rolling-moment coefficient at 10 angle of attack in Fig. 13 are similar in shape to the rolling-moment coefficient from the wind-tunnel data for 7 angle of attack in Fig. 12. E. Computational Efficiency The results presented for this study were obtained by static calculations. The typical static calculation required 24,000 time steps using three Newton sub-iterations. Each static calculation required approximately 19,000 CPU hours on 560 processors of an SGI Altix machine. This translates to less than two days per static calculation. VI. Transonic Roll Damping The results of the transonic roll-damping calculations are presented in this section. The calculations were conducted at Mach 0.8 for wind-tunnel scale and for the aircraft in flight. The accuracy of the results was assessed using wind-tunnel data and known trends of the aircraft in flight. This presents an opportunity to assess whether Kestrel is capable of accurately predicting the presence of AWS on the aircraft. Two different configurations were evaluated during this study. One of the configurations is known to have tendencies toward AWS while the other configuration does not. Previous calculations on the roll damping of the F/A-18E Super Hornet were conducted by Green 6. A. Geometry and Grid Descriptions Two different F/A-18E configurations were evaluated during this study. These configurations are shown in Fig. 14. On the left, the pre-production F/A-18E configuration is shown. This is the configuration that first experienced AWS in 1996 during the Engineering and Manufacturing Development phase of flight testing. This configuration was modeled with 6/8/4 flaps. The configuration on the right in the figure is the configuration from the Transonic Flying Qualities Improvement (TFQI) program. On this configuration, the flap setting has been changed to 10/10/5, the leading-edge snag has been replaced with a Sawtooth geometry and a wing fence has been added. These changes were made in an effort to eliminate the issues associated with AWS. Grids were generated on the pre-production geometry for wind-tunnel scale calculations and flight calculations. Each of these grids had 12.5 million cells. A grid for the TFQI configuration was generated for flight calculations. This grid had 18.7 million cells. B. Approach Calculations were conducted at Mach 0.8 in an effort to assess the ability of Kestrel to predict the roll-damping characteristics of the pre-production F/A-18E. For this study, Kestrel 1.1 was used with the SA turbulence model. To predict the roll-damping characteristics of the aircraft, several phases of calculations were required. Each set of calculations were conducted at a specific angle of attack. The first phase of calculations was conducted with the aircraft oriented in a wings-level position. In this orientation, the static solution was converged. A notional time step of Δt was used. Then, the grid was rotated about the body axis at the specified roll rate. The grid was rotated 270 at a time step of 10Δt and the solution was stopped. The solution was then restarted at that point at a time step of Δt and the aircraft was rotated at the specified roll rate through the wings-level position. When the aircraft rolled 7

8 through the wings-level position, the rolling-moment coefficient was used along with the roll rate to determine the roll-damping coefficient. C. Results Abrupt wing stall was documented on the pre-production F/A-18E in the wind tunnel and in flight. As a result, this tendency toward AWS should be manifested in the roll-damping characteristics of the aircraft at both conditions. The first set of calculations were conducted at Mach 0.8 at 8%-scale to simulate the wind-tunnel test. The CFD calculations were conducted at a roll rate of 250 degrees/second. These results are compared to the windtunnel data on the left side of Fig. 15. It should be noted that the roll-damping coefficient can be obtained directly from the CFD calculations. However, roll damping cannot be obtained directly from the wind-tunnel data. This is specifically noted in Fig. 15. As a result, the comparison between CFD and the wind-tunnel data is not one-to-one. It is also important to note that negative (stable) values of roll damping are plotted in the top half of the plot. The positive (unstable) values of roll damping are shown in the bottom half of the plot. Positive (unstable) values of roll damping imply that the aircraft could be susceptible to AWS. On the left side of Fig. 15, it can be seen that the roll damping from the wind-tunnel data tends toward being unstable between 8.1 and 8.8 angle of attack. In fact, between these two angles of attack, the model motion became too violent to take data. The roll-damping coefficient from Kestrel is also shown in the figure. These results indicate that the roll damping tends toward being unstable between 7 and 8 angle of attack. As a result, Kestrel appears to be capturing the correct trend of the roll damping at wind-tunnel conditions. On the right side of Fig. 15, the roll damping is plotted from Kestrel at flight conditions. These calculations were conducted at Mach 0.8 at 30,000 ft at a roll rate of 20 degrees/second. The non-dimensional roll rate at full scale is equal to the non-dimensional roll rate that was used for the calculations at 8%-scale. As a result, the comparison of the roll-damping coefficients between the wind-tunnel scale and full-scale CFD results should be one-to-one. The full-scale results from Kestrel show that the roll-damping coefficient is unstable (positive) between 7.5 and 9 angle of attack. This trend in roll damping is consistent with the fact that the pre-production aircraft was susceptible to AWS during flight. On the left side of Fig. 16, the results that are presented in the right side of Fig. 15 are repeated. These results show Kestrel s prediction of the roll damping of the pre-production F/A-18E at Mach 0.8 at 30,000 ft. On the right side of Fig. 16, the results for Kestrel on the TFQI configuration are plotted. All of the results in Fig. 16 are for a roll rate of 20 degrees/second at Mach 0.8 at 30,000 ft. Based on the results of the TFQI program, the TFQI configuration should not be susceptible to AWS. As a result, all of the roll-damping coefficients should be negative (stable). While Kestrel does predict some unstable values of roll damping for this configuration, the trend from the left side of the figure to the right side of the figure is correct. That is, Kestrel predicts that the TFQI configuration should be less susceptible to AWS than the pre-production F/A-18E configuration. This trend is correct and has been confirmed in flight. D. Computational Efficiency The results for the roll-damping calculations were obtained in several steps as described above. The computational efficiency of a typical case will be presented here. The initial static calculation required 14,000 time steps using three Newton sub-iterations. This calculation required approximately 6000 CPU hours on 576 processors of an SGI Altix machine. The second phase of the calculations required 4320 time steps using five Newton sub-iterations. This calculation required approximately 2000 CPU hours on 576 processors. The third phase of the calculations required 15,000 time steps using five Newton sub-iterations. These calculations required approximately 7000 CPU hours on 576 processors. In total, the roll-damping calculation at each angle of attack was completed in approximately one day. VII. Longitudinal Stability and Control in Flight at Mach 0.6 Calculations were conducted on the full-scale pre-production F/A-18E Super Hornet in flight at altitude to evaluate the longitudinal S&C characteristics of the aircraft. The calculations were done at Mach 0.6 and 30,000 ft with fixed flaps and control surfaces. The results were compared to data from the F/A-18E flight database. This represented the first attempt by NAVAIR to do calculations on the F/A-18E at flight and compare the results to the flight database. 8

9 A. Geometry and Grid Descriptions The geometry evaluated during this computational study was that of the full-scale pre-production F/A-18E Super Hornet. A picture of the computational geometry is shown in Fig. 17. The wing was modeled with 0/0/0 flaps, indicating that the leading-edge flap, trailing-edge flap, and aileron deflections were neutral. The geometry was modeled with both horizontal and vertical tails present. The horizontal tails and rudders had neutral deflections. A Sidewinder missile and launcher were modeled at the wing tip and the inlet was modeled as flow-thru. The computational grid associated with this geometry had 35.3 million cells. B. Approach Kestrel with the DDES/SA turbulence model was used to evaluate the longitudinal S&C characteristics of the F/A-18E at Mach 0.6 at an altitude of 30,000 ft. For this study, static calculations were conducted at specific angles of attack. Dynamic pitch sweep calculations were not conducted. C. Results The results for this study were compared to the results from the NAVAIR F/A-18E Flight Database. This database is comprised of wind-tunnel data that has been corrected with flight-test data. This database is the most accurate and extensive database that is available for the F/A-18E. The CFD results for the lift, drag, axial-force, normal-force, and pitching-moment coefficients are compared to those of the flight database in Fig. 18. The CFD results compare extremely well to the flight database up to an angle of attack of 35. Above 35, the Kestrel results deviate from the flight database. While not shown in the figure, a second similar flow solver was used to confirm the Kestrel results at an angle of attack of 50. It is believed that the deviation at the higher angles of attack occurs because the comparison is not one-to-one at the higher angles of attack. As mentioned above, the database is comprised of data from the wind tunnel and flight test. However, it is unlikely that the database contains any 0/0/0 flaps wind-tunnel or flight-test data at the higher angles of attack. Simply put, the aircraft does not fly with neutral flaps and ailerons at high angles of attack. Thus there is no reason to obtain data at these conditions. The forces and moments from the flight database at the higher angles of attack were likely obtained through a series of interpolations and extrapolations. On the other hand, the CFD calculations were conducted at 0/0/0 flaps for all angles of attack considered. D. Computational Efficiency The results presented for this study were obtained by static calculations. The typical static calculation required 30,000 time steps using three Newton sub-iterations. Each static calculation required approximately 62,000 CPU hours on 512 processors of a Cray XE6. This translates to approximately five days per static calculation. VIII. Constant-g Wind-Up Turn Maneuver In an effort to assess the ability of Kestrel to model a maneuvering Navy aircraft, Kestrel was used to model the pre-production F/A-18E doing a constant-g wind-up turn maneuver. The data for the maneuver was obtained from an actual flight test of the aircraft. This maneuver is significant in the history of the F/A-18E program. It was during a similar wind-up turn maneuver in 1996 that AWS was first discovered on the aircraft. As a result, it is a fitting maneuver to study with CFD to confirm that Kestrel is able to accurately predict the presence of AWS during this maneuver. While the actual maneuver with the aircraft involved moving flaps and control surfaces, these calculations were conducted with fixed flaps and control surfaces. This calculation represented the first attempt by NAVAIR to model a maneuvering aircraft. A. Geometry and Grid Descriptions The geometry evaluated during this computational study was that of the full-scale pre-production F/A-18E Super Hornet. A picture of the computational geometry is shown in Fig. 19. The wing was modeled with 6/8/4 flaps. The geometry was modeled with both horizontal and vertical tails present. The horizontal tails and rudders had neutral deflections. A Sidewinder missile and launcher were modeled at the wing tip and the inlet was modeled as flowthru. The computational grid associated with this geometry had 42.2 million cells. B. Approach Kestrel with the DDES/SA turbulence model was used to model the F/A-18E doing a constant-g wind-up turn maneuver. Using data from an actual constant-g wind-up turn maneuver of the F/A-18E, the motion inputs for Kestrel were determined to simulate this motion. The velocity, altitude, angle of attack, angle of sideslip, roll 9

10 orientation, pitch orientation and yaw orientation of the aircraft as a function of time were required. The entire wind-up turn maneuver consisted of approximately 43 seconds of data. The actual wind-up turn data obtained for the F/A-18E was very noisy. After a calculation was conducted using this noisy data, the data were smoothed and a second calculation was conducted. This resulted in smoother forces and moments. The goal of these calculations was not to compare the results to known data. The goal was to confirm that Kestrel was able to model a constant-g wind-up turn maneuver of the pre-production F/A-18E and predict the presence of AWS. Correctly predicting AWS on this configuration for this maneuver would go a long way to confirming that Kestrel was capable of screening aircraft for AWS early in the acquisition process. C. Results The angle of attack and angle of sideslip as a function of time for the constant-g wind-up turn maneuver are shown in Fig. 20. These plots show both the original (noisy) maneuver data and the smoothed maneuver data. The angle of attack of the maneuver increases to approximately 17. The sideslip angle is relatively small throughout the entire maneuver. Note that the noisiness of the data was eliminated by the smoothing. In Fig. 21, the lift, drag, axial-force, rolling-moment and pitching-moment coefficients are plotted as a function of time throughout the maneuver. Once again, results for the original and smoothed maneuver are shown. As expected, the forces and moments of the smoothed maneuver are better behaved than those of the original maneuver. The important result of these calculations is shown in the lift, axial-force and rolling-moment coefficients. In the vicinity of 35 seconds, the lift coefficient abruptly decreases and the axial-force coefficient abruptly increases. In addition, the rolling-moment coefficient changes character, indicating that lateral activity is present. For the F/A- 18E Super Hornet operating at or near transonic speeds, these changes are indicative of the presence of AWS on the aircraft. Thus, Kestrel has predicted the presence of AWS on the pre-production F/A-18E while doing a constant-g wind-up turn maneuver. This is consistent with the fact that the aircraft itself experienced AWS in a similar maneuver as early as Once again, it is important to mention here that the F/A-18E Super Hornet that is operating in the fleet today does not have issues with AWS. The results presented here are for the pre-production F/A-18E which did have issues with AWS. D. Computational Efficiency Two constant-g wind-up turn maneuver calculations were done during this study. Each calculation was done in the same way with approximately the same computational efficiency. Each calculation required 47,000 time steps using five Newton sub-iterations. Subsequently, each calculation required approximately 392,000 CPU hours on 2048 processors of a Cray XE6. This translates to approximately eight days per calculation. IX. Summary The goal of this study was to evaluate the ability of the Kestrel flow solver to accurately predict the stability and control characteristics of the pre-production F/A-18E Super Hornet. The Kestrel flow solver is currently in development and test cases on a variety of aircraft are being used to assess the accuracy of the code. The quantity of data and the challenge of predicting the forces and moments accurately on the F/A-18E made this aircraft the perfect choice for such a study. Data from wind-tunnel tests, the F/A-18E flight database, and known trends of the data were used to assess the accuracy of the results. Several configurations and conditions were analyzed at both windtunnel and full-scale conditions. While most of the calculations were conducted above Mach 0.6, low-speed calculations were also conducted at Mach The study looked at the longitudinal and lateral/directional stability and control characteristics of the aircraft as well as the evaluation of the transonic roll damping characteristics of the aircraft. Static, pitching, rolling and maneuvering calculations were all conducted during this study. All calculations were conducted with fixed flaps and control surfaces. In general, the Kestrel results correlated well with the truth data. Kestrel was able to predict the presence of abrupt wing stall on the pre-production F/A-18E at transonic speeds and during the maneuver. The low-speed high angle-of-attack calculations of Kestrel also correlated well with the data. The results at full-scale indicate that Kestrel is fully capable of accurately predicting the forces and moments on the aircraft in flight. This study seeked to answer the following questions: (1) Can Kestrel accurately predict the forces and moments of the F/A-18E at wind-tunnel scale for low-speed and transonic flow? (2) Can Kestrel accurately predict the forces and moments of the full-scale F/A-18E in flight at altitude? (3) Can Kestrel accurately predict the tendencies of the pre-production F/A-18E toward abrupt wing stall? 10

11 (4) Can Kestrel model a maneuvering F/A-18E? While a rigorous verification and validation study may be required to fully answer these questions, these results show that Kestrel is able to do a reasonable job of predicting the stability and control characteristics of the preproduction F/A-18E. With further development of the code, the accuracy of the results is expected to improve. Because of the success of this study, this work will be continued in the years to come. Within the next year, the maneuvering F/A-18E will be modeled with moving flaps. This work is also being funded by CREATE/Shadow- Ops. During this project, Kestrel will be used to model several basic F/A-18E maneuvers. While initially the maneuvers will be done with fixed flaps and control surfaces, the final goal is to model the aircraft with moving flaps, rudders and horizontal tails. During flight, the flaps and other control surfaces on the F/A-18E are constantly moving. A maneuver is no exception and modeling a maneuvering aircraft with fixed control surfaces is not representative of flight. This new capability will further enhance how Kestrel and the Naval Air Systems Command will be better able to support the F/A-18E program office and other Navy acquisition programs. Beyond this work, it is anticipated that Kestrel will be used to revolutionize the way the Naval Air Systems Command supports flight clearance requests. Calculations such as this will involve the fully-loaded aircraft doing maneuvers with moving flaps. With these types of calculations, it is possible for Kestrel to provide a significant amount of data in a reasonable amount of time. This data can then be evaluated for a flight clearance decision. This would be a drastic improvement to the current flight clearance process in terms of schedule, accuracy and quantity of data. Acknowledgments The author would like to gratefully acknowledge the funding and computer time that was granted for this study by the HPCMP through the CREATE/Shadow-Ops program. The author would like to specifically acknowledge Dr. Robert Meakin, Mr. Joseph Laiosa, and Dr. Nathan Hariharan of the Shadow-Ops program for their help during this work. In addition, the author would like to acknowledge Drs. Scott Morton and David McDaniel of the Kestrel development team for their guidance and advice during this project. The author would also like to acknowledge Mr. Dan King of F/A-18 Flight Dynamics for his input and support for this project. Finally, the author would like to thank Dr. David Findlay, head of the Applied Aerodynamics and Store Separation Branch at NAVAIR, for his continued support and encouragement. References 1 Morton, S. A., McDaniel, D. R., Sears, D. R., Tillman, B., and Tuckey, T. R., Kestrel: A Fixed Wing Virtual Aircraft Product of the CREATE Program, AIAA , January Morton, S., Eymann, T., McDaniel, D., Sears, D., Tillman, B., and Tuckey, T., Rigid and Maneuvering Results with Control Surface and 6DoF Motion for Kestrel v2, AIAA , January Chambers, J. R., Hall, R. M., Historical Review of Uncommanded Lateral-Directional Motions at Transonic Conditions, Journal of Aircraft, Vol. 41, No. 3, May-June 2004, pp Hall, R. M., and Woodson, S. H., Introduction to the Abrupt Wing Stall Program, Journal of Aircraft, Vol. 41, No. 3, May-June 2004, pp Green, B. E., Computational Prediction of Nose-Down Control for F/A-18E at High Alpha, Journal of Aircraft, Vol. 45, No. 5, September-October 2008, pp Green, B. E., Computational Prediction of Roll Damping for the F/A-18E at Transonic Speeds, Journal of Aircraft, Vol. 45, No. 4, July-August 2008, pp Green, B. E., and Chung, J.J., Transonic Computational Fluid Dynamics Calculations on Preproduction F/A-18E for Stability and Control, Journal of Aircraft, Vol. 44, No. 2, March-April 2007, pp Frink, N. T., Pirzadeh, S., Parikh, P., Pandya, M. J., and Bhat, M. K., The NASA Tetrahedral Unstructured Software System (TetrUSS), The Aeronautical Journal, Vol. 104, No. 1040, Oct. 2000, pp Samareh, J., "GridTool: A Surface Modeling and Grid Generation Tool," Proceedings of the Workshop on Surface Modeling, Grid Generation, and Related Issues in CFD Solutions, NASA CP-3291, 9-11 May, Pirzadeh, S., Three-Dimensional Unstructured Viscous Grids by the Advancing Layers Method, AIAA Journal, Vol. 34, No. 1, Jan. 1996, pp Löhner R and Parikh P. C., Three-dimensional Grid Generation by the Advancing Front Method, Int.J.Num.Meth. Fluids 8, pp (1988). 12 Spalart, P. R., and Allmaras, S. R., A One-Equation Turbulence Model for Aerodynamic Flows, AIAA Paper , Jan

12 Figures Figure 1. 15%-scale pre-production F/A-18E CFD geometry used for the low-speed, high angleof-attack longitudinal S&C calculations Figure 2. Internal duct geometry for the 15%-scale CFD model 12

13 C L Wind-Tunnel Data Kestrel C D C m C A C N Figure 3. Force- and moment-coefficients for Kestrel and the 15%-scale wind-tunnel data for the lowspeed, high angle-of-attack longitudinal S&C calculations at Mach for the configuration with full nose-down control 13

14 Figure 4. Tail effectiveness for Kestrel and the 15%-scale wind-tunnel data for the low-speed, high angle-ofattack longitudinal S&C calculations at Mach Figure 5. 8%-scale pre-production F/A-18E CFD geometry used for the transonic longitudinal and lateral/directional S&C calculations 14

15 Figure 6. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic longitudinal S&C calculations at Mach

16 Figure 7. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic longitudinal S&C calculations at Mach

17 Figure 8. Tail effectiveness for Kestrel and the 8%-scale wind-tunnel data for the transonic longitudinal S&C calculations at Mach 0.8 C l C n C Y Wind-Tunnel Data Kestrel Figure 9. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic lateral/directional S&C calculations at Mach 0.8 at an angle of attack of 5 17

18 C l C n C Y Wind-Tunnel Data Kestrel Figure 10. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic lateral/directional S&C calculations at Mach 0.8 at an angle of attack of 10 C l C n C Y Wind-Tunnel Data Kestrel Figure 11. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic lateral/directional S&C calculations at Mach 0.9 at an angle of attack of 5 18

19 C l C n C Y Wind-Tunnel Data Kestrel Figure 12. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic lateral/directional S&C calculations at Mach 0.9 at an angle of attack of 7 C l C n C Y Wind-Tunnel Data Kestrel Figure 13. Force- and moment-coefficients for Kestrel and the 8%-scale wind-tunnel data for the transonic lateral/directional S&C calculations at Mach 0.9 at an angle of attack of 10 19

20 Figure 14. Pre-production and TFQI F/A-18E CFD configurations used for the roll-damping calculations < 0 C lp C lp + C l. sin C lp > 0 WT Data Mach 0.8 p = 250 deg/sec at 8% scale Mach 0.8 p = 20 deg/sec at Full Scale/30k ft Figure 15. Roll-damping coefficients for the pre-production F/A-18E for Kestrel and the 8%-scale windtunnel data for the transonic roll-damping calculations at Mach

21 < 0 C lp > 0 6/8/4 Flaps Pre-Production 10/10/5 Flaps TFQI Figure 16. Roll-damping coefficients for the pre-production and TFQI F/A-18E configurations for Kestrel for the transonic roll-damping calculations at Mach 0.8 at 30,000 ft Figure 17. Full-scale F/A-18E CFD geometry used for the longitudinal S&C calculations at Mach 0.6, 30,000 ft 21

22 Figure 18. Force- and moment-coefficients for Kestrel and the F/A-18E flight database for the longitudinal S&C calculations at Mach 0.6 at 30,000 ft 22

23 Figure 19. Full-scale F/A-18E CFD geometry used for the constant-g wind-up turn maneuver Figure 20. Angle of attack and angle of sideslip from Kestrel for the original and smoothed constant-g windup turn maneuvers for the pre-production F/A-18E 23