VIEWS: 55 PAGES: 12 POSTED ON: 11/22/2010 Public Domain
UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools Carlos González Biedma Head of CFD Tools & Methods, Aerodynamics Department. AIRBUS MILITARY, Getafe (Madrid), 28906 SPAIN ABSTRACT For the EADS Airbus Military A400M aircraft program, the integration of a new powerplant into a new airframe, made largely of composite materials, has presented challenges in the thermal and ventilation design of the engine nacelle. High fidelity, fully coupled, “aero-thermal” models with internal and external components have been used as an aid in the design process. A large use of fully coupled models using CFD Commercial Of The Shelve (COTS) tools has been made for system and detail component analysis as well as a calibration tool for simpler 1D fast design models. Data from the model has also been used for component thermal sizing and primary structure fatigue calculations. Due to the nature of the problem, with an aircraft with a complex flight envelope, large variations of internal and external flow velocities, large temperature gradients and non-negligible radiative effects, the creation of such models with a “single” tool has been challenging. 1.0 INTRODUCTION Military transport aircraft like the AIRBUS MILITARY A400M (see Figure 1) presents very challenging design issues due to its novel configuration with large structures made of composites and a totally new highly efficient powerplant. The close coupling of the engine-propeller combination with the wing has presented new areas of research and a rethink of traditional or standard practice aerospace methods for the design and analysis of all the systems of the aircraft. The design of the nacelle ventilation and thermal management has been one of the complex issues that had to be rethought due to the time constraints. A heavy use of computer simulation and 3D design has been used extensively in order to speed up the design process while keeping cost to a minimum. Conjugated heat transfer modelling within COTS tools has been used to investigate thermal and ventilation aspects of the nacelle design. RTO-MP-AVT-178 30 - 1 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools Figure 1: AIRBUS MILITARY A400M. The ventilation concept follows the already proven simple designs from previous Airbus Military products such as the C-212, CN-235 and C-295. Two main ram ventilation scoops force air inside the fire zone with an exit to the exhaust which is used as an ejector. Gaps through the spinner region are kept to a minimum by means of a seal between the Propeller Gear Box (PGB) and the front support frame. Leakage through the doors is also keep to a minimum by the use of close tolerances and rubber seals. For ground and low dynamic pressure conditions (such as flight idle, take off and landing), the engine primary exhaust flow is used as a large ejector for the fire zone. The same ventilation flow is used to cool the engine hot components (burner and turbines) since there is no conductive insulation in these regions to save weight. In order to keep other fire zone systems cool there is a radiative shield around the engine. This radiative shield also couples as a double wall for the engine cooling. This device is known internally and in engine manufacturer jargon as a “Case Cooling Duct” or CCD. This untested highly coupled design concept demanded the use of more complex analysis techniques and coupling of radiative and convective tools to accurately determine the boundary layer growth within the CCD for the design of the fire zone and engine cooling as well as a good estimation of engine casing convective coefficients. Since the primary exhaust at low engine powers has a high low pressure turbine exit swirling flow, the eductor efficiency also yielded high uncertainties and a typical 1D approach was considered invalid for this task. Thus, a single 3D flow and heat transfer model was selected for all the ventilation and thermal tasks. Infra-red radiation effects have been included in all the models in order to accurately compute temperatures in the fire-zone and for the CFRP components. 2.0 CFD MODEL The CFD tool selected for this study was Fluent from ANSYS. Several versions of Fluent were extensively used: 6.2, 6.3 and within the last year version 12. Fluent is a general CFD tool based on a finite volume discretization of the Navier Stokes equations in its several forms with a Backward Euler time marching and stabilized through Multi-Grid techniques. 30 - 2 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools A Reynolds Averaged Navier Stokes approach (RANS) was selected for this project with closure by means of simple 1 and 2 equation turbulence models (Spalart-Almaras and k-w). Air is simulated as an ideal gas with specific heat, infrared absorptivity and conductivity as a function of temperature only. Viscosity is assumed to follow Sutherland´s equation. All schemes used are second order. A conformal “modular” mesh approach has been used in order to account for faster mesh modification turnaround time. This mesh has been divided in different regions so each one can be modified without having to modify the other (see Figure 2 for all internal regions). The interfaces between regions are conformal and all fluid and solid volumes are merged prior to calculation with the tool TGRID also from ANSYS. Figure 2: CFD Nacelle Model (Propeller and cowls not shown). The first mesh of the model only considered the internal flows and boundary conditions were imposed at the external interfaces (ventilation inlets, vents, cowl external surfaces and exhaust exit) while the second version simulates all the external flow in order to account for installation effects including propeller and wing interaction as well as uncertainties of the boundary conditions at the exhaust, ventilation scoops and spinner gap interfaces. This latter approach was found to be more useful since boundary conditions could be imposed much further away from regions of interest. From the CATIA full 3D model of the nacelle a triangular surface mesh has been created using I-DEAS mesher. All components above 1 inch in size were modelled. This patch surface has been imported into Icem CFD for the volume mesh generation, surface relabeling and volume labelling. The final tetrahedral volume mesh has been smoothed with T-Grid and “thick” boundary layers created for each of the fluid volumes. This thick boundary is usually made from 3 or 4 layers with a first layer height adjusted in each of the volumes to capture correctly convective effects is all components with the use of RTO-MP-AVT-178 30 - 3 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools wall functions with y+ objective values ranging from 60 to 100. This boundary layer height was extruded for each wall surface and at the interfaces, so boundary layer continuity is not satisfied at these interfaces. This discontinuity was in part offset by a higher resolution of the wall boundary conditions near these interfaces and the interfaces themselves. The final grid has a medium resolution with close to 5.5 million cells, near 11.5 million faces (both internal and boundary conditions) and 1.25 million nodes. The model is composed of over 150 volumes covering a total volume of over 2,000 cubic meters (including external and wing domain) and over 1,400 boundary condition surface zones. The external air domain has a meshed volume of little over 2,100 m3 with a total of over 1.2 million cells which is rather low for a mesh of external aerodynamics but sufficient for the purposes of the present work. Average edge size is approximately 250 mm. The wing portion of the model has a meshed volume of 16 m3 with a total of over 50 thousand cells. Average edge size is approximately 125 mm. The nacelle has a meshed volume of approximately 10 m3 (excluding external and wing volumes) with a total of over 4.0 million cells. Average edge size is approximately 30 mm. It must be noted that any component above one inch has been modelled. This included electrical harnesses, oil lines, fuel lines, fire extinguishing lines, small supports… Leakage from the fire zone to the exterior fluid flow has not been modelled. Since leakage areas can be large for prototypes and even production models, this effect will be included in future analysis iterations. Also, leakage from engine casing components has not been modelled and will be taken into account in future model updates. 3.0 MATERIALS, MASS AND THERMAL INERTIA MODELING The engine nacelle makes use of a very large set of different light weight materials with low conductivity (CFRP, Titanium), high temperature material such as Inconel, as well as other conventional materials such as steel or aluminium. For each of the surfaces of the model a separate material was used to correctly establish the density and heat capacity of the component. Structural (bolts, small stringers, latches,…) and non-structural (paint, insulations, …) components were added to the mass of each major component, since the thickness and surface area are known, the component density and specific heat could be adjusted. With this approach, the model accounts for 98.3% of the system weight provided by the weight and balance department. Although transient RANS analysis was not performed, an effort was put in computing accurately the thermal inertia of all the components in order to be used in simpler transient heat transfer calculations with other specific tools. 4.0 PROPELLER MODELING The usual approach to simulate the propeller in this project is by means of an actuator disk (left image in Figure 3) with a mapping of the pressure rise based on data from a “Propeller Deck” supplied by the propeller manufacturer. This deck provides static pressure rise and swirl data at different propeller radial stations for different altitudes, flight Mach numbers, RPMs and pitch. For low velocities, near Mach 0.0, the propeller decks do not provide accurate results and alternative modelling is required. Besides a constant radial pressure distributions, maps from other sources can be used, for example, maps from transient propeller calculations have been used in this project for testing purposes. 30 - 4 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools Figure 3: Actuator Disk and Frozen Rotor Propeller Model. 5.0 CONDUCTIVE MODELING Engine nacelle cowls make an extensive use of composites while the engine pylon is largely made of titanium. Low component conductivities make the accurate nacelle thermal computation a major objective of the program. CFRP sandwich with honeycomb core are modelled. The faces of the sandwich are modelled with constant bi-axial conductivity finite element thin shells and the core with a variable conductivity solid. Radiative effects are included in the core since radiation has been found not to be negligible in honeycomb cores. Since the conductivity of the CFRP is fairly low, many computations do not include the in-plane conductivity terms. A scheme of the 1D model used for validation of the CFRP conduction experiments is shown in Figure 4. This 1D model yields values of conductivity within 5% of the experimental values. The out of plane 3D conductive represents accurately this 1D modelling approach. Figure 4: Thick Core Conductive model. RTO-MP-AVT-178 30 - 5 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools The following conductivity values have been obtained with the 1D model: Hot side Cold side Ambient Theoretical Experimental Temp [ºC] Temp [ºC] Temp [ºC] K [W/mK] K [W/mK] 123.9 65.6 54.4 0.136 0.144 152.8 73.3 54.4 0.154 0.148 166.7 76.7 54.4 0.163 0.158 6.0 RADIATIVE MODELING Different radiation models are available in the tool: • P-1 Model. • Rosseland. • Discrete Transfer Radiation Model (DTRM) • Surface to surface (S2S). • Discrete Ordinates (DO). Both the S2S and DO models have been tried in this work. The P-1 model was not tried since the optical thickness in this simulation is rather low which is one of the shortcomings of this model. P-1 also tends to over predict heat fluxes and one of the main objectives of this program is to compute accurately the radiative heat fluxes. Rosseland model were not tried because of optical thickness issues. DTRM model is not available in parallel and was not tried. The use of the S2S model was attempted but the calculation of view factors was very time consuming and the tool provided little or no feedback or progress. After 2-3 days of view factor calculations the processes were killed and thus the usage of the S2S was discarded until more reliable and faster view factor calculations are available. This type of modelling also does not currently support non-conformal interfaces and mesh adaptation. The DO model was intensely used in this program. The radiative transport equation is discretized in a set of solid angles (usually 16 or 32) and for other angles, interpolation is used. Since the view factors do not need to be computed the penalty in computation time is not very large, however memory and storage requirements increase since an additional set of 16 or 32 scalar equations are solved. The usual approach to use this set of equations is to initialize the flow with all the variables to make sure there is enough memory and solver does not break out, compute the flow equations with a fairly good convergence and then add the additional set of DO equations. The DO equations are calculated every 4-5 flow iterations decoupled from the main energy equation iteration for the higher Mach number cases. For cases near Mach 0.0 the DO equations are solved every flow iteration. One of the problems found was the impossibility to make a Full Multi Grid initialization with the DO equations active. Also, interpolation between CFD meshes of DO data for initialization produced undesirable results and increased the computational cost of modified geometries or meshes. 30 - 6 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools Solar effects have been also simulated for the ground conditions. A ray tracing approach was used to compute the solar heat fluxes for the nacelle external surfaces as well as the ground increase in temperature. Fluent currently only computes solar effects in its serial version which lead to large computing times. 7.0 CONVECTIVE APPROACH All convective heat flux terms are computed from the wall function implementation of the turbulence models tested. The Spalart-Almaras (SA) and the k-w SST model have been tested. The k-w-SST model yielded higher convection fluxes than the SA but solution stability decreased and time requirements increased. The extraction of convective coefficients from the CFD tools was cumbersome since the estimation within the tool uses the air temperature from the first cell instead of the air temperature from outside the boundary layer. An alternative option of obtaining the wall function convective flux is available but it also suffers from this temperature issue. Thus, convective coefficient data has to be extracted from the convective heat flux, the surface temperature and air data extracted from the first cell outside the prismatic layer. 8.0 HEAT EXCHANGERS AND PIPING The model is divided into several air fluid volumes and other system volumes. Parts of the nacelle systems (like electronic components or valves) have been left in the model in order to accurately impose the heat dissipation of such components. Most of the system piping (oil, pneumatic) interiors are not modelled and the internal fluid temperatures have been imposed in the pipe walls. For those pipes filled with air, the volume has been left with a solid material that replicates the thermal behaviour of air and that is transparent to radiation in order to accurately simulate radiative heat fluxes within the pipe. Figure 5: Engine Exhaust. Each of the 3 heat exchangers (Engine Oil Cooler, Variable Frequency Generator Oil Cooler and Pneumatic Pre-Cooler) is modelled by means of independent fluid volumes representing the radiator volume. In these fluid volumes, volumetric heat dissipation is included to simulate the heat rejected by the RTO-MP-AVT-178 30 - 7 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools exchanger. Each heat exchanger system has also a detailed, more accurate, model in order to capture local effects with greater accuracy. The total pressure losses of the exchanger are simulated by “radiator” internal boundary conditions both at the entry and exit plane of the radiator control column. This type of boundary conditions provides a total pressure loss as a function of the dynamic pressure of the inflow. Since the total pressure losses in most heat exchangers are characterized as a function of mass flow, an iterative procedure is usually required to correctly mach the radiator pressure losses. This correction is required for a detailed analysis of the subsystem but since only overall nacelle ventilation was required, the iterations were not performed for this analysis. Another interesting characteristic of heat exchangers is the fact that the orientation of the core is very important since it locally redirects the flow. This flow redirectioning has been imposed with User Defined Functions (UDF). Two UDF options have been tried: momentum source terms addition and direct velocity components re-projection. The latter has been chosen due to its simplicity and robustness. Figure 6: Exhaust Plume. In order to obtain realistic temperatures in the heat exchangers, and since FLUENT can only limit temperatures globally, an UDF to locally limit the maximum temperature was introduced for each heat exchanger. With this approach the numerical solution became more stable to incorrect settings of the heat exchanger power dissipation. 9.0 ENGINE MODELING The engine is an integral part of the nacelle and it is the largest component of the system. All engine effects and subcomponents (intake, exhaust) must be taken into account for a representative modelling (see Figure 5 for an image of the exhaust region). A small air passage, about 1 inch in height, called Case Cooling Duct (CCD) exits the nacelle ventilation flow driven by the exhaust primary flow that acts as an ejector. The following items have been included in the model: • Engine casing temperatures have been provided by the engine manufacturer. These temperatures are computed iteratively since they are a function of the ventilation mass flow at the CCD and the convective and radiative fluxes. Usually 2 or 3 iterations are required for a good matching. 30 - 8 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools • Engine intake with a pressure outlet with target mass flow rate. • Engine exhaust with a mass flow inlet at the Low Pressure Turbine (LPT) exit plane. Flow exit direction has been imposed to account for exit swirling flow. • For compressor stability, the engine is equipped with Handling Bleed Valves (HBV) that extract air from the compressor. At low power settings these valves are open and eject air in the free stream. These valves are modelled and boundary conditions change depending on the engine power rating. • The Engine Turbine Exhaust Casing (TEC) acts as an LPT stator to increase the exhaust propulsive efficiency and is included in the model since nacelle ventilation is driven by exhaust gases. Figure 7: Main Vent streamlines. Since the complete ventilation model has a modular approach, the exhaust portion of the mesh has been used to compute performance maps to be used in 1D engine performance models. This model has been also used to introduce modifications in the exhaust region of the aircraft. Results from the CFD calculations have been also used to characterize regions of interest for 1D modelling. 10.0 CALCULATED CASES The model has been used for different design activities. • Ground conditions. • Critical Ventilation flight conditions. • Thermal fatigue conditions. • Miscellaneous. Ground conditions typically impose demanding requirements on the systems due to the lack of free-stream flow. The ventilation system needs to provide enough cooling for all components of the nacelle, including CFRP cowls, electronics, engine cooling, pneumatic systems, etc. Usually these conditions need to be RTO-MP-AVT-178 30 - 9 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools calculated for different propeller pitch settings (ground idle, ground feathered and maximum reverse) and engine mass flows. Since Mach 0.0 calculations are very difficult to compute with current COTS tools, small velocities (usually less than 5 knots headwind or 15 knots tailwind) had to be imposed in the model for faster convergence rates and lower CPU times. To determine critical ventilation conditions in flight the approach is different. The 3D CFD model is used to calibrate a simplified 1D model, which is then used to explore the complete flight envelope for critical ventilation conditions, both maximum ventilation and minimum ventilation. For thermal fatigue calculations, a typical aircraft mission is divided into several discrete steady flight conditions for which thermal characteristics are computed characteristics (temperatures, HTC, ventilation mass flows, radiative fluxes). From these calculations a mission was computed by direct interpolation of the steady state thermal results. This “typical” mission can be discretized in 15-20 steady state flight conditions Other ventilation conditions required modifications of the mesh in order to compute other critical aspects of the design. For fire simulations, a flame region following ISO2685 norms was included in the volume of the fire zone to simulate fires near critical components of the nacelle structure. The mesh was refined iteratively near these regions to decrease mesh diffusivity and to obtain better results. Since fires can be active for a long time (over 5 minutes) before the pilot reacts, the flame can achieve a steady state and no transient calculations are required for the ISO flames. For the nacelle intake anti-ice design, the mesh was modified in the intake region to simulate the effect of the hot de-icing flow in the surrounding components. 11.0 COMPUTING APPROACH All the calculations have been performed in parallel with a small 7 PC cluster running Linux. The cluster has the following characteristics: • Double core 64 bits Intel Pentium x86 at 3.40 GHz. • 2Gb per CPU. • Linux OpenSuse 10.1. • Pathscale and PGI for Fortran with MPICH library. • Grid Engine 6.0u7 for load balancing. • Inner connection by a 1Gbps switch. Figure 8: Static Temperature in Flight without Propeller Effects. 30 - 10 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools Principal axes or X-Coordinate Cartesian Partitioning was usually the preferred choice since METIS partitioning was unstable and yielded some problems in Fluent 6.2 & 6.3. With the release of Fluent 12.0 this problem was solved and the preferred partitioning method switched to METIS in its parallel form. For calculations in which in-plane conductivity is active through finite element shells, the walls boundary conditions could not be splitted across partitions. These restrictions yielded large discrepancies in the cell and face counts of the partitions and increased computation time accordingly. Typical flight calculations were performed in 3-4 days of work in 7 CPUS (around 650 CPU hours) with a pressure coupled solver with segregated energy, turbulence and radiation. For ground conditions approximately 6-7 days of computing time in 7 CPUs were required (around 1100 CPU hours) for the ventilation mass flow to stabilize. The addition of shells elements for in plane conduction required an additional 1 or 2 days of CPU time. The convergence rate for near Mach 0.0 was very slow and very low values of CFL (well below 1.0) were required in order to obtain a stable reduction of residuals. 12.0 RESULTS The results obtained with this model during the design phase are being compared to the data been gathered in the flight test phase. In general, satisfactory agreement in the mass flow estimations have been reached for the studied conditions. For maximum dynamic pressure conditions with engine stopped (fire extinguishing critical design point) and ground with propeller feather condition, the ventilation mass flow error is within “2%” of the flow determined from flight test data which is still ongoing with three aircraft prototypes. On the other hand, for larger power flight conditions the agreement was of only around 15%. This large discrepancy between the power on and power off conditions seem to indicate an inaccurate modelling of the propeller effects with the actuator disk or the low pressure turbine exit characteristics which are the only changing factors. This addresses the importance of the propeller and engine modelling and correct imposition of boundary conditions for this design. Currently, the propeller model is been upgraded to a blade element model as well as to a frozen rotor to isolate propeller modelling issues. Typical calculation pathlines are shown in Figure 7 and temperature maps at the middle plane of the model are shown in Figures 6 & 8. 13.0 CONCLUSIONS Current COTS CFD tools can be a useful tool to help in the design and analysis of complex aero-thermal systems like those involved in the ventilation of a turboprop. Very large models with several hundred volumes and thousands of boundary conditions can be prepared and successfully run. Nevertheless, it is the author’s opinion that several areas need to be improved, namely: • The modelling of complex systems in CFD is a very time consuming task due to the man time required to build such complex models and to coordinate all the data from the different departments. The use of “full 3D” models with all the data within single tools would greatly decrease the time required for such data mining. • User Interfaces need to be improved for easier use. • Meshing time needs to be reduced drastically by improved meshing techniques such as wrapping or cut-cell approaches. Resolutions of 5mm or better might be required in order to accurately simulate ventilation of nacelles including gaps and leakage, increasing mesh sizes to values over 10 or even 20 million cells. • Boundary layer creation needs to be improved for such large systems. Tools based in boundary layer inflation techniques have been proven not to be robust enough to cope with this kind of complex models and new ways need to be explored. RTO-MP-AVT-178 30 - 11 UNCLASSIFIED/UNLIMITED UNCLASSIFIED/UNLIMITED Fully Coupled Aero-Thermal Modeling of Aircraft Powerplant Installations with COTS Tools • Mesh smoothers need to be improved for this complex meshes with very large variations in cell sizes and close proximity of components. • CFD tools need to improve management of these large components since they are not designed to cope with so many boundary conditions and fluid zones. • All task and models should be available in its parallelised form in order to decrease overall project lead time. • The mixing of finite volume with finite element discretizations is not robust enough for this kind of problem. A finite volume approach for thin shell conduction is preferable. • Fast computing and switching from primal to dual mesh calculations is desirable. • Fast cell adaptations based on gradients or adjoin for primal and dual meshes are desirable. • CFD tool convergence needs to be improved for near Mach 0.0 conditions in order to accurately simulate ground cases. • All the CFD calculations required significant amounts of “baby-sitting” and changing of default options. More automated, stable, accurate and fast time advancement schemes without user intervention are required. Finally, the author wishes to thank all the Airbus Military Flight Physics and Powerplant departments for their help in the successful completion of this work. Thanks also to the Spanish costumer support team from ANSYS. 30 - 12 RTO-MP-AVT-178 UNCLASSIFIED/UNLIMITED