Drag Analysis for an Economic Helicopter S. Schneider, S. Mores, M. Edelmann, A. D'Alascio and D. Schimke
Content Numerical Simulation vs. Measurement Wind Tunnel Setup Numerical Simulation Setup Discussion of Results Drag Breakdown of EC135 Configurations with different level of complexity Structured Mesh Generation Unstructured Mesh Generation Case Description Discussion of Numercial Results Optimization of Fuselage Components Case Description Configurations Three modified backdoors Discussion of Numercial Results Conclusion 2
Numerical Simulation vs. Measurement Wind Tunnel Setup Wind tunnel measurements of a scaled EC135 model are being carried out at the Technical University of Munich between winter 2010 and spring 2011. The experimental investigation is performed in open test section mode. The model is equipped with 128 steady pressure ports. Unsteady Kulites (21) are placed on the backdoor, the upper rear part of the engine fairing and the vertical fin. The measurements are performed with rotating rotor head. Primary air inlets, outlets and the Fenestron duct have been closed. Presently only global loads can be compared. A detailed analysis of the unsteady pressure data and the PIV data is on going. Wind tunnel model of the EC135 3
Numerical Simulation vs. Measurement Numerical Simulation Setup Geometry simplification: Only the first element of the support strut is modeled. All simulations are performed without the rotor head. The hybrid mesh (commercial software: ICEM-Tetra) consists of 27 prism layers and the remaining volume is filled-up with tetrahedra. The unsteady numerical simulation have been performed with the DLR TAU code. Five different test cases have been analysed at constant angle of attack and different yaw angles. Cabin Engine deck Mast fairing Exhaust Tailboom Stabiliser with Endplates Fin, shroud, bumper Strut Initial conditions and test cases Condition Value Unit M 0.178 [-] p 94465 [Pa] T 284.85 [K] q 2107.72 [Pa] ρ 1.1555 [kg/m³] Δt 0.23*10-3 [s] Menter-SST unsteady [-] angle of attack 0 [ ] yaw angle -20 / -10 / 0 / 10 / 20 [ ] Landing skids CFD model of the scaled EC135 4
Numerical Simulation vs. Measurement Discussion of Results Both the force and the moment coefficients show a good correlation with the experimental data. The rotor head of the wind tunnel model exerts a higher drag and lift force which causes the almost constant discrepancy between the experimental and the numerical data. The maximum deviation of data lies at a yaw angle of +/- 20, which might be also due to the different model geometry (is currently under examination) A detailed analysis of the unsteady pressure data and the PIV data in on going. 5
Drag Breakdown of EC135 Configurations with different level of complexity This section relates the drag breakdown over several components of the full-scale EC135 helicopter only by means of CFD simulations. For this purpose, several configurations with different level of complexity will be investigated. Configuration 1 - Isolated fuselage with closed Fenestron duct and engine inlet and exhaust Configuration 2 - Based on configuration 1 including landing skid components Configuration 3 - Additionally simulation of air mass flow through the inlet of the engine fairing and out of the engine exhaust Configuration 4 - Additionally simulation of the influence of the main rotor on the fuselage by using an actuator disc approach Configuration 5 - Highest level of complexity 6
Drag Breakdown of EC135 Structured Mesh Generation Structured multi-block approach using the HEXA module of the commercial grid generator ICEMCFD. The structured mesh generation necessiates a different meshing strategy compared to the unstructured one. The landing skid components and the engine exhaust components are embedded in several sub-grids communicating with the fuselage mesh through Chimera interpolations. As the structured mesh generation of complex geometries is very sophisticated, only the first three configurations will be considered for comparison. Configuration Part Blocks Cells [Mio.] 1 Complete 135 8.270 2 Structured grid statistic Fuselage 135 8.383 LandingSkid (LK) (right) 48 2.037 LandingSkid (LK) (left) 48 2.037 LK-Connector (front) 24 0.274 LK-Connector (rear) 24 0.240 Complete 279 12.971 Fuselage 135 8.383 LandingSkid (LK) (right) 48 2.037 LandingSkid (LK) (left) 48 2.037 3 LK-Connector (front) 24 0.274 LK-Connector (rear) 24 0.240 Engine Exhaust (right) 18 0.274 Engine Exhaust (left) 18 0.240 Complete 315 12.971 Structured sub-grids of configuration 3 Mesh of landing skid (green) and front landing skid conncetor (red) 7
Drag Breakdown of EC135 Unstructured Mesh Generation The unstructured grids are prepared with the commercial software CENTAUR of CentaurSoft. Hybrid meshing technique using the four primarily element types (tetrahedra, hexahedra, prisms and pyramids) Unstructured grid statistic Configuration Blocks Points 1 1 5377118 2 1 7926226 3 1 11276581 4 1 11045683 5 1 32975146 Generation of a one block mesh without the need of applying the Chimera method. Structured hexahedra elements mainly used on the stator blades, the horizontal stabiliser, the backdoor and the landing skids This facilitate higher stretching ratio of the cells and therefore a reduction of mesh points. Additionally the grid and solution quality is improved. Surface mesh generated by CENTAUR Cut through volume mesh at position y=0 8
Drag Breakdown of EC135 Case Description and Discussion of Numercial Results Case Description: The considered flight state corresponds to a fast level flight at a TAS of 140kts and an altitude of 5000ft (ISA condition). Flight conditions Altitude and atmospheric condition 5000ft ISA True Air Speed (TAS) 140kts Helicopter pitch angle -1.5 Helicopter side slip angle (configuration 1 and 2) -1.5 Helicopter side slip angle (configuration 3, 4 and 5) 0.0 Discussion of Numerical Results: The total drag is divided into three parts: drag of the fuselage components, drag of the tailboom components and drag of the landing skid components Landing Skid Components: The drag analysis of the landing skid components results in a good correlation between the several configurations as well as the different applied flow solvers. Moreover the low RMS deviations, indicated by the black error bars, suggest converged drag values. EC135 drag breakdown 9
Drag Breakdown of EC135 Discussion of Numercial Results Tailboom Components: The massive drag increase can be explained with the additional Fenestron components and the flow separation in the front part of the Fenestron duct. The flow separation occurs since the Fenestron rotor, represented by an actuator disc, produces only sparse thrust in the fast level flight condition. In general the drag values of the tailboom components show a good correlation between the different configurations and the different flow solvers Flow separation in the Fenestron duct The drag value of configuration 3 (FLOWer) seems not to be fully converged, since the error bars (tailboom components) show a wider bandwidth compared to the other drag values. Fuselage Components: The results of the predicted drag of the fuselage components show the largest dispersion between the different configurations and the flow solvers. The increased drag of configuration 5 can be explained again by the additional components of the engine deck. Change of the unsteady flow field in and around the engine deck. 10
Drag Breakdown of EC135 Discussion of Numercial Results Fuselage Components: The integration of the windows and the more detailed floor of the cabin also affects the unsteady flow field and accounts for the drag increase. The landing skid components massively influences the flow field and the flow separation position at the backdoor. Unsteady flow field in and around the engine deck The flow field behind the rear bending tube possess an intense turbulent character and flow separation occurs more upstream. At each bending tube (configuration 5) the flow is interrupted which results in a completely different flow behaviour at the backdoor. There is a reverse flow beginning at the flange of the tailboom and going upstream to the rear cross tube. The flow field at the cabin floor and backdoor is very sensitive which also arises in larger RMS deviations of the drag An apparently contrary behaviour of the drag values between the configurations and the flow solver can be identified (is currently under examination). Flow field at the floor of the cabin and at the backdoor 11
Drag Breakdown of EC135 Discussion of Numercial Results The reduction of the drag between configuration 1 or 2 and 3 can be qualitatively explained by the different flow situation at the inlet of the engine deck. In configuration 1 and 2 the inlet of the engine deck is closed and a retention effect of the air is formed. This turbulent and unsteady air generates a vortex going downstream along the edge between the fuselage and the engine deck. Simulating an air mass flow (engine boundary condition) through the inlet of the engine fairing reduces this effect and therefore the drag. The assumed value for the mass flow is too small since the retention effect still can be observed. Only when the simulating the complete engine deck the retention effect vanishes. Introducing the main rotor represented by an actuator disc increases the drag mainly of the fuselage components. The downwash effect of the main rotor slightly changes the flow field around the engine deck and therefore also the flow field of the remaining fuselage components are affected. Different flow situation at the inlet of the engine deck between the different configurations 12
Optimization of Fuselage Components Case Description This last section will give an outlook towards an economic helicopter by disclosing the potential of aerodynamic improvements of selected components. For this purpose a study of passive shape modifications on the lightweight class helicopter EC135 was conducted. Detailed aerodynamic investigations were carried out with main emphasis on the drag reduction. Main focus was on the modification of the landing gear and the aft body region, which were identified as the main drag contributors. Case Description: The considered flight state is defined as a fast level flight at a true air speed of 140kts and an altitude of 5000ft (ISA condition). Both the rotor head and the components of the Fenestron anti-torque system are not considered. Altitude and atmospheric condition True Air Speed (TAS) Helicopter pitch angle Helicopter yaw/roll angle Flight State 5000ft ISA 140kts (72m/s) -1.5deg 0.0deg However each of the four computations includes an engine boundary condition to represent a more realistic airstream around the aft region of the fuselage All unstructured meshes for this study were generated using the grid generator CENTAUR of CentaurSoft. 13
Optimization of Fuselage Components Configurations Three modified backdoors In the context of the fuselage optimisation investigation three modified backdoors were investigated to determine the aerodynamic drag improvements. Configuration A sharp trailing edge closing the backdoor Configuration B truncated sharp trailing edge closing the backdoor Configuration C backdoor with defined flow separation edges Faired cross tubes (the modified cross tubes and steps are marked green) Baseline - EC135 14
Optimization of Fuselage Components Discussion of Numerical Results The main drag reduction contributors are the landing skids and the backdoor. Introducing faired bending tubes results in a reduction of the fuselage drag for all three configurations. Since the flow around the backdoor is significantly changed the tail unit is affected slightly negatively due to an increased dynamic pressure resulting from the separated vortices. Configuration C shows the smallest drag reduction improvement as a result of the reshaped engine deck fairing in the area of the modified backdoor. For the future development the engine fairing will be investigated in further studies. Modifying the backdoor and adding bending tube fairings an overall drag reduction benefit of approximately ~24% can be reached. Relative drag breakdown of the main components 15
DES of a helicopter fuselage (ATAAC) L. Paluszek, F. Le Chuiton
Experiment angle of attack = 0 degrees angle of side-slip = 0 degrees upstream velocity V = 40 m/s Mach number M = 0.1131 -- Reynolds number Re = 2.27 10 6 m -1 The experiment was carried out at the Technical University of Munich in 2009 Measured quantities: forces, unsteady pressures and averaged velocity components at 6 PIV windows behind the back door Location of the pressure taps and transducers (red) Transition line Experimental setup PIV windows
Numerical model 3 grids considered: 12.2 mln cells, block structured grid (mandatory for ATAAC) 9.9 mln cells, hybrid grid (mandatory for ATAAC) 12.9 mln cells, hybrid hexacore grid Solver settings (URANS): Central scheme with artificial dissipation Preconditioning Least Square gradient reconstruction Menter SST turbulence model Dual time stepping Implicit relaxation solver for inner iterations FAS Multigrid Computational domain Predefined laminar zones Geometry of the wind tunnel model of the EC145 helicopter fuselage
Mesh details Block structured hexahedral grid (12.2 mln cells) ICEM CFD Hybrid tetrahedral grid (9.9 mln cells) CENATUR Hybrid hexacore grid (12.9 mln cells) ICEM CFD
Preliminary URANS results, lambda-2 iso-surfaces
Preliminary URANS results, lambda-2 iso-surfaces Very strong mesh dependence observed
TAU averaging module Testing of the on-the-fly averaging option in TAU Means Variances Averaged surface streamlines A very useful tool for both steady and unsteady solutions Instantaneous (top) vs mean Cp Pressure variance
Numerical challanges in Tau Frequent divergence of the omega equation in the Menter SST model Divergence at coarse multigrid levels when using low dissipation schemes Slow residual convergence (or none at all) for the dual time stepping scheme Convergence was achieved when using ΔT ~ global convective CFL = 1 and at last 100 inner iterations Divergence after restarting from a solution file or after grid adaption Engine inlet boundary condition sometimes blows air into the domain Tau user guide does not mention that Reference bl-thickness parameter is used when initialising turbulent quantities in the solution and it s default value is 1e+22 (which means that all cells within 1e+22 metres from the laminar walls are initialised with TKE and TI = 0) Problems with the averaging module in TAU python
Suggestions Green-Gauss or TSL gradient calculation option for coarse grids CFL reduction factor for coarse grid levels (instead of a single value as it is now) Normalisation of all residuals Especially for DES Hybrid discretisation scheme (upwind for RANS, central for DES) Different numerical dissipation settings for RANS and DES zones