Main CfdTm Forum Saturne Aster ATAAC TWikiEditRaw editAttachPrint version

TWiki> CfdTm Web>InternalSeminars>InternalSeminar010 (2011-05-09, StefanoRolfo)

TWiki> CfdTm Web>InternalSeminars>InternalSeminar010 (2011-05-09, StefanoRolfo)

Internal Seminar Series 2011, 2011-05-11

C18, 15:30

School of MACE, The University of Manchester

C18, 15:30

School of MACE, The University of Manchester

neil.ashton@postgrad.manchester.ac.uk

**Files:** Abstract: Presentation:

The optimisation of helicopter design and aerodynamic performance has traditionally been achieved through experimental studies and empirical calculations, but over the past 30 years advanced numerical methods such as Computational Fluid Dynamics (CFD) have played an increasingly important role in the conceptual design phases of helicopter design. Simulating the flow over an entire helicopter fuselage is still however a challenging test case for any CFD code.
### Case setup

### Numerical and HPC setup

All calculations were performed using the open-source software, (Archambeau et al. (2004); Fournier et al. (2011)) which was developed by EDF. It is a 3D unstructured code based on a finite-volume numerical method capable of solving both steady and transient flows. A range of turbulence models are available including several RANS and Large Eddy Simulation (LES).
Currently 1st order spacial and temporal discretisation schemes have been used for the convective terms, however future simulations will investigate the effect of using higher-order schemes. Each simulation was run in a transient mode using a non-dimensional time step of .

### Initial Results

Initial results show that while the pressure coefficient distribution (Figure 3) using the $k-\omega$ SST model is in good agreement with the experimental data, the large scale separation observed in the experiments is under-predicted (Figure 4). The two velocity planes show this more clearly in Figures 5 \& 6 and suggest that the SST model using the current numerical settings is unable to capture the correct flow features. The full paper will show results from other URANS models and compare these against a wider range of experimental results (Velocity and vorticity at six windows behind the helicopter fuselage and steady pressure distribution on the backdoor as well as the top and bottom of the fuselage) as well as a more detailed analysis of the HPC performance of .

With the ever-increasing availability of High Performance Computing (HPC) facilities within the aerospace industry, transient calculations using novel Unsteady Reynolds Averaged Navier-Stokes (URANS) models are becoming increasingly useful tools for real-life industrial applications and are still preferred over more accurate but more costly methods such as Large Eddy Simulation (LES) and other time resolving approaches.

Hybrid RANS-LES methods have been used to simulate a helicopter fuselage but with mixed success Le Chuiton et al. (2008). While some of these methods did demonstrate an improved prediction for the drag, the results for the lift were not consistently better than the baseline RANS model and several models were shown to depart further from the experimental values. The authors of Le Chuiton et al. (2008) recommended investigating the effect of grid refinement and the choice of time step to improve the agreement with the experimental data and to ascertain what benefits these methods could have.

While the study of Le Chuiton et al. (2008) investigated the SST model using a steady algorithm on a relatively coarse mesh, a comprehensive review of URANS models using suitable meshes and numerics has not been conducted on this case. It is therefore the intention of this study to investigate a range of URANS models on suitably fine meshes to ascertain whether advanced URANS models can improve upon the standard RANS model used in Le Chuiton et al. (2008).

In this study the industry standard SST model of Menter (1994) is compared to several URANS models including a Reynolds-stress model so as to ascertain the advantage (if any) of employing such a scheme.

The geometry under investigation is that of a wind tunnel scale EC145 helicopter fuselage including a support strut that is mounted beneath the fuselage (Figure 1 (Left)). The model dimensions (excluding the strut) are 1.37m (length), 0.40m (width) and 0.41m (height). The flow is at a Reynolds number of 2.27 million, which is based on a free-stream velocity of = 40 m/s and a length scale m. The flow is at a zero angle of attack and has a zero side-slip angle.
Experimental data is available from Eurocopter Germany (ECD) Le Chuiton et al. (2008) which includes global force and moment coefficients as well as Particle Image Velocimetry (PIV) planes and surface steady pressures at several locations.
A multi-block structured grid of 12.2 million cells (Figure 1 (Right)) has been provided by ECD as part of the ATAAC project (Advanced Turbulence Simulation for Aerodynamic Application Challenges). The maximum non-dimensional wall-distance is below 1 which enables the use of low-Reynolds number RANS models (no need for wall functions).
This flow represents a complex case for URANS models as the flow undergoes transition on the top of the helicopter fuselage and also experiences massive separation from several surfaces that interact to form a complex flow pattern. Moreover, vortex shedding is observed from the support strut and from the backdoor of the helicopter.

Figure 1: EC145 helicopter fuselage geometry (Left). Multi-Block structured mesh for the helicopter (Right). |

A no-slip wall boundary condition was applied to the helicopter and inlet and outlet conditions were applied to the remaining boundary faces. The simulations were run using two computer facilities, a local cluster for initial calculations and debugging and then a HPC facility, HECToR, for the main calculations.

HECToR has two systems with a total of 44,544 cores, a Cray XE6 24-core system and a XT5h system - which includes a Cray XT4 quad-core system and a Cray X2 vector component. All calculations in this study were performed using the XE6 system.

On the XE6 system, each of the nodes has two 12-core AMD Opteron 2.1GHz Magny Cours processors. Each 12-core socket is coupled with a Cray Gemini routing and communications chip. Each 12-core processor shares 16Gb of memory, giving a system total of 59.4 Tb.
showed good performance up to 768 cores on the XE6 system(Figure 2).

Figure 2: Parallel performance on HECToR using . |

Figure 3: Pressure Coefficient distribution through the mid plane for the SST URANS model (red line) and experimental data . |

Figure 4: Streamwise velocity contours through the mid-plane. The five levels of lines represent the magnitude of the streamwise velocity. |

Figure 5: Streamwise velocity distribution at x=0.9 behind the helicopter fuselage. |

Figure 6: Streamwise velocity distribution at x=0.9 behind the helicopter fuselage. |

I | Attachment | Action | Size | Date | Who | Comment |
---|---|---|---|---|---|---|

png | A_exp.png | manage | 48.7 K | 2011-05-09 - 19:50 | StefanoRolfo | |

png | A_sst.png | manage | 49.7 K | 2011-05-09 - 19:51 | StefanoRolfo | |

png | B_exp.png | manage | 43.1 K | 2011-05-09 - 19:50 | StefanoRolfo | |

B_sst.pdf | manage | 85.5 K | 2011-05-09 - 19:49 | StefanoRolfo | ||

png | B_sst.png | manage | 47.1 K | 2011-05-09 - 19:50 | StefanoRolfo | |

NAshton_Abstract.pdf | manage | 1374.0 K | 2011-05-09 - 19:54 | StefanoRolfo | ||

png | cp_sst.png | manage | 59.7 K | 2011-05-09 - 19:33 | StefanoRolfo | |

png | heli_geo.png | manage | 115.9 K | 2011-05-09 - 19:12 | StefanoRolfo | |

png | heli_mesh.png | manage | 512.7 K | 2011-05-09 - 19:20 | StefanoRolfo | |

png | heli_shaded.png | manage | 119.3 K | 2011-05-09 - 19:39 | StefanoRolfo | |

png | speedup_hector.png | manage | 5.8 K | 2011-05-09 - 19:25 | StefanoRolfo |

Edit | Attach | Print version | History: r1 | Backlinks | Raw View | Raw edit | More topic actions

Topic revision: r1 - 2011-05-09 - 19:54:48 - StefanoRolfo

Copyright &© by the contributing authors. Unless noted otherwise, all material on this web site is the property of the contributing authors.