STAR Global Conference 2014 Vienna, March 17-19 Optimization of a Wing-Sail shape for a small boat G. Lombardi, F. Cartoni, M. Maganzi Dept. of Civil and Industrial Engineering of Pisa Aeronautical Section
34 th America s Cup A new concept of wing sail The capabilities of the new configuration were deeply analysed 2
Scientific Research and Sport The Sport at high level is a powerful engine for the Research, in particular for Aerodynamics Typical fields: Typical cycle: Race cars (F1 on the top) Motorbike Offshore Cycling Skiing Bob Kayak... AMERICA S CUP scientific researches implementation on sport development and testing It is an important way to connect the basic scientific research to the innovation in the industrial world application to commercial production 3
The Problem Realization of a Wing-Sail to be used on the boat that will take part to the next edition of the inter-university race «1001 vele» Huge range of Geometrical Parameters which can be modified in order to obtain an improvement in the performances Design of an Optimization Procedure The complexity of the flow that acts on the Wing requires the use of a sophisticated CFD solver to perform the Aerodynamic evaluations inside the optimisation loop 4
The boat Max Overall Length = 4,60 m Max Overall Width = 2,10 m 1 Centerboard 1 Rudder Mast Height = Free Sail Plan Max Surface = 33 m² 5
Work Strategy Multi-Step Wing-Sail Optimisation Procedure 2-D Airfoil Shape Optimisation at different spanwise sections The results obtained are used to fix the Airfoils Shapes at the analysed sections, in order to realize the parametric geometry of the 3-D Wing-Sail 3-D Wing-Sail Optimisation 6
2-D Airfoil Shape Optimisation Optimisation Procedure of the shape and the configuration of the Airfoils at any spanwise section such that, at a fixed value of the Cl coefficient, it is minimized the Cd coefficient, respecting a series of geometrical, structural, technological and regulatory constraints. 7
2-D Airfoil Shape Optimisation modefrontier Optimisation Software, developed by Esteco Intuitive management of the Logical Flow Set of Optimisation and DOE Generation Algorithms Statistical Analysis Tools 8
2-D Airfoil Shape Optimisation Optimisation Procedure Flow Diagram 9
Grid Sensitivity Analysis STAR-CCM+ Grid Sensitivity Analysis The Number of Cells inside the Calculation Domain is modified by controlling the Surface Resolution on the walls of the Airfoils Test Configuration : 10
Grid Sensitivity Analysis 0.32 0.31 0.3 0.29 0.28 0.27 0.26 Cl Cd 0.009 0.0085 0.008 0.0075 0.007 0 20000 40000 60000 80000 100000 120000 0 20000 40000 60000 80000 100000 120000 Number n. celle of Cells (2D) (2-D) n. celle (2D) Number of Cells (2-D) Compromise choice between solution stability and computational costs Number of Cells (2-D) ~ 50k 11
Grid Sensitivity Analysis 12
Geometry 13
Optimisation Parameters 14 Optimisation Parameters Global Parameters Chord GAP R Theta_2 Front Airfoil t_c_1 x_tc_1 3 Shape Parameters Rear Airfoil t_c_2 x_tc_2 3 Shape Parameters 14
2-D Airfoil Shape Optimisation The number of Parameters with respect to which we are interested to conduce the Optimisation needs a number of Initial Designs and Generations that is too large to ensure the proper development of the process. Optimisation-by-Steps 15
2-D Airfoil Shape Optimisation Design of Experiments creation : SOBOL Genetic Algorithm : MOGA II Population : 250 designs Generations : 20 16
Geometry Airfoils generation Script Use of Bézier Curves to realize the shapes 17
Geometry Constraints check Geometrical Structural Regulatory 18
Geometry Realization of the Calculation Domain 1/40 chord 25 chords 10 chords 14 chords 25 chords 19
CFD Analysis Grid Generation 2-D Conversion of the Calculation Domain Imposition of the Physics Models Initial Conditions and Boundary Conditions Calculation of Cd at a fixed value of Cl Authomatisation with Macro and Run on remote HPC 8 CPUs for each simulation (~ 15 minutes) 12 Concurrent Evaluations 20
2-D Airfoil Optimisation The genetic algorithm analizes the results and, when a population is completed, the successive is realized focusing the research in the range of each Parameter where statistically there is the highest probability to find the minimum of the objective function; this happens until the end of the procedure. 21
Root Airfoil Optimisation Chord = 2400 mm Cl = 0.3 Vapparent = 10 knots Standard Air, M.S.L. Step 1 : Preliminary Optimisation of the Front Airfoil Step 2 : Preliminary Optimisation of the Rear Airfoil Step 3 : Full Optimisation 22
Root Airfoil Optimisation Step 1 Preliminary Optimisation of the Front Airfoil 23
Root Airfoil Optimisation Step 1 Relative Optimum Configuration 24
Root Airfoil Optimisation Step 2 Preliminary Optimisation of the Rear Airfoil 25
Root Airfoil Optimisation Step 3 Complete Optimisation 26
Analysis of the Results Global Parameters 27
Analysis of the Results Front Airfoil Rear Airfoil 28
Analysis of the Results Design 2380 300-300 300 600 900 1200 1500 1800 2100 29
Analysis of the Results Validation with finer grid 30
Analysis of the Results Validation with finer grid 31
Analysis of the Results Validation with finer grid Theta_2 Theta_2 32
Analysis of the Results Validation with finer grid Theta_2 Theta_2 33
Analysis of the Results Validation with finer grid Optimum Theta - Cl Optimum Theta [deg] 34
Analysis of the Results Validation with finer grid Comparison between IDs 2380-255 Theta_2 GAP = 0.015 35
Analysis of the Results Validation with finer grid Cl = 0.3 36
Analysis of the Results Validation with finer grid Cl = 0.3 37
Analysis of the Results Validation with finer grid Cl = 0.3 38
3-D Wing-Sail Optimisation The 2-D Airfoil Shape Optimisation can be executed at any spanwise position The objective functions and the constraints imposed can be modified referring to the position considered The shape of the Airfoils at different spanwise positions is fixed, depending on the results of the 2-D Optimisation, in order to realize the 3-D wing 39
3-D Wing-Sail Optimisation Parametric representation of the Geometry New set of Optimisation Parameters New Objective Functions and constraints 40
7 Optimisation Parameters Geometry R_tip Wing Height Root Chord Middle Chord Tip Chord α δ ε Applied to the whole Main Wing with respect to the Wind Applied to the whole Flap, hinged at R on the Main Wing Twist Angle, opposite to δ and proportional to the Height R_middle R_root 41
7 Optimisation Parameters Geometry R_tip δ - ε Wing Height Root Chord Middle Chord Tip Chord α δ ε R_middle δ - ε/2 R_root δ 42
Physics Models Profile of the Apparent Wind 45 deg Vtw_Root Vwa_Root -Vb 30 deg Domain X Axis The boat moves on a 30 degrees direction with respect to the Global X Axis The Wind blows at 45 degrees with respect to the boat moving direction The True Wind Profile and the Boat Velocity are such that the Apparent Wind at the Root section makes a 0 degrees angle with the Main Wing Chord when α = 0 43
Physics Models Vb = 2 m/s Vtw@10m = 5,36 m/s (Von Karman Profile) Vwa components (Global Reference Frame) 44
Grid The Mesh Settings are chosen with the same criterions described for the 2-D Optimisation ~ 1,7 MLN cells 45
3-D Wing-Sail Optimisation Optimisation Parameters Maximum Thrust (Vb Direction) Objective Functions and Contraints Roll Moment on the Mast Root not greater than 1000 N m Plan Surface at rest not greater than 12 m 2 46
The HPC 1024 core cluster (~10 Tflops), AMD 1 GB Ram/Core Infiniband DDR fast network (20 Gb/sec) Authomatisation with Macro and Run on remote HPC 128 CPUs for each simulation (~ 30 minutes) 4 Concurrent Evaluations 47
3-D Wing-Sail Optimisation Design of Experiments: SOBOL Genetic Algorithm: MOGA II 50 Designs 20 Generations 444 Designs Analized 48
Analysis of the Results 49
Analysis of the Results Validation with finer grid Design 382 160 140 120 100 80 60 40 20 0 Thrust [N] 0 1 2 3 4 5 6 7 8 9 10 In the same way, an entire range of evaluations on the Wing-Sail behaviour can be done at any value of the flap and twist angle, for different Moving directions and True Wind Profiles; the Optimised Shape should in fact guarantee its performances in the broadest possible range of race conditions. Alpha 50
Analysis of the Results Validation with finer grid 6,5 MLN cells, Run on 256 CPUs 51
Analysis of the Results Validation with finer grid 52
Analysis of the Results Validation with finer grid 53
Analysis of the Results Validation with finer grid 54
Conclusions and Future Developments The Optimisation Procedure described in this work is a very powerful instrument to investigate on the influence of a wide range of geometrical parameters on the Wing-Sail behaviour. The Analysis can be conduced, thanks to the versatility of STAR-CCM+ and its strong possibility of customization of Macros, for many operative conditions and for different Objective Functions and Constraints, in order to increase the Wing-Sail performances in all the situations that can occur during a race. The next step of this study will be focused on the utilization of the CD-Adapco ADJOINT FLOW SOLVER tool 55
AKNOWLEDGEMENTS Thanks are due to A. Ciampa and E. Mazzoni (INFN of Pisa); through their huge research activity on computer networks, applied on our HPC, they made the computing system very efficient and easy to use, enhancing its performances and improving its reliability. 56
THANK YOU fair winds and calm seas! 57