Struct Multidisc Optim (2010) 41:637646 DOI 10.

1007/s00158-009-0437-6

INDUSTRIAL APPLICATION

Aeroelastic tailoring using lamination parameters

Received: 6 February 2009 / Revised: 15 June 2009 / Accepted: 5 September 2009 / Published online: 15 October 2009 c The Author(s) 2009. This article is published with open access at Springerlink.com

Abstract The aim of the present work is to passively reduce the induced drag of the rear wing of a Formula One car at high velocity through aeroelastic tailoring. The angle-ofattack of the rear wing is xed and is determined by the required downforce needed to get around a turn. As a result, at higher velocity, the amount of downforce and related induced drag increases. The maximum speed on a straight part is thus reduced due to the increase in induced drag. A bre reinforced composite torsion box with extension-shear coupled upper and lower skins is used leading to bendingtorsion coupling. Three-dimensional static aeroelastic analysis is performed loosely coupling the Finite Element code Nastran and the Computational Fluid Dynamics panel code VSAERO using ModelCenter. A wing representative of Formula One rear wings is optimised for minimum induced drag using a response surface methodology. Results indicate that a substantial induced drag reduction is achievable while maintaining the desired downforce during low speed turns. Keywords Aeroelastic tailoring Lamination parameters Drag reduction Automotive

1 Introduction In present Formula One (F1) racing, the continuous need for speed drives engineers to push both driver and the car to the limit. F1 cars nowadays regularly reach velocities up to 300 km/h during a race. However, the cars are not optimised for high speed on straight lines, but are mostly optimised to achieve high turning speeds (Katz 2006). In order to be able to turn as fast as possible, the car must have as much traction as necessary to pass the turn at the intended velocity. At low speeds, this traction is mostly a mechanical traction originating from the cars weight acting on the tyres. At high speeds, however, the grip is mostly generated by the aerodynamic downforce. This downforce is provided by the aerodynamic surfaces and aerodynamic shaping of the bodywork (e.g. diffuser, front and rear wing). These aerodynamic devices give the driver the opportunity to go faster in a turn. The disadvantage is that, when the car is not cornering, the larger grip given by the aerodynamic devices, together with the increase in aerodynamic drag, tend to slow the car down. At the dawn of Formula One racing, this problem was solved by means of moveable aerodynamic devices (Wright 1974). The driver had the ability to alter the angle-ofattack of the rear wing, such that on a straight line, the downforce and thus also the aerodynamic drag could be reduced. However, in the present regulations of the Fdration Internationale de lAutomobile (FIA 2007) it is stated that moveable aerodynamic devices are prohibited. The question arises whether it is possible to design a xed exible wing capable of providing the right amount of downforce needed to turn at a certain velocity, while reducing the downforce at higher velocity when compared to a rigid wing. One such solution is presented in Massegur et al. (2007), where a two-element rear wing assembly is

Part of this paper has been presented at the 2008 Sixth International Conference on Engineering Computational Technology in Athens (Thuwis et al. 2008). G. A. A. Thuwis (B) R. De Breuker M. M. Abdalla Z. Grdal Faculty of Aerospace Engineering, TU Delft, Kluyverweg 1, 2629 HS Delft, The Netherlands e-mail: [email protected]

638

G.A.A. Thuwis et al.

used with a exible upper element. Due to its exibility, the gap between both elements disappears at high velocities, causing the rear wing to reduce the produced downforce and induced drag. In the present work, a concept for the reduction of induced drag is proposed based on aeroelastic tailoring of the rear wing. The rear wing contributes to one third of the total downforce (Zhang et al. 2006) and associated induced drag. Furthermore, tailoring the rear wing has no effect on any part of the car behind the rear wing, as would be the case for the front wing. A denition of aeroelastic tailoring within the context of Aerospace Engineering is given by Shirk et al. (1986): Aeroelastic tailoring is the embodiment of directional stiffness into an aircraft structural design to control aeroelastic deformation, static or dynamic, in such a fashion as to affect the aerodynamic and structural performance of that aircraft in a benecial way. So, the target is to make use of the interaction between the structure and aerodynamics (Bisplinghoff et al. 1955; Dowell 2004) to design a rear wing such that at a high velocity the angle-of-attack, and as such the induced drag, of the rear wing are reduced. The wing design, at the same time, should still have the required angle-of-attack to generate sufcient downforce during cornering at low velocity. According to Lynch and Rogers (1976), aeroelastic tailoring creates the possibility of drag reduction compared to a rigid wing. More specically, induced drag reduction using aeroelastic tailoring has been applied to improve Unmanned Aerial Vehicle (UAV) performance by Weisshaar et al. (1998), Bramesfeld et al. (2008), and to Micro Air Vehicles (MAV) in Stanford and Ifju (2008). According to Shirk et al. (1986) and Weisshaar and Duke (2006), wash-out wings are traditionally identied with induced drag reduction. Composite laminate tailoring of the rear wing is proposed to obtain the envisioned result. The focus of this tailoring is on the torsion box of the rear wing (Reheld and Cheung 2003, 2004; Guo et al. 2005, 2006). The effect of washout can be achieved by using an extension-shear coupled upper and lower skin (Fukunaga and Sekine 1994; Diaconu et al. 2002). The top and bottom skin of the torsion box are constructed to generate a circumferentially asymmetric stiffness (CAS) lay-up (Patil 1997). The net result of this construction is a bending-torsion coupling of the torsion box, without having bending-torsion coupling at the individual skin laminate level. This eliminates the curing problems encountered when bending-torsion coupling is present at laminate level. It has been shown by Weisshaar (1981) that bending-torsion coupling using a CAS lay-up is also effective for low aspect ratio wings.

The goal of this paper is to nd a wing design which effectuates such a bending-torsion coupling to maximise the reduction in angle-of-attack of the rear wing at higher velocity while maintaining traction at low speeds. For this optimisation, the material properties are modelled using lamination parameters (Tsai and Pagano 1968), similar to previous work by Fukunaga and Vanderplaats (1991), Fukunaga and Sekine (1992), Miki and Sugiyama (1993), Abrate (1994), where lamination parameters are applied to model and optimise composite plates and shells, or wing structures in Liu and Haftka (2004). In Abdalla et al. (2007) and Kameyama and Fukunaga (2007), the lamination parameters are applied for aeroelastic tailoring of composite wing structures, and in Grdal and Olmedo (1993) and Setoodeh et al. (2006), they are used to model variable stiffness composite plates. In the following rst a closer look at the problem at hand is presented. The third section presents a 2D aeroelastic analysis of a typical section to show the envisioned effect of bending-torsion coupling. This effect is further analysed using a 3D model which is presented in the consecutive section. This 3D model is used in a uid-structure routine which is further described in section ve. Finally the results are presented in the sixth section, followed by the conclusions and recommendations.

2 Problem statement A Formula One car is a compromise between going as fast as possible on a straight track and achieving the highest possible turning speed while cornering. This problem is mostly solved, depending on the lay-out of the track, by optimising the car performance either to go fast through a corner or to go fast on a straight part of the circuit. It would, however, be a great achievement if the best of both worlds could be combined. In the following we analyse the cornering manoeuvre and the straight acceleration. 2.1 Cornering manoeuvre To give a general idea about the maximum turn velocity, the car is represented as a rigid body having a mass m . Consider the force required to turn at a certain velocity: F r eq = m V2 , R (1)

where V is the velocity along the path of the turn and R the radius of that turn. This required force is generated through contact with the ground, where the maximum generated force is dened as: F max = (m g + L ), (2)

Aeroelastic tailoring using lamination parameters

639

The drag force is dened as: D= 1 V 2C D S, 2 (6)

with C D the drag coefcient of the rear wing:

C D = C D0 +

where is the friction coefcient of the tyres, L is the downforce, and g the gravitational acceleration. Using 2 (1) and (2) and recalling L = 1 2 V C L S results in the following expression for the maximum turning velocity: 1 CL S V = g R 2mg
1 2

2 CL , e

(7)

(3)

where C L is the 3D lift coefcient and S the surface area of the rear wing. In order to increase the turning velocity, only C L can be changed in (3). The other constants are assumed to be xed either by regulations (m , S ) or by external and physical factors (, R ). From (3) it follows that the lift coefcient of the rear wing C L has to be increased in order to increase the turning velocity. Since in this work the focus is on the rear wing, only the rear wings contribution to the lift coefcient is considered variable. Thus increasing the overall lift coefcient can be done by increasing the rear wings angle-of-attack. 2.2 Straight acceleration Since the angle-of-attack is a xed value throughout a race, this high angle-of-attack also has an inuence on the straight acceleration performance. Consider a basic representation of a car as depicted in Fig. 1, the total acceleration a of the car is given by: Fa F f D a= , m (4)

where C D0 is the drag coefcient at zero angle-of-attack, is the aspect ratio of the rear wing, and e the Oswald factor. The increase in C L for the higher turning velocity will result in an increase of both the friction force F f in (5), and the drag force D in (7). This leads to a lower overall straight acceleration of the car. The objective of this work is thus to obtain a exible xed wing which reduces both the friction force F f , and the drag force D at high velocity, by reducing the lift coefcient C L at this higher velocity. At the same time, however, the magnitude of the lift coefcient C L should not be reduced at low velocity in order to maximise the low speed turning velocity as dened in (3).

3 Feasibility study In order to clarify the envisioned aeroelastic effect using bending-torsion coupling, a typical section is investigated using a static aeroelastic analysis. In Fig. 2 such a at plate section is depicted, with K h and K the bending and torsional spring stiffness, respectively, h the wing plunge, c the chord length, and other parameters dened as: = 0 + e , Cl = Cl , (8) (9)

where Fa is an acceleration force given by the engine, D is the aerodynamic drag force and F f is the friction force acting on the tyres. Since the engine is a xed given subject, the acceleration force is assumed to be xed as well. So to increase the acceleration of the car, both F f and D are looked at. The friction force can be dened as:1 1 F f = m g + V 2 SC L . 2 (5)

where 0 is the initial rigid angle-of-attack, e is the elastic angle-of-attack, Cl is the 2D lift coefcient, and Cl = 2/(1 + (2/ )) for a nite at plate. From this at plate assumption it also follows that the moment around the aerodynamic center m AC , and the lift coefcient at zero angle-of-attack Cl0 are assumed to be zero.

1 Note

that this is a basic representation of the friction force using a constant friction coefcient. In reality this relation is much more complicated. However, for the purpose of demonstrating the inuence on the acceleration a simple representation will sufce.

Fig. 2 2D section for aeroelastic analysis

640

G.A.A. Thuwis et al.

The moment about the elastic axis due to the aerodynamic loads is then dened as m y = Le. Note that until now, the model does not have a coupling between the bending and torsion. This coupling is introduced using K h in the potential energy equation. Using Lagranges equations (Trk 2000), assuming static aeroelasticity and recalling L = Cl (0 + e )cq , the equations of motion become: Kh K h K h 0 Cl + cq K 0 Cl e h Cl = cq 0 , e Cl e (10) in which q is the dynamic pressure. The behaviour of both systems is studied by using a parameter which describes the relation between the stiffnesses of the system. Similar to Weisshaar and Ryan (1986) and Weisshaar et al. (1998), the cross-coupling coefcient is dened as: = K h . Kh K (11)

at decreasing the angle-of-attack at higher dynamic pressure in order to reduce the induced drag. To facilitate such behaviour, a negative coupling between bending and torsion should be applied.

4 Three dimensional model The two-dimensional analysis has shown that the target behaviour can be achieved when using a negative bendingtorsion coupling . According to Katz (1989) low aspect ratio wings have to be designed using a three-dimensional model. Thus a 3D aeroelastic analysis model is developed to investigate the possible drag reduction achievable on an F1 rear wing. This section denes the wing geometry and the corresponding structural and aerodynamic models used in the analysis. 4.1 Model geometry According to the FIA Season 2008 regulations (FIA 2007), the rear wing can consist of two elements, a front and a rear element, as depicted in Fig. 4. By using a multi-element airfoil, the angle-of-attack of the rearward element can be increased, thereby increasing the overall downforce created by the assembly (Katz 1989). In the present study, the rear element is not considered to simplify the model. This simplication is reasonable for an initial proof of concept. The dimensions of the front rear wing element are found in Table 1 (FIA 2007). The wing is assumed to be attached to the car via central pillars, as shown in Fig. 4. This arrangement maximises the exibility of the wing, making aeroelastic tailoring more effective. To perform the 3D analysis, a wing cross-section should be known. In F1 racing, an airfoil is chosen for its property of being a high lift/low drag airfoil (Liebeck 1978; Katz 1989). Exact data about present F1 airfoils is hard to nd, however, some information is known about airfoils used in the past. One series of airfoils which were often used in F1, and nowadays still are used for airfoil optimisation, are the Benzing airfoils (Benzing 1992), designed by the Italian engineer Enrico Benzing. The chosen airfoil from the

The stiffness matrix has to be positive denite, (van Kan et al. 2005), so the determinant of the stiffness matrix should be larger than zero. Introducing this leads to the following requirement: 1 < < 1 . (12)

The inuence of a positive or negative cross-coupling on the angle-of-attack versus the dynamic pressure q is depicted in Fig. 3. It can be seen that at increasing dynamic pressure, the angle-of-attack will either increase or decrease for positive or negative bending-torsion coupling, respectively. In the present work, the main interest lies

1.15

1.1

<0 =0 >0

/ref []

1.05

0.95

0.9

0.2

0.4 0.6 q/qref []

0.8

Fig. 3 Inuence of cross-coupling on angle-of-attack

Fig. 4 Support method of the rear wing

Aeroelastic tailoring using lamination parameters Table 1 Dimensions of F1 rear wing Name b c Taper ratio Value 1 0.2 1 5 Dimension
z/c 0.2 0.1 0 0.1 0.2 0 0.2 0.4 x/c 0.6 0.8

641

m m

Benzing family is known as BE153-105.2 A cross section of the BE153-105 airfoil is found in Fig. 5. 4.2 Structural model The structural model used in the 3D analysis consists of several components, each of which are depicted in the schematic overview of the wing in Fig. 6. The inner voids of the wing are lled with a structural foam which is not depicted in the schematic overview. This foam is represented in the discrete model using a solid mesh as explained later in this section. The normal directions of the upper and lower skin are dened such that a CAS lay-up is created when the upper and lower skin are modelled using the same material properties. All components are given constant material properties, except for the upper and lower skin of the torsion box, which are described using lamination parameters. Symmetric laminates are used and the skin is assumed to be thin enough to be modelled as a membrane. The in-plane stiffness matrix A is expressed in terms of the lamination parameters: A=h(
0

Fig. 5 Cross-section of the BE153-105 airfoil

both rectangular (CQUAD4) and triangular (CTRIA3) elements for meshing the skin, spars, and endplate. The inner solid mesh consists of both ve sided (CPENTA) and six sided (CHEXA) solid elements. 4.3 Aerodynamic model An aerodynamic model is necessary to perform Computational Fluid Dynamics (CFD) calculations. The pressure forces are calculated by VSAERO (Maskew 1982), a commercially available panel code capable of calculating the non-linear characteristics of arbitrary objects. The use of a panel code for aerodynamic modelling has two limitations. First, viscous effects can only be taken into account using a post-processing step within VSAERO. It was found that the effect of viscosity is not signicant, and results with viscosity are not presented in this paper. Second, an initial wake position has to be provided to VSAERO to be used in its iterative wake relaxation routine. A representation of the entire aerodynamic model including the wake can be found in Fig. 8. A structured mesh is applied to the wing skin and the trailing edge of the endplate. The remaining part of the endplate is meshed with an unstructured mesh. The input for VSAERO is created using a Matlab routine, which uses the Matlab program DistMesh by Persson and Strang (2004) to create the unstructured mesh.

+ V1

+ V2

+ V3

+ V4

4) ,

(13)

with the stiffness invariants , dened in terms of the laminate invariants (Grdal et al. 1999; Diaconu et al. 2002). The four in-plane lamination parameters are dened as: V1 V2 V3 = V4
1 2 1 2

cos 2(z ) sin 2(z ) cos 4(z ) dz sin 4(z )

(14)

with z = z / h the normalised coordinate in thickness direction, h the total thickness of the laminate, and (z ) the bre angle at z . The wing is modelled in Nastran (MacNeal 1972). Using appropriate symmetry conditions, only half of the actual wing has to be modelled, as shown in Fig. 7. The mesh uses
2 The rst 2 digits refer to the maximum relative thickness of the airfoil,

the third digit refers to the position of this maximum thickness. The fourth and fth digit denote the maximum chamber, while the last digit refers to the position of this maximum chamber along the airfoil.

Fig. 6 Schematic overview of wing components

642

G.A.A. Thuwis et al.

Fig. 7 Structural model of F1 wing with BE153-105 airfoil

5 Wing simulation and optimisation For the purpose of analysing and optimising the 3D wing model, a multi-disciplinary routine is created in ModelCenter, coupling the commercial FE code Nastran to the CFD panel code VSAERO trough a uid-structure interaction controlled by Matlab routines. The rst section discusses the aeroelastic analysis, while the second section highlights the applied uid-structure coupling. In the third section the optimisation formulation is presented, and the last section explains the implementation in ModelCenter. 5.1 Aeroelastic analysis This work makes use of two commercial solvers, thus a partitioned method, as used by Farhat and Lesionne (1998) is applied to create the aeroelastic analysis. In a partitioned method, the coupling between uid and structure can be done either by advancing both meshes parallel or serial. The present work uses the serial advancing of both meshes, also referred to as the loosely coupled procedure. The most basic and common way to introduce this loose coupling is by using the Conventional Serial Staggered (CSS) procedure (Farhat and Lesionne 1998; Beckert 2000). The procedure for this scheme (see Fig. 9) starts by calculating the aerodynamic forces on the undeformed structure (step 1). Next, these aerodynamic forces are transformed to the structural mesh (step 2) to be able to calculate the deection of the structure (step 3). This deection is used to update the aerodynamic mesh (step 4) such that the entire procedure can restart on the deformed mesh.

Fig. 9 The Conventional Serial Staggered procedure

5.2 Fluid-structure coupling The structural and aerodynamic model have different meshes. A coupling procedure should be provided to enable the interpolation of forces and displacements between both models. The coupling method used in the present work is based on the principle of equivalence of work. The virtual work done by the aerodynamic loads acting on the aerodynamic model should be equal to the virtual work done by the equivalent aerodynamic loads on the structural model, thereby assuring conservation of energy. According to Beckert (2000) and Beckert and Wendland (2001), a coupling matrix H can be dened such that: u f = H us , fs = HT f f , (15) (16)

with us and u f the displacement of the structure and uid boundary, respectively, and fs and f f the aerodynamic forces on the structural and aerodynamic model, respectively. The coupling matrix H can be constructed using a radial basis function. According to Rendall and Allen (2007), a decaying radial basis function should be chosen. These functions put more emphasis on the local inuence, leading to a more physically meaningful force transfer. Moreover, to reduce the computational effort, a compactly supported radial basis function should be applied. The radial basis function used in this work satises both requirements, and is known as Wendlands C 2 -function (Wendland 1995; de Boer et al. 2007). 5.3 Optimisation formulation An optimisation is performed to be able to create a rear wing which passively reduces its induced drag at higher velocity. However, the wing should still be able to provide the required downforce to be able to pass a low speed turn. The idea is that the initial xed angle-of-attack of the wing is chosen such that the wing is able to provide the required

Fig. 8 Aerodynamic model of F1 wing with BE153-105 airfoil

Aeroelastic tailoring using lamination parameters

643

downforce in low speed turns. By using a bending-torsion coupling, at a higher velocity, the wing should reduce its angle-of-attack as much as possible, thereby reducing the induced drag at this higher velocity. Thus, the objective for the optimisation is to minimise the induced drag from the wing, calculated by VSAERO, by using the lamination parameters V1 , V2 , V3 and V4 as design variables. Since the lamination parameters are dened in terms of trigonometric functions in (14), they are bounded by constraints (Fukunaga and Sekine 1992; Miki and Sugiyama 1993). In order to nd a physically meaningful combination of in-plane lamination parameters, the following requirements (Setoodeh et al. 2005) are used as constraints:
2 2 2 2 (1 V3 ) + 2V2 (1 + V3 ) + V3 + V4 4V1 V2 V4 1, (17) 2V1 2 2 V1 + V2 1,

(18) (19)

1 Vi 1, for i =1,2,3,4. 5.4 Implementation in ModelCenter

ModelCenter (Phoenix Integration 2003) is a process integration environment enabling design and control using various programs from within one visual environment. The ow diagram of the aeroelastic routine using the CSS procedure is depicted in Fig. 10. The initial CFD and Finite Element Model (FEM) meshes are created rst. Using these

meshes, the interpolation matrix H, which is independent of any mesh deformation, can be calculated. The material properties are dened by the optimiser. These steps are completed before entering the iterative process, which consists of the CSS procedure, coupling Nastran to VSAERO. A displacement convergence check is performed to decide on the convergence of the uid-structure iteration. The optimisation is performed using Design Explorer (Cramer and Gablonsky 2004), a surrogate-model-based optimiser available in ModelCenter. Since no gradient information is required by the optimiser, it is highly suitable for usage with commercial codes such as Nastran and VSAERO. The Design Explorer will start by creating initial surrogate models based on an initial set of design variables. Using a sequential approximate optimisation method, these surrogate models are used to perform several local optimisations from different starting points. The local optimisation points and some additional points are then evaluated using the actual model. This procedure is denoted as the Search Step within Design Explorer. After such a search step, either the information from the actual model evaluations is used to update the surrogate models, or a local exploration around a base point is performed with the actual model (Poll Step).

6 Results In this section, the output of the optimisation is presented. The rst section discusses the optimisation results. The second section considers the results of a Multi-BodySimulation. 6.1 Optimisation results The optimisation is performed for three cases, where the upper and lower skin have a thickness of either 2, 3 or 4 layers. A single layer carbon bre ply has a thickness of 0.075 mm. During the optimisation, the wing is simulated at an angle-of-attack of 10 degrees, and at a velocity of 92 m/s. For the 2 layer model optimisation, a total of 113 function evaluations are performed, where each function evaluation needed typically 3 uid-structure iterations to converge. However, 42 of the function evaluations have a set of
Table 2 Optimal lamination parameters for 2/3/4 layer models N layers 2 3 4 V1 [] 0.40625 0.23437 0.13282 V2 [] 0.61328 0.56250 0.18359 V3 [] 0.15625 0.27734 0.35938 V4 [] 0.76563 0.03516 0.80078

Fig. 10 Flow diagram of 3D aeroelastic routine

644
1.02 1 0.98 [] 0.96 0.94 0.92 0.9 Low Speed 0.88 20 40 [] 0.95
ref

G.A.A. Thuwis et al.

ref

C /C 2 Layers 3 Layers 4 Layers Benchmark 60 V [m/s] 80

C /C

0.9 2 Layers 3 Layers 4 Layers Benchmark 60 V [m/s] 80

0.85 High Speed 100 20 Low Speed 40

High Speed 100

(a)

(b)

Fig. 11 Results 2/3/4-layer optimised wing [ = 0.1745 rad]. a Non-dimensional lift coefcient versus velocity. b Non-dimensional drag coefcient versus velocity

design variables which violate the constraints as dened by (17)(19). Therefore no actual function evaluation at these design points is performed. Thus 71 runs are performed inside the design space, consisting of 14 search steps and 3 poll steps. The optimal in-plane lamination parameters for the models having either 2, 3, or 4 layers are given in Table 2. For each of the optimised models, the lift coefcient C L and induced drag coefcient C Di can be plotted against velocity. Comparison of the lift and induced drag coefcient, respectively, against the results for a benchmark wing without bending-torsion coupling are shown in Fig. 11a and b. It can be observed that both the lift and induced drag decrease at higher velocity. This is a direct effect of the wing torsion, which is larger at higher velocity. The torsion induces

a reduction in angle-of-attack, thereby reducing the lift and associated induced drag. By increasing the number of layers, the drag reduction at 90 m/s reduces from about 16% for the 2 layer model to 10% for the 4 layer model. This is due to the increase of the bending stiffness of the wing by adding layers, thereby decreasing the amount of torsion through the bending-torsion coupling. A lower reduction in angle-of-attack leads to a smaller reduction of the induced drag. It is worthwhile to investigate whether the relation found by the 2D analysis is continued by the results from the 3D analysis. Therefore, the non-dimensional lift coefcients for both 3D optimised and benchmark wing (2 layers) are compared in Fig. 12 with the results obtained from a typical section analysis. From the gure it can be observed that the 3D result shows good correspondence with the 2D analysis. 6.2 Multi-body-simulation

1 []

C /C

0.95

The inuence of the exible optimised wing on the driving characteristics of a F1 car is investigated in a Multi-Body Simulation (MBS) using the software Simpack (INTEC GmbH 2004). A model having a benchmark wing and a model equipped with the optimised wing are used for two test cases. Both a 10 second acceleration along a straight
2D No Coupling 2D Coupling 3D No Coupling 3D Coupling 20 40 60 Velocity [m/s] 80 100

ref

0.9

Table 3 MBS: acceleration performance Variable Benchmark model Distance [m] 730.08 99.65 Optimised model 730.48 99.79 Absolute difference 0.40 0.14

Fig. 12 3D versus 2D comparison of non-dimensionalised lift coefcient

Velocity [m/s]

Aeroelastic tailoring using lamination parameters Table 4 MBS: turn performance Variable Benchmark model Distance [m] Velocity [m/s] 156.45 29.95 Optimised model 156.48 29.94 Absolute difference 0.03 .01

645

application of this concept is considered, current framework can be used, however some updates need to be made to the models: The full rear wing assembly needs to be taken into account. This can be done in two ways. The rear element can be considered to be rigid, thus only the aerodynamic inuence of it is taken into account while optimising the front element. Another option is to optimise the material properties of both elements simultaneously in order to increase the drag reduction. The airfoil type of the rear wing needs to be updated based on more recent data. The presence of the car in front of the rear wing needs to be taken into account, as it will have an inuence on the airow acting on the rear wing.

line, and a 1.8 g turn at a constant low velocity of 30 m/s are investigated. The results for the straight line acceleration are given in Table 3, while in Table 4 the turn performance is presented. From both tables is can be seen that the optimised model has an advantage during a straight line acceleration, without reducing the turn performance at lower velocity.

7 Conclusions & recommendations The goal of this work is to nd a rear wing for an F1 car capable of passively reducing its angle-of-attack to create a drag reduction at higher velocities. This goal is achieved by using an optimised circumferentially asymmetric stiffness (CAS) lay-up for the torsion box of the rear wing main element. Lamination parameters are applied to describe the variable material properties of the upper and lower skin of this torsion box. These lamination parameters are used as design variables in an aeroelastic optimisation framework, having a minimisation of induced drag as design objective. Based on the results of this optimisation, the following conclusions can be drawn: Application of bending-torsion coupling through the use of a CAS lay-up can lead to a substantial drag reduction of an F1 cars rear wing. A Multi-Body Simulation (MBS) indicates that the use of the optimised rear wing on an F1 car has an advantage compared to a benchmark car when acceleration in a straight line is considered. The simulation also indicates that the low speed turning performance of the car is maintained. The 2D predictions using a typical aeroelastic section show good agreement with the results from the 3D aeroelastic optimisation framework. The use of lamination parameters is valuable, since the continuous behaviour of these parameters is useful during optimisation. Once an optimal set of lamination parameters is available, the corresponding lay-up can be determined.

These updates are highly dependent on the actual F1 car and team, since each F1 car has slightly different characteristics with respect to the cars of competing teams. As such, a wing designed for the car of a particular team, might have a different lay-up and performance increase when compared to a wing which is designed for another teams car.
Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

References
Abdalla M, De Breuker R, Grdal Z (2007) Aeroelastic tailoring of variable-stiffness slender wings for minimum compliance. In: IFASD, Stockholm Abrate S (1994) Optimal design of laminated plates and shells. Compos Struct 29:269286 Beckert A (2000) Coupling uid (CFD) and structural (FE) models using nite interpolation elements. Aerosp Sci Technol 4:1322 Beckert A, Wendland H (2001) Multivariate interpolation for uidstructure-interaction problems using radial basis functions. Aerosp Sci Technol 5:125134 Benzing E (1992) Ali/Wings: studio per tecnici e piloti di auto da corsa/study for racing car engineers and drivers. Milano Bisplinghoff R, Ashley H, Halfman R (1955) Aeroelasticity. AddisonWesley, Reading de Boer A, van Zuijlen A, Bijl H (2007) Review of coupling methods for non-matching meshes. Comput Methods Appl Mech Eng 196:15151525 Bramesfeld G, Ironside D, Schwochow J (2008) Simplied modeling of wing-drag reduction due to structural dynamics and atmospheric gusts. In: 26th AIAA applied aerodynamics conference, Honolulu Cramer E, Gablonsky J (2004) Effective parallel optimization of complex computer simulations. In: 10th AIAA/ISSMO multidisciplinary analysis and optimization conference, Albany Diaconu C, Sato M, Sekine H (2002) Feasible region in general design space of lamination parameters for laminated composites. AIAA J 40(3):559565

The results show the capability to reduce the induced drag of an F1 car by tailoring the rear wing. When actual

646 Dowell E (2004) A modern course in aeroelasticity. Kluwer, Dordrecht Farhat C, Lesionne M (1998) Higher-order staggered and subiteration free algorithms for coupled dynamic aeroelasticity problems. In: 36th aerospace sciences meeting & exhibit, Reno FIA (2007) 2008 Formula 1 technical regulations. Federation Internationale de lAutomobile Fukunaga H, Sekine H (1992) Stiffness design method of symmetric laminates using lamination parameters. AIAA J 30(11):2791 2793 Fukunaga H, Sekine H (1994) A laminate design for elastic properties of symmetric laminates with extension-shear of bending-twist coupling. J Compos Mater 28(8):708731 Fukunaga H, Vanderplaats G (1991) Stiffness optimization of orthotropic laminated composites using lamination parameters. AIAA J 29(4):641646 Guo S, Cheng W, Cui D (2005) Optimization of composite wing structures for maximum utter speed. In: 46th AIAA/ASME/ ASCE/AHS/ASC structures, structural dynamics & materials conference, Austin Guo S, Cheng W, Cui D (2006) Aeroelastic tailoring of composite wing structures by laminate layup optimization. AIAA J 44(12): 31463149 Grdal Z, Olmedo R (1993) In-plane response of laminates with spatially varying ber orientations: variable stiffness concept. AIAA J 31(4):751758 Grdal Z, Haftka R, Hajela P (1999) Design and optimization of laminated composite materials. Wiley, New York INTEC GmbH (2004) Simpack users guide Kameyama M, Fukunaga H (2007) Optimum design of composite plate wings for aeroelastic characteristics using lamination parameters. Comput Struct 85:213224 van Kan J, Segal G, Vermolen F (2005) Numerical methods in scientic computing. VSSD Katz J (1989) Aerodynamics of high-lift, low-aspect-ratio unswept wings. AIAA J 27(8):11231124 Katz J (2006) Aerodynamics of race cars. Annu Rev Fluid Mech 38:2763 Liebeck R (1978) Design of subsonic airfoils for high lift. J Aircraft 15(9):547561 Liu B, Haftka R (2004) Single-level composite wing optimization based on exural lamination parameters. Struct Multidiscipl Optim 26:111120 Lynch R, Rogers W (1976) Aeroelastic tailoring of composite materials to improve performance. In: 17th AIAA/ASME/ASCE/ AHS/ASC structures, structural dynamics & materials conference, King of Prussia, pp 6168 MacNeal RH (1972) The Nastran theoretical manual. NASA (SP221(01)) Maskew B (1982) VSAERO A computer program for calculating the nonlinear aerodynamic characteristics of arbitrary congurations. NASA (CR-166476) Massegur D, Quaranta G, Cavagna L (2007) Race cars ex their muscle. ANSYS Advantage 1(1):911 Miki M, Sugiyama Y (1993) Optimum design of laminated composite plates using lamination parameters. AIAA J 31(5):921922

G.A.A. Thuwis et al. Patil M (1997) Aeroelastic tailoring of composite box beams. In: 35th AIAA aerospace sciences meeting & exhibit, Reno Persson P, Strang G (2004) A simple mesh generator in matlab. SIAM Rev 46(2):329345 Phoenix Integration (2003) Improving the engineering process with software integration Reheld L, Cheung R (2003) Some basic strategies for aeroelastic tailoring of wings with bend-twist coupling: part one. In: 44th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics & materials conference, Norfolk Reheld L, Cheung R (2004) Strategies for aeroelastic tailoring of wings with bend-twist coupling: part two. In: 45th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics & materials conference, Palm Springs Rendall T, Allen C (2007) Unied uid-structure interpolation and mesh motion using radial basis functions. Int J Numer Methods Eng Setoodeh S, Abdalla M, Grdal Z, Tatting B (2005) Design of variable-stiffness composite laminates for maximum in-plane stiffness using lamination parameters. In: 46th AIAA/ASME/ ASCE/AHS/ASC structures, structural dynamics & materials conference, Austin Setoodeh S, Abdalla M, Grdal Z (2006) Design of variable-stiffness laminates using lamination parameters. Composites: Part B 37:301309 Shirk M, Hertz T, Weisshaar T (1986) Aeroelastic tailoringtheory, practise, and promise. J Aircraft 23(1):618 Stanford B, Ifju P (2008) Aeroelastic tailoring of fixed membrane wings for micro air vehicles. In: 49th AIAA/ASME/ASCE/ AHS/ASC structures, structural dynamics, and materials conference, Schaumberg, IL Thuwis G, De Breuker R, Abdalla M (2008) Aeroelastic tailoring of a Formula One car rear wing. In: Proceedings of the sixth international conference on engineering computational technology, Athens Trk J (2000) Analytical mechanics. Wiley, New York Tsai S, Pagano N (1968) Invariant properties of composite materials. In: Composite materials workshop, pp 233253 Weisshaar T (1981) Aeroelastic tailoring of forward swept composite wings. J Aircraft 18(8):669676 Weisshaar T, Duke D (2006) Induced drag reduction using aeroelastic tailoring with adaptive control surfaces. J Aircraft 43(1): 157164 Weisshaar T, Ryan R (1986) Control of aeroelastic instabilities through stiffness cross-coupling. J Aircraft 23(2):148155 Weisshaar T, Nam C, Batista-Rodriguez A (1998) Aeroelastic tailoring for improved UAV performance. In: 47th AIAA/ASME/ASCE/ AHS/ASC structures, structural dynamics & materials conference, Long Beach Wendland H (1995) Piecewise polynomial, positive denite and compactly supported radial functions of minimal degree. Adv Comput Math 4:389396 Wright P (1974) Aerodynamics for Formula 1. Aeronaut J 78:226230 Zhang X, Toet W, Zerihan J (2006) Ground effect aerodynamics of race cars. Appl Mech Rev 59:3349