Aerospace Science and Technology 11 (2007) 396404 www.elsevier.

com/locate/aescte

Aeroelastic optimization of an aerobatic aircraft wing structure

Shijun Guo
Department of Aerospace Engineering, School of Engineering, Craneld University, Bedfordshire, MK43 0AL, UK Received 18 April 2006; received in revised form 4 January 2007; accepted 26 January 2007 Available online 8 February 2007

Abstract This paper aims at presenting an investigation into a minimum weight optimal design and aeroelastic tailoring of an aerobatic aircraft wing structure. Firstly, analytical and FE models were created for the original metallic wing and an improved composite wing box structures. In order to validate the numerical models, predicted vibration parameters of the metallic wing box were compared with the experimental results. Comparison was also made for the predicted stress results between the metallic and the composite wing box structures of different dimensions and laminate layups. Secondly, based on a minimum weight composite wing box model of adequate strength the investigation was focused on the aeroelastic tailoring of the wing box by employing the gradient-based deterministic optimization method. The study demonstrates that in addition to a signicant weight saving, up to 30% increase of utter speed for the composite wing box can be achieved by optimizing the bre orientations of the wing skin and spar web laminates. The optimized laminates are trimmed and reinforced to meet the manufacture and strength requirements with little compromise of utter speed and a minor weight penalty. 2007 Elsevier Masson SAS. All rights reserved.
Keywords: Aeroelastic tailoring; Optimization; Composite wing box; Laminate layup

1. Introduction In addition to the favorable high specic strength and stiffness, ber reinforced composite materials also offer great potential for designers to achieve desirable directional stiffness and aeroelastic behavior of a wing structure by optimizing the bre orientation with minimum weight penalty. Some previous work in this eld has demonstrated that the divergence speed of a forward swept composite wing can be increased by optimizing the laminate layup and restraining the wash-in stiffness coupling warping effect [1820,25]. The elastic or stiffness coupling produced by unsymmetrical laminate layups could also have signicant effect on the dynamic aeroelastic behavior of a composite wing [15,23]. Therefore investigation has been made into optimizing the laminate stacking sequence of composite wing structures for desirable aeroelastic behaviors [10, 13,21,22]. Previous research has also shown that an optimiza* Tel.: +44 (0)1234 754628; fax: +44 (0)1234 758203.

E-mail address: s.guo@craneld.ac.uk. 1270-9638/$ see front matter 2007 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.ast.2007.01.003

tion solution obtained by a gradient-based deterministic method (GBDM) depends on the initial design variables set at the beginning of the optimization process [11]. However, the optimum solution from this method is reliable based on a continuous and nite gradient of objective function at each step of the process. This method is computationally more efcient comparing with a genetic algorithm based on a stochastic procedure [12]. In the unconstrained optimization problem presented here, the boundary specied for the design variables virtually draws a line to limit the numerical variation of the ber orientations. The location of the design variables or optimum design relative to the boundary has no signicant inuence on the behavior and gradient of the objective function. A small deviation of the layup from the optimum solution will not lead to a signicant change of the objective function. This allows the optimized layup be trimmed within the manufacture constraint for a practical design option that is close enough to the optimum layup. However this may not be true for a general case where the sensitivity of an optimum design may depend on its location relative to the boundary.

S. Guo / Aerospace Science and Technology 11 (2007) 396404

397

Nomenclature Aij , Ae coefcient of the laminate axial stiffness matrix [A] and the enclosed area of a thin-walled anisotropic closed-section beam A(s), B(s), C(s) reduced axial, coupling and shear stiffness of the closed-section beam stiffness coefcients of the closed-section beam Cij Mx , My bending moment and torque applied to the beam EI, GJ, CK bending, torsion and bending-torsion coupling rigidities of the beam respectively h, transverse displacement and rotation of a thinwalled wing box beam m, Ip mass and polar mass moment of inertia per unit length of a wing box beam distance between the mass and elastic axes of the wing box cross-section b, , V wing semi-chord, air density and velocity , k = b/V oscillating frequency of the wing (rad/s) and reduced frequency utter speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m/s Vf , fv ( ) lamina ber orientation (deg); objective function [KD ()] frequency dependent generalized dynamic stiffness matrix {q}, [D] generalized coordinate and damping matrix of the structure [QA]R , [QA]I real and imaginary parts of the generalized unsteady aerodynamic matrix X

The objective of this paper is to present an investigation to achieve an optimal design for an aerobatic aircraft wing structure to meet the lightweight, strength and aeroelastic design requirements. The investigation was conducted in two stages. In the rst stage, effort was made to design and model a composite wing for a minimum weight structural option. An analytical method was used for vibration and aeroelastic analysis and the nite element method (FEM) was employed for structural stress analysis. Vibration test data of the original metallic wing was used to validate the analytical model by comparing the measured and predicted structural modes. In the second stage, the study was focused on aeroelastic tailoring of the basic composite wing model to achieve the maximum utter speed under the strength limit. To minimize the computing demand in aeroelastic optimization process, an analytical method was employed, which has been developed and used by a number of researchers to evaluate the structural rigidities of a composite thin-walled box beam [1,4,14,16,24]. This method was initially based on the two-dimensional shell theory for asymptotic analysis and further developed by Berdichevsky et al. to derive the governing equations of anisotropic thin-walled beams based on the variational asymptotical theory [4]. Later Armanios and Badir extended this theory to the free vibration analysis of anisotropic thin-walled beams [1]. In the present paper, the method has been adapted to determine the bending, torsion and bendingtorsion coupling rigidities of the thin-walled composite wing box structure. The stiffness and mass matrices in the governing equations were established by using the dynamic stiffness method [2,3]. The WittrickWilliams algorithm [26] was then used to calculate the wing structural modes for accuracy and computational efciency. Examples presented in this paper demonstrate four cases of aeroelastic tailoring in a condition of 88 kg fuel stored in the wing box. Additional case study was conducted to include an additional 40 kg external payload mounted below the middle of the wing. In this example, the lifting surface theory was used for the incompressible airow unsteady aerodynamic force calculation [6]. In the aeroelastic optimization stage, utter speed was chosen as the primary parameter in the objective

function and the ber orientation of the laminate was taken as the design variable. This has resulted in an optimized solution without weight penalty. Comparing with the metallic wing, it is noted that a composite wing box made of the widely used quasi-isotropic laminate layup is not necessarily better in terms of the strength, stiffness and utter speed apart from being a lighter structure. Instead of the usual constrained optimization [5,22], the unconstrained aeroelastic tailoring was conducted separately from the stress analysis in this current investigation. This approach has provided the exibility for a designer to make a suitable and necessary structural reinforcement in a post process to meet the strength requirement. Based on an example presented in this paper, it has been demonstrated that a composite wing box can be optimized to achieve the maximum utter speed and meet the strength requirement by aeroelastic tailoring and reinforcement. 2. Analytical and numerical methods 2.1. Analytical and nite element methods In this investigation, the wing box between the front and rear spar of a wing as illustrated in Fig. 1 was assumed to be the primary structure of the wing and the principal load carrier. The components and surface after the rear spar were only counted in calculating the mass, inertia and aerodynamic force of the wing. For stiffness and vibration analysis, the tapered wing box structure clamped at the root section was divided into a number of spanwise segments. Each of the wing box segments was modelled as a uniform thin-walled single-cell box beam and the whole wing box was modelled as an assembly of those beams. For a circumferentially asymmetric stiffness conguration and based on the analytical method by Armanios and Badir [1], the relationships between the bending moment Mx , torque My and transverse and twist deections at the end of an anisotropic thinwalled closed-section beam, as shown in Fig. 2, are expressed below. My = C22 + C23 h and Mx = C23 + C33 h (1)

398

S. Guo / Aerospace Science and Technology 11 (2007) 396404

Fig. 1. Planform and dimension of the aircraft wing (unit: m).

The stiffness coefcients Cij can be calculated based on the geometry of each segment by integration along its crosssectional circumference as follows C22 = C33 = A2 e (1/C(s)) ds A(s) C23 = Ae (B(s)/C(s))z ds (1/C(s)) ds (2)

Fig. 2. A spanwise wing box segment modelled as a beam element.

mode method, the utter equation for an oscillating wing can be written in generalized coordinates as: 1 KD () V 2 [QA]R + i [D] 2 1 + i V 2 [QA]I {q} = 0 2

[ (B(s)/C(s))z ds]2 B(s)2 2 z ds + C(s) (1/C(s)) ds

(6)

where Ae is the enclosed area of the cross section and parameters A(s), B(s) and C(s) are given below. A(s) = A11 (A12 )2 A22 (A26 )2 A22 B(s) = 2 A16 A12 A26 A22 (3)

C(s) = 4 A66

In the above equation, Aij are the coefcients of stiffness matrix [A] of the thin-walled anisotropic closed-section beam. For the laminated composite wing box structure as illustrated in Fig. 2, the stiffness matrix [A] is for the wing box skin and spar web laminate. It depends on the laminate properties and layup along the circumference of the wing box cross-section. According to the force-deection relationships shown in Eq. (1) and stiffness denition, the stiffness coefcients C33 , C22 and C23 actually represent the bending, torsion and bending-torsion coupling rigidities of the wing box beam, which are often indicated by symbols EI, GJ and CK respectively. The dynamic stiffness matrix (DSM) method [2,3] was subsequently used for vibration analysis. In this method, the equations of motion for each of the thin-walled box beams were represented as follows when the transverse shear deformation and warping effect were neglected. EI h + CK + m h m X = 0 GJ + CK h + m X h Ip = 0 (4) (5)

The unsteady aerodynamic forces were calculated by using the lifting surface theory [6] in incompressible airow. The above equation was solved in an iterative procedure by searching and matching the utter speed and frequency within subsonic range using the V-g method. For stress analysis, the MSC PATRAN/NASTRAN package based on the nite element method was employed. Additional structural components such as ribs, stringers and leading edge cell box were included in the wing structure model. The analysis produced stress distribution over the wing skin with the stringers contribution. For the composite wing, detailed stress and failure index for each layer of the wing skin laminate were analyzed. The FE stress results were also used to evaluate the load in the wing root section. Based on the load, the stress and failure index at the critical region of the skin laminate were calculated based on the thin-walled laminate theory [17] as an alternative and approximate approach in the optimization process. 2.2. Optimization method for aeroelastic tailoring In the optimization, effort is primarily focused on achieving a maximum utter speed by tailoring the ber orientations of the skin and spar web laminates. Since the wing weight will not be affected, the analysis can be expressed as an unconstrained optimization problem as follows: Minimize fv (x) = 1 Vf (x) Vf (0 ) 2 Vf (0 ) where x {A | (1 , 2 , . . . , n )}, n [90, 90]

where h = 4 h/y 4 , h = 2 h/t 2 , = 3 /y 3 and = 2 /t 2 . By solving the above differential equations, an exact solution for the displacement functions h(y) and (y) can be obtained. A dynamic stiffness matrix can be subsequently determined by relating the displacements to the bending moment and torque at both ends of each beam element. A dynamic stiffness matrix for the whole wing box structure is obtained by assembling the beam stiffness matrices along the wingspan. It is noted that the dynamic stiffness matrix is actually a combination of stiffness and mass matrices of the beams and is frequency dependent. This particular type of matrix produces a non-standard eigenvalue problem, and is solved by using the WittrickWilliam algorithm [26]. By employing the normal

(7)

where fv (x) is the objective function, Vf (x) the wing utter speed, x a vector containing the ber orientations {1 , 2 , . . . , n } set as design variables with a lower and upper bounds of 90 , 0 represents a set of specied ber orientations in the initial laminate layup of the wing box. To solve the optimization problem, the DavidonFletcher Powell (DFP) variable metric method is employed as the optimizer whereas the Golden Section method based on polynomial interpolation is employed for the one-dimensional search [8,9].

S. Guo / Aerospace Science and Technology 11 (2007) 396404

399

Fig. 3. Dimensions of the composite wing box root section. Table 1 Frequency comparison of the original wing obtained by different methods Symmetrical modes 1st bending 2nd bending 1st torsion 2nd torsion Natural frequency (Hz) Experiment 11.8 37.6 71.0 128.9 FEM 11.8 44.4 75.0 110.2 Analytical 12.7 49.0 74.9 117.9

very well in terms of the dynamic behavior. By using the analytical method described in Section 2.1, predicted utter speed without fuel is Vf = 506 m/s, which is very close to the original wing design calculation. These results allow the methods and structural model be validated before employing them in the subsequent design and analysis. To reduce the structural weight, rstly the original wing box was scaled down by reducing the wing box depth in terms of chord ratio from 15% to 12% at the root section and to 9% at the tip. Secondly a laminated composite wing structure made of carbon/epoxy skins, spars, ribs and stringers was selected for design improvement. Detailed stress analysis and aeroelastic tailoring are described in the following sections. 3.2. Loading and stress results To verify the structural design in terms of strength and stiffness, detailed FE analysis of structural stress and deection was conducted by using the NASTRAN code. Since the wing box carries the principal load on the wing, the metallic and composite wing structures were modelled initially as a single-cell wing box. For comparison, double-cell wing box models including the leading edge cell were also created and analyzed. In the FE model, a quadrilateral 2D shell element and beam element of size 2035 mm were chosen for the spar webs and skinsstringer panels. The material properties were listed in Table 2 for aluminum alloy 2024 used for the metallic wing and XAS Ciba-Geigy 913 carbon/epoxy unidirectional pre-preg used for the composite wing. For the composite wing skin reinforced by ve top-hat stringers, quasi-isotropic and orthotropic laminate layups were selected to compare the design results. Based on the ultimate wing load of 30.5 kN in transverse direction at 7g, an equivalent distributed aerodynamic pressure was applied over the wing surface. Six different wing box congurations listed below have been modelled. Model 1a single-cell metallic wing box made of 1 mm thick z-section stringers, skins and spar webs; Model 1b single-cell composite wing box made of 8-layer (0.125 mm ply thickness) carbon/epoxy quasiisotropic laminates for spars webs, ribs, stringers and skins; Model 1c single-cell composite wing box with 8-layer (0.125 mm ply thickness) carbon/epoxy quasiisotropic laminates for spar webs, ribs and stringers, and [(45)2 ]s layup for skins; Model 2a double-cell metallic wing box with 1 mm thick zsection stringers, skin and spar webs; Model 2b double-cell composite wing box with 8-layer (0.125 mm ply thickness) carbon/epoxy quasi-

3. Examples and results 3.1. Wing structural model and vibration results An aerobatic aircraft was taken as a study example. The aircraft has a maximum take-off mass 1000 kg, dive speed 122 m/s and cruise speed 103 m/s. The design ultimate load factors are +7g and 5g. Fig. 1 shows the planform and dimension of the wing, which has aspect ratio 6.7 and taper ratio 0.44. The original airfoil at the wing root is NACA 23015 and NACA 23012 at the tip. Fig. 3 illustrates the shape and geometry of the wing box section at the root. The original wing made of aluminum alloys had a predicted utter speed Vf = 500.7 m/s in empty fuel condition, which indicates a conservative design of the structure in terms of stiffness and weight. Before conducting a design improvement, a numerical model was created and validated. Based on the original wing data, a single-cell wing box structure made of aluminum alloy front and rear spars, skin-stringer panels (5 z-section stringers for each skin) and 13 ribs along the span was modelled by using the PATRAN/NASTRAN package. Vibration analysis was conducted by using FE and the analytical method presented in Section 2. The predicted rst two symmetric bending and torsion frequencies listed in Table 1 were compared with the available data from a vibration test conducted in the design of the original aircraft [7]. The results indicate that both the FE and analytical models represent the wing structure
Table 2 Mechanical properties of the metallic and composite materials used in this study Carbon/ Epoxy Al 2024 E1 (GPa) 130 70 E2 (GPa) 9 v12 0.28 0.33 G12 (GPa) 4.8 26.9 Xt (MPa) 1370

Yt (MPa) 42 427

Xc (MPa) 1000 420

Yc (MPa) 200

S (MPa) 60 250

(kg/m3 ) 1610 2820

400

S. Guo / Aerospace Science and Technology 11 (2007) 396404

Table 3 Maximum stress and deection of different wing box models Wing box models Model sk (MPa) st (MPa) dz (mm) Single-cell metallic 1a 181/175 202 129 Single-cell composite 1b 174/167 166 184 1c 167/143 272 279 Double-cell metallic 2a 142/144 167 124 Double-cell composite 2b 136/139 139 163 2c 142/155 205 275

isotropic laminates for spar webs, ribs, stringers and skins; Model 2c double-cell composite wing box with 8-layer (0.125 mm ply thickness) carbon/epoxy quasiisotropic laminates for spar webs, ribs and stringers, and [(45)2 ]s layup for skins. For each of the wing box models, the maximum transverse deection and stress results are presented in Table 3. In the table, sk and st represent the maximum von Mises stress in the skins and stringers of the metallic wing respectively. For the composite wing models, the stress results are the spanwise maximum average values through the laminate thickness. In addition, the stress in each layer of the laminate in the principal material directions and the failure index were also calculated. Taking the model 1b for example, the maximum Hoffman failure index (F.I.) in the skin laminate is 0.7. The result indicates that the composite wing box has sufcient strength despite the maximum stress of 335 MPa occurring in the 0 ber ply is much higher than the metallic wing. Regarding the structural weight saving, the composite wing box is 40% lighter than the original metallic wing box. Comparing with the single-cell wing box, the maximum stress in the double-cell wing box is reduced by 16% to 25%. For the metallic wing models, the stringers carry 918% higher level of stress than the skin. For the composite wing box models 1b & 2b made of the same quasi-isotropic laminates for skin and stringers, the stress level in the skins and stringers is the same. However comparing with the skins of 45 layup in the models 1c and 2c, the stringers carry 4463% higher normal stress and play a major role in the load transfer. This example shows a potential for the skin laminate layup to be optimized to obtain a desirable stiffness and aeroelastic behavior for the wing box without compromising the strength criterion. It is also noted that the composite wing in such layup has a smaller bending stiffness but greater torsional stiffness than the composite wing model 1b and its metallic counterpart. For both metallic and composite wing models, the double-cell wing box offers slightly higher bending stiffness than that of the single-cell box. However it produces signicantly greater torsional stiffness and normally a higher utter speed. The spanwise average stress through the skin thickness of the double-cell composite wing model 2b under the ultimate load of 7g is obtained by FEM. The result was used to calculate the maximum normal force intensity 139 kN/m and shear force intensity 67.4 kN/m acting on the skin laminates at the wing root section. Under this loading, the failure index (F.I.) in each ply of the skin laminate was calculated based on the laminate

theory. For the composite model 1b, the resulting maximum F.I. value 0.61 (by the Hoffman criterion) and maximum strain 2766 in the skin laminate indicate adequate strength of the wing box. For the model 1c however, the maximum F.I. value 1.71 and maximum strain 8199 indicate inadequate strength. The strength of optimized wing box by aeroelastic tailoring is analyzed in the next stage. 3.3. Aeroelastic tailoring of the wing box In this investigation, 88 kg fuel was stored in the wing box taking 25% of its internal volume along the span. In this case, the utter speed of the metallic wing is reduced to 288 m/s. Despite the dramatic drop of utter speed compared with the empty fuel wing, there is an adequate safe margin in terms of strength and utter speed for design improvement and further weight saving. In the aeroelastic tailoring process, the composite wing box model 1b, which has sufcient strength, was taken as an example. In the structure modelling based on the analytical method described in Section 2.1, the wing box was divided into six equal segments along the wingspan as shown in Fig. 1. In each of the segments, the wing box was further divided into four laminate panels along its cross section circumference representing the two skin covers and two spar webs as illustrated in Fig. 2. In the optimization process, each of the ply ber orientations of the total 4 6 pieces of laminate panels was taken as an independent design variable, which results in the maximum number of design variables of 8 24 = 192. In the case of a symmetric layup for all the panels, the total number of independent design variables was reduced to 96. Depending on the initial laminate layup, selection of layers and segments for tailoring, the number of design variables and optimization results are different for each case. In this example, four optimization cases and results are presented below. Case 1. In this case, each layer in all the six spanwise segments was set as an independent variable and different number of layers was selected each time in the optimization. Since the layup in each of the spanwise segments was tailored independently, the optimum layup in the segments along the span was different from each other. The number of design variables will vary from the minimum zero to the maximum 192 when eight layers were selected. As shown in Fig. 4 (case 1, 0 [0/45/90]s), the utter speed corresponding to the initial quasi-isotropic symmetric layup is Vf = 253.4 m/s. When two layers of zero ber orientation (2 [0-deg]) were selected and tailored independently, the utter speed was increased up to 281 m/s. When more number of layers was selected for tailoring, the utter results were obtained and shown in Fig. 4. As expected, when all the eight plies in each of the six segments

S. Guo / Aerospace Science and Technology 11 (2007) 396404

401

(a) Fig. 4. Flutter results corresponding to the selected number of optimization layers and laminate layups in the four study cases.

were tailored independently (8 [0/45/90]), the maximum increase of utter speed up to 309.9 m/s was achieved. It shows an increase of 22% comparing with the utter speed in the initial layup. It is noted that the optimized laminate layup in each of the segments may be non-symmetric and different. Table 4 shows the optimum laminate layup of the wing box at root section only and the utter results. A comparison between the initial rigidities (EI-0, GJ-0 and CK-0) and the optimized (EI1, GJ-1 and CK-1) values of the wing box segments along the span is shown in Fig. 5(a). Case 2. In this case, each layer of the laminates was taken as an independent variable, but the layups of the six spanwise segments were kept uniform during the optimization. Although the spanwise layup tailoring was limited and the maximum number of design variables was reduced to 4 8, the utter speed was steadily increased as more layers were being tailored as shown in Fig. 4. When all eight layers in each laminate were tailored (8 [0/45/90]), the maximum utter speed was increased by 36% (comparing with the initial layup) achieving 345.9 m/s. Further details of results are presented in Table 4. As shown in Fig. 5(a), the optimized wing box in this case has greater GJ values plus contribution from CK along the spanwise segments comparing with the results in case 1.
Table 4 Optimized laminate layups, stiffness and utter results in design cases 14 Cases Case 1 Laminate parts upper skin lower skin front spar web rear spar web upper skin lower skin front spar web rear spar web upper skin lower skin front spar web rear spar web upper skin lower skin front spar web rear spar web

(b) Fig. 5. (a) Rigidities from optimum non-symmetric laminate layups in cases 1 and 2. (b) Rigidities from optimum symmetric laminate layup in cases 3 and 4.

Case 3. Similar to case 1, each layer of the laminates in the spanwise segments was selected as an independent design variable. The difference in this case is that the laminate layup of skins and spar webs is restricted to symmetric. Under this condition, optimized utter speed results from selecting and tailoring different number of layers was obtained as shown in Fig. 4. The results indicate that the most effective option is to tailor just the rst pair of layers (2 [0 deg]) since the second and third layers with 45 bers are nearly optimal in the symmetric layup condition. The tailored layup details for the wing box root sec-

Optimum laminate layupa in ber orientations (degree) [57.2/48.3/45.8/87.6/87.3/47.1/55.2/41.5] [89.9/44.8/45.9/90.0/90.0/45.9/44.8/71.2] [84.9/44.5/47.6/89.8/89.7/47.3/44.1/76.4] [30.9/44.6/56.1/90.0/90.0/45.3/44.6/28.0] [49.2/44.2/49.5/45.6/48.2/49.7/43.2/49.0] [51.3/43.2/52.4/43.7/43.3/50.5/43.0/51.3] [87.5/44.1/48.7/89.2/89.0/49.9/43.3/47.8] [45.4/43.0/47.6/89.6/89.6/47.1/46.1/48.4] [5.7/45/45/90/90/45/45/5.7] [20.2/45/45/90/90/45/45/20.2] [4.1/45/45/90/90/45/45/4.1] [3.6/45/45/90/90/45/45/3.6] [49.0/39.4/62.7/73.3]s [41.5/51.0/52.8/54.4]s [26.9/44.1/51.1/89.3]s [62.2/44.0/45.9/89.9]s

Rigiditya and utter results EI = 335.4 (kN.m2 ), GJ = 1050.0 (kN.m2 ), CK = 6.2 (kN.m2 ) Vf = 309.9 (m/s) f = 122.6 (rad/s) EI = 261.0 (kN.m2 ), GJ = 1450.0 (kN.m2 ), CK = 46.5 (kN.m2 ) Vf = 345.9 (m/s) f = 136.4 (rad/s) EI = 729.3 (kN.m2 ), GJ = 971.4 (kN.m2 ), CK = 127.7 (kN.m2 ) Vf = 302.2 (m/s) f = 122.0 (rad/s) EI = 301.6 (kN.m2 ), GJ = 1290.0 (kN.m2 ), CK = 67.8 (kN.m2 ) Vf = 348.0 (m/s) f = 138.1 (rad/s)

Case 2

Case 3

Case 4

a Optimum layups in cases 1 and 3 and the optimum rigidities are for the root section.

402

S. Guo / Aerospace Science and Technology 11 (2007) 396404

Table 5 Trimmed optimum laminate layups for manufacture consideration Cases Case 2 Laminate parts upper skin lower skin front spar web rear spar web upper skin lower skin front spar web rear spar web Round-off optimum layup (degree) [50/45/50/45/50/50/45/50] [50/45/50/45/45/50/45/50] [90/45/50/90/90/50/45/90] [45/45/50/90/90/50/45/45] [50/40/65/75]s [40/50/55/55]s [30/45/50/90]s [60/45/45/90]s Flutter results Vf = 341.9 (m/s) f = 134.4 (rad/s) Vf = 340.5 (m/s) f = 136.1 (rad/s)

Case 4

Table 6 Strength and utter results for reinforced wing box Cases Case 1 Case 2 Case 3 Case 4 F.I. of the optimum skin laminates upper skin 2.5 1.9 0.7 1.5 lower skin 1.6 1.7 0.6 2.5 F.I. of the reinforced optimum skin laminates upper skin 0.7 0.4 0.6 0.5 lower skin 0.6 0.4 0.5 0.6 Flutter speed of the reinforced wing box Vf (m/s) 310.9 346.0 310.7 351.8

tion are presented in Table 4 and the tailored EI, GJ and CK values of the spanwise segments are shown in Fig. 5(b). Comparing with the initial layup values, the maximum utter speed was increased by 19% reaching 302.2 m/s. Case 4. In addition to the symmetric laminate layup as specied in case 3, the layups in the spanwise segments were kept uniform. In this case, the number of design variables was reduced to the minimum (4 4). Nevertheless greater utter speed increase than that obtained in case 3 has achieved when different number of layers is tailored as shown in Fig. 4. The optimum layup and utter results from tailoring all layers are given in Table 4 and the optimum EI, GJ and CK values of spanwise segments are shown in Fig. 5(b). Comparing with the initial layup result, the maximum utter speed was increased by 37% reaching 348.0 m/s. As indicated in Figs. 4 and 5, a few points are noted by comparing the optimization results in the four cases. Firstly the rigidities from the optimized layups in the wing box segments along the span have the same trend as the initial layup results. Secondly the optimized layups result in a greater GJ and smaller EI than the initial layups in the spanwise segments. Thirdly the increase of GJ and decrease of EI in the spanwise segments are in favor of the utter speed. For example, the rate for GJ increase and EI decrease in the segment 6 and the resulting utter speed in the cases 2 and 4 are clearly greater than that in cases 1 and 3. Finally a bending-torsion coupling rigidity from an optimized non-symmetric layup plays a benecial and important role in favor of the utter speed. 3.4. Sensitivity and strength analysis for the optimized wing box Comparing the results in the above cases, it is noted that the layup tailoring of the skin and spar web laminates along the wing box section circumference has more inuence on the utter speed than that along spanwise segments. The results shown

in Table 4, Figs. 5(a) and 5(b) indicate that an optimum layup in favor of aeroelastic stability is normally associated with a great torsional stiffness GJ combined with benecial bending-torsion stiffness CK. Such a CK for a wing box can be obtained from an optimum symmetric layup or a non-symmetric layups for the skins and spar webs. In the above optimization process, one decimal place for the design variables was kept for adequate numerical accuracy in the gradient calculation and sensitivity in the direction search towards the minima of the objective function. However it is impractical to take the optimized skin and spar web laminate layups as shown in Table 4 for the manufacture straightaway. Fortunately the sensitivity of utter speed to a small variation of the optimum layups obtained by the GBDM is very low. Consequently the optimum layups can be rounded off without having signicant inuence on the optimized utter result. For example, when the optimum layups in case 2 and case 4 as listed in Table 4 were rounded off to get a set of practical layups for manufacture, the resulting utter speed was reduced by only 1% and 2% respectively as shown in Table 5. Although the initial wing box model 1b has been proven to have adequate strength, it is not necessarily maintaining the strength after being optimized by the unconstrained aeroelastic tailoring method. Under the ultimate loading described in Section 3.2, the stress and F.I. in each layer of the optimized skin laminates in the four cases have been calculated and presented in Table 6. The results show that the optimized wing box design case 3 is the most favorable one because it maintains the same level of strength as the initial design and has 19% higher utter speed without reinforcement. For the other optimized wing box design cases 1, 2 and 4, the strength condition has been violated after aeroelastic tailoring. To meet the strength requirement, two 0 plies were added onto the optimized skin laminates of the wing box root section in each of the three design cases. As shown in Table 6, the reinforced wing box with a penalty of 0.5 kg extra structural weight will have

S. Guo / Aerospace Science and Technology 11 (2007) 396404

403

Table 7 Effect of additional payload and reinforcement on utter speed and strength Laminate layup Metallic wing [(45/45)2 ]s Case 3 optimum layup Case 4 optimum layup Reinforced [(45/45)2 ]s Reinforced case 3 layup Reinforced case 4 layup Vf (m/s) (original mass) 288.1 358.8 302.2 348.0 116.2 137.8 122.0 138.1 f (rad/s) Vf (m/s) (extra payload) 267.5 310.0 261.0 284.0 309.0 281.0 285.0 101.2 124.1 98.9 107.0 116.7 109.5 107.5 f (rad/s) R.F. & F.I. 0.4 1.7 0.7 1.5 0.4 0.6 0.6

higher strength than the initial design while maintaining its optimized high utter speed. By analyzing the laminate strength separately from the aeroelastic tailoring, a designer has more options and exibility to rene the optimum design to suit the requirements. 3.5. Additional payload and wing box reinforcement Finally a case study was focused on the aeroelastic tailoring of the wing box with an extra payload of 40 kg mass mounted 0.2 m below the elastic axis in the middle of the wing semispan. As expected, the resulting utter speed is reduced by 511% comparing with the original fuel case for the wing box models shown in Table 7. Without reinforcement, although 40% weight saving can be maintained only the composite wing box optimized in case 3 is competitive with the metallic wing in terms of both strength and utter speed. Therefore structure reinforcement especially near the wing root section is necessary to enhance the wing box stiffness and strength. For the optimum design case 3, a lower F.I. (Hoffman criterion) and higher utter speed were obtained when a pair of 45 plies was added onto the skin laminates at the wing root section. For the [(45)2 ]s laminate layup, adding a pair of 45 plies onto the skin laminates could make the maximum F.I. reduced from 1.7 to 1.2 but not offering an adequate strength. Hence a pair of 0 plies was added onto the skin laminates resulting in F.I. = 0.4 without compromising the utter speed as shown in Table 7. Similar results have also been obtained in the optimized design case 4. 4. Conclusions The investigation has demonstrated the advantages of employing laminated composite over its metallic counterpart in aeroelastic tailoring of a highly loaded wing box structure of an aerobatic aircraft. The study has shown that the analytical methods for calculating the stiffness of an anisotropic closed-section beam together with the DSM method can be used for modal analysis of a thin-walled wing box structure. This approach shows its advantage of high efciency in aeroelastic tailoring of a composite wing box structure. Although the GBDM has some disadvantages, it is favourable for the wing aeroelastic tailoring. This is mainly because that it can produce a reliable solution in a sense that the objective function has low sensitivity to a deviation of the layups from optimum solution. This deviation tolerance allows the optimum solution to be trimmed to

obtain a feasible and practical design option to meet the manufacture requirement. Based on the theory and methods, design improvement and aeroelastic tailoring of the wing box structure has been demonstrated in four design cases. Comparing with the original metallic wing box, 40% weight saving can be achieved by using a thinner composite wing box of adequate strength. Without constraining the aeroelastic tailoring to a strength condition, signicant increase of utter speed up to 37% has been achieved by using the GBDM. Stress analysis under the ultimate loading was performed in a subsequent process based on the laminate theory. This approach offers the designers the exibility to select and reinforce the optimized wing box laminates to meet the strength requirement without compromising the optimized aeroelastic stability. In this example, the wing box can be reinforced to gain adequate strength by adding a couple of plies onto the optimized skin laminate of the wing box root segment at a minor weight penalty. Acknowledgements The author is grateful to the EPSRC for supporting this work (through grant EP/C511190/1). References
[1] E.A. Armanios, A.M. Badir, Free vibration analysis of anisotropic thinwalled closed-cross-section beams, AIAA Journal 33 (1995) 19051910. [2] J.R. Banerjee, F.W. Williams, Coupled bending-torsional dynamic stiffness matrix for Timoshenko beam elements, Computers & Structures 42 (1992) 301310. [3] J.R. Banerjee, F.W. Williams, Free vibration of composite beams an exact method using symbolic computation, Journal of Aircraft 32 (1995) 636642. [4] V. Berdichevsky, E.A. Armanios, A.M. Badir, Theory of anisotropic thinwalled closed cross-section beams, Journal of Composite Engineering 2 (1992) 411432. [5] T.M. Clements, M. Rais-Rohani, Design optimization of a composite tiltrotor wing with stability approximations, in: Proc. of 41st AIAA/SME/ ASEC/AHS/ASC Structures, Structural Dynamics and Materials Conference, Atlanta, GA, April 2000, pp. 829838. Paper No. AIAA-2000-1454. [6] D.E. Davies, Theoretical determination of subsonic oscillatory airforce coefcients, Royal Aircraft Establishment Technical Report (RAE-TR76059), 1978. [7] M.A. El-Shinnawey, Ground resonance and normal modes of aerobatic aircraft Craneld A1, MSc Experimental Thesis, College of Aeronautics, Craneld University, December 1978. [8] R. Fletcher, M.J.D. Powell, A rapidly convergent method for minimization, Computer Journal 6 (1963) 163168.

404

S. Guo / Aerospace Science and Technology 11 (2007) 396404

[9] R.L. Fox, Optimization Methods for Engineering Design, AddisonWesley, 1971. [10] G.A. Georghiades, S. Guo, J.R. Banerjee, Flutter characteristics of laminated wings, Journal of Aircraft 33 (1996) 12041206. [11] S. Guo, J.R. Banerjee, C.W. Cheung, The effect of laminate lay-up on the utter speed of composite wings, Proc. of IMechE, Part G, J. of Aerospace Eng. 217 (2003) 115122. [12] S. Guo, W. Chen, D. Cui, Optimization of composite wing structures for maximum utter speed, in: Proc. of the 1st AIAA Multidisciplinary Design Optimization Specialist Conference, the 46th AIAA/ASME/ASCE/ AHS/ASC Structures, Structural Dynamics, and Materials Conference, Austin, TX, 1821 April 2005. Paper No. AIAA-2005-2132. [13] S. Guo, C.W. Cheung, J.R. Banerjee, R. Butler, Gust alleviation and utter suppression of an optimized composite wing, in: Proc. of Int. Forum on Aeroelasticity and Structural Dynamics, Manchester, UK, June 1995, pp. 41.141.9. [14] D.H. Hodges, A.R. Atilgan, M.V. Fulton, L.W. Reheld, Free vibration analysis of composite beams, Journal of the American Helicopter Society (July 1991) 3647. [15] S.J. Hollowell, J. Dugundji, Aeroelastic utter and divergence of stiffness coupled graphite/epoxy cantilevered plates, Journal of Aircraft 21 (1984) 6976. [16] C.H. Hong, I. Chopra, Aeroelastic stability analysis of a composite rotor blade, Journal of the American Helicopter Society (April 1995) 5767. [17] L.P. Kollar, G.S. Springer, Mechanics of Composite Structures, Cambridge University Press, ISBN 0 521 80165 6, 2003 (Chapter 6).

[18] L. Librescu, A.A. Khdeir, Aeroelastic divergence of swept-forward composite wings including warping restraint effect, AIAA Journal 26 (1988) 13731377. [19] L. Librescu, J. Simovich, General formulation for the aeroelastic divergence of composite swept-forward wing structures, Journal of Aircraft 25 (1988) 364371. [20] L. Librescu, S. Thangjitham, Analytical study on static aeroelastic behavior of swept-forward composite wing structures, Journal of Aircraft 28 (1991) 151157. [21] M. Lillico, R. Butler, J.R. Banerjee, S. Guo, Optimum design of high aspect ratio wings subject to aeroelastic constraints, in: Proc. of 36th AIAA/SME/ASEC/AHS/ASC Structures, Structural Dynamics and Materials Conference, April 1995, pp. 558566. [22] M. Lillico, R. Butler, S. Guo, J.R. Banerjee, Aeroelastic optimization of composite wings using the dynamic stiffness method, The Aeronautical Journal 101 (1997) 7786. [23] I. Lottati, Flutter and divergence aeroelastic characteristics for composite forward swept cantilever wing, Journal of Aircraft 22 (1985) 10011007. [24] L.W. Reheld, A.R. Atilgan, D.H. Hodges, Nonclassical behavior of thinwalled composite beams with closed cross section, Journal of the American Helicopter Society (May 1990) 4250. [25] T.A. Weisshaar, Divergence of swept-forward composite wings, Journal of Aircraft 17 (1980) 442448. [26] W.H. Wittrick, F.W. Williams, A general algorithm for computing natural frequencies of elastic structures, Quarterly Journal of Mechanics & Applied Mathematics 24 (1971) 263284.