You are on page 1of 81

Multidisciplinary Optimization Standardization Approach

for Integration and Configurability


MOSAIC Project

Task 6

WING–BOX STRUCTURAL DESIGN


OPTIMIZATION

Report 6 (Final Report)

Conceptual Design Optimization of an Aircraft


Wing-Box Structure and Generating a High Fidelity
Stick Model

By

Mostafa S.A. Elsayed, M.Sc.


Ph.D. Candidate

Amandeep Singh
M.Sc. Student

Ramin Sedaghati, Ph.D, P.Eng.


Associate Professor
Principle Investigator for Task 6

Department of Mechanical and Industrial Engineering


Concordia University

Sponsor’s Ref. No: CRIAQ 4.1-TASK 6

January 2007

1
Conceptual Design Optimization of an Aircraft
Wing-Box Structure and Generating a High
Fidelity Stick Model

Abstract
The current report is a summary of the work done in the four stages of
TASK 6 of the MOSAIC project. The objective of task 6 is to perform a structural
design optimization of the aircraft wing box by adding an optimization loop
around the analysis code and also reviewing the available design constraints
identified in the Bombardier Aerospace and developing new design constraints if
required. The initial dimensions of the wing-box components are obtained from
stages one, two and three where optimization routines are created in MatLab to
size all parts of an aircraft wing-box. The optimized dimensions of skin-stringer
panels in tension and compression are obtained in stages one and two while the
optimized dimensions of spars and spar’s caps are generated in stage three where
the incomplete diagonal tension theory is used. New design constraints as fatigue
constraint, inter-rivet buckling constraint and beam column eccentricity are
found to be important and added to the available design constraints in the
preliminary design optimization stage.
The dimensions obtained from stages one, two and three are employed to
generate 3D Finite Element Model (FEM) in the fourth stage. The DLR-F6
aircraft wing-box is used to implement the developed design optimization
methodology. MSC.PATRAN is used as a modeler while MSC.NASTRAN is used
as analyzer. The generated 3D FEM is validated by testing the performance of the
model due to the application of a set of aerodynamic loads representing normal
cruising conditions.
Subsequently, using the 3D FEM, the stick model of the aircraft wing-box
is created. This model is often used for static aeroelastic and flutter analysis and
is important element of multidisciplinary design optimization. Different

2
methodologies are used to extract the stiffness properties of the wing-box along
the span. All the methodologies in the literature and currently used in the
industry depend on some approximations which in some cases generate
undesirable results. A new approach for generating high fidelity stick models is
developed. The formulation of the new approach is explained. The results
obtained from the developed stick model of the DLR-F6 aircraft wing-box
structure are in excellent agreement with those obtained from the 3D FEM which
shows its high fidelity and accuracy.

Key Words: Wing-Box, Spars and Spar Caps, Stick Models, Diagonal
Tension, Multi Disciplinary Design Optimization (MDO).

3
Section 1 Overview
1-1 Description of the Project:
The objective of task 6 in the MOSAIC project is to improve the available
structural analysis modules in the Bombardier Aerospace and perform a
structural design optimization of the wing box by adding an optimization loop
around the analysis code. The objective is to design a wing-box more rapidly and
automatically. Task 6 is divided into four stages.

Stage I: Optimization of one skin stringer panel:


Stage I explained in details the procedure to optimize one skin-stringer
panel consists of one stringer with one stringer spacing (or pitch) of skin in the
chord wise direction and the distance between two ribs in the span wise direction.
Skin-stringer panels on the upper and lower wing covers are considered. The load
acting on the panels is taken to be constant (i.e. same load acting on all panels)
which resulted in identical dimensions for all panels. Stage-I provides a
methodology to obtain the optimum dimensions for a skin-stringer compression
panel (upper wing cover) with a minimum mass under six constraints namely
crippling stress, column buckling, up-bending at center span (compression in
skin), down-bending at supports (compression in stringer outstanding flange),
inter-rivet buckling and beam column eccentricity. It also provides optimum
design variables for panels under tensile loading (lower wing cover) with fatigue
life based on damage tolerance approach as a design constraint with the same
objective function (minimum mass for the panel). Please refer to reports I and II
for more information.

Stage II: Load Redistribution:


Stage II presented the methodology for calculating the actual load
experienced by each skin-stringer panel when arranged on the airfoil profile at
any span wise section of the wing. The number of stringers required on the upper

4
and lower wing covers is obtained by dividing the width of the wing-box by their
corresponding stringer pitch obtained from stage I. These panels are then re-
arranged on the actual airfoil profile at certain span wise section. Each panel now
experiences different magnitude of compressive or tensile load depending on its
relative location with respect to the centriodal axes of the section.
The optimum dimensions for panels on upper and lower wing covers are
thus obtained using the new calculated design load which resulted in a different
optimum dimensions for each panel according to its location. Please refer to
reports I and II for more information.

Stage III: Optimization of Spars and Spar Caps:


Stage III presents a methodology for the design optimization of spars and
spar’s caps. The developed optimization algorithm generates lighter design while
satisfying all constraints. The optimization algorithm is used to size web spar and
spar caps at each section of wing box. Furthermore, the comparison between T
type and L type section is made and the effect of numbers of uprights on the
optimum design is investigated. The optimum dimensions of spar web and caps
obtained from optimization process is used to build conceptual wing box model
in stage 4.

Stage IV: 3D FE Model of the wing box:


In this stage the Finite Element Model (FEM) of a full aircraft wing-box is
created based on the initial dimensions obtained from the preliminary design
optimization of all wing-box components obtained in the previous stages.
It is noted that the DLR-F6 aircraft has been chosen as a practical example
to apply the optimization methodology under investigation. Also, the process of
generating the stick model using the 3D FEM is investigated. A new approach is
developed for stiffness properties extraction from the 3D FEM which resulted in a
high fidelity stick model.

5
It should be noted that the current report provides final and detailed
information regarding stages three and four. Please refer to reports two and three
regarding stages one and two.

1-2 DLR-F6 Aircraft Geometry and Wing Details


The geometry and load details are taken from the DLR-F6 aircraft [1]. The
actual wind tunnel model and its overall geometry are shown in Figures (1) and
(2), respectively.

Fig. (1) DLR-F6 wind tunnel model [1]

In Figure (2), Axes x, y and z denote the coordinate system for the aircraft
body and axes x*, y* and z* refer to the wing coordinate system. The wing with
nacelle is defined in wing coordinate system and is placed in the body system
with x and z translations of 13.661 in. and -1.335 in., respectively with a dihedral
of 4.787 degrees.
The nacelle is located at 8.189 in. from the wing origin. The projected wing
semi-span is 23.0571 in. The wing is defined by a number of airfoil sections at
different stations along the wing span as shown in Figure (3). The shape of the
airfoil at each station is selected based on the aerodynamics and holds the shape
of the wing.

6
y*
y

Fig. (2) DLR-F6 wind tunnel model geometry [1]

Airfoil section

y*
η=
s*
Fig. (3) DLR-F6 wing showing different airfoil sections [1]

7
Figure (3) shows a number of airfoil sections that are defined at different η

y*
along the wing span, where η is the normalized coordinate defined as η =
s*
where s* is the semi-span of the aircraft wing. The front spar is usually positioned
at 15% of chord and the rear spar at 65% of chord measured from the leading
edge. The enclosed area between the spars is called the wing-box.
In order to test the optimization, the wind tunnel geometry is scaled by a factor of
λ=20 to build an approximately realistic aircraft model. The scaled model
dimensions of the wing are given below:
The wing reference area for the scaled model is S=90148 in.2, the semi-span in
wing coordinate system is s*= 463.3 in., the average chord length of the wing is
C av = 97.746 in.

8
Initial Sizing of Wing-Box
Section 2 Spars and Spar’s Caps

2-1 Introduction
Generally, there are two categories of approaches to deal with the design
and analysis of spars. One is the traditional method, which was introduced in
detail by Kuhn et al. [2], Bruhn [3], and Niu [4]. The dimensions of spar web and
spar cap can be optimized by using column buckling, plate buckling and crippling
analysis theories combined with empirical equations or curves. This approach is
simple and generates reasonable results. However, the main obstacle is that the
empirical equations and curves can only be used to the specific materials and the
configuration of structures. Therefore, the applications of this approach are
limited to certain type of materials. Another approach is the numerical approach
based on the finite element analysis and most commonly used for composite wing
structure. With the developments of integrated CAD/CAE software and reducing
cost of computer hardware, a considerable amount of research has been
conducted in this field, especially for Multi-Disciplinary Optimization (MDO).
The main obstacle of this method is the increased computational cost and the
computational time increases significantly when the finite element analysis is
integrated with optimization algorithm. The objective of stage 3 is the size
optimization of spar web and spar caps at each section by using traditional
method.

2-2 Design of Spar’s Web and Caps


Each wing box structure has a front spar assembly, a rear spar assembly
and the ribs. The spars are approximately located early in the design phase
during the selection and layout of the wing box size. A natural tendency is to
locate the front spar at a constant chord location, between 5% to 20% chord. The
front spar location should be selected to ascertain the space provisions in the
leading edge device and to maximize the box volume for fuel containment and
structural rigidity. The rear spar is usually located between 60% to 80% chord.
The rear spar location is subject to as many or more influences as the front spar.

9
Spar caps are used to connect spar web to the skin of the wing box as shown in
Figure (4).

Fig. (4) Spar (web and cap) assembly

Generally, T-type and L-type caps are used for the aircraft structures. The
cross sectional area and other design parameters for these sections are shown in
Figures 5 and 6.
For a T-section as shown in Figure (5), we have:

Fig. (5) T-Type spar cap

Acap = B1T1 + B2T2 (1)

Y=
(B T
1 1
2
+ T2 B2 (2T1 + B2 ) )
(2)
(2(T1 B1 + T2 B2 ))

B1 B3
Ix = ( B2 + T1 )3 − 2 ( B1 − T2 ) − Acap ( B2 + T1 − Y ) 2 (3)
3 3

10
where Acap , I x and Y are cap cross- sectional area, second moment of area about

centroidal axis and centroidal distance from the top of the cap.
For a L-section as shown in Figure (6), we have:
Acap = B1T1 + B2T2 (4)

B22 + ( B1 − T2 )T1
Y= (5)
2( B1 + B2 − T1 )

(T1 ( B2 − Y ) 3 + B1Y 3 − ( B1 − T2 )(Y − T1 ) 3 )


Ix = (6)
3

Fig. (6) L-Type spar cap

The effective depth of web measured between centroids of caps he and clear

depth of the web (depth of the web measured between bottom of both caps) hc
are calculated as:
3 ( B2 − Y )
he = h − 2Y , hc = he − (7)
2
where h is the depth of the web.
The air load act directly on the wing covers which transmits the loads to
the rib. The rib transmits the load in shear to the spar web and distributes the
load between them in proportion to the web stiffness. The spar web and caps are
mainly subjected to bending and shear loading. The depth of the web is usually
large as compared to depth of the cap, therefore bending stress in the web are
neglected. It is assumed that caps develop the entire bending resistance and

11
shear flow is constant over the web. The applied shear flow in the web can be
written as:
V
q= (8)
he
where V is the applied shear force on the spar. The applied shear stress in the web
can be calculated as:
q
fs = (9)
t
where t is the thickness of web.

2-3 Failure Constraints


In order to optimize the wing box structure, the design must satisfy a set of
constraints, e.g. material failure and buckling must not occur anywhere within
the configuration. In the present work, the incomplete diagonal tension failure of
spar web, and crippling and bending failure of spar caps are considered.

2-3-1 Spar Web Failure


The two basic types of web design are shear resistant type and diagonal
tension field type. A shear resistant web is one that carries its design load without
buckling of the web. The design shear stress should not be greater than the
buckling shear stress for the individual web panels and the web should have
sufficient stiffness to keep the web from buckling as a whole. It is realized that the
buckling web stress is not a failing stress of the web and the web can withstand
more load before collapsing. Thus based on this approach, the web is not fully
loaded. The diagonal tension type webs are generally used for the design of spars
of an aircraft. In the diagonal tension webs, buckling of the web is permitted with
shear loads being carried by diagonal tension stresses in the web. At the buckling
load, panel buckles into the diagonal folds and additional loading is taken by
diagonal tension produced in these folds. The equations required for the analysis
are presented here and the detail description of theory of incomplete diagonal
tension can be found in [2].

12
The stresses in the web subjected to incomplete diagonal tension depend
on the diagonal tension factor which is the measure of degree of loading of
structure above its buckling strength. The diagonal tensional factor can be
calculated as:
⎧⎪ ⎛ f ⎞⎫⎪
K = tanh ⎨0.5 log10 ⎜ s ⎟⎬ (10)
⎪⎩ ⎜F ⎟⎪
⎝ s_cr ⎠⎭
where Fs_cr is the shear buckling stress of the web. The shear buckling stress of

the web can be obtained by following steps [2]:

(i) Calculate the flat plate buckling coefficient K s for in-plane shear
loading using the following polynomial:
K s = 0.001 r 6 − 0.0293 r 5 + 0.3401 r 4 − 2.0808 r 3 − 13.668 r + 16.497 (11)

hc
where r = and d is the length of the spar web.
d

(ii) Calculate the following ratio:


2
⎛t⎞
FS_R = K S E ⎜ ⎟ (12)
⎝d ⎠

(iii) The shear buckling stress Fs_cr of the web is obtained from the

following polynomial approximations:


Fs_cr = 42.5 FS_R > 197.85
(
Fs_cr = −3 × 10 −12 Fs_R)6 + 2 × 10 −9 (Fs_R )5 − 8 × 10 −7 (Fs_R )4 FS_R > 0 and FS_R < 197.85
+ 0.0001(Fs_R )3 − 0.0141(Fs_R )2 + 0.8472Fs_R + 12.819
Fs_cr = 0 FS_R = 0
(13)
The maximum shear stress f s_ max in the web corresponding to the above

calculated diagonal tension factor can be evaluated as:


f s_ max = f s ( 1 + K 2 C1 )(1 + KC 2 ) (14)

13
Where C1 and C 2 are the stress correction factors. The factor C1 is to allow for
the fact that the angle α of the diagonal tension differs from 45 degrees and can
be obtained as:
1
C1 = −1 (15)
Sin ( 2α)

The factor C 2 is the stress concentration factor arising from flexibility of the cap
and can be obtained as:
C 2 = 0.0013(wd)6 − 0.0211(wd)5 + 0.1156(wd) 4 − 0.2434(wd)3
+ 0.2267(wd) 2 − 0.0713(wd) wd < 4 (16)
C2 = 1 wd > 4
Where wd is the cap flexibility factor and can be obtained as:
1
⎛ t ⎞4
wd = d ⎜⎜ ⎟⎟ (17)
⎝ 4 he (I T + I c ) ⎠
Where I T and I C represent the second moment of area of the tension and the
compression flanges respectively (with respect to their centroidal axis).
The allowable maximum shear stress Fs,all in the web can be obtained from
the following [5]:
For α = 40o
Fs,all = −52.612K 6 + 186.3K 5 − 243.12K 4 + 124.62K 3 + 4.2935K 2 − 27.272K + 34.85 (18)

For α = 45o
Fs,all = −1.1689K 6 + 41.756K 5 − 92.897K 4 + 53.707K 3 + 19.277K 2 − 28.05K + 34.924 (19)

The allowable maximum shear stress at α = 45o is good approximation for the
most of the problems. Now the optimization problem is constrained such that
maximum shear stress is less than the allowable maximum shear stress. In other
words:
f s, max − Fs,all ≤ 0 or (20)

f s, max
−1 ≤ 0 (21)
Fs,all

14
2-3-2 Spar Cap Failure
The crippling and bending failures are two main modes of failures in the spar
cap. Here both modes of failures are addressed.

a) Crippling Failure
The cap resists two types of axial compressive stresses which cause crippling
failure, compressive stress caused by bending moment and compressive stresses
caused by diagonal tension. The compressive stress f b in the cap caused by the
beam bending moment can be written as [2]:
M
fb = (22)
he Acap

The compressive stress f F caused by the diagonal tension in the web can be
written as [2] :
KV
fF = (23)
2 Acap tan α

where K is the diagonal tension factor and V is the applied shear force on the
web. The crippling failure in the cap is caused by combination of f b and f F .
To compute the allowable crippling stresses of the cap, the section is
broken down into individual segments and each segment n has a width of b and a
thickness of t and will have either one or no free edge. The allowable crippling
stress for each segment n is found from the applicable material test curve or from
the following empirical formulas [5]:
If segment n has free edge:
−0 .788
⎡b Fcyn ⎤
Fccn = 0.6424 Fcyn⎢ n ⎥ (24)
⎢⎣ t n E n ⎥⎦

If segment n has no edge free:


−0 .7882
⎡b Fcyn ⎤
Fccn = 1.1819 Fcyn ⎢ n ⎥ (25)
⎢⎣ t n E n ⎥⎦

15
Where Fcyn and E n are the allowable compression yield stress and the modulus of

elasticity of segment n. Now the allowable crippling stress for the entire section is
computed by taking a weighted average of the allowable for each segment:

Fcc =
∑b t F
n n ccn
(26)
∑b t n n

b) Bending Failure
In addition to the compressive stress, the cap is also subjected to secondary
bending moment generated by incomplete diagonal tension in the web. The
secondary bending moment can be obtained as [2]:
K C 3 f S t d 2 tan α
M max = (27)
12
Where C 3 is the stress concentration factor and can be obtained from the
following equation:
C 3 = 5 × 10 −5 (wd)6 − 0.001(wd)5 + 0.0084(wd) 4 − 0.0332(wd)3
+ 0.0341(wd) 2 − 0.0092(wd) + 1 wd < 4 (28)
C 3 = 0.6 wd ≥ 4

The secondary bending stress f sb in the cap can be obtained as:

M max ( B2 + T1 − Y )
f sb = (29)
I x ,cap

The cap is subjected to both compressive and bending stress simultaneously,


therefore the margin of safety of the cap is combination of both compressive and
bending failure. The margin of safety for the cap can be calculated as:
1
MS = −1 (30)
fb + f F f
+ sb
Fcc Ftu

Where Ftu is the ultimate tensile stress. Here the constraint on the optimization
problem is imposed such that margin of safety of the cap is greater than zero or
1
1− ≤0 (31)
fb + f F f
+ sb
Fcc Ftu

16
2-4 Optimization Problem
The objective of the optimization is to minimize the mass of the spar webs
and caps assembly while guarding against any type of failure. The design
variables are the thickness of the spar and dimensions of the caps. The design
optimization problem constituting all the constraints can be stated as:
Minimize
Mass= (t he + 2 Acap ) L (32)

f s, max 1
Subject to −1 ≤ 0 and 1− ≤0
Fs,all fb + f F f
+ sb
Fcc Ftu
The optimization problem stated above is nonlinear optimization problem where
constraints are nonlinear function of design variables. The Sequential Quadratic
Programming (SQP) [6] is used to solve the above optimization problem.
Sequential Quadratic Programming (SQP) method is mathematical programming
technique for solving Non-Linear Programming (NLP) optimization problems.
SQP is considered to be the most robust and powerful mathematical
programming optimization algorithm available today. In this method a Quadratic
Programming (QP) sub problem is constructed from the initial NLP problem. A
local minimizer is then found by solving a sequence of these QP sub problems
using a quadratic approximation of the objective function. At the end of the one
dimensional minimization, Hessian of Lagrangian, required for the solution of
the next QP problem is updated using Broyden-Fletcher-Goldfarb-Shanno
(BFGS) [6] update formula. The gradient of Lagrangian is required at the start of
iteration, and it consists of the gradient of both the objective function and the
constraint. Here, the MATLAB optimization toolbox is used to implement SQP
algorithm and gradient is calculated by using the finite difference technique.

2-5 Numerical Validation


The optimization problem formulated above is validated by comparison
with design example solved by Niu [4] and Abdo [5]. The design parameters for
the problem are:

17
h = 14in.,V = 30000lb, d = 8.0in and M = 560000lb.in
The T-type spar caps assembly as shown in Figure 7 is used for design. The
material properties of the web and caps are:

Fig (7) T-type spar cap assembly

Web-7075-T6 bare sheet Caps-7075-T6 Extrusion


E = 10.5 × 10 6 E = 10.7 ×10 6
Ftu = 80000 psi Ftu = 82000 psi
Fcy = 71000 psi Fcy = 74000 psi

In the present case, the optimization problem is solved by considering three


different cases:
Case 1: All design variables are independent (B1, B2, T1, T2, t)
Case 2: Lengths of both flanges in spar cap are constrained to be equal (B1=B2, T1,
T2, t).
Case 3: Length and thickness of both flanges in spar cap are constrained to be
equal (B1=B2, T1=T2, t).

18
Table 1: Optimization results
Total
Acap Cross. He Hc Fs, all fs, ,max MS MS
(in2) T (in) Area K (in) (in) Fs_cr (lb) (lb) (web) Cap
Niu 0.918 0.085 3.026 0.25 12.08 9.7 8772 29700 29387 0.01 0.03
M.Abdo 1.01 0.085 3.21 0.26 12.12 9.71 8145 29383 29242 0.0048 0.067
Case 1 0.7946 0.0855 2.626 0.2804 12.122 10.21 7680.1 29255.8 29273 0.0006 0.0002
Case 2 0.8177 0.0822 2.6869 0.3047 12.784 10.889 6702.2 28993.9 29009 0.0005 0.0004
Case 3 0.8181 0.0821 2.6873 0.3047 12.798 10.828 6706.3 28995 29994 0 0.0003

The results obtained from the present optimization algorithm are compared with
those obtained by Niu [4] and Abdo [5], and are presented in Table 1. It should be
noted the design obtained by Niu [4] and Abdo [5] is obtained by using ad-hoc
design iteration approach. The optimum dimensions of the spar cap are given in
Table 2.
Table 2: Optimal dimensions of the cap (in)
B1 B2 T1 T2
Case 1 1.0986 2.0491 0.1617 0.3011
Case 2 1.6315 1.6315 0.2396 0.2616
Case 3 1.669 1.669 0.2451 0.2451

Examination of the results reveals that the total cross-sectional area of the web
caps is considerably reduced by using present methodology. It can also be
observed that the diagonal tension factor obtained at the optimum design using
present method is more than that obtained by both Niu [4] and Abdo[5], and web
is subjected to large diagonal tension. Among all three cases, as expected the
minimum cross sectional area is obtained considering Case 1. For Case 2, an
additional constraint is imposed on the optimization algorithm such that lengths
of both flanges are equal. The additional constraint is imposed to obtain more
symmetrical design and results in decrease in the number of design variables.
The additional constraint in Case 2 results in a heavier design with respect to the
Case 1, however it is still much lighter than that obtained by Niu [4] and Abdo[5].
The thickness of the web is decreased and web is subjected to higher diagonal

19
tension field. To obtain more symmetrical design, Case 3 is considered. It is
assumed that both flanges have equal length and thickness. It can be realized that
very small increase in the mass of the structure is observed by imposing this
additional constraint, and insignificant change has been observed in the diagonal
tension factor and thickness of the web.

2-6 Uprights Design


Uprights are transverse members which prevent spar web from diagonal buckling
as shown in Figure (8).

Fig. (8) Spar web reinforced by an upright [2]

The correct evaluation of number of uprights and their sizes are pertinent
for optimum design. The average stress over the length of the double upright can
be calculated as [2]

K f s tan α
fU = (33)
AU
+ 0.05 (1 − K )
dt
Where K is the diagonal tension factor, d is the distance between successive
uprights, f s shear stress acting on the spar’s web, α is the diagonal tension

folding angle, t is the thickness of the spar’s web and AU is the upright cross-

20
sectional area. The stress fU is uniformly distributed over the cross-section of the

upright until buckling of the upright begins. For single upright, the stress fU is

AU A
obtained in the same manner, except that the ratio is replaced by Ue , where
dt dt
AU
AUe = 2
(34)
⎛e⎞
1 + ⎜⎜ ⎟⎟
⎝ρ⎠
Where AU e is the effective cross-section area of the upright, e is the upright

cross-section centroid eccentricity from the center of the web while ρ is the
upright cross-section radius of gyration.
AU e
An estimate for the value of can be made using Figure (9):
AU

Fig. (9) Ratio of the effective to the actual area of the upright [2]

In the current study the analysis is made for the L-shape upright. The stress fU is
the average stress along the length of the upright. However, it has been noted in
test that this stress actually has a maximum value fU max at the middle of the

upright and decreases towards the end. The maximum stress fU max can be
calculated by the following equation:

21
⎛f ⎞
fU max = fU ⎜⎜ U max ⎟⎟ (35)
⎝ fU ⎠b

⎛f ⎞
where ⎜⎜ U max ⎟⎟ is the value of the ratio when the web just buckled and can be
⎝ fU ⎠b
calculated as:
⎛ fU max ⎞
⎜⎜ ⎟⎟ = K (0.625 r1 − 0.8) − 0.64 r1 + 1.78 (36)
⎝ fU ⎠b
Where
d
r1 = (37)
hU

Where hU is the height of the upright. The two most common failure criteria’s for
the design of uprights are column failure and forced crippling.

Column Failure
When column bowing begins, the upright will force the web out of its
original plane. The web tensile force then develops component normal to the
plane of the web which tends to force the uprights back. This action is taken into
account by using a reduced effective column length Le of the upright, which is
given by the following empirical formula
hU
Le = (38)
⎛ d⎞
1 + K 2 ⎜3 − 2 ⎟
⎝ h⎠
The stress at which column failure takes place can be found for slenderness ratio
Le
of as:
ρ
2
Fcs2 ⎛ Le ⎞
Fc = Fcs − ⎜ ⎟⎟ (39)
4π E ⎜⎝ ρ ⎠
Where Fcs is the critical shear stress in the spar’s web. To guard column against
column failure, the average stress in the upright should be less than column

22
Le
failure stress at slenderness ratio of . A constraint on the optimization
ρ
algorithm is imposed such that:
fU
−1 ≤ 0 (40)
Fc

Where f u is the average stress in the upright.

Forced Crippling
Forced crippling is a major failure criteria of uprights. The shear buckles
in the web will force buckling of the upright in the leg attached to the web,
particularly if the upright leg is thinner than the web. These buckles give a lever
arm to the compressive force acting in the leg and therefore produce a severe
stress condition. The upright stress at which final collapse occurs is obtained by
the following formula:
1
⎛t ⎞
2 3
For 24S-T3 alloy FU = 26000 K 3 ⎜ U ⎟ (41)
⎝ t ⎠
Where tU is the upright flange thickness.
1
2 ⎛ tU ⎞ 3
For 75S-T6 alloy FU = 32500 K 3 ⎜ ⎟ (42)
⎝ t ⎠
To avoid force crippling, the maximum stress fU max should be less than the

allowable stress FU .

fU max
−1 ≤ 0 (43)
FU

2-7 Design of Spar and Upright Assembly


Before designing the spar cap assembly, the study is performed on the
number of uprights in the bay. The assembly is designed using L-shape upright
with different number ranging from one to three. For the same load condition
given in Section 2-5, the design for the spar-cap assembly and uprights for
different number of uprights are shown in Table 3 and Table 4, respectively. The

23
design variables of the upright are typically similar to those of the L-shape spar’s
cap since they have the same geometry of cross-section.

Table 3: Optimum design of spar (web and cap) assembly for different
number of uprights

Number Mass of Spar (Web and


of Cap) Assembly (lb)
Uprights B1(in) B2(in) T1(in) T2(in) t(in)
1 3.9391 3.2850 0.5869 0.4963 0.0986 243.80
2 4.1034 2.7316 0.6113 0.4070 0.0928 226.22
3 4.1324 2.0757 0.6179 0.4614 0.0882 219.2

Table 4: Optimum design of uprights


Number Total Mass of
of Upright mass Spar-Upright
Uprights B1u(in) B2u(in) T1u(in) T2u(in) (lb) Assembly (lb)
1 1.0861 31.1539 0.35 0.1494 9.6386 253.44
2 0.89 0.8388 0.35 0.0594 8.9434 244.08
3 0.7573 0.4288 0.35 0.051 8.5784 244.93

It can be seen from Tables 3 and 4 that the mass of the spar cap assembly and
upright decreases with increasing the number of uprights in assembly. As it can
realized from Table 4, the total mass of spar-upright assembly for two uprights is
less than that of one upright and the total mass of assembly for three uprights is
slightly higher than that of two uprights. Thus, it can be concluded that spar
assembly with two uprights is optimum design and this arrangement will be used
to design spar-upright assembly.

2-8 Comparison of L Shape and T Shape Cap


The comparison study is performed between L and T shape caps. The
objective is to choose a cap that result in minimum weight of the structure.

24
The optimum design for different cap shapes for spar (web and cap) assembly
with two uprights is provided in Table 4. Results show that the T shape design
result in lower weight with respect to the L shape design. It can also be seen that
the major weight reduction is due to the top flange of the T cap (T1 for T shape
cap is much smaller than that of L shape cap). It is noted that the crippling failure
is the deciding factor for the design of cap, and upper flange of T shape increases
the crippling strength of the cap.

Table 5: Design of spar assembly using T and L shape cap

Type of Upright t(inc) Mass of Spar (Web and Cap)


B1(inc) B2(inc) T1(inc) T2(inc) Assembly (lb)
T Shape 5.6463 2.8297 0.4254 0.4216 0.0922 224.9828
L Shape 4.1034 2.7316 0.6113 0.4070 0.0928 226.22

2-9 Conclusion
The optimization algorithm has been developed to optimize the design of the spar
(web and cap)-upright assembly while guarding against all types of failure. The
results of the optimization algorithm show better performance than the
procedures used in literature. Further, investigation is performed on the
optimum number of uprights in the spar assembly and it has been concluded that
best efficiency can be obtained using two upright. Also, the comparison is
performed between the T and L shape caps. The T shape provides more crippling
strength and result in lower weight of the structure. Finally, the optimization
algorithm is used to size the spar assembly for each bay. The final conceptual
design is used to build the finite element model.

25
3D Finite Element Model of DLR-F6
Section 3 Aircraft Wing-Box Structure
3-1 Introduction
The current section explains the methodology of creating the 3D finite
element model of the DLR-F6 wing-box structure based on the preliminary
design optimization of wing-box components conducted in stages I-III.

Rear spar
(Stage III)
Skin stringers
(Stages I & II)

Wing ribs
(Special Report)
Front spar
(Stage III)

Fig. (10) DLR-F6 wing-box structure created in CATIA

Figure (10) shows the different components of the DLR-F6 wing-box that
are employed to generate its 3D finite element model. The optimized dimensions
of these components are obtained from the preliminary design optimization
process performed in the previous three stages.
Since the thickness of the skin as well as the width of the skin-stringer
panels are the two design variables of the optimization process of stages one and
two, then the output of these two stages is the skin thickness along each panel
pitch at each wing station. These dimensions are determined through an

26
optimization process for mass minimization as an objective functions. Using the
stringer’s pitch along each skin-stringer panel, the number of stringers as well as
the coordinates of the stringers along the upper and the lower wing skin are
determined in wing coordinate system.
To insure moment of inertia maximization, a set of relations are adopted
to relate skin thickness and panel width with the other dimensions of the skin-
stringer panel (please refer to reports 3 and 4).
The dimensions of spars and spar caps are determined based on the
incomplete diagonal tension theory as explained in Section (2).

3-2 Building the Finite Element Model


The wing-box is divided into 20 bays extending between 21 stations as
shown in Figure (11).

Fig. (11) DLR-F6 wing lay-out

MSC.PATRAN is used as the modeler to build the finite element model.


Each bay is grouped into four groups namely the skin, the stringers, the rib and a
group for the load card and its rigid elements which are used to distribute the

27
load. Spar caps are included with the stringers’ group since they are modeled with
one dimensional beam elements while the spar webs are included with the skin’s
group since they are modeled with two dimensional shell elements.
Using the dimensions generated in the previous stages the geometry of the
wing-box is defined and then the finite element model is generated.
It is well known that increasing the number of elements in the model
enhances the accuracy of the results but it increases the computational cost.
Accordingly, it is required to keep the minimum number of elements necessary to
obtain acceptable accurate and efficient results. A tolerance error of 0.005% in
the FEM response compared with the analytical results is considered acceptable.
It has been found that a mesh of three elements in the span wise direction along a
wing-box bay results in acceptable results.
The stringers are modeled by beam elements with a Z shape cross-section. The
details of the Z-shape cross-section are shown in the Figure (12).

Fig. (12) The Z-shape cross-section of the beam element used in the PBEAML
card for stringer modeling [7]

A comparison between the dimensions of this Z-shape cross-section and the


dimensions obtained from the optimization process in stages one and two, shows
that DIM1, DIM2, DIM3 and DIM4 can be easily calculated from the following
equations
tw
DIM 1 = ba − (44)
2

28
DIM 2 = t w (45)

DIM 3 = bw − t a (46)

DIM 4 = bw + t a (47)

Where ba , bw , t a and t w are the stringer dimensions as shown in Figure (13).

Fig. (13) Panel geometry definition using ‘Z’ stringer [9]

Since the dimensions of the stringers vary from one station to the other, an
interpolation process is used to obtain the dimensions of all elements between
stations. This is done by defining the dimensions in PATRAN as fields, where a
local coordinate is created at each station with its Z-coordinate directed in the
span wise direction. Set of fields are created in PATRAN defined in the station
local coordinate, representing the variation of the dimension in the span wise
direction. As an example, consider the dimension DIM1 of the Z-shape stringer
extending between stations 21 and 20, this dimension is defined in PATRAN as a
field in the form of
DIM 1 = DIM 1 _ 20 + ( DIM 1 _ 21 − DIM 1 _ 20) Z (48)
Other dimensions are defined in a similar way. Figure (11) shows the stringers for
one bay extending between two stations created in PATRAN. As it can be see a
local coordinate system is defined for each stringer to model its taper shape in the
span wise direction.

29
Fig. (14) Stringers for one bay created in PATRAN

The spar caps are also modeled by beam elements but with L-shape cross-
section as shown in Figure (15).

Fig. (15) The beam element with L-shape cross-section used in the PBEAML card
for spar caps modeling [7]

30
The skin is modeled by SHELL elements, where the thickness of the shells
is also defined by fields representing the variation of the skin thickness in the
span wise direction.

Modeling of ribs:
A group of points is generated to represent the perimeter of the rib; these
points have the same y-coordinate of the wing-box station, its x-coordinate has
the same x-coordinate of the corresponding stringer, while its z-c00rdinate can
be defined by the following equation
z r = z s − (DIM 4)s (49)

Where z r is the z-coordinate of the rib perimeter point, z s is the z-coordinate of

the corresponding stringer while (DIM 4)s is a dimension belonging to the


corresponding stinger at this rib point.
The rib web is modeled by QUAD4 elements with PSHELL card for its properties.
While the perimeter of the rib and the lightening holes are reinforced by beam
elements, as shown in Figure (16). Figure (16) shows complete single wing bay
extending between two ribs modeled in PATRAN.

Fig. (16) Complete bay modeled in PATRAN

31
3-3 Model Verification
Early model verification is very important before proceeding for the whole
finite element model of the wing-box of the aircraft, since detection of errors in
advanced stages is very costy and time consuming.
The finite element model of the DLR-F6 wing-box bay, created in PATRAN
in the previous section, can be tested in NASTRAN for an arbitrary loading. Then
the result obtained from NASTRAN is compared with the analytical solution of
such model for the same loading. The complementary internal virtual work
theory of idealized beams is used to calculate the analytical deflection of this wing
bay under the effect of a flexural bending force.

NASTRAN Analysis
The model created in the previous section is loaded by an arbitrary load of
1000 lbs acting at the bay centroid. To apply the load at the centroid of the bay, a
grid point is created at the section centroid; this grid point is connected to the
skin-stringers connectivity grid points by a group of RBE3 elements with their
reference grid at the centroid and their independent degrees of freedom at the
skin-stringers connectivity grids. A load of 1000 lbs is applied in the negative z-
direction at the centroidal grid point. The model is fixed at all skin-stringers
connectivity grid points of the opposite bay station. The model is then submitted
to NASTRAN for linear static analysis, which resulted in a deflection of 0.0126 in.
at the centroid point. Figure (17) shows the deformed shape of the wing-bay
analyzed in NASTRAN.

32
Fig. (17) Deflection of wing-box bay due to applied load of 1000 lbs

It is clear from Figure (17) that the deflection is the result of superposition of a
combined loading. Since the station centroid does not coincide with its shear
center, the force applied at the centroid generates bending, shear as well as
torsion effect. The deflection obtained from the NASTRAN analysis agrees very
well with that obtained analytically.

3-4 Complete 3D Finite Element Model of the DLR-


F6 Wing-Box
Once the model is verified, the work can proceed towards creating the full
finite element model of the wing-box, as shown in Figure (18).

33
Fig. (18) DLR-F6 wing-box finite element model

The model is submitted to NASTRAN for linear static analysis. The wing-
box is subjected to static loading which represents normal cruising conditions
(detailed load calculations can be found in report 4), applied along the wing-box
elastic axis which produced a maximum deflection of 17.7 in at the wing tip.
Figure (19) shows the deformed shape of the full wing-box 3D finite element
model.

34
Fig. (19) DLR-F6 wing-box deflection due to cruising condition loads

Completing the entry of the material card in the bulk data file, to include
the ultimate stresses of the material in tension and in compression, generates the
margins of safety of the developed finite element model. The analysis showed that
the margins of safety are between zero and one which indicates proper sizing of
the model.

35
3-5 Post-processing of the Wing-Box Finite Element
Model
Deformation of an aircraft wing during flight has significant consequences
on the aerodynamic performance. Predicting an accurate value of the bending
and twisting of the wing in flight depends on the fidelity of the finite element
model of the wing-box. Validation of the finite element model means making sure
that the structural response of the model reproduces the structural response of
the real wing within an acceptable accuracy.
The deflections in x, y and z directions and the twisting angle experienced
by the developed wing-box FEM due to the applied aerodynamic loads are plotted
against the normalized span wise coordinate “η ” and are shown in Figures (20-
23)
Deflection in z-direction

η
Fig. (20) Deflections in z-direction (vertical) experienced by the DLR-F6 wing-
box elastic axis under the effect of cruising aerodynamic loads [in.]

36
Deflection in x-direction

η
Fig. (21) Deflection in x-direction (in-plane) experienced by the DLR-F6 wing-
box elastic axis under the effect of cruising aerodynamic loads [in.]
Deflections in y-Direction

η
Fig. (22) Deflections in y-direction experienced by the DLR-F6 wing-box elastic
axis under the effect of cruising aerodynamic loads [in.]

37
Twisting around y-axis

η
Fig. (23) Twisting angle around the y-axis (torsional) experienced by the DLR-F6
wing-box elastic axis under the effect of cruising aerodynamic loads [degree]

It would be more realistic to evaluate the developed wing-box FEM


stiffness properties. To evaluate the equivalent moment of inertia and torsional
rigidity of the model, two shear center nodes are created at the extremities of
each wing bay. These two nodes are attached to the structure using RBE3
elements, as previously explained. The next step is to load the node, which is
towards the wing tip, by three sets of unit load moments. The first set moment is
along the x-axis to calculate the vertical bending moment of inertia. The second
set is the moment along the y-axis to calculate the torsional rigidity and the third
set is along the z-axis to predict the horizontal bending stiffness. The wing-box
rotations in the x, y and z directions due to the corresponding applied moments
are computed using NASTRAN and then the corresponding values of the stiffness
are calculated using the following equations [8]:

(η j − η i )s *
( EI x ) i→ j =
(θ xj − θ xi ) (50)

38
(η j − ηi )s *
(GJ y ) i→ j =
(θ yj − θ yi ) (51)

(η j − η i )s *
( EI z ) i → j =
(θ zj − θ zi ) (52)

Where s * = 463.2" is the semi-span of the DLR-F6 wing. η i and η j are the

normalized coordinates of the two stations i and j respectively.


The stiffness properties of the 20 wing bays are calculated and plotted
against the wing normalized coordinateη and are shown in Figures (24-26).

EI x

η
Fig. (24) Distribution of the vertical bending stiffness of the DLR-F6 wing-box
along its span

39
GJ y

η
Fig. (25) Distribution of the torsional stiffness of the DLR-F6 wing-box along its
span

EI z

η
Fig. (26) Distribution of the horizontal bending stiffness of the DLR-F6 wing-box
along its span

40
3-6 Model Stiffness Validation
A methodology to estimate the stiffness distribution of a new wing using
the stiffness distributions of Bombardier’s existing wings was developed by Abdo
et.al. [8]. This methodology is based on a set of empirical relations that are
generated to predict the ideal stiffness of an arbitrary aircraft wing-box. To
obtain these empirical relations, the data of the stiffness properties of a group of
existing Bombardier’s aircrafts were normalized. One of the normalization
techniques used is that, the stiffness of the existing wing structure is divided by
the stiffness of a solid block material bounded by the leading and trailing edge of
the wing, which referred to as ( EI ) CATIA , as CATIA was used for the calculation of
this solid wing stiffness, as shown in Figure (27).

Fig. (27) DLR-F6 wing airfoil sections

41
Figure (27) shows airfoil sections of the DLR-F6 wing at 21 wing stations.
At each wing station, the airfoil section is padded, and then the measure tool bar
is used to calculate the stiffness of each wing section in CATIA.
The data obtained from CATIA are used along with the empirical relations
to predict the ideal stiffness properties of an aircraft wing-box.
For EI x the behavior of the normalized stiffness appeared to be different
outboard and inboard of the break in the plan form, consequently different
relations were used to fit the data of empirical relations, as follows [8]:
(EI x )FEM R Break − RRoot
= (η − η Root ) + RRoot for η Root ≤ η ≤ η Break
(EI x )CATIA η Break − η Root
(53)
(EI x )FEM
= RBreak for η Break ≤ η ≤ 1
(EI x )CATIA
y Break
η Break = (54)
S*
y Root
η Root = (55)
S*
where R Root is the (EI x )FEM (EI x )CATIA ratio at η = η Root

and R Break is the (EI x )FEM (EI x )CATIA ratio at η = η Break

It was determined that RRoot = 0.03 and R Break = 0.1 provides an acceptable fit for
the airplanes.
For GJ y the following empirical relations was developed

(GJ y )FEM
(GJ y )CATIA = 0.002 (56)

For EI z the following empirical relations was developed


(EI z )FEM
= 0.0103η + 0.007 (57)
(EI z )CATIA
The previous empirical relations are used to predict the stiffness properties of the
ideal wing-box of the DLR-F6 aircraft, and it is compared with the current FEM
stiffness properties, as shown in Figures (28-30).

42
+ + + Ideal FEM Model

o o o o Current FEM Model

EI x

η
Fig. (28) Comparison between the vertical bending stiffness of the DLR-F6 wing-
box along its span based on the ideal FEM and current FEM

+ + + Ideal FEM Model

o o o o Current FEM Model

GJ y

η
Fig. (29) Comparison between the torsional rigidity of the DLR-F6 wing-box
along its span based on the ideal FEM and current FEM

43
+ + + Ideal FEM Model

o o o o Current FEM Model

EI z

η
Fig. (30) Comparison between the horizontal bending stiffness of the DLR-F6
wing-box along its span based on the ideal FEM and current FEM

As it can be seen there is a very good agreement between the current NASTRAN
FEM and the ideal FEM (predicted by the empirical relations) of such a wing.

44
4- New Approach for Generating Stick Models of
Complex Aircraft Wing-Box Structures Used for
Aerodynamics-Structures Interaction

4-1 Introduction
Deformations of aircraft’s wing structure during flight have a significant
effect on the aerodynamic performance of the aircraft. On the other hand, the
aircraft’s wing structural deformations generate new aerodynamic load cases
which must be taken into account during design and optimization.
Different levels of structural models have been used for static aeroelastic analysis
and optimization, ranging from simple models based on analytical or empirical
expressions to complex finite element structural models.
The finite element analysis model is an idealization of a given structural
system which is used to determine the system’s response in the form of
displacements, stresses, internal loads …etc to a given load case. The complete
finite element model is always prepared once the wing’s layout and structural
details are obtained. Once the complete finite element model is prepared,
simplified beam finite element model of aircraft’s wing structure, also known as
stick model, can be generated. This model is often used for aerodynamics-
structure interaction, static or dynamic aeroelastic analysis or optimization and
flutter analysis.
The difficulty is to find or develop models that are sufficiently simple to be
called thousands of times during optimization, but are sophisticated enough to
accurately predict wing deformations in bending as well as twisting. The wing
stick model can be generated by extracting the stiffness properties of the three
dimensional finite element model in the form of bending, axial and twisting
stiffnesses and apply these stiffnesses to a set of simple beam elements which
form the wing’s stick model.
Abdo et. al. [8] presented two methodologies to create wing stick models
of known wing structures, where the end product of their methodologies was a

45
NASTRAN finite element beam model which can be coupled to aerodynamic CFD
models. In their methodologies the determination of the wing cross-section’s
principal axes is necessary to extract its stiffness properties with respect to its
principal axes. Those stiffness properties were then employed to generate the
beam stick model. The limitation of these two methodologies is their dependency
on the principal directions as a main condition for extracting high fidelity inertia.
Although principal axes can be easily found for simple geometries, it is very
difficult to identify these axes for complex structures such as aircraft wing-box.
Approximate methods are typically used in the industry to determine these
principal axes which may generate unsatisfactory results.
Here, a new approach for extracting wing-stiffness properties with respect
to a general user defined coordinate system is presented. The new approach is
validated by generating the stick model of a DLR-F6 wing three dimensional
finite element model. The deformations experienced by the developed 3D FEM
and the stick models generated by the new approach and other approaches in
literature are presented. The results obtained from the developed stick model are
in an excellent agreement with those of the 3D FEM. The results show the
simplicity and the high fidelity of the new approach.

46
4-2 Analytical Procedure to Generate Wing Stick Model
(method I)
Once the wing’s layout and structural details are obtained, the wing cross-
section properties can be calculated.

Consider an element of the wing of length ‘dy’ extending from station ‘j’, located
at an arbitrary normalized coordinate ‘η j ’ along the wing span, to station ‘j+1’.

bs
2 bs
zi, j

xi , j
+ + +
qc sc o

0.1c

Fig. (31) Airfoil cross-section at an arbitrary ‘ η j ’ along the wing span

(Station-j)

47
In Figure (31), three main points are defined; the quarter chord point (qc) which
c
is located at distance measured from the leading edge in the horizontal
4
direction, where ‘c’ is the local chord length at that wing station, the centroid of
the airfoil cross-section (o) and the shear center of the airfoil cross-section (sc).
Also there is a local coordinate system ( xi , j , yi , j , z i , j ) with its origin

−* −*
*
( X i, j , s η j , Z i, j ) located at the centroid of an arbitrary ‘Z’ stringer (i) in station (j)

as shown in Figure (31).

4-2-1 Calculation of Section Centroid:


Since the stringers are assumed to carry the bulk of bending in skin-stringer
sections, only the areas of the stiffeners are used in the calculation of the section
centroid.
Recalling Figure (13), the different dimensions of the stringer can be
calculated to maximize the moment of inertia of the stiffened panel using the
following empirical formulae
ba = 2.08t s + 068 if t s < 0.3 and ba = 1.312 if t s > 0.3
t a = 0.7t s
for equal flange atringers (58)
b f = ba and t F = ta

⎛b ⎞
bw = ⎜⎜ e ⎟⎟( Ast − 1.4ba t a )
⎝ ts ⎠
‘ Ast ’ is the stringer cross-section area, and it can be represented as

( ) ⎛ t ⎞
Ast = bw − t f t w + 2⎜ b f + w ⎟t f
2⎠
(59)

And ‘ be ’ is the effective width of the skin [5], where,

ηE sk
be = t s K (60)
σ sk
where ‘K’ is the skin buckling coefficient and it takes the values

48
bs bs
K = 3.62 for < 40 or K = 6.32 for > 110
ts ts
Between the above values there is a gradual transition. ‘η ’ in equation (60) is the
plasticity reduction factor which is determined using the following equation
Et sk
η= (61)
E sk

Where ‘ E sk ’ and ‘ Et sk ’ are the elastic and tangent modulii of the skin,

respectively while ‘ σ sk ’ is the skin axial stress.


For practical use, the design curves for the skin stringer panels can be used,
where
bs N
=1 for high value of load index
be L
(62)
bs N
= 1.1 : 1.3 for low values of load index
be L
Where ‘N’ is the axial load intensity and ‘L’ is the effective column length, or the
distance between two successive ribs.
The axial load intensity can be calculated as
Mb
N= cs hs
(63)

the cross-section area of the ith stringer at station j can be calculated as

(
Ast i , j = bwi , j − t f )t
i, j w i, j

+ 2⎜ b fi , j +

t wi , j ⎞
2
⎟t
⎟ fi , j
(64)
⎝ ⎠
For an airfoil cross-section with n-stringers, the centroidal coordinates will be
−*

n
−* A X i, j
i =1 sti , j
Xj = n
∑ Asti , j
i =1
(65)
−*

n
−* A Z
i =1 sti , j i , j
Zj = n
∑ Asti, j
i =1

49
−* −*
Where X i , j and Z i , j are the coordinates of of the ith stringer cross-section
centroid at the jth wing station.

4-2-2 Calculation of the Section Moments of Inertia


The inertia moment of the total section can be calculated from the following
equations
⎛ ⎛ ⎞ ⎞⎟
2
⎜ ⎜ −* * ⎟
n ⎜ −

I − = ∑i =1 ⎜ I − + Asti , j ⎜ Z i , j − Z j ⎟ ⎟
Xj ⎜ ⎟
⎜⎜ X i , j ⎜ ⎟ ⎟⎟
⎝ ⎝ ⎠ ⎠
⎛ ⎛ ⎞ ⎞⎟
2
⎜ ⎜ −*
n ⎜ −*⎟

I − = ∑i =1 ⎜ I − + Asti , j ⎜ X i , j − X j ⎟ ⎟ (66)
Zj ⎜ ⎟
⎜⎜ Z i , j ⎜ ⎟ ⎟⎟
⎝ ⎝ ⎠ ⎠
⎛ ⎛ ⎞⎛ ⎞⎞
⎜ ⎜ −* − ⎟⎜ − * − * ⎟⎟
I − − = ∑i =1 ⎜ Asti , j ⎜ Z i , j − Z *j ⎟⎜ X i , j − X j ⎟ ⎟
n

XjZj ⎜ ⎜ ⎟⎜ ⎟⎟
⎜ ⎜ ⎟⎜ ⎟⎟
⎝ ⎝ ⎠⎝ ⎠⎠
Where I − and I − are the local area moment of inertia of the ith stringer in
X i, j Z i, j

the jth station, with respect to its local centroidal axes. For ‘Z’ stringers, the
expressions for I − and I − are
X i, j Z i, j

I−
X i, j
=
1 ⎛⎜
6 ⎜⎝
ba j +
tw j ⎞ ⎛
2 ⎟⎠ a j 2 ⎜⎝ j
a
t ⎞
⎟t 3 + 1 ⎜ b + w j ⎟t b 2 + 1 t b − t
a w
2 ⎟⎠ j j 12 j
w wj fj ( )
3

2
(67)
3
I−
1 ⎛
= t a j ⎜ ba j +
6 ⎜⎝
tw j ⎞
2 ⎟⎠

j⎜ j
t w ⎞⎛ ba
2 ⎟⎠⎜⎝ 2
tw ⎞
⎟ + 2t a ⎜ ba + j ⎟⎜ j − j ⎟ + 1 bw − t f t 3
4 ⎟⎠ 12 j j wj
( )
Z i, j ⎝
It is noted that the local product moment of inertia is zero, since the ‘Z' stinger
has centroidal axes of symmetry.
Using these formulations the wing cross-section inertia properties can be
determined with respect to the wing coordinate system. It should be noted that

50
these inertia values put constraints on the orientation of the beam elements that
will be used to generate the stick model. In other words, the beam elements
which will be used to generate the stick model using the inertia values generated
by the analytical analysis, must be oriented along the wing coordinate system to
insure high fidelity stick model. Since the stick model is always generated along
the wing elastic axis (the stick model is represented by a set of beam elements
connecting the shear centers along successive wing stations which are not
oriented along the wing coordinate system), then using these analytical inertia
values to present section properties along the wing elastic axis will generate
approximate low fidelity stick model.
Table (5) provides the cross-section properties of the DLR-F6 aircraft
wing-box calculated using the analytical procedure described in this section. To
build the stick model of an aircraft wing-box, each wing bay is modeled by only
one bar or beam element representing the bay elastic axis with approximate
beam properties.
Following is the entry necessary for the CBAR card used to construct the stick
model in NASTRAN along with its PBAR card.

Where EID…element unique ID


PID…element property unique ID
GA, GB…grid point identification number, the two shear centers
identification number of the wing bay two stations.

51
X1, x2, x3…components of the element orientation vector
MID…material unique identification number
A…area of the element, equivalent area of the wing-box bay
I1, I2, J…equivalent bending stiffness in two planes, and torsional stiffness
of the wing-box bay
K1, K2…equivalent area factors for shear of the wing-box bay
Table (5) Cross-sections properties along DLR-F6 aircraft wing-box
Station S.C. (x,y,z) Ix Iz J Ixz A
1 (101.374107,68.212097,1.151940) 4.2767e3 3.4059e4 38e3 -2.8803e3 31.0505
2 (107.357979,88.957199,3.721518) 3.2794e3 2.8709e4 31e3 -2.1267e3 31.1960
3 (113.342010,109.702904,6.291171) 2.5021e3 2.3969e4 25e3 -1.5271e3 29.5022
4 (119.323738,130.440598,8.859831) 1.8439e3 2.0188e4 22e3 -1.0778e3 28.3718
5 (125.308037,151.187195,11.429596) 1.3849e3 1.7091e4 18e3 -674.1277 26.8366
6 (131.289825,171.925095,13.998281) 887.3440 1.2731e4 13e3 -236.1660 26.9245
7 (137.279602,192.690704,16.570400) 663.2558 1.1312e4 12e3 -59.1673 26.2704
8 (146.523422,213.498901,18.532312) 520.1212 8.6682e3 9e3 -16.3241 23.3872
9 (155.764862,234.301697,20.493717) 446.8150 7.5715e3 8e3 2.3469 20.9122
10 (165.011185,255.115494,22.456160) 371.0782 6.4595e3 6.7e3 6.6140 18.8147
11 (174.258926,275.932495,24.418901) 301.2099 5.2220e3 5.5e3 7.3983 16.3768
12 (183.503830,296.743134,26.381044) 236.3099 4.0677e3 4.2e3 4.3473 13.7504
13 (192.753647,317.564789,28.344229) 179.7233 2.9837e3 3e3 4.9814 11.6117
14 (202.000595,338.380005,30.306803) 140.4200 2.4172e3 2.5e3 12.9593 9.9506
15 (211.254105,359.209991,32.270771) 100.6141 1.6890e3 1.7e3 5.1014 8.9070
16 (220.503189,380.029999,34.233799) 77.7788 1.3394e3 1.34e3 5.1876 6.8692
17 (229.752258,400.850006,36.196823) 67.2339 1.1887e3 1.19e3 12.9359 5.7850
18 (239.005768,421.679993,38.160793) 53.4311 943.3881 0.14e3 17.6004 5.3652
19 (248.254852,442.500000,40.123821) 45.3523 863.1392 0.9e3 22.5334 5.0415
20 (257.486145,463.279999,42.083073) 35.0668 666.8249 0.7e3 23.9796 4.9649

The inertia values that are calculated in table (5) are employed to generate the
stick model of the DRL-F6 aircraft wing-box, as shown Figure (32).

52
Fig. (32) DLR-F6 aircraft wing-box stick model created in NASTRAN

The deformations of the generated stick model are compared with those of 3D
finite element model for the same load case, as shown in Figures (33-36).
Axial Deformations [inch]

Stick model (method I)


+ + + + 3D FEM

η
Fig. (33) Comparison between the axial deformation (y-direction) obtained from
3D FEM and the analytical stick model

53
Horizontal deformations [inch]

Stick model (method I)


+ + + + 3D FEM

η
Fig. (34) Comparison between the horizontal deformations (x direction) obtained
from 3D FEM and the analytical stick model
Twisting deformations [degree]

Stick model (method I)


+ + + + 3D FEM

Fig. (35) Comparison between the twisting angle deformations (around y)


obtained from the 3D FEM and the analytical stick model

54
Vertical deformations [inch]
Stick model (method I)
+ + + + 3D FEM

Fig. (36) Comparison between the vertical deformations (z direction) obtained


from the 3D FEM and the analytical stick model

As it can be realized, the stick model generated based on analytical equations (66-
67) is not accurately representing the 3D FEM. This is mainly due to the fact that
the torsional principal axis of the beam elements used to generate the stick model
does not coincident with the wing torsional elastic axis.

4-3 Wing Stick Model Extracted from the Full FEM


(method II)
The second methodology extracts the stiffness properties of the beam
model from the complete FEM of the wing-box. The method requires some
processing in PATRAN and NASTRAN and it requires that the user be
knowledgeable of both software packages. The stiffness extraction process begins
by the definition of the elastic axis at each wing station. At each wing station a
local coordinate system is created at the shear center along the elastic axis as well
as the axes of the principal inertia [8]. The limitation of this methodology is that
the analysis coordinate system is along the principal axes, which can not be

55
created unless an approximation is applied. This may generate undesirable
results in some cases.
The equivalent stiffnesses of the DLR-F6 wing-box, extracted using this
methodology, are already given in Eqs. (50-52) in section (3-5). Using these
relations, approximate equivalent principal moment of inertia can be obtained as:

1 (η j − η i )s
*

( I x )i → j =
(
E θ x j − θ xi ) (68)

1 (η j − η i )s
*

( J y )i → j =
(
G θ y j − θ yi ) (69)

1 (η j − η i )s
*

( I z )i → j =
(
E θ z j − θ zi ) (70)

The equivalent area and equivalent shear factors require further processing in
NASTRAN and PATRAN which is explained in the next section.

4-3-1 Equivalent EA’s and GK’s of the Wing-Box


The process of evaluating the equivalent area and shear factors span wise
distribution is different from that of the bending and torsional stiffness
evaluation. In this process NASTRAN is executed for three load subcases for each
wing-box bay. The three degrees of freedom related to rotation in all skin-
stringers connectivity grid points at all wing station are frozen. While the
translation degrees of freedom are kept free. The elastic axis nodes are connected
to the skin-stringer connectivity grid points by RBE2 elements with its dependent
degrees of freedom specified at the skin and its independent degrees of freedom
specified at the shear center grid points.
The first subcase, a unit force in the y-direction (the axial direction) is
applied at the bay shear center grid point. Thus the equivalent area can be
evaluated from the following equation as follows

1 (η j − η i )s
*

( A)i→ j =
(
E D y j − D yi ) (71)

56
In the second subcase, the structure is loaded by a unit load at the shear center in
the z-direction, to calculate the shear factor in the z-direction as follows

1 (η j − η i )s
*

(K1 )i→ j =
(
GA D z j − D zi ) (72)

Similarly, the structure is loaded by a unit force in the x-direction to calculate the
shear factor in the x-direction, as follows

1 (η j − η i )s
*

(K 2 )i→ j =
(
GA D x j − D xi ) (73)

where “D” denotes the displacement deformation in the relative direction. The
generated equivalent moments of inertia, equivalent cross-sectional areas and the
equivalent shear factors at each wing bay are summarized in Table (6).
The stiffness properties presented in Table (6) are used to generate the
stick model of the DLR-F6 wing-box. The generated stick model deformations are
compared with those of the 3D FEM for the same load case as shown in Figures
(36-39).

57
Table (6) PBAR card entries necessary to generate wing stick model
bay Ix Iz Jy A K1 K2
1 1.0e+003 * 4.2220 1.0e+004 *2.5894 1.0e+004 *1.0549 31.0505 0.2303 1.0244
2 1.0e+003 * 3.3290 1.0e+004 *2.1679 1.0e+004 *0.8692 31.1960 0.2423 1.0150
3 1.0e+003 * 2.5298 1.0e+004 *1.6881 1.0e+004 *0.6786 29.5022 0.2374 0.9807
4 1.0e+003 * 1.9032 1.0e+004 *1.3030 1.0e+004 *0.5353 28.3718 0.2356 0.8337
5 1.0e+003 *1.3495 1.0e+004 *0.8151 1.0e+004 *0.3961 26.8366 0.2311 0.8602
6 1.0e+003 *1.0322 1.0e+004 *0.7390 1.0e+004 *0.3099 26.9245 0.221 0.8203
7 1.0e+003 *0.7285 1.0e+004 *0.5305 1.0e+004 *0.2335 26.2704 0.2190 0.8243
8 1.0e+003 *0.5558 1.0e+004 *0.4432 1.0e+004 *0.1774 23.3872 0.2048 0.8379
9 1.0e+003 * 0.4592 1.0e+004 *0.3676 1.0e+004 *0.1477 20.9122 0.2075 0.8386
10 1.0e+003 *0.3814 1.0e+004 *0.3013 1.0e+004 *0.1213 18.8147 0.2056 0.8344
11 1.0e+003 *0.3071 1.0e+004 *0.2379 1.0e+004 *0.0958 16.3768 0.2014 0.8366
12 1.0e+003 *0.2367 1.0e+004 *0.1741 1.0e+004 *0.0716 13.7504 0.1935 0.8411
13 1.0e+003 *0.1850 1.0e+004 *0.1356 1.0e+004 *0.0553 11.6117 0.1879 0.8454
14 1.0e+003 *0.1440 1.0e+004 *0.1059 1.0e+004 *0.0431 9.9506 0.1890 0.8389
15 1.0e+003 *0.1156 1.0e+004 *0.0841 1.0e+004 *0.0348 8.9070 0.1827 0.8357
16 1.0e+003 *0.0812 1.0e+004 *0.0576 1.0e+004 *0.0233 6.8692 0.1792 0.8333
17 1.0e+003 *0.0610 1.0e+004 *0.0434 1.0e+004 *0.0177 5.7850 0.1788 0.8280
18 1.0e+003 *0.0497 1.0e+004 *0.0347 1.0e+004 *0.0145 5.3652 0.1790 0.7729
19 1.0e+003 *0.0406 1.0e+004 *0.0280 1.0e+004 *0.0119 5.0415 0.1814 0.7679
20 1.0e+003 *0.0329 1.0e+004 *0.0224 1.0e+004 *0.0098 4.9649 0.1728 0.7155

58
Axial deformations [inch]

Stick model (method II)


3D FEM

η
Fig. (37) Comparison between the axial deformations (y direction) obtained from
3D FEM and the stick model extracted from FEM model
Horizontal deformations [inch]

Stick model (method II)


3D FEM

η
Fig. (38) Comparison between the horizontal deformations (x direction)
obtained from 3D FEM and the stick model extracted from FEM model

59
Vertical deformations [inch]
Stick model (method II)
3D FEM

η
Fig. (39) Comparison between the vertical deformations (z direction) obtained
from 3D FEM and the stick model extracted from FEM model
Twisting deformations [degree]

Stick model (method II)


3D FEM

η
Fig. (40) Comparison between the twisting angle deformation (around y)
obtained from 3D FEM and the stick model extracted from FEM model

60
4-4 Wing Stick Model Generation Using Empirical
Relations (method III) [8]
The stiffness extraction method described in section (4-3) has been used to
extract the stiffness properties of four Bombardier wings. The results obtained
are employed to obtain empirical relations that can be used to estimate the
stiffness distribution of a new wing. These relations could also be used to
compare the stiffness behavior of newly designed wings with the existing ones or
check the stiffness predicted by other methods. In order to obtain the empirical
relations that apply to all wings, one has to normalize the data in order to
collapse the stiffness distribution of all existing wings into as tight a cloud of
points as possible. Several normalization techniques were investigated [8], and
the results of one of those normalization techniques were provided in section (3-
6). The empirical relations given in Eqs. (53)-(57) are used to generate the
stiffness properties of the DLR-F6 aircraft wing box.
Table (7) provides stiffness properties of the DLR-F6 wing-box obtained
using these empirical relations. The evaluated stiffness properties are used to
generate the stick model of the wing-box. The deformations experienced by this
stick model are compared to the deformations experienced by the 3D FEM model
of the wing-box as shown in Figures (41-44).

61
Table (7) stiffness properties of the DLR-F6 wing-box extracted using
empirical relations
bay Ix Iz Jy
1 1.0e+003 * 2.7969 1.0e+004 *2.6480 1.0e+003 *6.4023
2 1.0e+003 *2.5223 1.0e+004 *2.0401 1.0e+003 *4.6691
3 1.0e+003 *2.0480 1.0e+004 *1.5250 1.0e+003 *3.3121
4 1.0e+003 *1.5213 1.0e+004 *1.1006 1.0e+003 *2.2734
5 1.0e+003 *1.0362 1.0e+004 *0.7609 1.0e+003 *1.4980
6 1.0e+003 *0.6457 1.0e+004 *0.4991 1.0e+003 *0.9382
7 1.0e+003 *0.4484 1.0e+004 *0.3524 1.0e+003 *0.6344
8 1.0e+003 *0.4273 1.0e+004 *0.3148 1.0e+003 *0.5444
9 1.0e+003 *0.3642 1.0e+004 *0.2792 1.0e+003 *0.4645
10 1.0e+003 *0.3085 1.0e+004 *0.2455 1.0e+003 *0.3936
11 1.0e+003 *0.2595 1.0e+004 *0.2142 1.0e+003 *0.3312
12 1.0e+003 *0.2166 1.0e+004 *0.1851 1.0e+003 *0.2766
13 1.0e+003 *0.1793 1.0e+004 *0.1585 1.0e+003 *0.2290
14 1.0e+003 *0.1470 1.0e+004 *0.1342 1.0e+003 *0.1878
15 1.0e+003 *0.1194 1.0e+004 *0.1124 1.0e+003 *0.1524
16 1.0e+003 *0.0958 1.0e+004 *0.0930 1.0e+003 *0.1223
17 1.0e+003 *0.0761 1.0e+004 *0.0759 1.0e+003 *0.0969
18 1.0e+003 *0.0599 1.0e+004 *0.0610 1.0e+003 *0.0756
19 1.0e+003 *0.0466 1.0e+004 *0.0481 1.0e+003 *0.0581
20 1.0e+003 *0.0360 1.0e+004 *0.0373 1.0e+003 *0.0438

62
Axial deformations [inch]

Stick model (Method III)


+ + + + 3D FEM

Fig. (41) Comparison between the axial deformations (y direction) obtained from
the 3D FEM and the stick model based on empirical relations
Horizontal deformations [inch]

Stick model (Method III)


+ + + + 3D FEM

η
Fig. (42) Comparison between the horizontal deformations (x direction)
obtained from the 3D FEM and the stick model based on empirical relations

63
Twisting deformations [inch]

Stick model (Method III)


+ + + + 3D FEM

η
Fig. (43) Comparison between the twisting angle deformations (around y)
obtained from the 3D FEM and the stick model based on empirical relations
Vertical deformations [inch]

Stick model (Method III)


+ + + + 3D FEM

η
Fig. (44) Comparison between the vertical deformations (z direction) obtained
from the 3D FEM and the stick model based on empirical relations

64
4-5 New Approach for Generating Stick Models of
Complex Aircraft Wing-Box Structures Used for
Aerodynamics-Structures Interaction

Consider the fixed-free beam element shown in Figure (45), where the
coordinate system O p is located at its fixed root cross-section Center of Gravity

(CG) and its axes showing the principal directions of this beam, where x p and

z p show the two bending principal directions while y p shows the torsional

principal direction.

zp
yp
Op Tip

xp
Fig. (45) Schematic drawing of a fixed-free beam element with a coordinate
system at its fixed root Center of Gravity (CG) with three principal axes

To extract the inertial moments of this beam, the method described in


section (4-3) can be used, which will generate accurate results in this case. For
example, the second moment of inertia of this beam around the x p -axis can be

determined by applying a unit moment at its tip around x p and calculating the

corresponding angles of rotation of the tip CG, which will be a unique angle θ x p

around the x p since this is a principal direction. Thus, the moment of inertia

around the x p -axis can be calculated as

1 L
I xp = (74)

xp

Similarly, the moment of inertia around y p and z p can be calculated as

1 L
I zp = (75)
E θ
zp

65
1 L
J yp = (76)

yp

or the torsional moment of inertia can be calculated as


J yp = I xp + I z p (77)

Where G is the beam shear modulus of elasticity, E is the modulus of elasticity


and L is the length of the beam element and θ y p and θ z p are the rotational angles

around y p and z p respectively.

z
yp
o 'p Tip

Fig. (46) Schematic drawing of a fixed-free beam element with a coordinate


system at its fixed root CG with two general bending axes and a principal
torsional axis

Now, consider another analysis coordinate system o 'p with its axes (x, y p ,z)

where x and z are two general bending axes while maintaining the torsional axis
as principal axis. A bending moment acting on this beam element will generate a
normal bending stress which can be interpreted in the form of internal force per
unit length in the y p - direction as shown in Figure (47)

x yp

Fig. (47) Forces acting on a cross-section area of a beam subjected to bending

66
The moment balance between the internal forces and the external force generates
the following set of equations
E
Mx = (θ x I x + θ z I xz ) (78)
L
E
Mz = (θ z I z + θ x I xz ) (79)
L
Assuming that this beam element is loaded by a unit bending moment around the
x-axis, we can write
Mx =1 My =0 Mz = 0

and Eqs. (78) and (79) yield


E
1= (θ x I x − θ z I xz ) (a)
L
E
0= (− θ z I z + θ x I xz ) (b)
L
The reason of the negative sine that introduced in the θ z angle is that a positive
moment around the x-axis will generate a negative rotation angle around the z-
axis. Similarly a unit moment around the z-axis will make the load case as
Mx = 0 My =0 Mz =1

and Eqs. (78) and (79) yield

0=
E
L
(
− θ x' I x + θ z' I xz ) (c)

1=
E '
L
(
θ z I z − θ x' I xz ) (d)

The unknowns in these four equations are three second moment of inertia,
namely I x , I z and I xz , then equations (a), (b) and (d) can be arranged in matrix
form as
⎡1 ⎤ ⎡θ x −θz 0 ⎤⎡ I x ⎤
⎢0 ⎥ = E ⎢ 0 θx − θ z ⎥⎥ ⎢⎢ I zx ⎥⎥ (80)
⎢ ⎥ L⎢
⎢⎢1⎥⎥ ⎣⎢ 0 − θ x' θ z' ⎦⎥ ⎢⎢ I z ⎥⎥
or

67
−1
⎡ Ix ⎤ ⎡θ x −θz 0 ⎤ ⎡1⎤
⎢I ⎥ = ⎢ 0L
⎢ zx ⎥ E ⎢ θx − θ z ⎥⎥ ⎢0 ⎥
⎢ ⎥ (81)
⎢⎢ I z ⎥⎥ ⎢⎣ 0 − θ x' θ z' ⎥⎦ ⎢⎢1⎥⎥

where θ x and θ z are the angles of the rotation of the beam tip center of gravity

due to applying a unit moment at the tip around the x-axis while θ x' and θ z' are
the rotational angles due to applying a unit moment around the z-axis. The
torsional stiffness can be calculated from Eq. (77).

O Tip

y
x
Fig. (48) Schematic drawing of a fixed-free beam element with a coordinate
system at its fixed root CG with three general axes

Now, let us consider a general analysis coordinate system O, where non of its
axes, (x, y, z), is a principal axis as shown in Figure (48). In other words, a
moment acting around the x-axis will generate rotational angles around the x, y
and z axes, because that moment will have components around the beam three
principal directions. To calculate the bending stiffness with respect to this general
coordinate system, a new approach is developed.
The new approach assumes two axes in the general coordinate system as
pure bending axes while the third axis is the torsional principal axis. Assume a
moment M x is applied around the x-axis at the tip of the beam, then the tip point

of the beam will experience three rotational deformations θ x , θ y and θ z . Assume

that the original coordinate of the tip point with respect to coordinate system O is
( xt , yt , z t ) then its new coordinate after deformation can be calculated as

68
⎡ xt' ⎤ ⎡1 0 0 ⎤ ⎡ cos(θ y ) 0 sin(θ y ) ⎤
⎢ '⎥ ⎢ ⎥ ⎢ ⎥
⎢ y t ⎥ = C * ⎢0 cos(θ x ) − sin(θ x )⎥ * ⎢ 0 1 0 ⎥
⎢z' ⎥ ⎢⎣0 sin(θ x ) cos(θ x ) ⎥⎦ ⎢⎣− sin(θ y ) 0 cos(θ y )⎥⎦
⎢ t⎥ (82)
⎡cos(θ z ) − sin(θ z ) 0⎤ ⎡ xt ⎤
* ⎢⎢ sin(θ z ) cos(θ z ) 0⎥⎥ * ⎢⎢ yt ⎥⎥
⎢⎣ 0 0 1⎥⎦ ⎢⎢ z t ⎥⎥

where “C” is a correction factor to evaluate elastic deformations using rigid body
motion. It is found that “C=1/2”.
To illustrate this equation, a simple example will be introduced. Consider
the cantilever beam element illustrated in Figure (45). Assume that this beam
element is loaded in pure bending with a moment “M” around the x p -axis at its

tip. Thus we can write

d 2zp
EI x p =M (83)
dy 2p

Integrating Eq. (83) and applying the boundary conditions to this equation
generates the deformation equation of the beam as
1
EI x p z p = My 2p (84)
2
So the tip deflection at y p = L can be expressed as

1 ML2
z p ( L) = (85)
2 EI x p

Since the x p - axis is a principal axis, then the CG point of the tip will experience

one rotation angle of θ x p .

As shown in Figure (49) a point in the z p − y p plane with coordinate (x, y,

z) experiences rotation around the x-axis. The new coordinate ( x ' , y ' , z ' ) can be
obtained from simple transformation as

69
zp
P’(x’, y’, z’ )

P(x, y, z)

o yp

Fig. (49) Schematic drawing of rotational motion of a point in a plane

⎡ x ' ⎤ ⎡1 0 0 ⎤⎡ x⎤
⎢ '⎥ ⎢ ⎥⎢ ⎥
⎢ y ⎥ = ⎢0 cos(θ x ) − sin(θ x )⎥ ⎢ y ⎥ (86)
⎢ z ' ⎥ ⎢0 sin(θ ) cos(θ ) ⎥ ⎢ z ⎥
⎢ ⎥ ⎣ x x ⎦⎢ ⎥

For a point located at x p = z p = 0 and y p = L , after a rotation of θ x , it will go to

the coordinates
⎡ x ' ⎤ ⎡1 0 0 ⎤⎡0⎤
⎢ '⎥ ⎢ ⎥⎢ ⎥
⎢ y ⎥ = ⎢0 cos(θ x ) − sin(θ x )⎥ ⎢ L ⎥ (87)
⎢ z ' ⎥ ⎢0 sin(θ ) cos(θ ) ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎣ x x ⎦⎢ ⎥

For small rotation angles, cos(θ x ) ≈ 1 and sin(θ x ) ≈ θ x


Then, the change in the point location can be expressed as

z' = θxL (88)

where θ x can be expressed as

ML
θx = (89)
EI x
Substituting Eq. (89) into Eq. (88) yields
ML2
z' = (90)
EI x

70
Dividing Eq. (90) by Eq. (85) results in the ratio between the deformations
experienced by the rigid body motion and the deformations experienced by the
elastic motion of the beam element. Then the deformation of the beam can be
expressed as
⎡x' ⎤ ⎡1 0 0 ⎤⎡0⎤
⎢ '⎥ 1 ⎢ ⎥⎢ ⎥
⎢ y ⎥ = ⎢0 cos(θ x ) − sin(θ x )⎥ ⎢ L ⎥ (91)
⎢ z ' ⎥ 2 ⎢0 sin(θ ) cos(θ ) ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎣ x x ⎦⎢ ⎥

which is Eq. (82) for the special case of in-plane rotation about the x-axis.
Since the new approach assumes that the x and z axes are pure bending
axes, then applying a moment around a pure bending axis must not generate any
deformations around the torsional axis. In other words the degrees of freedom of
Eq. (82) can be reduced to only two degrees of freedom by equating θ y to zero.

Now, considering x and z as pure bending axes and y as torsional principal axis
and applying a moment around the x-axis, yields
⎡ xt' ⎤ ⎡1 0 0 ⎤ ⎡cos(φ z ) − sin(φ z ) 0⎤ ⎡ xt ⎤
⎢ '⎥ 1 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ y t ⎥ = * ⎢0 cos(φ x ) − sin(φ x )⎥ * ⎢ sin(φ z ) cos(φ z ) 0⎥ * ⎢ y t ⎥ (92)
⎢ z ' ⎥ 2 ⎢0 sin(φ ) cos(φ ) ⎥ ⎢ 0 0 1⎥⎦ ⎢⎢ z t ⎥⎥
⎢ t⎥ ⎣ x x ⎦ ⎣

Equating Eqs. (92) and (82) enables the calculation of φ x and φ z which are the
bending angles necessary to place the tip point in its location after beam
deformation. Repeating the calculations by applying the moment around the z-

axis will generate the angles φ x' and φ z' . Now with substituting φ x and φ z for θ x

and θ z and φ x' and φ z' for θ x' and θ z' in Eq. (81) we can obtain the second
moments of inertia.

71
4-5-1 Using the New Approach to Generate the Stick Model
of the DLR-F6 Wing-Box
Figure (50) shows different levels of structural modeling where the real
structure of the wing is modeled by a 3D finite element model. The 3D FEM is
then reduced to the stick model which is useful for multidisciplinary design
optimization and flutter analysis. The stick models are always generated through
the aircraft wing elastic axis.

Aircraft

Wing stick model

Wing 3D FEM

Fig. (50) Different levels of structural modeling

The high fidelity of the structural modeling means that the model is representing
the real structure with an acceptable tolerance error. In other words, an arbitrary
point Pr with coordinates ( x r , y r , z r ) on the elastic axis of the real aircraft wing
structure is represented by point Pfem of coordinates ( x fem , y fem , z fem ) on the

elastic axis of the 3D FEM and point Ps of coordinates ( x s , y s , z s ) on the stick

model. These three points Pr , Pfem and Ps must have the same coordinates with

respect to the same frame of reference when no load is applied. Under external

load these points transform to ( x r' , y r' , z r' ), ( x 'fem , y 'fem , z 'fem ) and ( x s' , y s' , z s' )

respectively and they should be equal with respect to the same frame of reference.

72
Extracting the proper stiffness properties from the 3D finite element
model and applying it with the right orientation to the stick model assures high
fidelity of the stick model.
The new approach is used to generate the stiffness properties of the DLR-F6
aircraft wing-box using its 3D FEM. The process needs some processing in
NASTRAN and PATRAN. Each wing bay is modeled by one beam element
extending along its elastic axis. The steps of extracting the stiffness properties of
each wing bay are summarized as
• A local coordinate system is created with its origin located at the shear
center of the wing bay station towards the tip. The z-axis of this local
coordinate system is placed along the elastic axis of the wing bay while the
x-axis is located in the chord wise direction and the y-axis is located in the
up direction.
• All the grid points at the bay station towards the wing root are fixed
• The analysis coordinate system of the shear center of the bay station
towards the tip is modified to be the local coordinate system generated in
step one, so the deformation angles experienced by the shear center can be
retrieved in the proposed local coordinate system.
• Three load cases are created in NASTRAN of unit moments around the x, y
and z axes and the corresponding rotational angles are obtained.
• Equations (82) and (92) are used to calculate the proper deformation
angles in the assumed orientation of the equivalent beam element (stick
model).
• The deformation angles obtained are used to calculate the stiffness
properties of the stick model in the assumed orientation from Eq. (1).

Table (8) provides the moments of inertia of the DLR-F6 aircraft wing-box
based on the proposed approach.

73
Table (8) Stiffness properties of the DLR-F6 wing-box extracted using
the proposed approach
Station Ix Iz Jy Ixz=I12 A K1 K2
1 6.4459e+003 3.3682e+004 4.0128e+004 -5.1023e+003 4.6286e+001 1.5273e-001 5.4340e-001
2 4.8661e+003 2.8821e+004 3.3687e+004 -4.0291e+003 4.5083e+001 1.6581e-001 5.4752e-001
3 3.5989e+003 2.3657e+004 2.7256e+004 -3.0167e+003 4.2674e+001 1.6231e-001 5.4802e-001
4 2.6330e+003 1.9530e+004 2.2163e+004 -2.2158e+003 4.0825e+001 1.6191e-001 5.4730e-001
5 1.6750e+003 1.5627e+004 1.7302e+004 -8.9016e+002 3.7393e+001 1.6499e-001 5.9805e-001
6 2.6318e+003 1.1176e+004 1.3807e+004 3.6839e+003 3.2634e+001 1.9079e-001 5.6987e-001
7 9.4449e+002 1.0636e+004 1.1581e+004 -6.3648e+002 3.8255e+001 1.4898e-001 5.3243e-001
8 7.6597e+002 8.2006e+003 8.9666e+003 -4.8764e+002 3.2187e+001 1.4803e-001 5.7027e-001
9 6.3685e+002 6.9971e+003 7.6340e+003 -4.1283e+002 2.9415e+001 1.4672e-001 5.5611e-001
10 5.2715e+002 5.9276e+003 6.4548e+003 -3.3987e+002 2.6813e+001 1.4346e-001 5.4531e-001
11 6.8708e+002 6.1681e+003 6.8552e+003 -8.3770e+002 3.3404e+001 9.8483e-002 4.0342e-001
12 3.2606e+002 3.7539e+003 4.0800e+003 -1.9538e+002 2.0101e+001 1.3174e-001 5.3482e-001
13 2.5122e+002 2.8852e+003 3.1364e+003 -1.5099e+002 1.7108e+001 1.2674e-001 5.3273e-001
14 1.9586e+002 2.3165e+003 2.5124e+003 -1.1977e+002 1.4965e+001 1.2488e-001 5.1638e-001
15 1.5540e+002 1.8722e+003 2.0276e+003 -8.9971e+001 1.3489e+001 1.1989e-001 5.1073e-001
16 1.0962e+002 1.3451e+003 1.4547e+003 -6.6677e+001 1.0687e+001 1.1441e-001 4.9429e-001
17 8.2216e+001 1.0331e+003 1.1153e+003 -4.5459e+001 9.2060e+000 1.2074e-001 4.8233e-001
18 6.7170e+001 8.5472e+002 9.2189e+002 -3.1691e+001 8.6849e+000 1.1776e-001 4.7454e-001
19 5.5144e+001 7.3826e+002 7.9341e+002 -2.1855e+001 8.4451e+000 1.1566e-001 4.5627e-001
20 4.3715e+001 6.1898e+002 6.6269e+002 -1.1611e+001 7.9225e+000 1.1669e-001 4.5498e-001
station Beam orientation vector
1 <-0.13101196 -0.45419693 3.9719698>
2 <-0.65505981 -2.2709808 19.859846>
3 <-0.49129486 -1.7032318 14.894886>
4 <-0.49129486 -1.7032318 14.894885>
5 <-0.29706573 -3.0567245 24.810644>
6 <4.4569626 -3.8405609 13.798026>
7 <-0.49128723 -1.7032318 14.894885>
8 <-0.27882385 -0.62763977 7.9704666>
9 <-0.27882385 -0.62763977 7.9704666>
10 <-0.24397278 -0.54917908 6.9741573>
11 <-0.27882385 -0.62763977 7.9704666>
12 <-0.17427063 -0.39227295 4.9815407>
13 <-0.1394043 -0.31381226 3.9852333>
14 <-0.1394043 -0.31381226 3.9852314>
15 <-0.24397278 -0.54919434 6.9741573>
16 <-0.1394043 -0.31381226 3.9852333>
17 <-0.24397278 -0.54919434 6.9741592>
18 <-0.348526 -0.7845459 9.9630814>
19 <-1.394104 -3.1381836 39.852329>
20 <-0.1394043 -0.31381226 3.9852333>

74
These stiffness properties are used to generate the stick model. The performance
of this stick model is compared with the performance of the 3D FEM for the same
load case and are shown in Figures (51-54).

Axial deformations [inch]

Stick model (proposed)


+ + + + 3D FEM

η
Fig. (51) Comparison between the axial deformations (y direction) obtained from
the 3D FEM and the stick model based on proposed method
Horizontal deformations [inch]

Stick model (proposed)


+ + + + 3D FEM

Fig. (52) Comparison between the horizontal deformations (x direction) obtained


from the 3D FEM and the stick model based on proposed method

75
Twisting deformations [degree]

Stick model (proposed)


+ + + + 3D FEM

η
Fig. (53) Comparison between the twisting angle deformation (around y)
obtained from the 3D FEM and the stick model based on proposed method

Stick model (proposed)


+ + + + 3D FEM
Vertical deformations [inch]

η
Fig. (54) Comparison between the vertical deformations (z direction) obtained
from the 3D FEM and the stick model based on proposed method

76
As it can be realized excellent agreement exists between the proposed stick model
and the full 3D FEM model.

4-5-2 Comparison Between the Performance of the


Stick Models (SM) Generated by the Different
Methodologies and the 3D FEM
Deformations obtained by stick models based on methods I-III and the
proposed stick model and also the 3D FEM are shown in Figures (55-58) for easy
comparison. As it can be realized the results obtained from the proposed stick
model agrees very well with those of the 3D FEM confirming its high fidelity. The
developed stick model is efficient and accurate enough to be used in
multidisciplinary design optimization
Axial deformations [inch]

SM (method II)
+ + + + 3D FEM
SM (proposed approach)
. . . . . . . . . . . SM (method I)
SM (Method III)

Fig. (55) Comparison between the axial deformations (y direction) obtained from
the 3D FEM and the stick Models

77
Horizontal deformations [inch]

SM (method II)
+ + + + 3D FEM
SM (proposed approach)
. . . . . . . . . . . SM (method I)
SM (method III)

Fig. (56) Comparison between the horizontal deformations (x direction) obtained


from the 3D FEM and the stick Models
Twisting deformations [degree]

SM (method II)
+ + + + 3D FEM
SM (proposed approach)
. . . . . . . . . . . SM (method I)
SM (method III)

Fig. (57) Comparison between the twisting angle deformations (around y)


obtained from the 3D FEM and the stick Models

78
SM (method II)
Vertical deformations [inch]

+ + + + 3D FEM
SM (proposed approach)
. . . . . . . . . . . SM (method I)
SM (method III)

Fig. (58) Comparison between the vertical deformations (z direction) obtained


from the 3D FEM and the stick Models

5- Conclusion
In task (6) of the MOSAIC project a methodology of the design optimization of
aircraft wing-box is developed. Optimization of all components of the wing-box is
taken into account. The objective function is the mass minimization while
different constraints are considered for each wing-box component. The
optimization routines are created in MATLAB. The sized dimensions are used to
generate a 3D FEM of the whole wing-box. PATRAN is used as modeler while
NASTRAN is used as analyzer. The performance of the 3D FEM is tested for
normal cruising conditions load case which generated a reasonable deformations
compared with deformations experienced by previously designed aircrafts. Also
the 3D FEM is tested for margin of safety which showed a conservative design.
The process of generated wing stick model is investigated. A new approach for
generating high fidelity stick models is introduced which showed excellent results

79
compared with stick models generated using the previous methodologies
available in literature.
The stick model generated is ready for aero-structure interaction which is
necessary for the multi-disciplinary design optimization process.

6- Future Work
Although the objectives of the project have been achieved, there are important
tasks which will be studied in the future
1. Developing aerodynamic model and its integration with the generated
stick model.
2. Flutter analysis and design optimization for the whole wing-box under
flutter constraint.

80
References
1. http://aaac.larc.nasa.gov/tsab/cfdlarc/aiaa-dpw, 3rd AIAA CFD Drag
Prediction Workshop, San Francisco, 2006.
2. Kuhn P., Peterson J. and Levin L., “A summary of diagonal tension, Part1-
methods of analysis,” NACA Technical Note 2661
3. Bruhn E.F., Analysis and Design of Flight Vehicle Structures, Jacobs &
Associates Inc., June 1973
4. Niu M., Airframe Stress Analysis and Sizing, Hong Kong, Conmilit Press
Ltd., 1997.
5. Abdo M., Piperni P., Isikveren A.T., Kafyeke, F., “Optimization of a
Business Jet,” Canadian Aeronautics and Space Institute Annual General
Meeting, 2005.
6. Arora, J.S., Introduction to optimum design, McGraw-Hill, New York,
1989.
7. http://www.mscsoftware.com/support/online_ex/Library.cfm
8. Abdo M., L’Heureux R., Pepin F. and Kafyeke F., “Equivalent Finite
Element Wing Structural Models Used for Aerodynamics-structures
Intraction”, Canadian Aeronautic and Space Institute 50th AGM and
Conference, 16th Aerospace Structures and Materials Symposium 28-30
April 2003.
9. Abdo M., Piperni P., Kafyeke F., “Conceptual Design of Stringer Stiffened
Compression Panels”, Canadian Aeronautic and Space Institute 50th AGM
and Conference, 16th Aerospace Structures and Materials Symposium 28-
30 April 2003.

81

You might also like