You are on page 1of 9

Composite Structures 57 (2002) 415423

www.elsevier.com/locate/compstruct

Parametric optimisation of composite shell structures


for an aircraft Krueger ap
Javid Bayandor a, Murray L. Scott a,*
, Rodney S. Thomson b

a
The Sir Lawrence Wackett Centre for Aerospace Design Technology, Department of Aerospace Engineering,
Royal Melbourne Institute of Technology, G.P.O. Box 2476V, Melbourne, Victoria 3001, Australia
b
Cooperative Research Centre for Advanced Composite Structures Limited, 506 Lorimer Street, Fishermans Bend, Victoria 3207, Australia

Abstract
This paper details the conceptual design optimisation of the conguration and composite lay-ups used to replace the conven-
tional honeycomb stiened structure of a Krueger ap. The multiple composite laminates selected for redesigning the lay-ups within
an initial symmetrical quasi-isotropic ply conguration of [0/45/45/90]s , had to demonstrate full orthotropic characteristics. In
order to construct a numerical process to optimise the required multi-layered composite shells, a commercial nite element code,
Ansys, was used to develop a parametric analysis le. This analysis subroutine was then integrated into an Ansys Parametric Design
Language code embedding the objective of the optimisation processmass minimisationas well as all the constraints and the
allowable domains of the parameters. The paper, in its conclusion, presents a comparison between the original product and the
optimal design, and reviews the advantages of the future implementation of this design.
2002 Elsevier Science Ltd. All rights reserved.

Keywords: Composite shells; Ply optimisation; Krueger ap; Stiened panels; Conceptual design

1. Introduction strains and the shear stresses also promote only shear
strains. In orthotropic materials, the response directions
In a number of previous case studies, the design of the are representative of the axes of symmetry, the number
primary and secondary structures of commercial aircraft and the directions of which are normally controlled by
has been carried out using isotropic and anisotropic the microstructure of the material. As an example out-
materials. Recent increasing demands within the avia- lining the complexity involved in the orthotropic struc-
tion industry however, aim to benet from the sub- tures versus isotropic ones, a simple unidirectional
stantial mass savings and the high structural integrity composite can be considered. This simple structure has
oered by utilising advanced composite materials in the orthotropic property with more than three axes
commercial eets. It has therefore become exceedingly of symmetry including the longitudinal axis and all
important for manufacturers to account for orthotropic other directions perpendicular to it. Therefore, due to
properties of such material and the challenges they the existence of these additional axes of symmetry,
present during the optimisation process. unidirectional composites are only transversely isotropic
One such challenge posed by these characteristics [1] requiring an extensive series of modications within
includes the deformation response of orthotropic ma- the nite element (FE) and optimisation codes for the
terials, which has the following signicance in compar- structural analysis to hold.
ison to those with isotropic properties: the general Due to such complications, the experimental investi-
response in orthotropic materials is direction dependant, gations of particular responses of aircraft orthotropic
and normal and shear stresses instigate various combi- structures to external loading conditions, as opposed to
nations of normal and shear strains. In isotropic mate- those of the isotropic ones, are considered a high add-on
rials however, the normal stresses only produce normal cost to production of optimised aircraft components.
This however, is in contrast with the advantages that
applications of advanced composite materials in aero-
*
Corresponding author. Tel.: +61-3-9925-8064. space structures oer. Thus, the present study is aimed
E-mail address: m.scott@rmit.edu.au (M.L. Scott). at introducing an automated numerical algorithm with
0263-8223/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 3 - 8 2 2 3 ( 0 2 ) 0 0 1 0 9 - 5
416 J. Bayandor et al. / Composite Structures 57 (2002) 415423

which the process of optimisation of such orthotropic


multi-layered structures can be facilitated, and the cost
associated with the required extensive testing reduced.

2. Krueger ap

The Krueger ap is a deployable leading edge lifting


component of the wing. It is located between the rst
engine and the root of the wing on each side of the fu-
selage in heavy transport aircraft. When idle, it rests on
the front part of the lower surface of the wing, con-
nected to the primary structure of the wing via two al-
uminium gooseneck ttings and their hinges (Fig. 1).
The gooseneck ttings are positioned on each side of the
middle transverse rib. On activation, the ap rotates Fig. 2. Modelled load cases for opening ap (Case 1) and fully de-
about the gooseneck hinges and forms an extension to ployed ap with the applied point loads (Case 2).
the upper surface of the wing with its bull nose creating
a new leading edge. tribution prole over the ap surface, where the
In conventional designs of Krueger aps, a honey- deployed structure had a linear pressure prole (Fig. 2).
comb stiened skin is supported by ribs and spars. This This pressure was applied over the ap surface in ad-
type of conguration results in low impact tolerance of dition to the point loads imposed on the ap by the
the structure and, over time, possible delamination of leading edge bull nose.
the skin at points of contact.
On the basis of the demands expressed by industry to
3.1. Procedures map
overcome these problems, a large commercial transport
aircraft Krueger ap was chosen to be optimised and
Modelling displacements, applying constraints and
reproduced entirely from orthotropic laminated struc-
pressure distributions to the model, calculating the
tures. The composite laminates were to replace the
failure indices and failure criteria within the layer ele-
honeycomb structures of the ap bays as well as the
ments, and optimising the laminates for a given design
aluminium ribs and spars.
objective were carried out using the pre- and post-pro-
cessing modules of the FE software Ansys. Parametric
solutions for global stiness and buckling eects, given
3. Modelling and numerical procedures the predetermined material properties and subject to
volume reduction were then obtained through the Ansys
In modelling the ap, idealised ttings were assumed solver and post-processing modules, providing the sim-
to be the sole bearers for bringing the actuator loads ulation with the required initial sets of data for pursuing
into the ap structure. the optimisation phase. This phase, based on predened
Two critical scenarios were considered for the model: sets of analysis variables and objectives, enabled the
opening ap and fully deployed ap. The load case de- determination of a series of feasible responses within the
ned for the opening ap was a uniform pressure dis- design domain of the problem. Of the solutions ob-
tained, the set with the lowest objective function value
that most closely met all the constraints, represented the
optimised shape for the composite structure lay-up of
the ap. The process also identied the infeasible an-
swers which violated the boundaries of the design space
as, due to the nature of the problem, some of the re-
sponse sets occurred outside the design domain.

3.2. Optimisation method

One of the two optimisation methods that Ansys


provides is a procedure based on the Ansys Parametric
Design Language (APDL). All the user dened para-
Fig. 1. Layout of the Krueger ap and its gooseneck ttings. meters or functions introduced in this module using
J. Bayandor et al. / Composite Structures 57 (2002) 415423 417

APDL commands, according to the objectives of the


problem, can be constrained or minimised. In this
method, constrained problems are transformed into
unconstrained ones, which are to be minimised in con-
verging processes. The optimum design, which is the end
result of such processes, has to meet all the predened
requirements with a minimum expense of certain factors
such as mass, dimensions, shape and volume, positions
of supports, stress or strain, cost and so on, making the
design as eective as possible [2].
The standard optimisation techniques used in this
module consist of a number of design tools which do not
directly perform the minimisation, but only oer an al-
ternative means for the system to identify the feasible
domain and recognise the characteristics of the prob-
lems variables. These tools include: single loop analysis, Fig. 3. Topology optimisation resultslight areas indicate where
random, sweep, factorial and gradient tools as well as, material is needed.
subproblem approximation and the rst-order optimi-
sation methods.
The subproblem approximation method, due to its feeding of the optimisation methods with the calculation
independency from using derivatives of the problem results of randomly selected initial data which serve as
variables, is the rst candidate used in the optimisation starting points for the optimisation processes.
subroutine. This advanced zero-order tool replaces the
dependant variables through the least square tting 3.3. Topology optimisation
approximation process. The constrained problem is then
converted to an unconstrained problem using penalty The early design of the ap constituted a heavy sec-
functions (the subproblem). The penalised functions are ondary wing structure. Therefore, an overall revision of
then minimised until the convergence is reached or the the topology of the component was required prior to
iterations are terminated. performing any lay-up optimisation on the ap.
The rst-order method is based on design sensitivities Using Ansys, topology optimisation was performed
and provides high accuracy. In this method, the problem on a grid stiened conguration of the ap determining
is again converted to an unconstrained one by adding the primary load paths under the dened load cases. The
functions to the objective function. However, unlike the results suggested an initial conguration consisting of
subproblem method, it is the actual FE representation two to three spars and three ribs, by revealing the areas
of the problem that is minimised, not an approximation. within which material was required (Fig. 3).
The method uses gradients of the dependant variables. These results were used as the basis for the parametric
In each iteration, using a steepest descent or conjugate design, to optimise the structural conguration, and the
direction method, a search direction is selected and the later lay-up optimisations.
unconstrained problem is minimised along the specied To carry out the parametric design optimisation, the
direction [2]. As is expected, each iteration consists of a thicknesses of the skin, thicknesses and positions of the
large number of subiterations carrying search directions ap ribs and spars, as well as possible tapering of
as well as gradients. Hence, the rst-order method for the spars height and width were considered as design
optimisation performs several analysis loops and re- variables (DVs). Constraints were the maximum de-
quires substantially longer CPU time in comparison to ection, given strain allowables, and a buckling factor,
that required by the subproblem method. versus minimised mass dened as the objective function.
Both of the optimisation methods perform a series of The optimisation enabled determination of the opti-
analysisevaluationmodication loops. They start by mum stiening arrangement as well as optimum thick-
analysing the initial design, proceed to performing nesses of the modelled ap.
evaluation of the results against specic design criteria,
and nally modifying the design as required. This multi- 3.4. Initial parametric model
functional cycle is repeated until the design specica-
tions and criteria are met. The composite lay-up of the ap (Fig. 4) was eec-
Additionally, Ansys provides a number of tools ca- tively optimised in two stages. In the rst stage, the ply
pable of increasing the eciency of the above mentioned stacking sequence was determined where the second
optimisation methods; as such, the random design iter- stage dealt with optimising the ply thicknesses for a
ation technique can be named. This technique enables given stacking sequence.
418 J. Bayandor et al. / Composite Structures 57 (2002) 415423

In the initial modelling stage, the orthotropic skin


model was originally set up with Shell99 elements, while
Shell63, a lower order nodal element type, was used to
mesh the spars, ttings and ribs. Consequently, prob-
lems were encountered in matching the mid-side nodes
between the two element types. This was due to the fact
that the removal of the mid-side nodes from Shell99
elements, as permitted in other element types, was not
possible.
To overcome this problem, the way and order in
which the structure was meshed became critical. To
ensure that the nodes of the two elements coincided at
the right locations, dierent element sizes for the con-
cerned regions were dened. Alternatively, where re-
Fig. 4. Skin lay-up of the modelled Krueger ap. quired, the Shell93 elements replaced the Shell63 type
elements providing the equivalent mid-side nodes for the
matching regions of the model. Due to the application
To simulate the complex reaction of the composite of a higher order element type however, this method
skin to external pressure elds, Ansys Shell99 option, increased the computer resource requirements consid-
covering the principles of the linear layered structural erably, which has the potential to adversely aect the
shells, was used as the primary candidate. This option accuracy of the solution in extreme cases [4].
can perform analysis on up to 250 constant thickness or The real attributive constants (properties that depend
125 tapered layers, sucient to correspond with the on the element type) for Shell99 are the key elements for
projects manufacturing requirements. Nevertheless, in constructing the multi-layered composite skin. The ini-
special cases, where greater numbers of layers are re- tial number of constants supplied for the analysis was
quired, user input constitutive matrices can overwrite 12 3  NL, with NL 8, where NL represented the
the analysis input. number of plies used to construct the laminate [5]. For
Each Shell99 element used in the analysis has six each element the real constant table was specied con-
degrees of freedom at each node constituting x, y and z stituting the material property, ply orientation angle and
directional nodal translations, as well as rotations about layer thickness. For orthotropic materials, densities are
the nodal x, y, and z axes. The element is fully dened dened by the software. As their material properties are
by its eight nodes (mid-side and corner nodes), average direction dependant, they had to be applied in the model
or corner layer thickness, orthotropic material proper- such that the required properties face the correct direc-
ties and its ply orientations. For the initial parametric tions. To ensure this, the element coordinate system was
run, as the lay-up material was orthotropic, it was re- used, rather than the default system. The local coordi-
quired to have the values of Youngs modulus in all nate system was created such that its x-axis was aligned
three coordinate system directions, the three dimen- with the bre direction of the skin.
sional Poissons ratio, the shear modulus, as well as the For the ap composite skin lay-up optimisation, the
3-D failure criteria dened using the manufacturers ply conguration was dened layer by layer from the
product specications. The 3-D failure criteria were bottom surface of the component (layer 1) to the top one
limited to ve main strengths in the principal direc- (layer 8), with additional layers stacked in the same as-
tions of the material including the longitudinal tensile cending order from the bottom to the top in the positive
strength, transverse tensile strength, shear strength, lon- normal direction of the element coordinate system. The
gitudinal compressive strength, and transverse compres- numbering sequence of the lay-up was found dependant
sive strength [3]. Also, the number of plies, their on the surface normal of the elements. Where the surface
symmetry status and orientations were to be provided normal faced outwards with reference to the structure,
from the earlier optimisation phase. This information the bottom layer became the designated layer 1, and the
was added to the program using constitutive matrices layer numbers increased as the layers approached the
(Keyopt options) that relate generalised forces and mo- top layer. If the surface normal was facing the structure
ments to generalised strains and curvatures. The con- however, the opposite held. Therefore, if two elements
stitutive matrices allow for incorporating an aggregate had an opposing element normal, their respective lay-
composite material behaviour, supplying a thermal load ering sequence would then be in a reverse orderthis
vector if required and, dening an unlimited number of was avoided using the Enorm command only when the
layers. However, the matrix approach does not provide elements were located in the same plane. Nevertheless,
detailed results in each ply, as individual layer details are in a symmetrical lay-up, where the thicknesses above
not specied. and below the centreline are optimised to the same
J. Bayandor et al. / Composite Structures 57 (2002) 415423 419

sign space of the problem. This le was integrated with


the analysis code.
In the optimisation module, the DVs and the state
variables (SVs), as well as the objective function were to
be declared within the Opt processor [7]. The unique
dependant variable, the optimisation objective function
to be minimised during the optimisation loop, for the
ap design was chosen to be minimised volume, as vol-
ume is directly proportional to the mass of the structure.
Therefore, minimising the total volume, in eect, repre-
sented the mass minimisation for the composite ap.
The required DVs were assigned to the individual ply
thicknesses of the multi-layered structure with both their
upper and lower limits derived from extreme experi-
Fig. 5. Initial layer stacking sequencethe hatch lines are indicative of mental values. As the lay-up was symmetrical about the
the orientation angle in each ply. centreline, the DV count reduced from eight to four. For
each value, a convergence tolerance of 0.005 was as-
values, this does not seem to create a concern. However, sumed.
even for such cases, care must be taken to ensure the For this case, the SVs, being response quantities,
accuracy of the results. which are functions of the DVs, were selected to be the
During the optimisation process, an initial symmet- stresses limited by the maximum stress failure criteria in
rical ply stacking sequence of [0/45/45/90]s was used each ply. Each criterion was required to be less than 1.0
followed by further alternate sequencing (Fig. 5). This for the given limit loadings. Deection allowables and
allowed the embedment of laminate symmetry in the eigenvalues identifying the buckling thresholds com-
design, reducing the need for investigating a large prised other SV parameters.
number of coupling eects exhibited by composite The maximum global deections and the lowest
structures. Of these eects, coupling between twisting buckling loads used ensured that the entire ap could
and bending or bending and extension can be named, not deect beyond 6.6 mm, and buckling did not
which can be produced by unsymmetric laminates. occur below the limit load.
After producing the geometry of the ap encom- The optimisation commands were executed using
passing the composite shell lay-up, the design was me- both the subproblem and the rst-order method sub-
shed using eight-noded quadratic elements [6]. routines. This was conducted by varying the input
For pursuing the analysis, it was critical to have the within the range of given extremes to achieve the opti-
failure criteria established for the composite layers, mum design.
constituting the structure of the ap. The maximum The DV parameters were initially introduced within
stress failure criterion, with nine specied failure stres- the analysis le. The initial values represented a poten-
ses, was applied to the problem which showed whether a tial design solution for the starting point of the analy-
certain layer had failed under external applied loadings. sis and assumed dierent values as the optimisation
Since the failure criteria are orthotropic, the failure progressed. In every stage of the solution, the failure
stress and strain values had to be provided for all di- indices for the maximum stress failure criteria were
rections. Introducing the constraints of the problem and calculated at the top and the bottom of each laminate at
the pressure distributions over the structure allowed the all of the aps in-plane integration points. The maxi-
simulation to proceed with nalising the parametric re- mum stress endured in dierent directions within each
sponse. layer was evaluated in these points by using up to six
Following the solution, in the post-processing module, criteria (Table 1).
the required parametric responses of the system, such as The stored optimisation database in each step of this
maximum stresses and strains, and the required tabular process contained the optimisation environment, which
and graphical result sets, were extracted and viewed. This included the optimisation variable denitions, the pa-
concluded the cluster of commands in a batch run com- rameters, all the optimisation specications, and the
prising the analysis le, saved under inp (or ans) exten- accumulated design sets. Executing the command layer
sion, and used during the optimisation process. with a non-zero layer number before the command
Rsys, Solu [2] brought up the result display in the layer
3.5. Ply optimisation coordinate system. Up to a total number of 130 opti-
mised data sets could be saved by the optimiser and
At this stage, the parametric model was solved using displayed. Any data set beyond 130 replaced the worst
an ADPL optimisation le developed to embed the de- design set.
420 J. Bayandor et al. / Composite Structures 57 (2002) 415423

Table 1
Maximum stress failure criteria
rxt rxc
or f whichever applicable
rxtf rxc
ryt ryc
or f whichever applicable
rytf ryc
rzt rzc
fcmax max of f
or f
whichever applicable
r
rzc
zt
rxy
f Fig. 6. Plan view of the panel assembly of the Krueger ap concept.
rxy
ryz
rfyz
jrxz j
rfxz

rit the greater of 0 or ri (i-directional stress in a layer) and i x, y, z.


ric the lesser of ri or 0 and i x, y, z.
rfit i-directional failure stress in layer under tension.
Fig. 7. Cross-sectional details of the Krueger ap.
The analysis was continued until either the results
converged or the optimiser reached the maximum
number of loops allowed by the program. Encountering properties throughout. Figs. 6 and 7 display the plan
the maximum number of sequential infeasible answers and the cross-sectional views of the concepts of the ap
could also terminate the analysis. structural conguration consisting of a number of DVs
used to construct the parametric model. Two concepts
were considered: blade stiened spars and C-spars. The
4. Results and discussion included the positions of the stiening elements (Fig. 6)
and the dimensions and thicknesses of the skin and
At the end of the optimisation process, the result sets, stieners.
a combination of feasible and infeasible data, were lis- Table 2 provides an overview of the component mass
ted. The best design had to satisfy all the constraints as breakdown of the Krueger ap C-spar concept versus
well as producing the minimum volume. Where all sets the blade stiened concept. Note that the leading and
were infeasible, the best design set was declared as the the trailing edge spars consist of the web and the upper
one closest to the frame of the design space, irrespective ange, whilst the lower ange masses are included in the
of the value of the volume. skin. From Table 2, it is evident that the C-spar model
Deriving the maximum indices in each ply, the layer provides a more ecient design compared to that of the
in which the material failure would occur was identied. blade stiened spars, with a 2.0 kg mass saving. Hence,
Accordingly, the thickness of each laminate was opti- the subsequent optimisation of the aerodynamic skin
mised to withstand the ultimate loads and avoid fail- was performed on the former, using an APDL code
uretaking into account the pre-dened allowables and based analysis in order to replace the quasi-isotropic
the boundary condition constraints. This was carried laminates of the skin with a fully optimised, laminated,
out using the Etable command, which retrieved the re- composite lay-up.
quired values from the calculation and assigned them to In arriving at the indicated solutions, it was found
the dened parameters such as maximum failure indices. that Ansys could not eectively handle a high number of
In this case, the Etable command operated in conjunc- DVs. Solutions to parametric models with over 12 DVs
tion with the programs Shell option, capable of storing often led to convergence problems.
the corresponding results from the bottom of the bot-
tom layer, the top of the top layer and the layer with the Table 2
maximum failure criterion value. Mass breakdown for main components in blade stiened spar concept
This enabled undertaking of a number of design versus C-spar concept
studies that allowed the work to establish the governing Component Blade stiened spar C-spar conguration
eects of the layer thickness and the lay-up sequencing conguration (kg) (kg)
on the structural behaviour of the ap. Leading edge spar 1.30 2.84
Mid spar 2.06 1.54
Trailing edge spar 2.85 3.29
4.1. Mass optimisation
Mid-rib 0.18 0.16
Aerodynamic skin 15.2 11.8
Prior to performing the ply optimisation, the struc-
Total 21.6 19.6
tural lay-out was optimised assuming quasi-isotropic
J. Bayandor et al. / Composite Structures 57 (2002) 415423 421

4.2. Eect of layer thickness Table 4


Eect of layer sequence
An understanding of the eects of load transfer Lay-up Umax1 Umax2 Unmax1 Unmax2
through the orthotropic material was gained by keeping sequence (mm) (mm) (mm) (mm)
the lay-up sequencing unchanged, while varying the [90/0/45/45]s 6.83 4.06 2.77 7.85
[90/45/45/0]s 6.83 3.79 2.49 8.69
thickness of the plies in each direction. The optimised
[90/0/45/45]s 7.29 4.01 2.64 9.30
layer thicknesses, calculated using this approach, can be [0/90/45/45]s 8.76 3.94 2.57 8.20
found in Table 3. Note that Umax and Unmax are the [45/45/90/0]s 9.22 3.40 1.75 8.81
structures maximum positive and negative deections, [45/45/0/90]s 10.74 3.51 2.08 8.26
respectively. The numbers correspond to each of the [45/45/0/90]s 10.90 3.25 1.55 9.98
[0/45/45/90]s 15.27 3.51 2.06 11.25
considered load cases.
Run A was an initial optimised conguration, whilst
the other runs were created by arbitrarily swapping the
thicknesses of each ply so that the total thickness was
of [45/45/0/90]s did not exhibit as high structural e-
constant. The initial design, Design Set A, produced the
ciency as that of [90/0/45/45]s . Accordingly, since the
lowest deections for the given mass. Subsequently, in-
design was being driven by deection, the 90 plies,
creasing the bre thickness in the 0 or 45 directions
which provide the highest overall stiness, were placed
and reducing the thickness in the 90 direction to as far from the centreline of the lay-up in order to en-
maintain the mass, resulted in a dramatic increase in the hance the stiness. Therefore, given its deection char-
deection response. Hence, as the optimum result de- acteristics, the lay-up sequence of [90/0/45/45]s was
picts, for the particular lay-up sequence used, employing
found to be capable of providing the highest structural
a greater ply thickness in the 90 direction proved ad-
integrity for the ap structure under investigation.
vantageous. This direction may well change if the
It needs to be noted that since the objective function
loading, position and direction of the internal structural
is independent of the orientation of the layers, some
members, or the lay-up sequence alter.
simulation runs may experience convergence diculties
[8]. As an alternative to the mass minimisation objective
4.3. Eect of lay-up sequence function, the laminate strength maximisation option can
be considered where required.
A model with a constant set of layer thicknesses
highlighted the eect of lay-up sequence on the stiness 4.4. Ply optimisation
of the structure. This was achieved through varying the
lay-up sequence during each run of the simulation. All Based on the outcomes of the investigation into ply
the layer thicknesses were assumed constant with the
thickness and lay-up sequence, the composite laminate
value of 0.32 mm. A summary of the results is provided
optimisation of the aerodynamic skin of the Krueger
in Table 4.
ap was performed. The skin was selected as it repre-
From Table 4, it is evident that the appearance of the
sented the element of the design with the highest mass
90 plies at the centreline of the lay-up resulted in the
(refer to Table 2). Table 5 compares the initial design
highest deections, whilst the sequences with the 90 options and the skin optimised C-spar model. According
layers positioned furthest from the centreline displayed to Table 5, an optimised composite skin replacing the
the lowest structural responsee.g. the lay-up sequence quasi-isotropic skin can oer a signicant mass saving,
where the cross-sectional areas of the ribs and spars
Table 3
and their structural specications remain unchanged.
Eect of layer thickness Placing the 90 plies on the surfaces of the laminate led
Parameter Design set to greater mass savings for the reasons discussed in
Section 4.3.
A B C D E F
45 ply (mm) 0.246 0.817 0.201 0.246 0.817 0.082
0 ply (mm) 0.201 0.201 0.246 0.817 0.201 0.201
Table 5
90 ply (mm) 0.817 0.246 0.246 0.201 0.246 0.817
Optimised specications of the Krueger ap components
45 ply (mm) 0.246 0.246 0.817 0.246 0.817 0.164
Conguration Opt. mass (kg)
Umax1 (mm) 6.43 8.94 7.30 8.28 4.39 8.33 Blade stiened spar concept 21.6
Umax2 (mm) 3.78 3.20 3.53 2.99 3.40 4.12 C-spar concept 19.6
Unmax1 (mm) 2.29 2.07 2.22 1.66 2.12 2.49 C-spar with ply optimised skin [45/45/0/90]s 18.3
Unmax2 (mm) 5.84 11.20 5.82 6.58 4.19 7.57 C-spar with ply optimised skin [90/045/45]s 16.8
Total V (m3 ) 0.012 0.012 0.012 0.012 0.015 0.011 Mass saving through ply optimisation 2.8
422 J. Bayandor et al. / Composite Structures 57 (2002) 415423

4.5. Performance

A number of post-processing graphs (Figs. 811)


detail the performance of the optimised Krueger ap
under the two primary load cases considered. A scaling
factor of 6.0 has been applied to all plots to aid visu-
alisation.
The results, as expected, show that the deection is
the driving factor for the design. The failure indices on
the plies are well below critical, and the buckling loads
are suciently higher than the limit.
The use of a tailored composite skin has increased the
structural stiness, as opposed to isotropic or quasi-
isotropic skin lay-upsalso investigated in detail during
the course of the study. Where the latter only allows for
up to 14.5% of the material to be removed, 29.5% mass Fig. 10. Y deectionap fully deployed.
reduction is readily achievable using orthotropic mate-
rials.
Further mass savings would be possible by altering
the lay-up sequencing or ply count, and re-optimisation

Fig. 11. Bucklingap fully deployed.

of the stieners in order for the simulation to reect the


Fig. 8. Y deectionap opening. latest skin changes. Optimisation of spar and rib lami-
nates would also lead to improved structural perfor-
mance.

5. Conclusion

A procedure for a commercial aircraft component ply


optimisation was developed using the optimisation ca-
pabilities of the software Ansys. The study provides a
means to facilitate and economise the recently started
practice of replacing the conventional materials used in
aerospace structures with advanced optimal multi-lay-
ered composite materials to reduce mass in conventional
designs as well as in future concepts.
By using an orthotropic skin for a heavy transport
aircrafts secondary component, a Kreuger ap, the
study achieved a higher structural stiness for the ap as
Fig. 9. Bucklingap opening. well as a considerably lower mass, compared to those of
J. Bayandor et al. / Composite Structures 57 (2002) 415423 423

the components original skin lay-up consisting only of Acknowledgements


isotropic and quasi-isotropic materials. The new lay-up
resulted in a mass saving of 1.3 kg for the [45/45/0/90]s The authors would like to acknowledge the collabo-
lay-up sequence, and 2.8 kg for [90/045/45]s . The rative eorts of the Hawker de Havilland Engineering
parametric study indicated that further mass savings on Team on this project: Chris Howe, Andrew Bergen, and
the stiening members were also possible. Ben Saunders.
It was found that by moving the 90 plies away from
the centre of the laminate, the structural deections can
dramatically decrease. However, other considerations References
such as applicability of this option, and most impor-
tantly, requirements for using particular structural bre [1] Niu CY. Composite Airframe StructuresPractical Design Infor-
mation and Data. Hong Kong: Conmilit Press Ltd.; 1992.
orientations on the outer surface to avoid premature
[2] Ansys Inc. Ansys advanced analysis techniques guide. 4th ed.
damage, need also be accounted for. Canonsburg, USA: SAS IP Inc.; 1999.
The next focus of this research would be directed [3] Agarwal B, Broutman L. Analysis and performance of ber com-
towards two main areas. Firstly, introducing the ply posites. 2nd ed. New York, USA: John Wiley and Sons Inc.; 1990.
bre orientation as a DV so that, for a given number of [4] Thomson RS, Scott ML. Experience with the nite element
plies, determining the best position of each ply within modelling of a full-scale test of a composite aircraft control
surface. J Compos Struct 2000;50:33145.
the multi-layered stack would be possiblewith or [5] Ansys Inc. Shell99-linear layered structural shell. Ansys elements
without overwriting the problems symmetry status. The reference. Canonsburg, USA: SAS IP Inc.; 1999.
second focus would be dening additional DVs enabling [6] Ansys Inc. Shell99-element library. Ansys theory reference. 11th ed.
optimisation of the individual ply thicknesses for Canonsburg, USA: SAS IP Inc.; 1999.
achieving improved load capability in the layered com- [7] Ansys Inc. Design optimisation. Ansys theory reference. 11th ed.
Canonsburg, USA: SAS IP Inc.; 1999.
posite structures. Further development of the Ansys [8] Gurdal Z, Haftka RT, Hajela P. Design and optimization of
optimiser to improve the solution robustness for a larger laminated composite materials. New York, USA: John Wiley and
number of DVs is also required. Sons Inc.; 1999.

You might also like