Escolar Documentos
Profissional Documentos
Cultura Documentos
Laminate Boundaries
Marc Funnell
Airbus UK
New Filton House, Bristol, BS99 7AR
marc.funnell@airbus.com
Abstract
This paper investigates the application of newly available optimization functionality available in OptiStruct [1] to
provide design guidance to generate innovative laminate composite solutions. Due to the flexibility of laminate
composites, it has great potential to exhibit displacement characteristics that could significantly increase the
aerodynamic performance. Free element sizing technology is used to determine concept lay-up solutions. These
solutions determine the laminate make-up (i.e. 0, 45, 90), thickness and the various laminate boundaries of
an aircraft wing covers under multiple loading conditions which meet the required displacement targets whilst
also minimising mass. These preliminary studies demonstrate that the technology can successfully achieve
displacement targets for multiple load cases. Each analysis study can be completed within minutes and
consequently can be utilised as a valuable concept design tool.
Keywords: Laminate Boundaries, OptiStruct, Free Element Sizing
1.0
Introduction
The aerospace industry provides many challenges which require the use of leading edge technologies to keep
up with increasing performance demands. In the field of composite design there are an almost unlimited number
of ways in which a laminate can be constructed. This is because for each composite patch it is potentially
possible to vary:
i.
ii.
iii.
iv.
Coupled with this, the final design is required to meet structural and manufacturability targets such as stiffness,
buckling, stress & strain limits and lay-up rules. For example, if a single composite patch has three ply
orientations (e.g. 0, 45 & 90) each containing 10 ply layers (30 layers in total), the total number of available
stacking sequences is over five thousand billion. Traditional optimisation methods cannot cope with problems of
this scale and it is not practical to attempt to optimise for all of the required targets in one step.
The development of Altairs optimization technique Free Element Sizing [2] (FES) has made it not only possible
to optimise complex composite structures but also complete the optimisation in a short time frame. Free
Element Sizing is a new optimisation technology implemented for the conceptual design of metallic or composite
shell structures. As opposed to topology optimization where the density is utilized as design variable, in free
sizing the element thickness for shells and ply-thickness for composite lay-ups are the design variables. This is
shown to be highly beneficial in determining the optimum locations for composite laminate boundaries but does
require some design interpretation.
This paper details the application of Free Element Sizing to generate an optimised composite wing skin lay up
for the concept phase of the development.
Altair Engineering 2007
5-1
2.0
2.1
Introduction
An existing model of the aircraft wing was analysed under a number of in flight and landing loading conditions.
The results of this baseline analysis were used to provide optimisation targets for the Altair OptiStruct [3]
optimisation. The targets extracted from the baseline model took the form of static displacements from the wing
tip. This will ensure that the optimisation produces a model which retains the same stiffness properties as the
existing model.
The objectives of the optimisation are to:
i.
ii.
iii.
iv.
v.
vi.
2.2
Use Free Size Sizing optimisation at the concept design phase to lead the design of the plies used
in the upper and lower composite covers of an aircraft wing
Save significant model set-up & run time compared to other optimisation methods
Determine the size, position & orientation of the ply patches
Visualise which ply orientations are doing the most work
Estimate the required number of plies per patch
Mass prediction
Free element sizing allows the thickness of individual shell elements to be varied independently. In the case of
composite structures, the thicknesses of each ply in each element are varied independently. The method is
based upon similar principles to topology optimisation as opposed to having actual independent design
variables for each element thickness. This has the following advantages:
i.
ii.
iii.
Easy setup one line added to model file to vary all elements independently
Single design variable per component - independent thickness changes handled internally
Fast solution time converges in 10-20 iterations
The concept of super-ply is to group plies of the same orientation together such that the number of plies in the
model is significantly reduced. The thickness of each super-ply can then be varied which simulates adding or
removing laminate plies. Altair OptiStruct allows an element formulation to be used which smears the available
stiffness from the plies uniformly throughout the element thickness. This is similar to dividing each ply into a
number of infinitely thin plies and mixing them evenly, allowing the composite to be modelled as a super-ply but
simulated as if it were a uniformly shuffled stack.
Super-Ply Stack
Smear Formulation
(layers uniformly mixed)
5-2
2.3
The wing was modelled in isolation from the remainder of the aircraft by applying constraints at the wing /
fuselage interface. The model is constructed from a total of 12,000 elements (8,000 shells and 4,000 beams)
using approximately 8,000 nodes. The study included ten loading cases which simulate the main loads
encountered during operation. The top and bottom skins of the wing were modelled using standard shell
element properties with orthotropic materials (i.e. the baseline model is not modelled using composites). The
orthotropic material zero direction was aligned with the wing length.
90
45
0
-45
Figure 2: Initial Wing Model
The baseline model uses macro mechanical properties for every element which have been determined from a
standard Airbus laminate sequence. Each element has been assigned different properties to produce a
representation of the composite and the initial thickness variation of the top cover is shown below:
5-3
5-4
Max
Min
Max
Min
0.09
0.08
1.00
-0.02
0.38
0.10
0.13
0.65
0.69
0.71
1.00
0.96
0.03
0.92
0.27
0.55
0.59
0.06
0.08
0.06
0.51
0.44
0.58
0.46
0.37
0.44
0.54
0.06
0.21
1.00
0.17
0.29
0.04
0.25
0.26
0.79
0.09
0.48
1.00
0.74
3.0
3.1
Introduction
The elements in the upper and lower skin were arranged into a new component group which was assigned a
super-ply composite property. The composite is made up of four super-ply layers (0, 45, -45 & 90 degrees) each
of which has an orthotropic material definition. The super-ply uses the SMEAR option to simulate an evenly
shuffled laminate stack. This property was used by the free element size optimisation to determine the optimum
thicknesses of each element in each of the four plies. This gives the optimisation a large amount of flexibility in
producing an optimum design as it has the potential to vary 11282 thickness vales, equivalent to the number of
elements in Free Size component multiplied by the number of independent composite layers. The composite
stack is simulated as symmetric, which is implied by the SMEAR formulation. The composite region which is to
be optimised is coloured yellow in Figure 6.
5-5
The bottom cover was also included in the optimisation. For illustrative purposes only the top cover results are
shown throughout this report, although similar results were obtained for the bottom cover simultaneously. All
remaining data with regard to the wing structure and loading conditions was left unchanged from the baseline
model.
Design constraints:
Achieve the same stiffness as the baseline model for all of the loading conditions, defined by the
wing tip displacement
Optimised 45 and -45 layers must be identical
3.2
An initial Free Size optimisation (Model 1) was set up which was required to calculate the thicknesses for each
of the four super-ply orientations (i.e. 0, 45, 90) which give the minimum mass whilst achieving the same
wing tip displacement as the baseline analysis. The +45 and -45 degree layers were automatically linked such
that their resulting thicknesses are identical. The optimised thicknesses for each of the ply layers are shown
below:
5-6
The optimization took approximately six minutes on a laptop PC and converged in around ten iterations. The
set-up time for the optimisation study is also minimal, around half an hour.
The initial free element sizing results show that the 0 ply requires the highest number of layers. This is because
the 0 ply is doing the most work as it is orientated in the direction of the load path. The 45 & -45 layers are
identical, meeting the manufacturing requirement, and show that a number of 45 /-45 layers are required in the
centre of the top cover.
The 90 ply, being orientated out of plane to the loading, is not being worked and consequently requires very
few layers. The mass of the optimised composite skin wing assembly using the above thickness was determined
and the predicted masses for all additional studies will be shown as a percentage of the Model 1 mass.
The Free Size optimisation for this model took five minutes to run on a standard laptop, requiring only 250Mb of
memory.
3.2
In order to investigate the sensitivity of the optimised design to the material orientation, two analysis models
were generated using the optimised ply thicknesses from Model 1 but with adjusted ply angles:
This will not affect the mass of the design; the mass of these two analysis models is the same as that of the
Model 1. The reorientation is shown in the figure below.
9
0
4
5
9
0
4
5
0
-10
+10
Model 1a
Model 1b
Figure 8: Ply Orientation Study Models
This resulted in the following percentage changes in the wing tip displacement:
Load
Case
1
2
3
4
5
6
7
8
9
10
5-7
It can be seen from the above results that changing the ply orientation has a negative effect upon the stiffness of
the wing. A change in the ply orientation of 10 affects the wing stiffness by up to 8.3%.
The composite material is orthotropic and has a longitudinal/transverse stiffness ratio of 2.4 (=E1 / E2). This
means that the material is 2.4 times stiffer in the 0 direction than it is in the 90 direction. Hence, the change in
stiffness due to the ply orientation change was expected.
In order to investigate the effect of ply orientation upon the optimisation results, two additional Free Size
optimisations were carried out using 10 ply offsets. A comparison of the initial optimisation results to the 10
degree offset optimisation results are shown below:
5-8
Comparing the mass of the individual plies makes it easier to determine where the material has been
redistributed.
The table below shows that the +10 change in ply orientation causes:
5-9
3.3
In order to achieve the aerodynamic characteristics of the wing it is very important not only to maintain the
structural stiffness but also to be able to tune the deformation to a required shape. An additional study, Model 3,
was carried out to investigate the effectiveness of tuning the composite ply lay-up to obtain a design which
meets specific displacement and twist angle targets.
The change in angle of the baseline wing model under each loading condition was calculated at specific
locations along the wing length. The measured angles were used as optimisation constraints to ensure that the
optimised design had the same behavior under the applied loading as the original model. A small deviation
(0.05) in the allowable angle was introduced to prevent over-constraining. The optimisation target for torsional
stiffness is given by the following equation:
(Model 0 - 0.05) Model 3 (Model 0 + 0.05) for all load cases
Model 0
Model 3
Baseline Model
Model 3, Composite optimisation model
Figure 11: Torsional Stiffness Calculation
In addition to the angle constraints, the displacement of the wing tip was constrained using the same target as in
the baseline free element size study for each load case.
5-10
Model 3, Optimised
Mass Change (%)
-5.6
17.3
134.7
9.6
Optimised Angle
(% difference from target)
-1.2
0.0
0.0
-0.9
-4.3
0.0
0.0
-5.3
-4.5
-7.7
Optimised Displacement (%
difference from target)
-0.9
-0.3
-0.6
-0.8
-1.0
-0.1
-1.5
0.0
-0.5
0.0
5-11
The design targets also specify that the Normal X direction strain in the skin must be less than 6,000. As well
as providing the optimum ply lay-up, OptiStruct also produces stress, strain and failure index results for each
ply. A comparison of the strain contour from Model 3 to Model 1 for the top cover 0 ply under the highest
loading is shown below.
Results Interpretation
In order to take the design further it is necessary to interpret the optimisation results into a number of discrete
composite patches. This involves a conversion from theoretical optimum laminate boundaries with continuous
thicknesses into manufacturable laminate patches with discrete numbers of layers.
This conversion process was carried out using an automatic interpretation procedure. This process involves
rounding the optimised thicknesses by a user defined tolerance and grouping them together into new composite
patches. This has the following benefits:
i.
ii.
iii.
iv.
Allows visualisation of ply boundaries, making it a much simpler task to design the composite
patches in the best locations
Generates a new model which could be used for size optimisation. This would then allow the
number of plies in each patch to be adjusted accounting for lay-up rules and stress, strain,
stiffness & failure index targets
Ply percentage lay-up visualisation
Estimates the required number of layers of each ply orientation per patch
5-12
The Free Element Size interpretation procedure was carried out on Model 1 and Model 3 using a coarse
grouping range. This resulted in the following models:
0%/
45%/90%
40/50/10
50/40/10
80/10/10
70/20/10
40/40/20
20/60/20
5-13
3.6
Discussion of Results
The optimisation studies have shown that the composite skin lay-up has a direct effect upon the global wing
stiffness and torsional stiffness. The stiffness was found to be sensitive to the accuracy of orientation of the
material longitudinal direction. In order to produce a robust design, it may be necessary to asses the
manufacturing tolerances on the ply orientation and investigate their effect on the design using stochastic
analysis methods. It was found that the stiffness could be improved by aligning the material longitudinal direction
with the main load path.
It is possible to use Free Size optimisation to perform aero-flexibility tailoring, tuning the composite lay-up
design to achieve a specific deformed shape. The effect of adding additional design constraints to the
optimisation is an increase in the required amount of material needed to meet those targets. This is shown by
the increase in mass from the baseline optimisation when the angular constraints were added.
The main objective of the optimisation was to determine the optimum composite patch locations. This has
proved to be successful. In addition to this, the optimised design met the strain targets and estimated the
required number of plies though the results interpretation process.
After the ply patches have been designed using the above results as a guide, a second optimisation study is
recommended to establish the exact number of plies that are required in each patch and the stacking sequence
required to meet Stress / Strain targets, Failure index targets, Ply lay-up rules, Buckling targets and Ply
percentage rules.
4.0
Conclusion
The outer skin composite lay-up has a direct effect upon the wing bending and torsional stiffness. It was found
that the composite lay-up can be tailored for aero-elastic tailoring to meet specific deformation and twist angle
targets. This is possible due to the orthotropic material in the plies, allowing the longitudinal and transverse
stiffness of the composite to be tuned almost independently. This also means that the stiffness is sensitive to the
accuracy of orientation of the material longitudinal direction.
The Free Size optimisation shows where in the model each ply orientation is needed to obtain the required
deformation profile. The optimum locations of the ply patches are typically along the load paths although this
was shown to change significantly when required to resist torsional deformation as well as bending. This shows
that the optimum location for the plies cannot be determined from the loading conditions alone; an optimum
design is dependent upon a combination of the applied loads and the design constraints.
The stresses, strains and failure indices in the plies can be assessed for the optimised design. The strains in the
new design were found to be within the required limits.
An automatic interpretation procedure was used to determine the optimum ply patch locations and also
estimated the required number of orientation in each patch. This will allow the ply patches to be designed in the
best possible locations. The orientation percentages in each patch can also be assessed.
A minimum feasible mass value for the composite design was determined, although this is likely to increase
after applying constraints to meet additional targets for lay-up rules, strains and failure indices.
The design can now be finalised through size optimization to meet additional criteria such as detailed structural
constraints (i.e. stress, failure index, buckling), manufacturing constraints and ply lay-up rules. The Free Size
optimisation has allowed for the development to be taken forward in a very short time frame.
5.0
References
[1]
5-14
[2]
Optimization Driven design of shell structures under stiffness, strength and stability
requirements P Cervellera, M Zhou, U Schramm 6th World Congresses of Structural and
Multidisciplinary Optimization Rio de Janerio 30 May - 03 June 2005 Brazil
5-15