Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United

Kingdom









Optimization Driven Design of a Composite
Underbelly Fairing with HyperWorks




Altair Engineering, Inc
June 2009

by
Dr. Robert N. Yancey
Warren Dias







Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
























Challenge:
To reduce the weight of a
composite underbelly
fairing.



Introduction
Larger aircraft, longer flight ranges, higher fuel costs and
increased overall operating expenses are some of the
challenges that have been posed to the aerospace industry
recently. This has resulted in a need for more efficient design
processes and more innovative designs. Weight has always
played a critical role in the design of aircraft which has led to
increasing use of lighter weight materials such as composites.
Composites do offer lighter weight but they pose many design
challenges due to non-isotropic material properties, a large
number of design variables, and unique failure modes.
Computer based design optimization methods provide an
efficient method to develop optimal designs while taking into
account the many design and manufacturing variables that need
to be considered.
Free-Size optimization is used to identify the optimal ply shapes
and locations as well as the most efficient ply drop-off zones.
Size optimization followed by a ply stacking sequence
optimization is used to obtain the optimal laminate configuration
in terms of ply angle thicknesses and stacking sequence. It is the
unique combination of these optimization techniques strung
together in a process that allows for more efficient and innovative
composite design. This finite element based composite design
and optimization process also incorporates manufacturing
constraints thereby leading to better designs in shorter design
cycle times.
Problem Definition
Composite optimization methods were employed for a composite
underbelly fairing as shown in Figure 1.



Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
2
Figure 1. Depiction of an aircraft
underbelly fairing












Process:
Design Synthesis, Fine
Tuning, and Stacking
Sequence Definition

This paper discusses the Optimization Driven Design approach
for the design of an aircraft underbelly fairing using Altair
OptiStruct. Applying this design paradigm is a 3-phase approach:
Phase 1: Concept Design Synthesis though Free Size
Optimization.
Phase 2: Design Fine Tuning using Ply Sizing
Optimization.
Phase 3: Detailed Design through Ply Stacking
Sequence Optimization.
The fairing has been designed considering two major
performance criteria: 1) the 1
st
natural frequency is at least 20Hz,
and 2) the maximum strain is less than 1000 micro-strain. The
underbelly fairing is considered as a secondary structure and
unlike a primary aerospace structure, doesnt serve any critical
load bearing requirements.
To represent operating conditions, different load cases were
setup. These included an internal uniform pressure loading of
0.02MPa and an external gravity loading of 6.75gs. Additionally,
the fairing was considered to be riveted along its edges to the
surrounding structure.
Design Process and Results
The 1
st
phase of this 3 phase design process is applying Free
Sizing Optimization techniques to establish the optimal ply
shapes for each angle orientation to be included. The user can
define the ply angles to be used up front.
Free Size Optimization applied to shells works on the concept
that the thickness of each designable element is defined as a
design variable. Applying this concept to the design of
composites implies that the design variables are now the
thickness per fiber orientation, per element. Therefore, the
thickness of a Super-ply (total designable thickness of a ply
orientation) of allowable orientations is defined as the design
variable in a free sizing optimization. This allows for a truly
concept level design synthesis of composite laminates,
generating ply shapes and optimizing ply drop-off zones per fiber
orientation.
Free size optimization is also recommended to be used when
trying to optimize material fiber orientation. While generating the
starting/baseline laminate stack, all manufacturable orientations
can be specified and the regions or zones of those orientations


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
3













Design Synthesis:
Determining the general
make up of the laminate
configuration
that arent structurally required will be removed (have their
thicknesses sent to 0).
Phase 1 Concept Design Synthesis Free Size
Optimization
The geometry of the fairing is shown in Figure 2. The idealized
finite element representation of the underbelly fairing was
generated using HyperMesh. The model also represented 2
equipment masses, weighing 2Kg and 3Kg each, which were
connected to their mounting locations through the use of RBE3
elements.

Figure 2: Geometry model of the underbelly fairing

Material properties for carbon-fiber were considered and the
same was represented using an orthotropic material model for
two-dimensional elements (MAT8).
Four Super-Plies 0, +/- 45 and 90 degrees were defined on
the PCOMP card with a thickness variable between 0 and 2mm
each. 2 load cases were considered to drive the optimization at
the concept design stage: 1) a load case representing a 6.75g
loading and 2) an internal pressure loading of 0.02MPa.
The SMEAR option was applied on the PCOMP card to make
the behavior of the laminate independent of the ply stacking
sequence. This is a mathematical approximation.
While running an optimization study, it is important to define
certain manufacturing constraints in order to produce results that
conform to manufacturing requirements. OptiStruct includes
several different manufacturing constraints such as ply
percentage, ply thickness balancing, ply thickness, and laminate
thickness, that can be defined at the concept level free sizing


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
4











Results:
Optimized Thickness and
Ply Distribution for
Laminate
stage. For the underbelly fairing optimization discussed in this
paper, 2 manufacturing constraints were incorporated:
1. Ply percentage for the 0s and 90s such that no less than
10% and no more than 60% exist.
2. A balance constraint that ensures an equal thickness
distribution for the +45s and -45s.
An important point to note here is that the manufacturing
constraints are carried through all phases of the design process,
with additional constraints that can be defined at the ply shuffling
phase.
The free size optimization problem was setup to use a 30%
volume fraction constraint of the designable volume while
minimizing the weighted compliance of the 2 load cases. An
equal weight of 1 was considered for both load cases to
eliminate any bias in the solution. This setup will try to identify
the stiffest design for a given fraction of the material.
Figure 3 shows the result of the thickness distribution achieved
from the free size optimization, post processed in HyperView:

Figure 3: Element thickness distribution from free size
optimization

The regions indicated in red or in colors tending towards red
(from the legend) can be interpreted as thicker regions, while
those in blue or tending towards blue are thinner regions. The
contour plot above shows the total thickness distribution that
includes contributions from each defined superply orientation, i.e.
a thickness contribution from the 0s, +/-45s and the 90s. The
overall thickness distribution can be post-processed into
individual superply thicknesses as shown in the following figures.


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
5












Angle Distributions:
Determine how much of
each angle ply is required
throughout the laminate


Figure 4: Optimized thickness distribution of 0 degree superply


Figure 5: Optimized thickness distribution of 45 degree superply


Figure 6: Optimized thickness distribution of 90 degree superply


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
6










Multi-Step Approach:
Superply sizing, ply bundle
sizing, ply lay-up


Due to the balance manufacturing constraint, the thickness
distribution of the +45 degree superply and the -45 degree
superply are the same.
From a free size optimization, the thickness of a superply is
optimized and subsequently subdivided into a user-defined set of
Ply Bundles. Each ply bundle would have a specific pattern
and the ply bundles for a given orientation would be built up to
approximate the thickness distribution of the superply. More ply
bundles will result in a final thickness profile of the superply that
more closely matches the free size optimization results. On the
other hand, more ply bundles increases the number of ply cut-
out patterns which can increase manufacturing costs. An
illustration of the concept of going from a superply to ply bundles
to individual plies is shown in Figure 7. For this study, 4 ply
bundles per fiber orientation (superply) are defined.


Figure 7. Illustration of Superply, Ply Bundles, and Individual
Plies
These ply bundles represent the shape and location of the plies
per fiber orientation. With 4 ply bundles per fiber orientation, this
study produced a total of 16 ply bundles. Since the +45 and -45
angle orientations are constrained to be identical, the study
produced a set of 12 unique ply cut-out patterns.
Figure 8 represents the results for one ply bundle. The left hand
images show the optimized ply shape for each angle considered
in the ply bundle sizing optimization and the right hand images
show the interpretation of these results taking manufacturing
constraints into consideration. The goal is to match the results
as closely as possible in the context of the manufacturing
operation. In this case, small patches of plies are not practical
so the interpretation generates larger ply patches that resemble
the raw optimization results.


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
7













Results Interpretation:
Intepret Results to meet
other design or
manufacturing
requirements not
considered in the
optimization set up
Example of 0 Degree Ply
Example of 45 Degree Ply
Example of 90 Degree Ply
Optimization Results Manufacturing Interpretation
Figure 8. Results and Interpretation for One Ply Bundle
This process is repeated for each ply bundle to come up with all
12 ply shapes required for the full laminate. This process is
currently done manually by the user but this could be automated
based on rules established by the manufacturer.
Phase 2 Design Fine Tuning Ply Bundle Sizing
Optimization
Phase 2 of this process involves a sizing optimization of the ply
bundles generated from phase 1 of the design process.
Having established the optimal ply shapes and patch locations,
the next step is to fine tune this design for thickness. Phase 2
involves identifying the optimal thickness of each ply bundle. A
choice of running the optimization with the thickness as discrete
variables or continuous variables is available. A minimum
manufacturable ply thickness can be specified, thereby running a
discrete optimization and allowing for the calculated


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
8














Ply Bundle Sizing:
Determine how many plies
of each pattern are
needed to meet all
performance requirements
optimal ply bundle thickness to be a multiple of the minimum ply
thickness value. This helps in calculating the total number of
plies required per fiber orientation.
Additional performance criteria can be incorporated into the
problem formulation to ensure that the optimized design meets
the necessary design requirements. In this case, an additional
load case was added to calculate the natural frequency of the
fairing under assembled conditions. The optimization setup was
also modified to factor in these additional performance targets,
among others. :
Total designable volume
Natural frequencies (1
st
5
th
modes)
Composite strain
Constraints were defined on the frequencies and strain, i.e.
Natural frequency > 20Hz
Composite strain in the fairing < 1000 micro-strain
The objective of the problem was to minimize the volume
Manufacturing constraints previously defined in the free sizing
phase are automatically carried over into phase 2, preserving
manufacturability across the process. These include:
1. Ply percentage for the 0s and 90s such that no less than
10% and no more than 60% exist.
2. A balance constraint that ensures an equal thickness
distribution for the +45s and -45s.
The results are presented in Figure 9 that shows the overall
thickness distribution map of all 16 ply bundles before the ply
bundle sizing optimization stage. Figure 10 shows the final ply
bundle sizing results where each ply pattern is sized to achieve
the overall performance specifications. The ply bundle size
optimization yielded a design that led to a further weight
reduction of 23% from the free size results, without violating any
of the prescribed design constraints.


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
9





Ply Bundle Sizing:
Determine how many plies
of each pattern are
needed to meet all
performance requirements

Figure 9: Element thickness contour before size optimization
.

Figure 10: Element thickness contour after size optimization

Phase 3 Detailed Design Ply Stacking Sequence
Optimization
While shuffling the stacking sequence it is important that
behavioral and design constraints are preserved. Additionally, it
is required that certain ply book rules be applied to guide the
stacking of plies based on specific requirements. Some of the ply
book rules that control the stacking sequence are:
Maximum number of successive plies of a particular fiber
orientation
Pairing of plies
Identification of a sequence for the core and cover
regions


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
10











Shuffling Optimization:
Determine the optimal
stacking sequence the
meets ply book rules

For this case, the optimization problem as previously formulated
in the sizing phase is retained and additional ply book rules are
applied. They are:
The maximum successive number of plies does not
exceed 4 plies
That the +45 and 45 plies be reverse paired
Figure 11 shows the history of the shuffling optimization. 0
degree plies are represented in blue, 90 degree in red, +45
degree plies in green, and -45 degree plies in yellow.

Figure 11. History of Ply Shuffling Optimization


Altair Engineering: United States, Brazil, Canada, China, France, Germany, India, Italy, Japan, Korea, Sweden, United Kingdom
11










Benefits:
Design laminate structures
in less time, achieve
greater performance, and
use less material

The final result produces a design with the following
performance:
Minimum natural frequency = 21.2Hz (>20Hz)
Maximum strain = 9.5e-4 (<1e-3)
The design therefore meets its performance requirements, is
feasible and manufacturable.
Summary and Conclusions

The composite optimization process detailed in this paper
expands upon advanced optimization techniques including
Phase 1 - free size optimization to determine ply
thickness per angle orientation
Phase 2 - size optimization to determine ply bundle
patterns and required number of plies per bundle
Phase 3 shuffling optimization to determine ply
stacking sequence
By stringing these 3 techniques together, OptiStruct offers a
unique and comprehensive process for the design and
optimization of composite laminates. Free size optimization for
composites allows a true concept level design synthesis of plies.
A new PLY and STACK based modeling technique that simplifies
laminate representation and facilitates the ply bundle sizing
optimization followed by the ply shuffling optimization make the
process unique. The process is automated and integrated in
HyperWorks by generating the input data for a subsequent
phase automatically from the previous design phase. The
process also allows flexibility in case any modifications are
required. Throughout the design process, manufacturability
constraints and behavioral constraints are preserved to arrive at
a feasible design and ensure a meaningful process.
Applying this process to the design and optimization of an
aircraft underbelly fairing resulted in a design that made optimal
use of material. Additionally, it resulted in a light weight design
that met all prescribed design and manufacturing constraints. It
is also important to note that the process was carried out in a
short span of time (< 1 week). Incorporating an optimization
driven design process (as described in this paper) can be a
major enabler to a more efficient and less costly design process
of composite structures.