You are on page 1of 15

AIAA 2016-1246

AIAA SciTech
4-8 January 2016, San Diego, California, USA
57th AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference

Wing Design Utilizing Topology Optimization and Additive


Manufacturing

David Walker1 David Liu2 and Alan Jennings3


Air Force Institute of Technology, Wright-Patterson AFB, OH, 45433

This research is a follow-on effort to a topology optimization (TO) study evaluating the ability to
computationally reduce mass of a wing structure while maintaining structural performance. The
static loading conditions were obtained through CFD analysis of the wing and the results were
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

applied to calculate the baseline displacement, von-Mises stresses, and buckling conditions. TO
was then performed both locally on the ribs of wing and on a global scale where the entire
internal structure of the wing acted as the design space. Additionally, a skin-thickness
optimization was performed, as well as the integration of a fuel-tank as a functional component
and as a load bearing structure. The local rib design was then manufactured in both plastic and
aluminum to show the capabilities of additive manufacturing with aircraft components.

Nomenclature

Ci Incremental compliance
Cw Global compliance
F Force vector
fi Applied elemental force
K Global Stiffness matrix
Ke Local stiffness matrix
̅ Penalized stiffness
N Total number of Elements
p Penalization power
U Global displacement vector
ue Local displacement vector
ui Elemental displacement matrix
Xmin Minimum relative density
wi Weight factor
ρ Density fraction

I. Introduction
This research is a continuation effort of a previous study on the feasibility of utilizing Topology Optimization
(TO) and Additive Manufacturing (AM) to design and fabricate an aircraft structure. Through traditional
manufacturing means, TO designs are difficult to manufacture. However, AM allows for more radical design
features due to its build-up nature. With the previous research from Walker et. al., the selected TO process was
verified through examination via Altair’s Optistruct software1. Additionally, the wing from a Van’s RV-4 homebuilt
aircraft was used as a baseline structure for the optimization. Finally, a plastic 3-dimensional wing section was
designed and built through AM. In this case, a thin skin was applied to a design space with two structural constraints
running laterally through the wing. The skin was subjected to two-dimensional forces in the x and z direction based
on the 212 mph (94.8 m/s) maximum cruise velocity at a cruise altitude of 10,000 ft (air density = 0.958 kg/m3). The

1
Graduate Student, Department of Aeronautical and Astronautical Engineering, Student Member AIAA.
2
Assistant Professor, Department of Aeronautical and Astronautical Engineering, Senior Member AIAA.
3
Senior Systems Engineer, System Dynamics and Performance, Raytheon Missile Systems.
1
American Institute of Aeronautics and Astronautics

This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States.
optimization was based on minimizing compliance constrained to a maximum volume fraction of 38 percent1. The
resulting design and fabrication are seen in Figure 1 and 2, respectively.
This continuation effort enhances the intent of the previous research by establishing objectives to meet the
overall intent of this study; develop a procedure which is used to reduce the weight of an aircraft while maintaining
structural integrity of the wing. Since the intent of this study is to utilize AM, radical design attributes often seen
from TO are realistically manufacturable. Through TO, it is important to formulate a means to apply realistic
loading conditions on a model, determine what attributes are important for an optimization, and find a means for
verification. Additionally, AM enables further design considerations when not constrained by traditional
manufacturing means. In this case, a dual-purpose structure which serves as a load bearing component was
considered as a way to fully utilized AM capabilities. Therefore, the following five objectives were established:

1. Determine pressure loading on the wing for critical phases of flight and apply the values towards the
analysis and optimization processes
2. Perform a computational analysis on the baseline aircraft wing to determine localized stress and
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

displacement values
3. Generate a computational analysis on the optimized designs used for comparative purposes
4. Integrate a traditionally independent component into the optimized wing as a dual-purpose structure
5. Interpret full-scale design to meet AM constraints and produce the model

Figure 1. Post-process topology optimized RV-4 wing segment with density fraction set to 38 percent.

Figure 2. 3D printed wing section for optimized RV-4 wing segment.

As discussed in the previous paper, TO is the mathematical approach of finding the optimal material distribution
over a given design space. For nearly all of this research, the designs were constrained to a set volume fraction of the
design space with the objective of minimizing compliance. The TO method used is the power-law approach,
otherwise referred to as the Simple Isotropic Material with Penalization (SIMP) method. The SIMP method
considers all material properties of a discretized design space to be constant. However, the elemental density factor,
ρ, varies as a value between 0 and 1. Elements of ρ=0 have no density and elements of ρ=1 are 100 percent dense.
Intuitively, this is not manufacturable since most material density is constant. Therefore, final designs are often
interpretations of the TO results; a threshold is set in which values below are considered a void and values above are
considered solid material2. The SIMP method uses a penalization power, p, on the density fraction to steer the value
towards a solid or a void. Applying a power to ρ reduces ambiguous moderate density areas near the chosen
threshold value by steering the density fraction lower3. The procedure is deemed acceptable for any material with a
Poisson’s ratio of 1/3 as long as the penalization power is ≥3 2. Mathematically, the SIMP method for minimum
compliance problem is shown in Equation 1. In this case, U is the global displacement vector, F is the force vector,
K is the global stiffness matrix, ue is the local displacement vector, ke is the local stiffness matrix, X min is minimum
relative density, and N is the total number of elements.

2
American Institute of Aeronautics and Astronautics
(1)

The Altair Engineering suite of computational software was the primary tool used for this research. Particularly,
the Optistruct optimization software was used for all topology and sizing optimizations. Optistruct utilizes the SIMP
with a penalization power of 1.0 for 2D elements and 2.0 for 3D. Under certain settings, the penalty is increased to
3.0. If is the real stiffness matrix of the element, then the penalized stiffness is, ̅ , is seen in Equation 24.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

(2)

For this research, all optimizations were subjected to an objective of minimum compliance constrained by a set
volume fraction. The selected volume fraction value was chosen through an iterative process and was dependent on
the specific problem setup. Multiple loading conditions were applied and the solver simultaneously calculated each
incremental optimization condition independently during each iteration. This was done as a means to ensure the
solution satisfied each condition independently. This was done through the weighted compliance formulation shown
in Equation 3. If Cw is the global compliance, wi is the weight factor, Ci is the incremental compliance, ui is the
elemental displacement matrix, and fi is the applied elemental force, global compliance, Cw, is shown in Equation 3.
Even though it is possible to vary the weighting for each load condition, all were weighted evenly for this research4.

(3)

II. Methodology

The first step in analysis and optimization was to calculate realistic dynamic pressure values for the surface of
the wing. These values were then set as the loading conditions for the TO. In order to do so, Computational Fluid
Dynamics (CFD) was utilized. A Computer Aided Design (CAD) model of the wing was developed and inserted
into Altair’s Virtual Wind Tunnel (VWT) CFD program. VWT allows a user to set inlet and environmental
conditions which are applied to the model. In this case, a generic fuselage was used to simulate the variation in
lateral surface pressure experienced on an aircraft wing. An image of the model in VWT is shown in Figure 3.

Figure 3. Van’s RV-4 CFD wing model attached to generic fuselage for surface pressure analysis.

3
American Institute of Aeronautics and Astronautics
The desire of this research was to develop a wing which would satisfy the loading conditions for all phases of
flight. In order to develop an optimization suitable for all flight profiles, the aerodynamic conditions for the most
severe cases were used5. For an aircraft, this occurs at the limits of its structural performance 5. In the case of the RV-
4, the envelope is at -3G, +6G, and +6G with full aileron deflection for both a left and right roll5. Using the RV-4
gross weight of 1,500 lbs, the -3G and +6G angle of attack was calculated at -10.80 degrees and +15.56 degrees,
respectively. The CFD model was then modified for an aileron deflection of ±25 degrees to account for maximum
roll. VWT results output localized surface pressure at all points along the wing. These values were then easily
interpreted into Optistruct as loading conditions for the wing model. Loading was applied on each element on the
discretized surface of the wing as a pressure value.
The next step was to develop a proper baseline model for comparative analysis with the final optimized wing.
Therefore, a simplified RV-4 wing was created and analyzed using Finite Element Analysis (FEA). The wing design
was adapted from schematic drawings provided by Van’s Aircraft 6. The baseline wing design consists of a series of
ribs with a main and trailing spar, surrounded by a thin sheet-metal skin. Each rib of the baseline wing has the same
design and consists of a series of holes along the centerline to save mass. The sheet-metal skin surrounding the ribs
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

provide the aerodynamic shape. This practice is extremely common for aircraft wing design, especially amongst
low-cost general aviation aircraft. All components are comprised of 2024-T4 aluminum ranging between 0.813 mm
(0.025 in) to 1.016 mm (0.032 in) thick. For this study, flanges and connectors were disregarded. Initially, structural
performance was calculated without considering the fuel tank. The FEA model was built using two-dimensional
surface structures with an applied thickness. Figure 4 is the baseline wing model. The wings was analyzed for
structural stress, displacement, and buckling. In addition, the mass and volume properties for the wing are shown in
Table 1.

Figure 4. Baseline RV-4 wing without fuel tank for FEA analysis.

Table 1. Baseline RV-4 wing without surface area, volume, and mass properties.

The resulting tip displacement, total displacement, and peak von-Mises stresses of the baseline wing analysis are
summarized in Table 2. Total displacement is the maximum displacement seen at any point on the wing whereas tip
displacement is the maximum displacement seen at the tip of the wing. This was seen with skin material between the
ribs which was displaced more from the starting point than the tip of the wing. An example displacement profile is
seen in Figure 5a for the +6G, roll right loading condition and a stress profile in Figure 5b. Displacement is in
meters (m) and von-Mises stress is in Pascals (Pa).

4
American Institute of Aeronautics and Astronautics
Table 2. Displacement and stress of baseline wing.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

(b) Displacement Profile (a) von-Mises Stress Profile

Figure 5. Wing Stress contours for the +6G, roll-right flight profile of the baseline wing.

In addition to displacement and stress, the first three modes of buckling were analyzed. In this case, buckling all
occurred near the root of the wing on the skin surface. Figure 6 is an image of the buckling which occurred in the
wing for the +6G, roll right flight maneuver. Figure 6 is an image of the Buckling Load Factor (BLF) and Table 3 is
a summary of BLF for all tested flight conditions. BLF is the ratio of the localized applied loading to the load at
which buckling would occur. In other words, higher BLF values show a higher resistance to buckling. BLF over 1.0
means a structure will not buckle.

Figure 6. Buckling analysis for +6g, rolling right for baseline wing.

5
American Institute of Aeronautics and Astronautics
Table 6. Buckling Load Factor (BLF) values of baseline wing.

Once the baseline wing was analyzed, the TO was setup and calculated. Optimizations were performed using a
variety of methods. The first optimization was performed as a local design, only concentrating on individual ribs. In
this case, all other structures remained constant. Figure 7 is the design space of the local TO. Once again all
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

optimizations were performed with an objective of minimizing compliance while constrained to a maximum volume
fraction. The specific volume fraction chosen was an iterative process specific to each optimization type. This
allowed the design to be fine-tuned using engineering judgment to provide the best output. In addition, there was a
focus on optimizing with and without a pattern repetition. Pattern repetition ensures each rib is of similar design.
This was done to provide a direct comparison with the baseline rib. When pattern repetition is not used, each rib is
uniquely designed. The local design is also not necessarily limited by traditional manufacturing considerations, but
is simply analyzed for comparative purposes. As with all optimizations, the pressure loading obtained from the CFD
analysis was applied only to the skin of the wing. The wings were structurally constrained at the spars and along the
entire skin surface on the wing root.

Figure 7. Local to 2D rib problem setup.

A global optimization considers the entire internal structure as the design space. Doing so allows the optimizer to
calculate the entire internal structure of the wing. This procedure was initially done without any structural material
along the edges of the wings, as shown in Figure 8. It was additionally performed, with a structural support rib and
rear spar to maintain the aerodynamic shape of the wing. Since TO utilizes variable density material, areas of low
density are disregarded in the final design when using the threshold method. Therefore, the edges of the wings were
left without support material. Figure 9 shows the design space of the global optimization with the tip rib and rear
spar. Note in both cases, the spars at the root of the wing were maintained in the same location as the local
optimization.

6
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 8. 3D global optimization problem setup.

Figure 9. 3D global optimization problem with tip ribs and rear spar.

Another design consideration as a means to reduce mass was to implement a dual-purpose structure which serves
as both a functional component and a load-bearing structure. In this case, the fuel tank was used. For comparative
analysis, the baseline wing performance was calculated with the fuel tank. For the baseline FEA, the fuel tank was
built as an independent component consisting of an additional spar running the length of the tank and a slightly
different rib design. The baseline wing with a fuel tank is shown in Figure 10. Once again, this wing was analyzed
for stress, displacement, and buckling. Table 4 outlines the surface area, volume, and mass of the wing.

Figure 4. Baseline RV-4 wing with fuel tank for FEA analysis.

7
American Institute of Aeronautics and Astronautics
Table 10. Baseline RV-4 wing without surface area, volume, and mass properties.

Similar to the baseline wing without an integrated fuel tank, structural properties were calculated through FEA.
Table 5 shows the displacement and stress of the baseline fuel tank. Table 5 summarizes the BLF for this wing.
Stress, displacement, and buckling occurred in a very similar manner to the baseline wing. Therefore, these figures
are not shown. Note the slightly higher levels of von-Mises stress in the wing. These peak levels occurred at the root
primarily along the main spar. However, the von-Mises stress and BLF values remained relatively similar when
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

compared to the baseline wing.

Table 5. Displacement and stress of baseline wing with fuel tank.

Table 6. Buckling Load Factor values of baseline wing with fuel tank.

Finally, a skin thickness optimization was performed on both the local and global designs. With a skin
optimization, material is distributed in a manner which satisfies the objective most appropriately. In this case, the
elemental thickness at each point on the skin and spars were used as a variable. Even though this is not a TO, it was
used as another means to improve final structural performance. Once again, the optimization was performed with an
objective to minimize compliance while constraining the maximum volume. The volume for all skin optimizations
was constrained to 0.00523 m3, which is the volume of material for the baseline wing. Therefore, minimal mass
savings are seen using this method. However, structural performance did change. Minimum and maximum skin
thickness at any point on the skin was set to 0.635 mm to 2.500 mm, respectively. The minimum thickness value is
the thinnest found at any location on the RV-4 aircraft and the maximum limit is approximately four times that
value. This optimization for the skin and spars performed concurrently to streamline the optimization process. The
ribs were optimized prior and completely independent of the thickness optimization.

III. Results
The first optimization result is shown in Figure 11. The variable density values between 0 and 1 are displayed.
Note the blue areas are approaching a density fraction of 0, and the red areas are approaching 1. Prior to FEA
analysis, this design was post-process to provide a manufacturable part. In this case, a volume fraction of 0.15 was
used. Overall, there was a total mass savings of 18.5 percent amongst the ribs, which equates to a 3.7 percent total
mass savings in the wing. Table 7 outlines the surface area, volume, and mass of this optimized design compared to
8
American Institute of Aeronautics and Astronautics
the baseline wing. The design which utilized pattern repetition is shown in Figure 12 and the surface area, volume,
and mass details are outlined in Table 7. The images of displacement, von-Mises stress, and buckling are not shown
here, but are summarized later in this section.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 11. Element density for 2D rib optimization without pattern repetition.

Table 6. TO without pattern repetition surface area, volume, and mass compared to baseline wing.

Figure 12. Element density for 2D rib optimization with pattern repetition.

Table 7. TO with pattern repetition surface area, volume, and mass compared to baseline wing.

9
American Institute of Aeronautics and Astronautics
An individual redesigned rib was also designed as a means of comparing structural performance for a TO design
compared to the baseline rib. The practice of using large holes to reduce weight is common in most traditionally
manufactured aircraft due to its simplicity. This is a heuristical process which aircraft designers use as a fast and
efficient way to design the ribs. However, this is not necessarily the case for more radical wing shapes. An
established TO process provides an opportunity to quickly develop a better design which can be manufactured
through traditional means. Figure 13 is a comparison between the individual redesigned rib for this study and the
baseline rib. This rib is 9.5 percent lighter than the baseline rib and does not exceed stress limits from the baseline
FEA.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 13. Shape comparison between baseline rib and to rib.

The skin thickness optimization resulted in a thick region of material near the root of the ring, as seen in Figure
14. In this design, the optimized rib was used for the problem setup and the skin and spars were optimized
concurrently. Red regions of this result approach the thickness limit and the blue regions approach the minimum.
This material distribution is expected since a stronger root intuitively resists bending on the wing in-line with the
minimizing compliance objective. Overall, this wing experienced similar peak stress levels compared to the baseline
wing, but with much less displacement. Once again, this design is extremely difficult to manufacture without AM.
However, it used the same amount of material as the baseline.

(a) Optimized Skin (b) Optimized Spars

Figure 14. Local TO results for spar and skin free-sizing optimization of local wing
design.

Since buckling of the baseline wing occurred on the skin at the root of the wing, the skin thickness optimization
had a significant impact on the buckling location and the BLF. As seen in Figure 15, buckling shifted laterally
towards the wingtip. In addition, the BLF values decreased significantly. This is a significant result since it shows
buckling location is controlled through the optimization process. It is important to consider this during problem
setup and analysis.

10
American Institute of Aeronautics and Astronautics
Figure 15. Buckling analysis for +6g, rolling right for spar and skin free-sizing optimization of local wing
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

design.

For the global optimization, the same objective and constraints were used as for the local designs. The difference
in these designs was the entire internal structure was considered a design space. Figure 16a-16d is a series of views
of the TO for the global optimization with a tip rib and rear spar for skin support. Even though other global TO
designs were calculated, they yielded similar results. For this design, a threshold of 0.20 was used to easily view the
internal structure. For all global results, an I-beam like structure was formed laterally on the wing. This makes sense
since I-beam structures oppose bending. These designs also show a need for very little support near the tips of the
wing. Intuitively, some support is required. This is most likely a function of variable density TO. Unfortunately, a
software fault prevented an FEA or a post-processed design.

(a) Upper Trailing Edge Iso (b) Lower Leading Edge Iso

(c) Top (d) Top, Sliced


Figure 16. Various view orientations for global TO with density fraction threshold of 0.20.

11
American Institute of Aeronautics and Astronautics
In order to integrate a dual-purpose component into the wing, the fuel tank design was influenced by previous
optimizations. Areas where the TO created high-density elements, notably near the root of the wing along the main
spar, were considered a template for the fuel tank. Loading conditions on the wing surface remained the same.
Pressure from the fuel was not considered for this optimization. Figure 17 is the resulting TO for the baseline wing
with the integrated fuel tank. In this optimization, only the ribs were considered a design space. After this design
was post-processed, a skin and spar thickness optimization was also conducted. Surface area, volume, and mass is
compared to the baseline wing in Table 8.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 17. Results for fuel tank to with rib only design space.
Table 8. TO of ribs with integrated fuel tank surface area, volume, and mass compared to baseline wing.

Finally, the fuel tank was integrated into the global design. The fuel tank for this design is similar shape to the
local TO, but with slight shaping to mimic the I-beam like structure. The results in Figure 18 show high
concentrations of material near the tank. Once again, there is obviously a need for support material near the tip of
the wing. However, the low density fraction indicates only minimal support is needed for these loading conditions.
As with the earlier global design, these optimizations could not be exported into a working CAD file to be analyzed
or manufactured.

Figure 18. Results for fuel tank global to with density fraction threshold of 0.20.

A relatively recent concept for TO is the use of lattice structures. Lattice structures replace a partial dense
element with a series of lattices to represent the comparable density fraction of the element. Figure 19 is the global
12
American Institute of Aeronautics and Astronautics
TO with a fuel tank processed to consider lattice structures. This capability is new in the Optistruct optimization
software and did not support CAD export at the time of this research. FEA analysis was conducted on this design.
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 19. Lattice structure design for fuel tank Global TO.

Table 9 outlines the performance this design along with the other optimized wings and the baseline wing. All of the
localized optimizations at least partially reduced mass compared to the baseline wing. The best improvement was
seen with optimization using the skin thickness optimizations. This indicates a significant advantage in manipulating
thickness to provide support in critical areas. Peak von-Mises stresses remained similar but the displacement was
significantly decreased and the BLF was significantly increased. Without the thickness optimization, there was
improvement in performance, but to a lesser degree. The only global design examined through FEA was the lattice
structure design. Even though the mass of this design was significantly increased, performance was greatly
enhanced. The inability to modify this design limited the engineering to compare mass and structural performance.

Table 8. Performance Summary of all TO Results.

The final step of this research was to verify the feasibility of utilizing AM for production. Even though AM is
often regarded as an ability to manufacture any design, this is not necessarily the case. All forms of AM have their
limitations. The first AM model, shown in Figure 20, is a photopolymer print of the TO rib design with integrated
fuel tank. Note the top skin of the wing was removed to show the internal structure. Detail in this part is relatively
high, with a build layer of 16-microns. All models for this research were printed at 1/12th scale. In order to build the
part without structural failure, the thickness was significantly increased relative to the actual design. Even so, great
care was taken during post-processing to avoid damaging the part. Overall, post-processing consisted of simply
removing the soft support material with a water-jet.

13
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 20. Plastic 3D printed model of local to integrated with fuel tank with top skin removed.

An aluminum model of the same wing was printed using Direct Metal Laser Sintering (DMLS). In this case, the
build layer is 40-microns. A significant disadvantage to DMLS is the inability to have sharp overhangs in the
material. This made applying the top skin difficult since using removable support material was not feasible for the
internal structure. Post-processing of the metallic AM parts is also more involved. The aluminum AM parts are
shown in Figure 21.

Figure 21. Aluminum am model of local TO integrated with fuel tank with top skin removed.

As previously discussed, a lattice structure model was not exportable with the available software. However, a
lattice structure model was designed by Within Engineering based on the loading conditions discussed earlier. Based
on the initial conditions provided, Within was able to generate a global TO solution and then incorporate their lattice
structure. This model was also built out of aluminum using AM. The AM model is shown in Figure 22. Notice the
lattice structure has varying lengths and diameters based on the structural requirement in the local region.
Unfortunately, an FEA analysis on this design was not completed at the time of this research.

14
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF CINCINNATI on January 9, 2016 | http://arc.aiaa.org | DOI: 10.2514/6.2016-1246

Figure 22. Aluminum AM model of lattice structure design.

IV. Recommendations and Future Analysis


Through this research, it was quickly realized the significance of mesh size for any TO. The chosen mesh size
for all three-dimensional optimizations was 1.5 cm on each side of the cube. With this size, computational run time
was on the order of 36 hours per optimization. A true, high-fidelity optimization would require much finer mesh but
the computation power required would increase exponentially. It is recommended to establish a solid strategy in
setting up the model, and then decreasing the mesh size to allow for more accurate results. This is especially
important when determining the placement of thin structures, such as the ribs in this research. It is possible a finer
mesh would result in significantly different results for the global optimization indicated by the somewhat sporadic
material placement near the tip of the wing. In addition, it is recommended to consider a broader spectrum of
loading conditions to best represent the stresses experienced by the aircraft. The wings in this research only
considered static surface pressure conditions. Dynamic loading and forces created from control surface deflection
will most likely impact the final results.
It is recommended future work should focus on lattice structures as a means to interpret variable density TO.
Lattices are an excellent way to represent the true intent of a variable density element. AM allows for manufacturing
of these designs which were not previously feasible. Finally, future work should consider a broader variety of
objectives and constraints. Even though the designs in this research showed improvement in the regions tested,
greater focus on the real objective of the optimization may yield superior results.

Acknowledgments
The authors would like to thank Mr. Dennis Lindell at the Joint Aircraft Survivability Program Office (JASP) for
supporting this research.

References
1
Walker, D., Liu, D., Jennings, A., “Topology optimization of an aircraft wing,” in 56th AIAA/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, 2015.
2
Sigmund, O., “Design Of Multipysics Actuators Using Topology Optimization; One Material Structures,” in Computer methods
in Applied Mechanics and Engineering,” Vol. 190, No. 49, pp. 6577, 2001.
3
Sigmund, O., “A 99 line topology optimization code written in Matlab,” Structural and Multidisciplinary Optimization, Vol. 21,
Springer-Verlag, 2001, pp. 120-127.
4
Altair Engineering Inc., “Optistruct 12.0 User's Guide,” 2013.
5
Raymer, D. P., “Aircraft Design: A Conceptual Approach.” American Institute of Aeronautics and Astronautics, Reston, VA,
2012.
6
Van’s Aircraft, Inc., “RV-4 Construction Manual,” RV-4 Preview Plans, Aurora, Oregon, Jan. 2005.

15
American Institute of Aeronautics and Astronautics

You might also like