Topologically optimized quad rotor UAV
- Arnab Chattopadhyay, Sanat Munot, Vishal Bagthalia,
Sumod Nandanwar
MIT,Pune
Abstract
Topology optimization is an effective tool for designing light weight and high
performance structures which for a considerable time been applied successfully in
the automotive industry, but it has not become a mainstream technology in
aerospace industries. Meanwhile Additive manufacturing has also advanced
significantly, resulting in more efficient techniques and better materials. When these
two technologies are used in synergy, aerospace structural design can be carried out
in a unique way, yielding high strength and least weight structures. Through this
paper we have demonstrated the potential of this synergistic design methodology by
totally redesigning the traditional quad rotor unmanned aerial vehicle (UAV).This
paper describes in detail the multiple iterations of design carried out, various
optimization considerations and analysis of the resulting airframes. The final
optimized multirotor airframe that we have designed is effectively stronger yet lighter,
hence showing that this design process can be used to design significantly better
airframes for larger tactical UAVs, and in near future, for manned aircrafts too.
Introduction
Ever since the Wright brothers first took flight in 1903, building a stronger and faster
yet lighter aircraft has always been the greatest challenge faced by any aerospace
engineer. And for that purpose aerospace engineers have recently started adopting
topology optimization, which is a mathematical technique based on the Wolffs law
on bone growth in different environments. Topology optimization essentially
optimizes the material distribution of a structure in a given domain, to reduce its
weight and maximize its performance. Topology optimization results in complex 3D
geometries which cannot be manufactured using subtractive manufacturing. To get
the full benefit of topology optimization, additive method of manufacturing is the right
method for fabrication of the optimized structure. Hence we carried out a design
process that uses these two technologies synergistically to produce a robust but
lightweight airframe.
The design process implemented by us mainly comprises of three phases: The first
phase involves the real time acquisition of data to accurately model the system and
the environment in which the drone is going to operate. The second phase involves
the feeding of this data to the computer and setting up of the model for accurate
topology optimization. The third and final step of the design process comprises of
the highly iterative process where the different optimization parameters are tweaked
to obtain the fully optimized structure. The final optimized structure is also
smoothened to remove any unwanted stress concentration points.
Additive manufacturing is used to prototype the actual part and to see the benefits of
the optimized airframe in real-time.
Topology optimization can help us build stronger yet lighter structures and this can
exponentially increase the performance of dynamic systems, especially in the domain
of aerospace engineering where weight reduction without compromisation of the
strength, is the greatest challenge. Drones are ideal systems to test the potential of
this design methodology because drones are unmanned, and that removes a lot of
human related design constraints, and also a need of an elaborated safety testing.
Quad rotor drones also have a simpler control system with lots of available
documentation which helps us essentially focus more the structural optimization of
the system.
Thus using the quad rotor drone as an example we want to show that topology
optimization coupled with additive manufacturing holds the key to design and
fabrication of the most optimally designed aerospace structures .This design
methodology not only helps us design stronger and lighter airframes but also helps
us design sustainable aircrafts that carry out less carbon emissions and hence hold
key to a sustainable future.
Theory
Force Analysis of generic Quad rotor UAV
In this section, a detailed static force and torque analysis of a traditional quad rotor
UAV, considering generic actuator configuration is done. Different in-flight loading
scenarios are also analyzed in detail.
A quad rotor UAV mainly comprises of four equally spaced coplanar rotors that provide
the body with six degrees of freedom (three translational and three rotational).
Understanding the dynamic relation between these four rotors whose varying speed
helps the drone move seamlessly is also key to understanding the different loading
diagrams.
Quad copter dynamics and loading scenarios
To optimally design a quad copter a detailed analysis of the different loading
conditions it will face is essential. The thrust vector which is directed perpendicular to
the plane of the rotors along with the torque due to the drag force generated are the
key forces experienced in flight hence we will systematically study them
Thrust
Thrust is the main cause of bending moment in the airframe and it is maximum at the
time of lift off. Hence maximum bending moment due to thrust experienced by the
airframe is at the time of lift off.
Thrust generated is mainly dependent on two major factors
1) Speed of the Brushless DC motors
2) Profile of the propellers
The payload that we wanted to lift was estimated at 1.8 kg and considering a power to
weight ratio of 2 a thrust close to or slightly higher than 3.6kg is necessary.
Propeller Volts (V) Amps (A) Watts (W) Thrust (g)
10 11.0 110.14 760
8045 11 12.5 137.50 920
12 14.5 174.13 1050
10 13.5 135.10 870
1047 11 16.0 176.17 1050
12 18.0 216.30 1140
Hence we decided to go ahead with a thrust generating system which comprised of a
propeller with 8 diameter and 4.5 pitch and a 1100 kV motor with a 11.1 A battery
pack and after reading the specifications of the motor we decided to use a 30A
Electronic speed controller to actuate our Brushless DC motor.
Hence based on this actuation system the loads of thrust was calculated.
Torque
The thrust is generated by pushing the air downwards and in that process as a reaction
the air exerts a reaction force which is essentially the drag force. Due to this drag force
torque is experienced at every motor mounting. The torque is directly proportional to
the drag force generated. Which can be given as:
Torque=FD*(propeller diameter)/2
The 8 propeller used by us has an effective area of 200 mm2 calculating at a
maximum rpm of 8490 we get a Torque of about 1.3N-mm.
Landing
Rough landings also cause a lot of force to be transmitted to the airframe of the drone.
Hence designing and optimization were to be done according to these static loads too.
While designing our airframes the main purpose was surveillance hence an
assumption was made that a pilot with considerable experience would be flying it.
Keeping that assumption in mind the load cases of landing and crashing were
formulated.
For the landing case a free fall from a height of 5m using the inbuilt accelerometer and
gyroscope an impact time of 0.2s was calculated at an 8 kHz sampling rate which
results in an impact force on the landing gear of about 60N (with a 1.5 factor of safety)
Crash
For understanding the loading scenario of crashes we carried out multiple flight tests
and crashing scenarios on a standard test airframe, in most of the crash scenarios
experienced the load was transmitted essentially as bending load.
And the crashes to which the optimized drone is resistant is essentially when the
crashes occur at top speed of the drone which is 72 kmph or 45 mph which translates
to 20 m/s and with the measured impact time of 0.2s gives an impact load of 100N
hence the drone will experience a max bending load of 100 N when a head on crash
occurs and the propellers are hit.
Topology optimization
Topology optimization essentially helps us come up with the most optimum material
layout, which makes our drone stronger but also lighter at the same time.
The key concept behind topology optimization is discretization. Rather than dealing
with the system as one single unit, the domain is divided into a discrete number of
points by doing so a the local effects of the loading scenarios are captured and
countered effectively based on the fundamental properties of the materials and a
highly optimized solution is obtained.
The optimization problem can be formulated as follows using the SIMP approach:
The design variable is density of every discrete element which can be represented by
D= (d1, d2, d3, [Link]); d varies between dL di dU theoretically the lower bound is zero
and the upper bound is the full density of the material. The Objective function is to
maximize the stiffness which can be carried out by minimizing the objective
compliance i.e. a product of the displacement vector and force vector. The stiffness of
the material is linearly dependent on the density of the material.
Once the optimization problem is defined it can be solved in a number of ways the
most common way is the gradient search algorithm with the use of advanced
approximating concepts.
Additive manufacturing
Additive manufacturing at its core is very similar to the Topology optimization
algorithms we used, to further expand on that, additive manufacturing also approaches
manufacturing in a different way, rather than manufacturing the part at once, in
additive manufacturing the part is divided layer by layer and is deposited one layer at a
time, in a heated environment. The heated environment enables proper fusion of the
layers which helps build a fully functional product.
We used Fused Deposition Modelling or FDM technology in our additive manufacturing
process. We used fused deposition modelling over other additive manufacturing
technique for a number of reasons.
1) Availability: We had access to a Wanhao 3D printer and also a Stratasys u-print
which helped us build prototypes faster.
2) Material: The drone does experience significant forces but they are
comparatively lot lower when compared to forces experienced by larger systems
like car chassis etc. Hence a polymer with good mechanical characteristics was
the ideal candidate, keeping that in mind we used ABS (Acrylonitrile Butadiene
Styrene) initially and later on Polycarbonate.
3) Infill: FDM gives us a lot of control over the infill of the part to be fabricated,
which is a huge advantage when designing parts that need to be lightweight.
Hence after we had tested and determined FDM to be our fabrication process and ABS,
Polycarbonate to be our materials we went ahead with the design process.
Optimization of a Quad-rotor UAV
Space Frame Structure
A traditional Quad-rotor UAV is a Space Frame type structure comprising of number of
components assembled together. The very first iteration for our study was a Space
frame structure.
Geometry:
CAD model of Arm CAD model of the Top Plate
The geometry of airframe we have opted for is simplistic and mainly comprises of two
parts in its assembly which is basically the arms of the drone and the central
electronics housing. This will enable us to get a more detailed comparison between
real-time data of our optimized frame and a similar un-optimized frame. The CAD
modeling of the geometry was done on SolidWorks.
Design Space:
Before the topology optimization study can begin, it is first necessary to determine the
maximum amount of volume that the geometry can safely occupy. This is known as
the design space and it represents the volume that will be iterated upon while the
optimization algorithm is working. Non-design parameters are those that are pre-
defined and may not be altered during the optimization. In case of the UAV, the areas
where the bolting points are located are to be treated as Non Design Space.
Design Space Mesh:
The design space volume was filled with Hex-Meshing (Brick) using Solid Map feature
of HyperMesh. This was done with element size of 2 mm. The resulting mesh that
was used as the design space for the topology optimization study can be seen in
Figure.
Material Definition:
The material chosen for the manufacturing was Wanhao ABS. The material was
selected on the basis of availability and the good rigidity offered by it. The material was
assumed to be linear isotropic (MAT1) and the values for Youngs modulus, Density
and Poissons ratio were entered.
The values taken from the data sheet were:
Youngs Modulus 2320 GPa
Poissons Ratio 0.35
Density 1040 kg/m3
Boundary Conditions:
1. Treatment of Bolt Holes
As in any finite element analysis, proper boundary conditions (BCs) are crucial if the
results are to be of any significance, and this is especially true for topology
optimization studies since these BCs will be the basis for the resulting distribution of
material. An example of this lies in the treatment of the boundary condition around the
region of a bolted or pinned joint. Bolts and pins can only transfer compressive load
unless they are bonded in place, thus they can only push on another surface that is in
contact. In a finite element model, however, this phenomenon is somewhat difficult to
capture as it requires the use of non-linear gap elements which have a very low
stiffness in tension to simulate the lack of connectivity. For the sake of time and
simplicity, the bolt holes in this model were filled with rigid (RBE2 elements) at the
acknowledged cost of reduced accuracy of topological results in the vicinity of the bolt
holes. These rigid connect all the perimeter nodes of the bolt hole to a single node in
the center at which the loads and constraints are applied.
2. Loads:
The loads were applied separately on both the components.
Load case for the Arm: A bending force is generated due to the Thrust of the motors.
Even considering a crash scenario, the force transmitted towards the motor mounting
points is generally in Bending. A total force of 100N was applied on the mounting
points considering a crash at a speed on 15.5m/s with an impact time of 0.8s.
Load case for the Central Frame: Since the central body houses the most expensive
control system of the UAV, it should take an impact force in all directions. Hence, a
force of 25N was applied on each edge of the frame.
3. Constraints
Constraints for the Arm: The arms of the UAV are connected to the central housing
with the help of bolts. Thus, the bolting points are to be constrained in all degrees of
freedom.
Constraints for the central housing: The frame was to be considered fixed at the
central portion in all degrees of freedom.
Finite Element
model of the Arm Finite Element model of the Central plate
4. Optimization.
The objective function of any optimization problem is to minimize or maximize a
certain response while meeting a prescribed set of constraints. For this it is necessary
to program the software to solve for the desired responses, and then choose limits to
these responses. In the case of the UAV optimization project, the objective function
was to minimize the mass while maintaining the integrity of the original design;
consequently, the response to minimize was chosen as the volume fraction.
The design constraint is put in order to satisfy the problem. The design of a highly
stressed component it would be ideal to define the elemental stress level as a
response and limit the stress to a set factor of safety below the material yield stress.
Considering a factor of safety of 4, the stress is to be limited to 9.25N/mm.
The Objective function is defined as Minimum mass target.
Stress Plot of the Arm, Base Model vs Topology Optimized Model
Stress Plot of the Plate, Base Model vs Topology Optimized Model
Results:
Component Stresses Mass of Percentag
s in Stresses in Mass of the the e Factor of
Safety of
the Initial The Initial model Optimized Weight the
Design Optimized Model Reduction Optimized
Componen
Design t
2.93
Arm N/mm2 7.66 N/mm2 125.35g 77.5 g 38.2% 4.8
Central 0.8
Housing N/mm2 5.28 N/mm2 81g 52g 43.54% 7
Observations and Conclusions for the first space frame iteration
Limitations of this Design:
The optimization workflow used while designing this model involved the use of
HyperMesh, wherein the model was meshed and analyzed on HyperMesh first to come-
up with the various Optimization parameters and targets. Then the model was set-up
using the OptiStruct solver to carry out the Topology on it.
Since HyperMesh has a very steep learning curve, setting up a finite element model
requires some time. The optimization was also quite time consuming even if a slight
change in the optimization parameter was to be done.
The complexity of the model made the meshing of the model unstructured. Due to this,
the results obtained after optimization had irregular contours. Due to the lack of any
post-processing tool, no changes on the results were possible.
The factor of safety kept for the model was kept higher than it should be.
Topology Optimization: Revised Workflow
The previous workflow adopted, had a few limitations which slowed down the overall
designing process. This was done by using the software solidThinking Inspire in place
of HyperMesh. To overcome these limitations, the optimization workflow was modified.
Altair, one of our sponsors, provided us with solidthinking Inspire and Evolve and high
RAM workstations and access to their tech team. These new software tools helped
increase our speed of iterations because of its ease to set the model up.
Monocoque Structure
The results obtained out of the Space-frame iterations were quite satisfactory. There
was still a scope for improvement for this type of UAV. The Space-frame comprised of
a total of 7 individual components and number of nut-bolts in order to connect the Arms
with the central plates. The space-frame approach was still quite traditional and there
was a need to reduce the number of components in order to achieve a greater structural
integrity. Since the arms and the central plates were analyzed and optimized separately,
the forces acting on them werent fully defined.
This led to the inception of a Monocoque i.e. a Unibody structure. The Monocoque would
be a single piece structure with four arms and a central body integrated.
There were two approaches while designing the Monocoque structure of our Quad-rotor
UAV keeping the overall dimensions the same.
Monocoque: Iteration 1
The first design concept was to come up with a minimalistic design based on
fundamentals of mechanics. The minimalistic design space obtained was then directly
optimized on Inspire.
The main object of this design was to build a more robust airframe with a highly stable
hovering for positional surveillance and photography.
Geometry:
CAD Model: Monocoque 1
Material:
The material chosen for the manufacturing was ABS M-30 Thermoplastic. This being
an Industrial grade material, its Structural integrity and functionality is far more
superior than any other ABS. The material was assumed to be linear Isotropic. The
values taken from the data sheet were:
Youngs Modulus 2320 GPa
Poissons Ratio 0.35
Density 1040 kg/m3
Model Set-up:
Model Setup: Monocoque 1
De-featuring the part into Design Space and Non-Design Space
Apply Loads at the arm supports
i. A downwards force of 100 N on each motor mounting points (Bending)
ii. A torque of 1.2 N-m on each motor mounting points
iii. An upward force of 25N on each Landing gear mounting points (Landing)
Apply fixtures in the central region
Apply mass of various components mounted on the UAV.
Symmetrical Design space about Axes.
Optimization:
Once the model is completely defined, the optimization parameters are to be set. This
is done by defining the allowable element size, allowable stress value and Mass
targets.
Topology optimization of the model is then carried out and the results are analyzed
and interpreted.
Multiple optimization iterations are carried out in order to obtain a practical and desired
design by tweaking the optimization parameters.
Optimised Model: Monocoque 1
After a satisfactory topology model is obtained, the model is to be refined and prepared
for manufacturing by removing the sharp contours and stress concentration points. The
PolyNURBS tool on solidThinking Inspire and Evolve is used to get a solid model closest
to the topology optimized results with precision and efficiency.
The model is Re-analyzed to check its Stress and Displacement plots.
Stress Plot: Monocoque 1
Displacement Plot: Monocoque 1
Finalized Design and Conclusion:
Final Model: Monocoque 1
The magnitude of the Stresses induced in the final optimized model was 11.67
MPa. There was an increase of 34.31% in the magnitude of stress when compared
to the Initial Base model.
The target Factor of Safety for this model was of 2.5. The achieved F.O.S. was 2.6,
which was quite satisfactory.
The mass of the optimized structure obtained was 546 grams.
Percentage weight reduction from the Base model: 29.5%
Monocoque: Iteration 2
In the second design concept, the design started by providing Inspire with the
maximum design space and the interpretation of its results constituted the design
space of the next iteration. This iterative process is referred as Multi-phased
optimization process.
The objective of this design approach was to build a UAV for on-the-go First person
view surveillance while making it highly maneuverability by keeping a low centre of
gravity.
Geometry:
CAD Model: Monocoque 2
The design started off with a complete box according to the UAV dimensions with the
various packaging considerations such as Motor mounting, Battery Slot, Electronics
assembly, etc. The model is exported to Inspire.
Material: ABS M-30 Thermoplastic
Model Set-up:
The box model is defined by:
Initial Model Setup: Monocoque 2
De-featuring the part into Design Space and Non-Design Space
Apply Loads at the arm supports:
i. A downwards force of 100 N on each motor mounting points (Bending)
ii. A torque of 1.2 N-m on each motor mounting points
iii. An upward force of 25N on each Landing gear mounting points (Landing)
Apply fixtures in the central region
Apply mass of the various components mounted on the UAV.
Symmetrical Design space about Z Axis.
Optimization:
Once the model is completely defined, the optimization parameters are to be set. This
is done by defining the allowable element size, allowable stress value and Mass
targets.
Based on the results obtained are interpreted and the initial box modeled is redesigned
on SolidWorks.
Revised Model Setup: Monocoque 2
The newly obtained model is again taken to Inspire for another stage of optimization.
After fairly optimized base model is obtained, final topology of the model is carried out.
Final Optimised Model: Monocoque 2
After a satisfactory topology model is obtained, the model is refined by using PolyNURBS
on Inspire and Evolve.
The model is Re-analyzed to check its Stress and Displacement plots.
Stress Plot: Monocoque 1
Displacement Plot: Monocoque 1
Finalized Design and Conclusion:
Final Model: Monocoque 2
The magnitude of the Stresses induced in the final optimized model was 15.63
MPa. There was an increase of 97.35% in the magnitude of stress when compared
to the revised base model.
The target Factor of Safety for this model was of 1.8. The achieved F.O.S. was 2;
hence the design was safe according to our considerations.
The mass of the optimized structure obtained was 494 grams.
Percentage weight reduction from the Revised Base model: 35.76%
Fabrication
To see the results of our design in real time fabrication of our part was necessary. We
used additive manufacturing for fabricating the part because additive manufacturing
deposits layer by layer and hence highly complex structures can also be obtained with
minimum waste of materials. With advancement in 3D printing materials and
technology building of functional parts has become possible. In this section we will
deal in detail with fabrication of our drone airframe.
Space-Frame Structure:
When the first iteration was finalized and to be manufactured, due to lack of
inexperience in 3D printing a 50% scaled down version of airframe was first printed.
50% Scaled Down Model: Space Frame
Post printing of 50% scaled down version of our airframe model a close inspection was
done and different support removal processes and post processing processes were
tried and tested following which the full scale model was printed. The model was
printed on Wanhao i3 V2.
Different printing orientations were also tried out and after testing it for a variety of
infill % we decided upon the final printing configuration for the full scale model.
Full Scale Model: Space Frame
The major limitations in the first space-frame iteration were as follows: ABS plastic
used was not an industrial grade ABS hence its structural integrity was uncertain.
The FDM printer used was a desktop printer hence it had a limited accuracy, chances
of warping at higher layers and low quality printing.
High amount of post processing was [Link] overall assembly was quite bulky and
could be compressed drastically.
Monocoque Structure:
Monocoque 1
Monocoque 2
The dimensions of the designed Monocoque airframes were large and being one
structure 3D printing it on a standard desktop printer was not possible hence a larger
dimension 3D printer was need for fabrication, we used a Stratysys Fortus 450mc
which has a higher bed size of 406*406*305 mm and a higher accuracy. We processed
the file on Catalyst Ex and carried out the print in ABS M30. But before printing, to get
an experience of use of soluble support we carried out printing of a miniature model in
ABS plus in a nearby 3D printing facility using a Stratasys u-print and post dissolution
of the support in the Wavewash machine a very fine print was obtained hence we
decided to ahead with the 3D print in the Fortus 450 mc.
Future scope for improvement
1) Testing: a detailed failure testing could not be carried out mainly due to the high
cost of fabrication of the prototype. Tests were cautiously planned, this was cost
effective but we could not generate enough data to truly validate the simulation
data.
2) Treatment of the body as a laminar body: In this iteration the body is treated as a
single solid part whereas due to the use of additive manufacturing the modelling
of the body as a laminar body will make it more accurate.
3) Anisotropic materials: In the current analysis ABS is considered to be an
isotropic material for ease of calculation and use but ABS is an anisotropic
material and hence this assumption shall be dealt with in the next iteration.
4) Lower Factor of Safety (F.O.S): For this current iteration factor of safety was
taken high as a measure of precaution towards the unknowns of the
manufacturing process and due to a cost constraint in the future iterations the
F.O.S shall be reduced to give a lot lighter yet equally strong design.
CONCLUSION
The strength and weight of this airframe demonstrates the potential of this design
methodology. In an era where Artificial Intelligence and Machine learning is developing
at an extremely fast pace, this design offers a small glimpse at how designing will be
done in the future. In the future the computer will actively be taking part in the design
process along with humans, rather than just being a documenting device, the computer
will be suggesting ways to better and optimize our designs.
Topology optimization uses the computational power available to us most effectively
to help the engineers come up with a truly optimized design. With the use of additive
manufacturing, new gates of design freedom are opened which have the power to
bring this design into reality. Together these processes show how technology
combined with human cognitive power can yield way better results than the traditional
methods of designing.
By redesigning a traditional quadcopter we have tried to show how effective this
design process is, even when it is applied to the most elementary unmanned system.
But having said that, the true potential of this design process can only be reached
when airframes of tactical UAVs can be redesigned in this way. But we do recognize
that for that to happen, a lot more detailed testing and research is needed. UAVs are
the ideal systems to test the prowess of this design methodology on, mainly because
of the absence of man related constraints and lack of the need of meticulous safety
testing. By optimizing and testing UAVs designed by this process we could pave a way
to similarly designed manned vehicles too, which are stronger, faster yet lighter. The
time when all cars could become electric is very close but for planes to be powered by
some form of electric propulsion we need to technologically advance in a lot of
domains. Hence this is a small attempt to make this future a reality, a future which is
sustainable for humanity.