Machines 10 00149
Machines 10 00149
Machines 10 00149
Article
The Design Process of an Optimized Road Racing Bicycle Frame
Alexandros G. Kitselis 1,2, * , Chrysavgi S. Nikolakea 2 and Dimitrios E. Manolakos 1
Abstract: This paper recommends an alternative designing process for a superior road racing bicycle
frame manufactured from composite materials that is much faster than typically used design pro-
cesses. The main design goal is for the rider to be faster under the same riding conditions and with
the same effort made. This performance gain is the result of a combined structural and aerodynamic
optimization process used during the design process along with the selection of the materials. As the
needs of the rider are the focus of this design proposal, the optimization can be carried out only after
they are understood. The main difference in this approach compared to the typically used method-
ology is that, instead of analyzing the frame as a whole from the beginning of the design process
and the CFD and CAE iterations, we examine each candidate part of the frame separately. After
evaluating the parts’ performances, we select those that performed better to create a single frame.
This final frame design is used to choose the appropriate layup that would meet the performance
needs of the riders and the necessary safety regulations. The benefit of this approach is that the design
time is reduced, allowing the product to reach the market faster. Furthermore, it is more convenient
and easier to make any modifications required by marketing or regulations.
Keywords: bicycle frame design; spaceframe; carbon fiber composites; computer-aided design; finite
element analysis; computational fluid dynamics; laminate optimization; quasi-isotropic layup
Citation: Kitselis, A.G.; Nikolakea,
C.S.; Manolakos, D.E. The Design
Process of an Optimized Road Racing
Bicycle Frame. Machines 2022, 10, 149. 1. Introduction
https://doi.org/10.3390/
After their introduction and mass production, composite materials such as polymer
machines10020149
matrix composites have been increasingly used in mass transit applications [1] and con-
Academic Editor: Xiaosheng Gao sumer products. This was because they possessed superior properties, such as strength,
fatigue life, stiffness, low weight [2], and sometimes, aesthetics. New recycling methods
Received: 9 November 2021
Accepted: 4 February 2022
help with the further expanding use of these materials [3]. Pre-Impregnated Carbon Fiber
Published: 18 February 2022
Reinforced Plastic (CFRP) sheets, in particular, are a revolutionary material for composites
manufacturing. They are lighter, more durable in harsh conditions, and allow more aero-
Publisher’s Note: MDPI stays neutral dynamic shapes to be manufactured. Non-CFRP frames are comprised of many separate
with regard to jurisdictional claims in
tubes, which are then united by using different methods, such as soldering (if they are
published maps and institutional affil-
metal frames). This creates a “bumpy” surface at the tubes’ joints, which in turn may
iations.
create turbulence in the flow when riding. CFRP frames may also have separate tubes,
but their joints are not visible over the paint. The smooth surface allows the air to flow
around it without causing turbulence. Furthermore, CFRP can be used to create a plethora
Copyright: © 2022 by the authors.
of shapes that would not be easily manufactured with traditional materials, eliminating
Licensee MDPI, Basel, Switzerland. any manufacturing restrictions that non-CFRP frames face. In addition, the correct use
This article is an open access article of their orthotropic properties by the engineer can create products that are stiff in one
distributed under the terms and direction and more compliant in another, leading to great design freedom. This makes
conditions of the Creative Commons them an ideal material for bicycle frame manufacturing by offering the advantages of high
Attribution (CC BY) license (https:// strength and stiffness to weight ratios, long fatigue life, low density, corrosion resistance,
creativecommons.org/licenses/by/ wear-resistance, and environmental stability [4] along with low weight. For example, a
4.0/).
carbon/epoxy bicycle frameset may weigh less than 1 kg, and this is much less than the
5 kg weight of a same steel frame [4].
Advanced knowledge of modern CAD modeling programs, such as SolidWorks,
CATIA, Creo, Siemens NX, etc., is necessary to make the design and evaluation of the
initial frame easier and its modifications more flexible until a final shape is attained [5].
The creation of CAD models instead of physical prototypes during the design process is
faster, costs less (both because of time saved and material waste), and allows the engineers
to test much more virtual solutions compared to actual prototypes. The above factors make
the creation of CAD models a much more efficient method to design products [6,7].
For an evaluation of the aerodynamic performance of the resulting bicycle frame
from each iteration step during the design process, CFD is used to easily set up and solve
the complex equations that describe the airflow around the bicycle [8,9]. Depending on
available computing power, many different scenarios can run simultaneously, resulting in
a very fast bicycle frame in less time compared to physical prototypes and wind-tunnel or
real-life testing. A frame that is manufactured by taking into consideration aerodynamic
principles is faster under the same conditions than a non-aerodynamic frame because it
requires less effort (and therefore, energy output) by the rider [10]. These CFD tools are the
same for the ones used in the analysis of various other applications, such as aircrafts [11,12],
wind turbine blades [13], and underwater vehicles [14].
The design of a state-of-the-art bicycle frame also requires a precise understanding of
the applied loading conditions along with a deep knowledge of the stress development and
failure analysis of composite materials [15]. Based on theoretical or numerical calculations,
the strength and stiffness of the bicycle structures can be predicted and modified to be made
optimal before the manufacturing of the prototype takes place. Finite element analysis
(FEA) is used to calculate stress/strain fields and fatigue life of the structures [16]. It is
a computational tool that is widely used in the automotive industry and manufacturing
of composite leaf and coil springs [15,16], suspension lower arms [17], in hovercrafts [18],
bus manufacturing [2], wheelchair wheels for athletes [19], aircrafts [20], and wind tur-
bine blades [21]. In bicycle manufacturing, application examples include the design of a
composite crank [22] and frames [7,23–27].
The efficient selection of materials and layup for the frame requires the evaluation of
its performance under different loading scenarios and material choices. This evaluation
includes a variety of composite materials, a wide range of ply angles, and a different
number of plies in each part of the frame [28].
This paper presents the whole process of such a project by making an introduction
to all the steps needed to design and test the mechanical and aerodynamic properties of
a composite road racing bicycle frame by using computational tools with user-friendly
environments (with the aid of pre-processors, solvers, and post-processors). The results are
visualized for faster and better reviewing.
What makes our designing process different is the manner of the approach. In such a
design scenario, the designer would typically start with a given geometry for the frame
(typically from bike fitting statistics) and then make a full 3D model of it. The shapes of the
tubes are initially chosen for aesthetical reasons and in a way in which all the necessary
components can be assembled on the frame. In the CAD environment, the components are
fitted on the frame. Then, a typical-sized rider for the given bicycle frame size is fitted on
the assembly of the frame and the various components. Next, CFD and FEA simulations are
executed to improve the shape of the tubes. This can be performed either manually by using
a parametric optimization tool or with the use of a topological optimization tool. Due to the
complexity of the bicycle tube connections and their interaction with the other components
and the rider, a lot of effort is needed to redesign or modify the existing parts. In addition,
it is very demanding on computer time, and especially on pre- and post-processing during
computational fluid dynamics analysis and finite element analysis.
Instead of using this strategy, our approach was to start by using a spaceframe model
from a bike fitting analysis, which was gradually modified to be able to accommodate the
Machines 2022, 10, 149 3 of 41
necessary parts, such as the wheels. Following this, a fast FEA analysis was conducted
using beam element spaceframe models to determine which part of the frame received
loads in the different loading scenarios. The next step was to design various cross-sections
for each tube that seemed to be able to receive the loads that each part of the frame bore.
Then, these shapes were tested in CFD for aerodynamic performance as well as in bending
and torsional loads for structural evaluation. The selection of the most suitable shape
for each tube was made by comparing the advantages of each cross-section model and
the load requirements of each part of the frame. The 3D design that derived from the
combination of the selected spaceframe model and cross-sections for each tube was used to
find the appropriate laminate layup that would satisfy the security regulations set by the
ISO standards and would have the required lateral stiffness for performance and vertical
compliance for comfort.
This approach is less time-consuming because the simulation of separate single tubes
is much faster compared to designing and simulating a full model in each design iteration.
Compared to full-frame analysis in each iteration, we expect that the result in CFD may
be a little different, but the actual percentage of drag for the frame alone compared to the
full bicycle rider combo is too small for this to matter. Any possible structural issue, for
example, a stress raiser point, can be fully covered by changing the layup or changing
the composite material. The above means that this method can bring the product to the
market faster and with less cost for the designer and/or manufacturer, thus making it more
competitive on the market.
Figure 3. Basic tube diagram of a typical frame [30]. No changes were made to the original illustration
(Keithonearth, CC BY 3.0).
Firstly, we wanted the resulting frame to comply with the standards set by the UCI.
From the regulations set by the UCI, the following must be true:
• The frame must be of a traditional pattern, i.e., built around the main triangle, which
consists of the top tube, head tube, down tube, and seat tube.
• It shall be constructed out of straight or tapered tubular elements (which may be
round, oval, flattened, teardrop-shaped, or otherwise in cross-section) such that the
form of each element encloses a straight line.
• The elements of the frame shall be laid out in such a way that the joining points shall
follow the following pattern: the top tube connects the top of the head tube to the top
of the seat tube; the seat tube (from which the seat post shall extend) shall connect to
the bottom bracket shell (which houses the bearings for the crankset); the down tube
shall connect the bottom bracket shell to the bottom of the head tube.
• The rear triangles shall be formed by the chain stays, the seat stays, and the seat tube,
with the seat stays anchored to the seat tube at points falling within the limits laid
down for the slope of the top tube.
• The maximum height of the elements shall be 8 cm and the minimum thickness 1 cm.
• The minimum thickness of the elements of the front fork shall be 1 cm; these may be
straight or curved.
• The top tube may slope, provided that this element fits within a horizontal template
defined by a maximum height of 16 cm and a minimum thickness of 2.5 cm [31].
Machines 2022, 10, 149 6 of 41
Figure 4. Cont.
Machines 2022, 10, 149 7 of 41
As shown in Model 1, we started with a basic bicycle frame that lacked shape and
included only the basic tubing (i.e., top tube, down tube, seat tube, chain stays, and seat
stays). Then, the bottom bracket was added to accommodate the crankset in Model_2.
The headtube was added to house the steerer tube of the fork in Model_3. The seat stays
were lowered to lower the center of gravity in Model_4 and then shaped to accommodate
the wheel in Model_5. The chain stays were shaped to create clearance for the wheel and
the crankset in Model_6. These first six models were each an evolution of the previous
spaceframe model by adding the necessary geometries and accounting for the various parts
of the bicycle. The next four models, mono-stay, dual-stay, complex, and optimized, were
variations of Model_6 with small differences between each other. They all presented a
curvature on the middle and bottom part of the seat tube to create the necessary space for
the rear wheel to fit. In the mono-stay model, the two seat stays were united into a single
tube before coming into contact with the seat tube. In the dual-stay model, the two seat
stays tubes were directly connected to the seat tube. The complex model had the same
set-up for the seat stays as the dual-stay model, but the bottom bracket was lowered. The
optimized model further shortened the bottom bracket and the head tube lengths, while
uniting the seat stays before meeting the seat tube in a mono-stay set-up. The different
spaceframe models we developed and examined are shown in Figure 4. In each picture
of the various models, the red line encircles the main difference compared to the previous
model. As far as the complex and optimized models are concerned, the variations were
many and relatively small, therefore no highlights are shown on them.
conducted on the frame were: a horizontal forces scenario as shown in Figure 5a (with a
dynamic loading force of 600 N at a frequency of 25 Hz applied on each end of the head
tube); a static start-up in a hill climb scenario as shown in Figure 5b (with an overweight
rider in mind, and axial loading forces of 143.6 N inwards and 1090 N downwards on
each side of the bottom bracket shell); a riding scenario as shown in Figure 5c (with stable
chain and seat stays and axial loading forces of 20 kgf at each side pedal connected to the
bottom bracket, 70 kgf at the end of the seat tube, and two 20 kgf forces on each end of a
handlebar connected to the head tube); and in the end, two frame out-of-plane loading
scenarios (one with stable seat, bottom bracket, chain stays, and seat stays and axial loading
forces of 600 N at the head tube, as shown in Figure 5d, and one with stable seat, chain
stays, seat stays, and head tube and axial loading forces of 600 N at the bottom bracket,
as shown in Figure 5e). It should be noted that each spaceframe design model scenario
presented previously was tested separately for the first three loading scenarios, these being
the horizontal forces, the static start-up, and the riding scenarios. Only the last four models
(mono-stay, dual-stays, complex, and optimized) were tested on all five loading scenarios
because they were the only viable choices. In Figure 5, the first three loading scenarios
applied to the Model_6 spaceframe and the two out-of-plane loading scenarios applied
to the optimized geometry spaceframe are presented. Fixed constraints and loads are
represented with green and red colors, respectively.
Figure 5. Cont.
Machines 2022, 10, 149 9 of 41
Figure 5. Tested loading scenarios: (a) horizontal forces scenario with fixed constraints on the rear
ends of the stays, simple support on the bottom end of the head tube, and forces on either end of the
head tube; (b) static start-up scenario with fixed constraints on the rear ends of the stays and either
end of the head tube, and loads on either end of the bottom bracket shell; (c) riding scenario with
fixed constraints on the ends of stays, simple support on the bottom end of the head tube, and forces
on the top of the seat tube, either end of the handlebar and the crankset; (d) out-of-plane load to head
tube scenario with fixed constraints on the ends of the stays, top of seat tube, and bottom bracket
shell; (e) out-of-plane load to bottom bracket shell scenario with fixed constraints on the ends of the
stays and head tube.
Figure 6. Top tube cross-sections from left to right: circular 40 mm, rectangular 35 × 35 mm, triangular,
and complex.
Figure 7. Down tube cross-sections from left to right: oval 39 × 79, oval 45 × 79, oval 51 × 79, circular
60.7 mm, rectangular 50 × 50, complex 1, and complex 2.
Machines 2022, 10, 149 10 of 41
Figure 8. Seat tube cross-sections from left to right: complex 1, complex 2, complex 3, complex 4, and
circular 34.9 mm and 53 mm.
Figure 9. Seat stay cross-sections from left to right: oval 10 × 30, oval 17 × 30, rectangular 10 × 30,
and circular 12 mm.
Figure 10. Chain stay cross-sections from left to right: oval 15 × 45, oval 20 × 60, rectangular 15 × 45,
rectangular 20 × 60, and circular 18 mm.
The cross-sections designed for the top tube, shown in Figure 6, were:
• A circular cross-section, with 40 mm diameter.
• A square cross-section, 35 mm × 35 mm.
• A triangular cross-section with 37.69 mm height and 34.61 mm width.
• A complex geometry cross-section with 38 mm height and 40 mm width.
The cross-sections designed for the down tube, shown in Figure 7, were:
• Three oval-shaped cross-sections with 79 mm height and width 39, 45, and 51 mm,
respectively.
• A circular cross-section with 60.7 mm diameter.
• A square cross-section, 50 mm × 50 mm.
• Two complex geometry cross-sections with 79 mm height and 45 and 48 mm width,
respectively.
The cross-sections designed for the seat tube, shown in Figure 8, were:
• Four complex geometry cross-sections with the same height (60 mm) and various
widths, as shown in Figure 8.
• Two circular cross-sections with 34.9 mm and 53 mm diameter, respectively.
The cross-sections designed for the seat stays, shown in Figure 9, were:
• Two oval-shaped cross-sections, 10 × 30 mm and 17 × 31 mm, respectively.
• A rectangular cross-section, 10 × 30 mm.
• A circular cross-section with 12 mm diameter.
The cross-sections designed for the chain stays, shown in Figure 10, were:
• Two oval-shaped cross-sections, 15 × 45 mm and 20 × 60 mm, respectively.
• Two rectangular cross-sections, 15 × 45 mm and 20 × 60 mm, respectively.
• A circular cross-section with 18 mm diameter.
In order to run the simulations, all the cross-sections were positioned vertically. A
fixed constraint was applied to the bottom end of the tube and the loads (a bending load of
250 N and a torsional load of 50 Nm) were applied to the top end of the tube.
Machines 2022, 10, 149 11 of 41
The next step was to select the wind velocity Vw that a cyclist may drive through.
Using a wind atlas [33] similar to the one shown in Figure 12 as a guide, we can estimate
that the most common wind velocity is between 3 m/s and 8 m/s, which is the equivalent
of 3 and 4 Beaufort. We decided to use 5 m/s or 18 km/h wind velocity in our simulations
because it is in the middle of the range.
Figure 12. Wind atlas of Europe [33] 1989 DTU Wind Energy, formerly Risø National Laboratory.
The velocity of the rider is chosen to represent a typical velocity he might achieve in a
race. Under still wind conditions, a cyclist may develop a speed of 45 km/h; therefore, this
Machines 2022, 10, 149 12 of 41
would be one of our design points in the CFD analysis. In the case of a resisting wind of
18 km/h, like the one we chose earlier, we considered a rider velocity of 32 km/h.
The yaw angle of the total perceived wind is the next variable we needed to specify to
fully describe our CFD analysis. From Figure 13, it is obvious that the most common yaw
angles are between −12 and 12 degrees. We decided to use three different yaw angles in
our simulations, 2◦ , 7◦ , and 12◦ .
Table 1 lists the four design points we used for the CFD analysis as well as the values
of the different variables. We also listed the weight for the drag and the side force, this
being the same for each design point. This is because we wanted to account for both of
these forces and not solely for the drag in our analysis, as the side force impacts the overall
stability of the bicycle and rider and also the sense of security of the cyclist.
We know from Figure 12 and the wind atlas that a zero-wind velocity is not commonly
encountered, therefore we only gave this a 10% weight in the overall analysis. The rest of the
design points were given weights according to their probability, as shown in Figure 13. For
yaw angles 2◦ and 7◦ , the probability is almost the same, whereas, for 12◦ , the probability
is more than half of that for 2◦ and 7◦ . That is why design point 2 was given a 20% weight
and both design points 3 and 4 were each given a 35% weight.
Each of the cross-sections for the tubes was given the appropriate length and direction,
as indicated from a bike fitting analysis. For each cross-section, we created a file with four
design points where the computational area, the mesh, and the orientation of the model
were the same but the velocities in the × direction (driving direction of the cyclist) and the
Y direction (vertical to the driving direction) changed according to the conditions specified
in Table 1.
The computational domain was checked to ensure that it encompassed the entire
tube; however, as the orientation of the various tube models varies, so does the size of the
domain. The mesh we used was a structured Cartesian mesh that had a basic mesh size of
5 × 5 × 5 mm with a level four refinement level, which means that the mesh size near the
surface of the tube is 0.625 × 0.625 × 0.625 mm.
Machines 2022, 10, 149 13 of 41
The final step involved the start of the simulation and monitoring the process until its
finalization. Figure 14 shows the flow lines around chosen cross-sections of the seat tube in
design point 1 CFD analysis.
Figure 14. Flow and pressure area lines on seat tube cross-sections in design point 1. From left to right:
the two left pictures show the flow lines around the complex 1 cross-section from different views. The
next three pictures show the flow lines around the complex 4, complex 3, and complex 1 cross-sections
as viewed from above. The different colors represent the variation of pressure caused when the air
flows around the models (red indicates high pressure, blue is low, and green is atmospheric pressure).
The joint areas where two or more tubes meet were created by initially joining the
tubes and then creating smooth transitions between them. To avoid stress raising areas,
flow separation, and detachment of the boundary layer from the surface, the joints were
selected to have large curvature and G2 continuity with the surrounding surfaces.
Machines 2022, 10, 149 14 of 41
Figure 16. (a) FE method preview of drop mass test, (b) FE test method preview of the falling frame,
(c) FE test method preview of horizontal forces, (d) FE test method preview of vertical force through
the seat tube, (e) FE test method preview of pedaling forces.
Machines 2022, 10, 149 15 of 41
The red arrows represent the direction of the loads applied in each scenario.
The drop mass test, depicted in Figure 16a, is an impact test wherein a dummy fork is
assembled on the vertically supported frame and a 22.5 kg mass is dropped 212 mm above
the end of the fork. The mass is released and allowed to bounce on the end of the fork until
it comes to rest on it. The wheelbase is measured before and after the impact. We used a
non-linear dynamic simulation with an implicit solver, where the falling mass was placed
directly above the end of the fork just before the contact point. The load is the velocity that
the mass acquires because of the drop.
The falling frame test, as shown in Figure 16b, is also an impact test. The frame with
the assembled fork is constrained and able to rotate around the axle of the rear wheel.
Three masses are placed on the frame: a 30 kg mass on the seat post, a 10 kg mass on top
of the steering head, and a 50 kg mass on the bottom bracket. The wheelbase is measured
with the three masses and the end of the fork resting on the floor. The lower end of the
fork is then raised 200 mm above the floor, and the assembly is allowed to fall freely to
impact the floor; the wheelbase is remeasured when the assembly comes to a rest. In this
case, we again used a nonlinear dynamic simulation with an implicit solver, where the
assembled frame fork was placed just before the lower part of the fork comes into contact
with the floor. The loads are the angular velocity due to the rotation of the assembly under
the gravitational acceleration and the static forces due to the masses fastened on the frame.
Figure 16c features the horizontal forces test, which is a fatigue test wherein a horizon-
tal force of 600 N in a forward and a rearward direction is applied at the lower part of the
fork for 100,000 cycles, with a maximum frequency of 10 Hz. In this case, the load applied
is a nonlinear dynamic force and the simulation takes into account the large displacements
that occur.
The test of the vertical force through the seat post is another fatigue test and is
presented in Figure 16d. A dynamic vertical force of 1200 N is applied on the dummy
seat post 70 mm behind the intersection of the ax of the seat post and the flat part for
50,000 cycles, with a maximum frequency of 10 Hz. In this case, the load applied is a
nonlinear dynamic force and the simulation takes into account the large displacements
that occur.
The pedaling forces test, shown in Figure 16e, is also a fatigue test wherein dynamic
forces of 1100 N are applied on each side of a dummy crankset, whose geometry is specified
by the ISO. Both of the cranks are 175 mm long and are inclined forwards and downwards
at an angle of 45◦ to the horizontal. The crank is secured by a rod that connects the top of
the dummy crankset to the rear axle replacing the chain. Each pedal spindle is subjected
to a repeated downward force at a position 150 mm from the centerline of the frame in a
vertical, transverse plane and inclined at 7.5◦ to the fore/aft plane of the frame. The forces
are applied to the pedals for 100,000 test cycles. In this case, the load applied is a nonlinear
dynamic force consisting of two components, one downward and one inward, to account
for the 7.5◦ angle; the simulation takes into account the large displacements that occur.
For the simulations, the fork and all the peripheral parts were modeled as solid parts
made of steel with a 3D CTETRA (10) mesh. Their element size varied according to their
dimensions, curvature, and overall geometry. The frame was modeled as a sheet part where
each tube had a different 2D mesh to account for the various layup recipes and orientations
of each section of the frame. The 2D mesh in every case was a CTRIA3 mesh with a 1.5 mm
element size, which was able to accurately represent all the small features of the frame
without increasing the computational time disproportionally. To simulate the overlap of
the layup at the junctions between the tubes, a different 2D mesh was created in each case
where the layup recipe contained the plies of both the recipes of the respective tubes taking
into account their different orientation.
These quasi-type stacking scenarios were the benchmark to check further stacking
scenarios, under the same conditions as the FE analysis, to better understand how material
selection, sequence, and orientation affect the final results. Each tube of the frame has its
own orientation, therefore the angle of each layer is relevant to the element orientation of
each tube.
The two quasi-isotropic scenarios were used as a basis for the development of the
stacking scenarios. Depending on the distribution of the strain energy in the optimized
geometry spaceframe model, extra plies were gradually added on the tubes that bore most
of the stresses. Next, the load that each ply bore as well as its direction were evaluated. For
instance, if a woven ply bore stressed only in direction 11, it was replaced by a unidirectional
ply in the direction that was loaded to remove any unnecessary mass. In some stacking
scenarios, each woven ply was replaced by two unidirectional ones to check for any
differentiation between the use of the woven and their equivalent unidirectional materials.
The difference between the effects of the two sets of materials, T300 Woven with T700 UD
and M46J Woven with M46J UD, on the overall behavior of the frame was also tested. The
next step was to check the displacements for the various loading cases, which we wanted
to minimize. In the case of the pedaling test, an attempt was made to ensure that the
displacements caused by applying the forces on either side of the bottom bracket were
similar; this is why the layup recipe on the chain stays was differentiated.
Table 4 shows stacking scenario 1. The material used in this scenario was the T300
Woven fabric. This time, there was a differentiation in the layup recipe of the various tubes
Machines 2022, 10, 149 17 of 41
of the frame, creating two groups: the first group consisting of the headtube, downtube,
seat tube, and chain stay right with a [(45F/0F)3 ]T layup recipe, and the second group
consisting of the top tube, seat stay left and right, and chain stay left with [(45F/0F)2 ]T .
T300 Woven was also used in stacking scenario 2, as shown in Table 5. We now had
three different groups; the first with a [(45F/0F)4 ]T layup recipe on the headtube and chain
stay right, the second with a [(45F/0F)3 ]T recipe on the downtube and chain stay left, and,
finally, the third group with [(45F/0F)2 ]T on the top tube, seat tube, and both seat stays.
In stacking scenario 5 of Table 8, the only material used was the M46J UD. The
headtube followed a [(±45/0/90)3 ]T recipe, the top tube and both seat stays a [±45/0]T
recipe, the downtube a [(±45/0)3 ]T recipe, the seat tube a [±45/0/90/03 ]T recipe, the left
chain stay a [(±45/0)2 ]T recipe, and lastly, the right chain stay a [(±45/0)3 ]T recipe.
Stacking scenario 7, in Table 10, was also very similar to stacking scenario 4. The only
difference here was that instead of the T300 Woven, we used the M46J Woven, and instead
of the T700 UD, we used the M46J UD.
Table 11 shows stacking scenario 8, which also used the M46J Woven and M46J UD
materials and had many similarities to stacking scenario 7. The layup recipes for the down
tube, seat tube, and both the chain stays remained the same, whereas the headtube had
[(45F/0F)3 ]T layup and the top tube and both seat stays had [45F/0F]T .
In stacking scenario 9 of Table 12, there was a significant increase in the material in
each tube. Specifically, the headtube had the [(45F/0F)6 ]T layup recipe, the top tube and
both seat stays had [(45F/0)3]T , the downtube and left chain stay had [(45F/0)4 ]T , the right
chain stay had [(45F/0)5]T , and the seat tube had [45F/0/0F/02 /45F/0/0F/0]T .
Finally, stacking scenario 10, presented in Table 13, was basically the same as stacking
scenario 5, but instead of the M46J UD, the material used was the T700 UD. The headtube
followed a [(±45/0/90)3 ]T recipe, the top tube and both seat stays a [±45/0]T , the down-
tube a [(±45/0)3 ]T , the seat tube a [±45/0/90/03 ]T , the left chain stay a [(±45/0)2 ]T , and
lastly, the right chain stay a [(±45/0)3 ]T recipe.
In the beginning, all the spaceframe geometries were tested in three loading scenarios
and the final four models in five loading scenarios. The results of displacements, stresses,
and strain energy distribution are shown in Figures 17–25.
Figure 17. Maximum displacement per model in horizontal impact, in static start-up, and in
riding scenarios.
Figure 18. Maximum displacement per model in horizontal impact, static startup, riding, out-of-plane
load in the headtube, and out-of-plane load in the bottom bracket shell scenarios.
Figure 19. Maximum stress of horizontal impact, static start-up, and riding scenarios per model.
Machines 2022, 10, 149 22 of 41
Figure 20. Maximum stress per model in horizontal forces, static start-up, riding, out-of-plane load
in the headtube, and out-of-plane load in the bottom bracket shell scenarios.
Figure 21. Absorption of strain energy and its distribution in out-of-plane loads on head tube scenario.
Figure 22. Absorption of strain energy and its distribution in out-of-plane loads on bottom bracket
shell scenario.
Machines 2022, 10, 149 23 of 41
Figure 23. Absorption of strain energy and its distribution in a static start-up scenario.
Figure 24. Absorption of strain energy and its distribution in a horizontal forces scenario.
Figure 25. Absorption of strain energy and its distribution in the riding scenario.
Machines 2022, 10, 149 24 of 41
As the 2D bike fitting sketch result, from which we start, becomes an actual 3D
spaceframe road bicycle frame, it is obvious that the maximum displacements change a lot
in the beginning as the frame takes shape, and less in the final models, in which changes
do not have such a big impact (Figure 17).
The last four models (mono-stay, dual-stays, complex, and optimized) were also tested
on the out-of-plane load in the headtube and the out-of-plane load in the bottom bracket
shell tests. The results of all the loading cases tested on the last four models are presented on
Figure 18. As the design evolves, small but measurable gains are noticed. The redesigned
optimized model had the least displacement in all tests. Less displacement in the frame
means that the wheels do not lose their alignment during heavy pedaling sessions, which
translates into less watts needed from the rider under the same speed and riding conditions.
Another important factor is the max stress that comes as a result of the loading. With
less stress, less material is needed in the actual frame, which makes it lighter. The evolution
of the stresses as they change from the beginning to the final model is presented in Figure 19.
Following the same procedure as with the maximum displacement figures, Figure 20
presents the results of the maximum stresses for the last four spaceframe models in all the
loading cases, including the scenarios with the out-of-plane loads to the head tube and
bottom bracket shell. The optimized model did not only have the least displacements, as
noted before, but was also better in all the loading scenarios.
The end of this phase of the design process involved the examination of the energy
that each part of the frame receives during the various loading cases. These results would
be used during the selection of a suitable cross-section for each part of the frame.
In the out-of-plane load to the head tube case, it is obvious that most of the load is
absorbed by the downtube and top tube, which, when combined, are actually part of the
head tube. This means that the ends of the top and down tubes that connect to the head
tube need to be reinforced to account for such loads, such as when sprinting out of the
saddle. The strain energy distribution in the out-of-plane load to the head tube case is
presented in Figure 21.
In Figure 22, the results of the out-of-plane loading of the bottom bracket shell indicate
that most of the load is carried by the chain stays. This means that the chain stays have to
be stiff enough so as to not deform during hard pedaling.
The static start-up scenario simulates the start of a ride from zero speed where the
rider pushes at the pedals and most of the load is carried through the bottom bracket shell
and the chain stays, as shown in Figure 23. Again, this means that these parts have to
be reinforced.
In the horizontal forces, as shown in Figure 24, the headtube received most of the load,
with the top tube and downtube receiving less. Similar to the out-of-plane headtube loading
scenario, the area between the headtube, downtube, and top tube must be reinforced.
Finally, in Figure 25, a realistic riding scenario depicts the load distribution during
normal riding. The optimized model had a good distribution of strain energy among the
different tubes. It is interesting to see that the optimized and the complex geometries have
a much different distribution, although the actual loads are not so different. Specifically, the
strain energy percentage that is absorbed in the seat stays of the complex geometry seems
to be absorbed by the seat tube of the optimized geometry. This is due to the mono-stay
formation of the seat stays in the optimized spaceframe model, which lets the seat tube
carry most of the load. This depicts the importance of the design process.
After reviewing the results of displacements, stresses, and strain energy distributions
for all the spaceframe models, it is obvious that the optimized geometry model offers the
lowest displacements paired with the lowest stresses and the evenest distribution of strain
energy compared to all the other models. This geometry was used as the basis for the
creation of the final CAD model of the frame.
Machines 2022, 10, 149 25 of 41
Table 14. Static analysis of all tubes under bending and torsional loads.
From the static loads’ tests results, we can select the tubes that combine the least mass,
which is related to the surface area, the least developed stresses, and displacements. We
mostly used the bending load test results because the loads that appear on the various
tests that we tested the different spaceframe models on caused bending on the tubes of
the frame.
As far as the top tube models are concerned, the cylindrical and the rectangular
models have the best stresses and displacements, even though they are heavier than the
triangular and complex models. From the down tube models, the oval 51 × 79 model
seems to have the best structural behavior while being the heaviest of the cross-sections.
The cylindrical and the rectangular models are the lightest and have the best results in
Machines 2022, 10, 149 28 of 41
torsion but not in bending loads. From the tests on the seat tube models, the cylindrical and
the complex 1 models seem to have the best structural characteristics while also having the
biggest surface area and, subsequently, the biggest mass. For the seat stays, the models
that show the lowest stresses and displacements are firstly the oval 17 × 31 followed by
the rectangular 10 × 30, which are also the two heaviest cross-sections. Finally, the oval
20 × 60 and the rectangular 20 × 60 cross-sections are the structurally best choices for the
chain stays.
The results of the CFD analysis for each tube of the frame are presented using the
graphs of Figures 26–30. For each design point, the resulting value of each model was
0.85× (Drag) + 0.15× (Side Force). The weighted drag results utilized the resulting values
of each design point following this formula: 0.1× (design_point1) + 0.2× (design_point2) +
0.35× (design_point3) + 0.35× (design_point4).
From the results of the aerodynamic tests on the top tube cross-sections shown in
Figure 26, the two models that show the best behavior are the cylindrical and the triangular
models. While the triangular model may have a lower value in design point 1 and design
point 4, the overall weighted drag of the cylindrical cross-section is lower.
The aerodynamic results for the tests on the down tube cross-sections are presented in
Figure 27. The oval 39 × 79 cross-section has the best aerodynamic behavior of the down
tube models, as stems from the CFD analysis. The models that follow closely behind are
the other two oval cross-sections and the two complex ones. This remains the same over
the four design points.
In the seat tube CFD analysis, whose results are shown in Figure 28, the model that
shows the worst behavior is the cylindrical 53 cross-section; the rest seem to have small
differences across the four design points, and so any of them is a viable option.
From the results of the aerodynamic tests on the seat stays cross-sections, which are
shown in Figure 29, the one that appears to have the lowest overall weighted drag is the
oval 10 × 30, followed by the rectangular 10 × 30.
Lastly, from the results of the aerodynamic tests on the chain stay cross-sections
presented in Figure 30, the cylindrical 18 model has the least weighted drag for the chain
stay models, followed by the oval 15 × 45.
Table 15 shows an evaluation of each cross-section of the various tubes of the frame
based on the results of the structural and aerodynamic tests and their comparison.
The symbols of Table 15 have the following meanings:
• 4 indicates a good behavior–result
• 44 indicates a very good behavior–result
• 8 indicates a bad behavior–result
• 88 indicates a very bad behavior–result
• 888 indicates an unacceptable result.
After evaluating the results of all the tests, both aerodynamic and structural, we were
able to select the most suitable cross-section for each part of the frame. For the top tube, the
obvious solution is the cylindrical 40 mm as it performs well in all the tests. For the down
tube, which is one of the frame parts with the biggest amount of airflow, we decided to
combine the three oval-shaped cross-sections in order to benefit from the great aerodynamic
behavior and low mass of the oval 39 × 79 while also keeping the low displacements, and
therefore increased stiffness, of oval 51 × 79. The seat tube is a very important element of
the frame because of its angle and position. From the structural tests, two cross-sections,
complex 1 and the cylindrical 53, showed promise. However, in the CFD analysis, the
cylindrical 53 showed the highest drag in all the cases, therefore it had to be discarded as a
possible solution. This means that complex 1 was the model we decided upon, despite its
higher mass. The seat stay geometry was selected to be the oval 30 mm × 10 mm because it
showed the least drag in all the design points of the CFD analysis. The high displacements
of this model are actually welcome because this part of the frame has to absorb the vibration
while riding. As for the chain stays, a very good solution was the combination of the oval
and the rectangular cross-sections to achieve reduced aerodynamic drag force by designing
Machines 2022, 10, 149 29 of 41
the oval 20 mm × 60 mm as the outside diameter of the chain stay. Necessary stiffness was
achieved by designing the inside diameter as a rectangular model with a 20 mm × 60 mm
cross-section.
Table 16. Allowable tensile and compressive stresses for each material.
The results of the various stacking scenarios for each loading case were produced
using the maximum stress failure criterion. Tables 17–28 present the results of the different
stacking scenarios across the five loading cases of ISO and Table 29 shows the weights of
each stacking recipe.
Table 17. Stress and displacement results of M46J Woven quasi-isotropic scenario.
Table 18. Stress and displacement results of T300 Woven quasi-isotropic scenario.
The resulting stresses and displacements of the M46J Woven quasi-isotropic scenario
for all the loading scenarios are shown in Table 17. No stress exceeded the allowable
stresses of Table 16.
In Table 18, the stresses and displacements listed are the results of the T300 Woven
quasi-isotropic scenario for the five loading cases. Again, all the stresses were below the
allowable levels of Table 16. The displacements had small variations from the M46J Woven
quasi-isotropic scenario results.
Machines 2022, 10, 149 38 of 41
The results of stacking scenario 1 for each loading test are presented in Table 19. Once
again, all the stresses were within the allowable values for each material and significantly
lower compared to the two previous stacking scenarios. The displacements also improved,
especially in the drop mass, falling frame, vertical forces, and pedaling tests.
The results of stacking scenario 2 are presented in Table 20. There were some small
variations from the previous scenario and all the stresses were allowable.
In the results of stacking scenario 3, presented in Table 21, there was a slight increase
in the resulting stresses without any of them exceeding the allowable stresses of Table 16.
After evaluating the results of stacking scenario 4, presented in Table 22, the stresses
appeared to be elevated compared to the results of stacking scenario 3 while still remain-
ing within the allowable limits. There was also a slight elevation on the displacements,
especially in the drop mass, vertical forces, and pedaling scenarios.
In the results of stacking scenario 5, shown in Table 23, there was a slight decrease
in all the displacements, with a simultaneous increase in the stresses. While the resulting
stresses were still in the allowable limits, the minimum stress of 11 in the pedaling forces
test was approaching the limit of 600 MPa.
The results of stacking scenario 6, shown in Table 24, were similar to those of stacking
scenario 5 in most of the loading cases. The most apparent differentiation is in the vertical
forces and pedaling tests, where the displacements were lower and the high stresses were
diminished in most cases. The exception was located in the Min 22 and Max 22 stresses
in the pedaling forces test where there was an increase; however, these values were well
within the allowable range.
In stacking scenario 7, whose results are presented in Table 25, there was a further
decrease in the displacements and stresses of the drop mass, falling frame, and horizontal
forces tests and an increase in the displacement and stresses of the vertical forces test. The
results for the pedaling test were similar between stacking scenarios 6 and 7.
After reviewing the results of stacking scenario 8, shown in Table 26, and comparing
them to those of stacking scenario 7, the displacements in drop mass and falling frame tests
were lower and the stresses were slightly elevated. The horizontal forces and pedaling
forces tests had very similar results in both scenarios. The vertical forces test showed a slight
increase in the displacement and stresses. All the stresses were below the allowable limits.
In stacking scenario 9, whose results are displayed in Table 27, there was a further
decrease in the displacements and the stresses in the drop mass, horizontal forces, and
pedaling forces tests. The vertical forces test showed similar stresses, with a slight increase
in the displacement, while the falling frame results presented a decrease in the stresses and
a minor increase in the displacement. All the stresses were allowable.
The results of stacking scenario 10, shown in Table 28, when compared to those of
stacking scenario 9, showed a significant increase in most of the stresses and displacements
in all the loading cases. Nevertheless, the stresses did not approach the limits set by the
materials presented in Table 16.
The final comparison between the various stacking scenarios was regarding the weight
of the frame that each lay-up recipe would result in. Table 29 presents the weights of the
various resulting frames.
The lightest stacking recipes were the two quasi-isotropic recipes with M46J Woven
and T300 Woven. However, the displacements they exhibited were not as low as those of
other tested recipes. The second-lightest scenario was the 8th stacking recipe. The stresses
that occurred in the five loading cases are all kept in the mid-to-low range. In addition, the
displacements were some of the lowest across all the test recipes. The only scenario that
displayed lower displacements in some of the tests (drop mass and pedaling tests) was the
9th stacking recipe, but this comes at the cost of additional mass (almost 50% more mass).
There were other recipes that had relative low weight, such as stacking scenarios 1, 4, 6,
and 7, but the displacements they developed were inferior to those of stacking scenario 8.
When further investigating the results of stacking scenario 8, we can see that the
developing stresses in the various loading cases were nowhere near the limit set for the
Machines 2022, 10, 149 39 of 41
materials used. When we tried to remove some of the plies that did not contribute much
to the overall structure, meaning that the load they bore was kept at low levels, the
displacements that occurred were significantly higher. Therefore, we decided that this was
the best solution for our frame.
4. Conclusions
In this paper, the main project was to indicate the design method of manufacturing a
competitive road bicycle frame with a combination of lightness, strength, and aerodynamic
efficiency using carbon/epoxy composite sheets. This superior blend of materials transmits
its mechanical attributes to the bicycle by applying it with a proper sequence and direction
on each part of the frame to pass approved tests of static and dynamic loading scenarios.
Firstly, through the use of spaceframe models, the basic geometry of the frame was
determined to be the optimized geometry because this produced reduced displacements
and stresses compared to the other models and the absorption of the strain energy was well
distributed across the different tubes of the frame.
The next step was to determine the shape of each tube of the bicycle. To this end, both
structural (bending and torsional) and aerodynamic tests were simulated for various cross-
sections for each part of the frame. After evaluating the results, the cross-sections that were
chosen were: the circular cross-section with 40 mm diameter for the top tube, a combination
of two oval cross-sections (39 × 79 and 51 × 79) for the down tube, complex 1 for the seat
tube, the oval 30 mm × 10 mm cross-section for the seat stays, and a combination of the
oval and rectangular 20 mm × 60 mm cross-sections for the chain stays.
Having selected the appropriate geometry and the cross-sections for each tube, the
final model of the frame was created. With the use of the five tests set by the ISO standards,
various stacking scenarios were tested to determine the layup recipe that best combined
low weight, high stiffness along the down tube, bottom bracket, and chain stays, and
absorption of vibrations while riding. The laminate recipe that was chosen was stacking
scenario 8, which presented the second lowest mass and some of the lowest displacements
among the other layup recipes.
The joints of the tubes on the frame were created as smooth surfaces connecting the
surrounding tubes and were not a primary subject of this study. Spending more time on
the evaluation of their shape would be counterintuitive to the development of this new
time-saving design process when considering the small gains the change in those areas
would offer to the overall performance of the frame. In the future, the shape of the joints
could be the parameter of a new aerodynamic and structural study.
Author Contributions: Conceptualization, A.G.K.; data curation, A.G.K. and C.S.N.; formal analysis,
A.G.K.; funding acquisition, A.G.K.; investigation, A.G.K. and C.S.N.; methodology, A.G.K. and
D.E.M.; project administration, A.G.K. and D.E.M.; resources, A.G.K. and D.E.M.; software, A.G.K.;
supervision, A.G.K. and D.E.M.; validation, A.G.K., D.E.M. and C.S.N.; visualization, A.G.K. and
C.S.N.; writing—original draft, A.G.K.; writing—review and editing, C.S.N. All authors have read
and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The raw/processed data required to reproduce these findings cannot
be shared at this time due to technical or time limitations.
Acknowledgments: In this section, we need to acknowledge the help provided during this research
by two NTUA students, Alexandros Chronis and Evangelos Goulas, who helped process the data as
well as write and proofread the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.
Machines 2022, 10, 149 40 of 41
References
1. Sajan, S.; Selvaraj, D.P. A review on polymer matrix composite materials and their applications. Mater. Today Proc. 2021, 47,
5493–5498. [CrossRef]
2. Ning, H.; Vaidya, U.; Janowski, G.M.; Husman, G. Design, manufacture, and analysis of a thermoplastic composite frame
structure for mass transit. Compos. Struct. 2007, 80, 105–116. [CrossRef]
3. May, D.; Goergen, C.; Friedrich, K. Multifunctionality of polymer composites based on recycled carbon fibers: A review. Adv. Ind.
Eng. Polym. Res. 2021, 4, 70–81. [CrossRef]
4. Daniel, I.M.; Ishai, O. Engineering Mechanics of Composite Materials; Oxford University Press: New York, NY, USA, 2006.
5. Fuerle, F.; Sienz, J. Decomposed surrogate-based optimization of carbon-fiber bicycle frames using Optimum Latin Hypercubes
for constrained design spaces. Comput. Struct. 2013, 119, 48–59. [CrossRef]
6. Covill, D.; Allard, P.; Drouet, J.-M.; Emerson, N. An Assessment of Bicycle Frame Behaviour under Various Load Conditions
Using Numerical Simulations. Procedia Eng. 2016, 147, 665–670. [CrossRef]
7. Khutal, K.; Kathiresan, G.; Ashok, K.; Simhachalam, B.; Jebaseelan, D.D. Design Validation Methodology for Bicycle Frames
Using Finite Element Analysis. Mater. Today Proc. 2020, 22, 1861–1869. [CrossRef]
8. Fintelman, D.M.; Hemida, H.; Sterling, M.; Li, F.-X. CFD simulations of the flow around a cyclist subjected to crosswinds. J. Wind.
Eng. Ind. Aerodyn. 2015, 144, 31–41. [CrossRef]
9. Malizia, F.; Blocken, B. Bicycle aerodynamics: History, state-of-the-art and future perspectives. J. Wind. Eng. Ind. Aerodyn. 2020,
200, 104–134. [CrossRef]
10. Malizia, F.; Blocken, B. Cyclist aerodynamics through time: Better, faster, stronger. J. Wind. Eng. Ind. Aerodyn. 2021, 214, 104673.
[CrossRef]
11. Della Vecchia, P.; Nicolosi, F. Aerodynamic guidelines in the design and optimization of new regional turboprop aircraft. Aerosp.
Sci. Technol. 2014, 38, 88–104. [CrossRef]
12. Panagiotou, P.; Kaparos, P.; Yakinthos, K. Winglet design and optimization for a MALE UAV using CFD. Aerosp. Sci. Technol. 2014,
39, 190–205. [CrossRef]
13. Idahosa, U.; Golubev, V.V.; Balabanov, V.O. An Automated Optimal Design of a Fan Blade Using an Integrated CFD/MDO
Computer Environment. Eng. Appl. Comput. Fluid Mech. 2008, 2, 141–154. [CrossRef]
14. Joung, T.-H.; Sammut, K.; He, F.; Lee, S.-K. Shape optimization of an autonomous underwater vehicle with a ducted propeller
using computational fluid dynamics analysis. Int. J. Nav. Archit. Ocean. Eng. 2012, 4, 44–56. [CrossRef]
15. Shokrieh, M.M.; Rezaei, D. Analysis and optimization of a composite leaf spring. Compos. Struct. 2003, 60, 317–325. [CrossRef]
16. Choi, B.-L.; Choi, B.-H. Numerical method for optimizing design variables of carbon-fiber-reinforced epoxy composite coil
springs. Compos. Part B 2015, 82, 42–49. [CrossRef]
17. Kim, D.-H.; Choi, D.-H.; Kim, H.-S. Design optimization of a carbon fiber reinforced composite automotive lower arm. Compos.
Part B 2014, 58, 400–407. [CrossRef]
18. Amiruddin, A.K.; Sapuan, S.M.; Jaafar, A.A. Analysis of glass fiber reinforced epoxy composite hovercraft hull base. Mater. Des.
2008, 29, 1453–1458. [CrossRef]
19. Kalyanasundaram, S.; Lowe, A.; Watters, A.J. Finite element analysis and optimization of composite wheelchair wheels. Compos.
Struct. 2006, 75, 393–399. [CrossRef]
20. Hansen, L.U.; Horst, P. Multilevel optimization in aircraft structural design evaluation. Compos. Struct. 2008, 86, 104–118.
[CrossRef]
21. Song, F.; Ni, Y.; Tan, Z. Optimization Design, Modeling and Dynamic Analysis for Composite Wind Turbine Blade. Procedia Eng.
2011, 16, 369–375. [CrossRef]
22. Chang, R.R.; Wu, W.J.; Jia, S.Y.; Tan, H.M. Design and Manufacturing of a Laminated Composite Bicycle Crank. Procedia Eng.
2013, 67, 497–505. [CrossRef]
23. Lin, C.-C.; Huang, S.-J.; Liu, C.-C. Structural analysis and optimization of bicycle frame designs. Adv. Mech. Eng. 2017, 9,
1687814017739513. [CrossRef]
24. Sani, M.S.M.; Nazri, N.A.; Zahari, S.N.; Abdullah, N.A.Z.; Priyandoko, G. Dynamic Study of Bicycle Frame Structure. Mater. Sci.
Eng. 2016, 160, 012009. [CrossRef]
25. Chung, C.-P.; Lee, C.-F. Parameters Decision on the Product Characteristics of a Bike Frame. Procedia—Soc. Behav. Sci. 2012, 40,
107–115. [CrossRef]
26. Covill, D.; Begg, S.; Elton, E.; Milne, M.; Morris, R.; Katz, T. Parametric finite element analysis of bicycle frame geometries.
Procedia Eng. 2014, 72, 441–446. [CrossRef]
27. Devaiah, B.B.; Purohit, R.; Rana, R.S.; Parashar, V. Stress Analysis of a Bicycle Frame. Mater. Today Proc. 2018, 5, 18920–18926.
[CrossRef]
28. Liu, T.J.-C.; Wu, H.-C. Fiber direction and stacking sequence design for bicycle frame made of carbon epoxy laminate. Mater. Des.
2010, 31, 1971–1980. [CrossRef]
29. Aceves, C.M.; Skordos, A.A.; Sutcliffe, M.P.F. Design selection methodology for composite structures. Mater. Des. 2008, 29,
418–426. [CrossRef]
30. Keithonearth, CC BY 3.0. Available online: https://commons.wikimedia.org/w/index.php?curid=4914154 (accessed on
12 January 2022).
Machines 2022, 10, 149 41 of 41
31. Clarification Guide of the UCI Technical Regulation—05.10.2021 version. 2021. Available online: https://www.uci.org/
equipment/bh2JJzw1eB0n876rX2iB1 (accessed on 12 January 2021).
32. Peterson, L.A.; Londry, K.J. Finite-element structural analysis: A new tool for bicycle frame design. Bike Tech 1986, 5, 2.
33. European Wind Atlas. Available online: https://www.wasp.dk/wind-atlas/european-wind-atlas (accessed on 12 January 2022).
34. ISO 4210-6, Cycles-Safety Requirements for bicycles Part 6, ISO 2014. Available online: https://www.iso.org/standard/78081.html
(accessed on 12 January 2022).
35. M46J Data Sheet, TORAYCA. Available online: https://www.toraycma.com/wp-content/uploads/M46J-Technical-Data-Sheet-1.
pdf.pdf (accessed on 12 January 2022).
36. T300 Data Sheet, TORAYCA. Available online: https://www.toraycma.com/wp-content/uploads/T300-Technical-Data-Sheet-1.
pdf.pdf (accessed on 12 January 2022).
37. T700S Data Sheet, TORAYCA. Available online: https://www.toraycma.com/wp-content/uploads/T700S-Technical-Data-Sheet-
1.pdf.pdf (accessed on 12 January 2022).