CFD of Drone Propeller
CFD of Drone Propeller
CFD of Drone Propeller
1
Student, Mechanical Engineering ,Natick, MA, 01760, and AIAA Student Member.
2
Student, Mechanical Engineering, Cincinnati, OH, 45242, and AIAA Student Member.
3
Student, Mechanical Engineering, Massillon, OH, 02120, and AIAA Student Member.
4
Student, Mechanical Engineering, Meriden, Ct, 06450, and AIAA Student Member.
5
Professor, Aerospace, Address/Mail Stop, and AIAA Member.
Nomenclature
A = amplitude of oscillation
a = cylinder diameter
Cp = pressure coefficient
Cx = force coefficient in the x direction
Cy = force coefficient in the y direction
c = chord
dt = time step
Fx = X component of the resultant pressure force acting on the vehicle
Fy = Thrust produced by propeller
f, g = generic functions
h = height
i = time index during navigation
j = waypoint index
K = trailing-edge (TE) nondimensional angular deflection rate
A. Abstract
A three-dimensional, steady, incompressible finite difference Navier-Stokes solver has been coupled with
experimental data to study the effects different drone propeller configurations have on producing thrust. The
Navier-Stokes solver was validated by comparing the results to those obtained through experimental means. The
solver was proven to be valid within a reasonable margin of error. It is within this margin of error that the initial
CFD analysis that compared the configurations on geometry alone, neglecting mass, can be held true. The analysis
concluded that based on geometry alone the four blade Qtip configuration would produce the most thrust.
The control design is a two blade propeller, based off of a standard drone propeller. The two blade propeller
will set standard measurements for each modification, allowing for a comparison to these measurements and
deciphering whether or not each design change has increased or decreased the thrust, while also helping understand
why the thrust has changed.
The three blade propeller is the first modification to the control. Another blade was added based off of the
idea that most propellers used are three blades. While the extra weight is thought to decrease the thrust, it is believed
the extra blade will help cause more air to be propelled back, thus creating more thrust.
The next design is the four blade propeller, which is the addition of two propellers to the control. As
previously stated, the extra blade will add weight, but it is thought more airflow will occur, helping in creating a
larger thrust.
The first modification made to the three propeller blade is adding a Q-tip to the end of each propeller blade.
The Q-tip shape is an a 90 degree curved bend added to the end of the blade, keeping the same airfoil geometry of
the blade. The purpose of the Q-tip is similar to the duct ring, as it is to help in the thrust generation of the propeller.
As most airfoils work, high pressure builds up below the blades of the propeller and low pressure is created above
the blades, which causes the high pressure to want to move up to the lower pressure. With the Q-tip, the high
pressure will not be able to move to the low pressure, causing the high pressure to push on the bottom of the plate,
thus helping in the generation of thrust. The downside to the Q-tip propeller is the production of drag, therefore
affecting lift. In adding the Q-tip modification, it is hypothesized that the thrust generated will be greater than the
reference three propeller blade.
The first modification that will be made to the Four Blade Propeller is the alteration of the winglet tip. A
Q-tip design will be added to the end of each propeller blade. The Q-tip shape is an a 90 degree curved bend added
to the end of the blade, keeping the same airfoil geometry of the blade. The Q-tip modification will be the exact
same as the 3 blade modification for consistency. The addition of a fourth blade should increase the thrust, and
paired with the Q-tip design it will reduce the effects of vortex shedding.
The ductile propeller is modified off of the 3 propeller base. The reason I chose this modification was
because there is a theory that a vortex is created at the end of a propeller blade which decreases the low pressure
created on top of the wing. Therefore by putting the duct around the propeller it makes it more efficient by
increasing the amount of low pressure on top of the propeller, along with stopping the vortex form forming. So, to
make my wing I added a very small outer circle that is connected right on to the end of the propeller blades. Also, I
added a radius to the intersecting points to help with the amount of surface area that is holding the duct on and to
make it look smoother.
The alternative to the winglet tip modification that will be made to the 4 blade propeller is the alteration of
offset angle and length blade pairs. A shallow angle of attack allows the propeller blades to spin at higher rpms
offing less drag however, this all means less lift. There for a more aggressive angle of attack for a given propeller is
needed in order to provide sufficient lift for the aircraft. This design modifies the 4 blade design such that it has a
pair of shallow angle full length blades and a pair of aggressive angled half length blades. The idea is that even
though the aggressively angled pair of blades might increase drag, they will also increase lift While making them
half the length the design saves on rotational mass, allowing it to spin higher speeds.
To test all the blades and modifications in a standard and controlled way, a modular fixture was designed to
mount sensor motors and various blade designs. The test stand is composed of a 3D printed box framed base, with a
mount for a load cell. In the figures there are two perpendicular load cells. The load cells are fastened together using
their mounting holes and a 3D printed adapter. Atop the 2nd load cell is a motor mount, where the motor is secured
in place by screws. Finally the test blades can be fastened to the motor with a screw and washer. The load cell is a
piece of metal of known young's modulus fixed with two opposing strain gauges. As the load cell experiences a
force and deflects the grid pattern in the strain gauges either expand or contract based on the orientation. This
deflection captured by the strain gauge is measured by a variable in resistance of the gauge. The change in resistance
is captured through an analog input on a microcontroller where the actual force is calculated and recorded. The
reason there are two load cells perpendicular to each other is because in addition to the axial force generated by the
blades a torque is also exerted due to the rotation of the mass of the blades. This torque is also part of the net overall
force produced by our designs and can be observed using the proposed test stand and sensors.
Experimental data was collected using a load cell fixed to a motor stand to record the force produced by a
propeller as it spun. A stroboscope was used to record the RPM of each blade during each run. The experimental test
setup was built from motor, controller and battery from an RC car.
The procedure for each blade was as follows; battery was charged for 1hr, connected to the RC circuit board,
then the propeller was mounted to the test stand, the controller was turned on along with the RC board. Once the RC
board established a connection with the controller, the arduino microcontroller was connected to a PC. The arduino
would launch the testing program and the test would begin. Each test captured ten seconds of data from the load cell
while the motor operated at the maximum power output.
Due to the draining power of our battery and method of capturing angular velocity there are some discrepancies
in RPMs recorded for each blade. However, in the long run this did not invalidate our final results. Table 1
showcases, at the observed RPM how each propeller performed in terms of thrust the error between the CFD results
and experimental thrust values.
Before data was experimentally collected, a CFD analysis for each design at a fixed RPM was completed. Table
1 presents the tabulated values for the thrust produced by each design at 7000 RPM. This graph presented some
insight as to some rough expectations for thrust performance based on geometry alone. Mass, materials, and stress
limitations of the designs were not taken into account for these Solidworks flow simulations.
Figure 9 graphs the experimental data for thrust performance of each design. When visually comparing Figure 9
and Figure 10, it is easy to see that the expected CFD performance does not match up with the experimental values.
This is due partially to the fact that in the experimental setup, the RPM was not held constant due to the limitation of
the testing equipment and other variables thoroughly explained in the conclusion. However, the important thing to
note is that when the CFD is re-run with the observed RPM during the experiment for each respective design, the
thrust outputs match with the CFD findings with relatively low and consistent margins of error. This means that
although the experimental data did not fit precisely into the CFD theoretical values, the experimental values are
accurate in the thrust measurements recorded.
From the experimental data, the design that performed best based on modification is the 4 blade rotor with
q-tipped blades. This makes sense for a few reasons, one is the q-tipped blades significantly reduce vortices that
occur at the blade tips which increase drag and reduce efficiency. Another reason is seen as the increase in total
amount of blades on the propeller itself produces an increase in thrust. This idea can be observed by Figure 9, where
it is evident the four blade propeller produces a significant amount more thrust than the two blade propeller. By
having the increase in thrust from the q-tip geometry and the four blade sequence, the thrust produced was vastly
larger neglecting mass.
Lastly from our data we can clearly take away ideas to improve not only our propeller design but also our
experiment method. From here the top three designs can be chosen for the next iteration of modifications.
Additionally a tachometer can replace the stroboscope to increase the accuracy of the RPM measurement, and
possibly integrate an anemometer to measure intake and exit velocity as a redundancy to confirm the thrust
measurement.
10
11
The partial differential equations used when solving CFD (Computational Fluid Dynamics) simulations with
Navier-Stokes solver (NSS) that govern fluid flow typically can not be solved by analytical means, except for very
simple cases. This is due to the fact that the NSS uses 4 values to compute a CFD analysis, velocity of fluid flow in
the x, y, z directions and pressure. Each having its own equation, this system is very hard to compute by hand and
thus is commonly used in algorithms by solvers to compute complex simulations.
Theses solvers available in different simulations tools utilize a mesh to define elements to step through the
algorithm solving large systems element by element. In order to analyze fluid flows, flow domains are split into
smaller subdomains and meshes are described globally and locally for optimization of limited computational
resources. The governing equations are then discretized and solved inside each of these subdomains.
These subdomains are typically referred to as cells and the entire collection of these cells is referred to as the
mesh. Creating this mesh, meshing, is a critical part of numerical solutions because it determines the accuracy of the
solution. For this reason refinements in mesh are applied along the geometry for which solutions are being
calculated. In order to achieve this for the propeller, a local meshing consisting of tetrahedron elements, tet mesh,
was applied to the surface of each blade (see figures 11 - 13 for meshing). The local meshing was refined for cells
that were in contact with the working fluid as well as cells that were on the surface of the geometry. This was done
to reduce the number of cells which were in partial contact with the surface. For each propellor configuration there
were a total of 301,266 cells and of those 46, 904 fluid cells contacting solids. The global meshing required no
refinement because nothing outside of the surface of the propeller was being calculated for this simulation (see
figure 14 meshing).
12
13
14
15
Figure 15. Chart of iterations vs force for 11400 RPM (Max RPM).
Table 3. Force produced at 11400 RPM (Max RPM) for 4 blade Qtip propeller.
16
17
18
19
20
21
22
23
24
25
26
27
28
30
31
32
33
34
void loop() {
cell.set_scale();
cell.tare();
Serial.println( cell.get_units(10) ); // average of 10 values
}
35
Yilmaz, Erdem, and Junling Hu. CFD Study of Quadcopter Aerodynamics at Static Thrust Conditions. University of Bridgeport,
1 Oct. 2018.
Kutty, Hairuniza Ahmed, and Parvathy Rajendran. “3D CFD Simulation and Experimental Validation of Small APC Slow Flyer
Propeller Blade.” Www.mdpi.com, 10 Apr. 2017, www.mdpi.com/2226-4310/4/1/10/pdf.1
36