You are on page 1of 3

L3 Applications Group 14: Final Report

Design Process and Performance Modelling

An important design consideration was to allow for a simple assembly, thus reducing potential errors by
limiting the number of parts and manufacturing stages. Properties of materials and parts were measured
so that the baseplate configuration accurately achieved a centre of mass which complied with static
stability requirements. Therefore, the wheel placement was chosen to be aft of the motor. The coefficient
of friction between the tyres and the runway was experimentally obtained as 0.9.
The frontal area design minimised profile drag; this requirement also influenced the chassis profile while
interference drag was minimised by placing the wheels within the chassis, so that there were no protruding
components. To reduce the number of parts, the chassis extends all the way to the back and integrates the
vertical tailplane. For stability control, an adjustable horizontal tailplane was added, which was achieved
with a slotted interference fit with the vertical tail.
Both a rectangular wing and an inverse Zimmerman cambered plate were shortlisted. However, the
rectangular airfoil section wing configuration with balsa and solar film was estimated to be too heavy.
The inverse Zimmerman configuration had superior performance predictions, particularly at low Reynolds
numbers [1], and also allowed a lighter design with simpler fabrication processes. The wing and tailplane
were enlarged for the final design as our Matlab model predicted a greater flight distance despite the
small weight penalty.
The following improvements were made to the provided Matlab skeleton code:

1. An enhanced free-body diagram representation of the glider including relative positions, both hori-
zontally and vertically, of components such as the wing, tail, wheels etc. This not only allowed for
improved modelling of the glide phase, but also for the drive train as a more accurate maximum
torque condition was found.
2. A component-wise analysis of the aerodynamic characteristics of the body. This involved finding,
digitising and implementing empirical lift-drag and moment data for the inverse Zimmerman wing
profile (of same aspect ratio) and the elliptical tail profile. Most importantly however, the change in
the position of the aerodynamic centre with respect to the angle of attack, due to the highly 3D flow
effects around profiles such as the inverse Zimmerman, was included. Simple empirical models
were used to analyse the bluff body drag characteristics of the chassis.
3. A detailed optimisation of the drivetrain. This involved solving for the normal force on the driven
wheel and hence computing the torque without the wheel-slip. The voltage required to achieve
this torque was computed by assuming a linear relationship between the maximum torque and
the maximum angular velocity of the motor. The power of the motor utilised was compared to the
maximum power available for a particular voltage to ensure a suitable motor choice. The voltage
applied was also tracked to ensure that it did not exceed the maximum of the battery. This data
was stored to be inputted into the microcontroller for testing.
4. An important part of the analysis was investigating the transition region where only the rear wheel
remains on the ramp. Similarly to the trajectory model, three degrees of freedom were used, and
moments were taken about the rear wheel. Care was taken to ensure the correct moment of inertia
about the rear wheel was used. The importance of this analysis comes not from the small change
in pitch, but from the large pitch rate used to initialise the trajectory modelling.

This model was used to find an optimal configuration thus constraining the critical dimensions of the robot.

1
Mechanical Implementation

The interference fit to integrate the tailplane was achieved by sanding down intentionally oversized parts.
The vertical tailplane in the final model was extended to allow for a larger range of tail setting angles. The
baseplate was attached to the chassis sides using a puzzle fit with glue, while the wing was attached using
laser cut balsa strips positioned in areas of high pressure distribution and concentrated masses. The
final model excluding the wing is presented in Figure 4. A multidirectional three layer balsa laminate was
considered for the baseplate, but was unsuccessful due to it bowing and delaminating during prototyping.
After the regions of structural failures were observed in prototype tests, the chassis sides were extended
to wrap around the front of the wing to prevent wing detachment, and their front was padded with foam for
impact protection. The baseplate was thickened as well. The motor was secured to the baseplate with a
bracket and two bolts to ensure proper gear meshing. The front wheels were chosen to be in the D-shaft
configuration to transfer motor torque efficiently, suitable for the previously selected 2:3 gear ratio. Laser
cut acetal rings were interference-fit on the front and back carbon rod axles to limit lateral movement and
to eliminate contact between the wheels and the baseplate.
Usage of a circuit board eased soldering and kept the circuit neat, compact, and sturdy. The circuit board
was secured to the baseplate with thin wires which enabled clear frontal access to recharge the battery.
Motor power was increased gradually to avoid slipping. An improved Arduino code allowed us to change
all motor power settings and run times from the remote. The microcontroller was also programmed to
switch the motor off before touchdown in order to minimise damage to the model from rolling on the
ground and hitting objects.

Robot Performance

The average ramp speed achieved in testing was 2.86 0.19 ms1 which is close to the value of 3.51
ms1 predicted by the Matlab model. Examining Figure 1, the flight distances of 3.10 0.08 m (maximum
3.18 m) achieved by the glider are reasonably close to the predicted value of 3.66 m.
There is clearly some lateral translation in the model during flight which had not been anticipated (Figure
2). On inspection of the model, it was clear that the wing was off-centre. This asymmetry is what is
believed to have caused the model bank left and consequently lose height faster compared to the pre-
dicted response. Another influencing factor may have been the tailplane, as there was a small amount
of warping when it was attached to the model which could have led to an asymmetrical loading in flight.
Despite this, the predicted and actual responses generally agree well. In some tests, the glider seems to
flare slightly on landing relative to the prediction. It is hypothesized that this was due to neglecting ground
effect in the model. If the transition region is neglected the flight distance remains similar, however there
is a pronounced difference between the predicted and the real flight paths (Figure 3).
The pitch during flight was slightly harder to predict owing to the fact that the stability of the model is ex-
tremely sensitive to tailplane angle, wing and c.g. position. However, predicted and achieved responses
are widely similar. There is a well-defined stall in test 3, as seen by examining figure 1. This can be
attributed to setting too steep a tailplane deflection angle.
One of the strengths of our glider was its consistency in both the ramp and glide phases of testing. This is
predominantly due to the the gliders impact resistance, meaning no physical degradation occured during
testing. There was also a marked increase in glide performance when the wing area was increased in
the final design, which suggested that maximizing the wing area should be a priority. Finally, our design
didnt stall or nose dive during most of the flight due to the tail plane, which was a major factor in achieving
a good performance. To further improve our design we would reduce weight by conducting an impact
analysis on the chassis to determine redundancies. Increasing the wing area would likely also improve
our performance. Finally, shielding the inside of the chassis would likely reduce drag.

2
Figures

Figure 1: Side view of the test flight trajectories Figure 2: Top view of the test flight trajectories

Figure 3: Comparison of the best test trajectory and the modelled trajectories

Figure 4: Final design excluding the wing Figure 5: Final Design bottom view

References

[1] T.J Mueller and G. E. Torres, Aerodynamics of Low Aspect Ratio Wings at Low Reynolds Numbers
With Applications to Micro Air Vehicle Design and Optimization, University of Notre Dame, Indiana, 2001.
Link to webpage: http://l3groupfourteen.weebly.com/

You might also like