Amro Alshareef, Dongjoon Lee, Salem Ali, Sai Ravela
ONR Project -- add supporting material

Background

The IcyQ project emerged from research experience developing and deploying autonomous unmanned aircraft systems (UAS) for environmental observation at the Earth Signals and Systems Group (ESSG) at MIT [1-5]. In prior work, low-cost UAS were rapidly demonstrated for tracking and mapping coherent atmospheric structures such as volcanic plumes and shallow clouds. Subsequently, systems to track wildlife, disaster reconnaissance, and mapping the retreat of Glacier were of interest.  Currently, in an Office of Naval Research (ONR) award, aircrafts that rapidly change dynamic regimes are being developed.

There are a few common issues that emerge in the aforementioned applications. In most cases access to runways for takeoff and landing is difficult. The aircraft must therefore be small, and limits develop on the payload. Features such as undercarriages are eliminated, then batteries powering the aircraft become smaller.  Batteries are selected as the power source to operate effectively in various sampling environments, as well as dynamic regimes. However, small batteries limit endurance, which typically must span hours, and further reduces range, which is typically required to be around 5-20 km. The aircraft must often transition from glider modes to agile behavior to respond and track environmental flow fields.  For example, when observing wildlife, it is necessary to not frighten them.  In other applications, the aircraft are often part of a closed loop observing system, gathering data to estimate parameters and states of environmental models. Numerous examples of other environmental applications exist that motivate our work [6]. In these applications, they may be called upon to implement varied and non-uniform sampling trajectories, in contrast to simple deterministic ones like a fixed, repetitive pattern. 

Considering all these requirements, it is difficult to use battery-operated fixed-wing or multi-rotor (e.g. quadcopter) systems. However, a design that would address most of these issues is a Vertical Take-Off and Landing (VTOL) aircraft system that can transition from hover to forward flight and back and employs a gas-electric hybrid engine to provide the endurance. This architecture can enable range, endurance, agility, glide and other elements of performance that have been lacking in small unmanned aircraft systems for environmental observation. Presented here is the design and fabrication of such an aircraft and the instruments needed to fabricate it.

Design

Wings:

• 3 m wingspan
• tapered leading and trailing edge: 4 degree angle of attack
• dihedral angle of 5 degrees
• 0.95 m2 cross-section

Fuselage:

• nose-cone front with and a tapered tail cone
• cylindrical center body to host the power plant and payload of the system
• detachable from the wings (fixed via four support beams)
• connected via a support beam to the tail of the aircraft, supporting two horizontal and a vertical stabilizer with control surfaces.

Figure 1: 3D visualization of the Aerodynamic surfaces of the aircraft

Figure 1 depicts the initial 3D model of the aerodynamic surfaces of the aircraft described above. Preliminary Computational Fluid Dynamics (CFD) testing on the model resulted in a drag coefficient for the wings of 0.09 at a flow speed of 30 m/s. The loads due to drag and lift informed the following design of the mechanical support structure and mechanisms controlling the aircraft.

Design Requirements

1. The entire structure must be lighter than 10 kg, including payload, to increase the flight duration to fuel consumption ratio and maneuverability.
2. The propeller arms must rotate 90 degrees from propeller horizontal to vertical, allowing the system to transform between a VTOL to a forward-flight aircraft.
3. The Center of Gravity (CG) of the system must be at quarter cord for stable flight.
4. CG must be in line with motors in forward flight to prevent any destabilizing moments from affecting flight control.
5. The structure must be able to be disassembled into five sections, including a main fuselage, two wings sides, a wing center, and a tail section.

Methods

Due to the large size of the aircraft (3 m wingspan with a 2 m nose-to-tail length) along with its mass constraints, the ideal material for the mechanical structure was Carbon Fiber (CF) for its strength to weight ratio. Figure 2 illustrates the simplified system model of the propeller support beam that was used to derive the desired CF tube size in order to minimize vibration and deflection under the loads experienced during flight. 

Figure 2: Propeller support beam model. The Young’s modulus is that of 45o twined CF tubing. The maximum expected acceleration of the aircraft during forward flight is ‘a’.

The effective force acting on the propeller beam due to the weight of the aircraft along with the thrust of forward acceleration was evaluated. The deflection of the of the propeller beam was then derived in terms of the dimensions of CF tubing.

A 3D plot in MATLAB with possible combinations of inner and outer CF tube diameters and the corresponding deflection associated with such a combination, as shown in Figure 3 below. The plot was filtered such that only realistic combinations remained, such as ones where the outer diameter was greater than the inner diameter, ones where the diameters were between zero and one inches, and ones that existed for commercial purchase. 

Figure 3: Combinations of inner and outer CF tube diameters and resulting beam deflection under associated flight loads. The OD and ID axis are in inches for easy cross-reference with US-based CF vendor catalogs. The green combinations represent ≤0.25 mm deflections, yellow represent between 0.25 and 0.5 mm deflections, and red represent >0.5 mm deflections.

In order to minimize beam deflection and eliminate the cost of having to source a custom sized CF tube, a tube with an outer diameter of 0.875 in and an inner diameter of 0.75 in was chosen, which results in a propeller deflection of approximately 0.24 mm under max load. 

The next structural members of concern were the four support beams connecting the fuselage to the wing. Figure 4 models the load on these beams. Note the axial load due to gravity was neglected because of the extremely lightweight foam composition of the wings compared to the high compressive strength of CF tubes. The only relevant load on the system is, therefore, the drag force on the wing.

Figure 4: Simplified model of the wing-fuselage connection. The Young’s modulus is that of 45o twined CF tubing. Awing refers to the frontal cross-section of the wing, and v is the maximum expected velocity for the aircraft during flight.

The drag force using the coefficient of drag determined by the CFD simulations was calculated, and the resultant beam deflection were derived. Due to the fact that 0.875 in by 0.75 in propeller tubes were already being purchased for the propeller beams, it was more cost efficient to add to a bulk purchase of the same tubing size rather than order a different size for the wing attachment. The deflection for this tubing size was calculated to be negligible at .1 mm.

For the main structure in the aircraft fuselage, thin walled, 1 in tubing was chosen to act as the structural connecting members of the mechanical structure.

The propeller motor mounts would be 3D printed with ABS plastic, which was accessible on campus, and mounted with epoxy on the ends of the propeller beams. The design shown in Figure 5 was chosen in order to withstand the torque and upward force imparted on the mount by the motor, and to facilitate the aircraft’s assembly process. 

Figure 5: Model of the initial propeller motor mount design. SolidWorks FEA simulation was run with a 5 Nm torque on the screw holes in the motor central axis due to the motor torque imparted during rotation, and a 100 N upward force caused by the thrust force generated by the propellers.

The resultant deflection of the motor mount was less than 0.005 mm, with stress well below yield. The design also allows ready access to the screw holes for motor mounting after the mount is already epoxied, and it covers the end of the CF tube such that no dust, dirt, or moisture can accumulate inside the tube.

Initially, the mechanism by which to rotate the propeller beams was a geared system with supporting ball bearings on either end of the structural base, as depicted on the left side of Figure 6. These ball bearings were fairly large and heavy, with a steel casing; they were, however, the only ball bearings available from our main vendor, McMaster, that fit the CF tube with a 0.875 in outer diameter. These bearings were also fairly expensive at approximately 15 USD apiece. A DC motor with a hall effect sensor encoder would be used and connected via two custom aluminum gears to the propeller beam, as depicted in Figure 3-6. These gears can be waterjet using an OMAX waterjet on campus, as opposed to an extremely expensive and heavy commercial gear that would fit the CF tube.

Figure 6: The rotation mechanism for the propeller beam that allows it to rotate between 0 and 90 degrees. On the left is the initial design of the mechanism, and on the right is the final redesigned mechanism.

Due to the fact that this motor mount rod does not have high radial velocities and only rotates between 0 and 90 degrees, however, a ball bearing is unnecessary and could instead be exchanged for a frictionless sleeve bearing. This is a more cost efficient design strategy, at about a dollar apiece, and could be mounted with 3D printed mounts, cutting down costs from 30 USD to about 2 USD per beam, and making the system about 200 g lighter per bearing. Furthermore, the limited necessary range of motion and low torque required to rotate the beam was an indication that a lighter, cheaper servo connected to a carbon fiber bar linkage could be utilized. This eliminated the weight and need for fabrication of the aluminum gears, and exchanged the motors for limited range servos that are approximately 150 g lighter.

For the aerodynamic shell of the structure, 20 kg/m^3 EPP foam due to its light weight and shock absorption characteristics. Due to restrictions on the maximum sized foam blocks that can be cut in a foam cutter, the fuselage needed to be split into three sections and the wings into two segments. The fuselage also needed to be cored so the mechanical structure could be embedded within it, with openings for the all extruding support beams. This ultimately resulted in two halves of the fuselage, upper and lower, as shown in Figure 7. These two halves would be epoxied together, with a door chamber on each side of the top section that could be opened to access inside components, including the power plant. 

Figure 7: 3D model of the mechanical structure embedded in the aerodynamic foam shell. The cored inside and three subsections of the fuselage are indicated through the transparent upper shell.

3.3 Power Plant

The hybrid power plant consisted of an engine, a battery, and a fuel tank. The engine and the battery had a stable mass, but the fuel tank had a variable mass as the fuel in the tank depleted from an initial 2.5 kg. The Desired Center of Mass (CM) of the system, therefore, had to be achieved by the placement of the battery and the engine; the CM of the fuel tank needed to be placed coincident with that of the resultant aircraft so as to prevent movement of the aircraft CM through the duration of its flight, resulting in unstable control dynamics. The CM of the system was set at the quarter cord of the wing, in line with the propeller propulsion force, and symmetrically between both propellers (X, Z, and Y-axis constraints in Figure 8, respectively) in order to prevent any undue moments on the system and for improved aerodynamic stability during flight.

Figure 8: The final mechanical structure of the system with the CM of the battery, engine, and fuel tank indicated with yellow symbols and the CM of the system indicated in white.

The final aircraft, including the foam shell, mechanical support structure, motors and power plant, and 3 kg payload was approximately 9.8 kg. This aircraft included control surfaces on the wings and horizontal and vertical stabilizers, and was detachable into five segments: three wing sections, the tail section, and the main fuselage. The dimensions of the aircraft are indicated in meters in Figure 9.

Figure 9:  3D rendering of the final aircraft, including the mechanical structure and foam Aerodynamic surfaces. The units of the dimensions indicated are meters.

Tools

Figure 10: Mechanical structure of the aircraft assembled with epoxy curing.

4-axis CNC Foam Cutter

In order to fabricate the foam wings and fuselage, we designed and fabricated a 4-axis CNC hot wire cutter. Designing the cutter was fairly simple, as I could base my design on existing CNC cutters on the market. I designed the main framing for the four axes (X1, X2, Y1, and Y2) to be assembled in a jigging fashion. Each supporting aluminum member was modeled with either keys or slots to fit into other members. This key slot design was meant to ensure accuracy and ease during the welding process to fabricate each axis.

Figure 11: 3D model of the 4-axis CNC foam cutter.

Each axis was driven by a Nema 23 Stepper motor attached via a shaft coupling to a threaded rod. The X1 and X2 axes utilized a ½” - 13 threaded rod and a flanged nut to move the platforms connected to the Y1 and Y2 axes. The nichrome cutting wire attached to the spring on one side of the cutter and a mounting hole on the other to account for the varying wire length needed in the event of cuts with two varying cross-sections.

The aluminum components of the structure were waterjet and welded, and the railings were cut and assembled to create the resultant four axes, consisting of two symmetric X-Y axis pairs. The control boards, power supply, and dead-man switch were all neatly organized in one control box shown in Figure 5-8. The wire connectors to each of the motor axes extended out of the control box and wrapped around the outside of the cutter so as to not interfere with the movement of the cutter along its axes or possibly come in contact with the hot wire. Each axis was also equipped with limit and homing switches, and the motor wiring was coiled and wrapped for clean operation.

Results