Design and Test of a 24 Hour Fuel Cell Unmanned Aerial Vehicle (FCUAV) Derek Keen, Grant Rhoads, Tim Schneider, Brian Taylor, Nick Wagner Colorado State University Faculty Advisor: Dr. Thomas Bradley Abstract Long endurance unmanned aerial vehicles (UAVs) have increasing value as a low cost, autonomous reconnaissance and remote sensing platform for research, commercial and military missions. Current multidisciplinary optimization techniques and fuel-cell technologies have the potential to increase the endurance of such systems significantly. Research performed by Dr. Thomas Bradley while at Georgia Tech. University showed that significant gains over current systems were possible. This aircraft, powered by a polymer electrolyte membrane (PEM) fuel cell, with compressed hydrogen storage, and integrated conditioning systems, is an effort to verify and continue his research. The flight test results will be compared with the optimization research leading to this aircraft design and flight tests, as well as to published results of similar 0.5 1kW longendurance unmanned aircraft. As per the research analysis, the flight tests will verify the increased endurance of greater than 24 hrs of flight time. Further improvements to the system and planned future work will possibly include switching to a liquid hydrogen storage system for greatly increased endurance. The practical implications of this effort are wide reaching and pertinent both to further research work and current UAV customers. 1 Airframe Design The research and aircraft demonstrator undertaken by Dr. Thomas Bradley at Georgia Tech University provided the set point for the airframe that was constructed during the summer of 2009. As noted above the goal of this aircraft is to demonstrate the use of a gaseous hydrogen supplied PEM fuel cell system. Based on Dr. Bradley s research, and a custom designed 600 W fuel cell from United Technologies Research Center (UTRC), we had an optimal threshold in terms of weight, size, and aerodynamics that had to be met in order to achieve the predicted 24 hour flight later on[3]. The design decisions made as a result are discussed in the following sections. 1.1 Wing Assembly All of the lifting surfaces on this aircraft are originally from the Blue Explorer 5m composite sailplane sold by Northeast Sailplane Products. This approach allowed for a shorter development time, while providing a high quality, aerodynamically efficient and stable wing to begin the design process. To maintain the aircraft stability, care was taken to ensure that the center of gravity was directly beneath the quarter chord of the wing. The quarter chord refers to the position one quarter of the distance between the leading and trailing edges. The existing fastener attachment points were used to connect the wing to the carbon
fiber spine via custom ASTM 6061 aluminum mounts. This carbon fiber spine is discussed in further detail below. The wing is a three piece spar and monocoque composite structure, with eight internal servo motors controlling split ailerons, flaps, and spoilers. The airfoil is a modified HQW 2.5 for high lift at moderate speeds and low Reynolds numbers. The lifting capacity of this wing was determined sufficient based on the coefficient of lift and wing area as compared with the computational design tool [2,3] developed at Georgia Tech as well as the published metrics of the acceptable G-loading. 1.2 Tail Assembly The empennage of this aircraft was taken from the Blue Explorer sailplane mentioned above. It utilizes a traditional configuration with the elevator positioned very close to the horizontal datum plane of the main wing making an upside down T with the rudder. Using a traditional configuration allowed for the application of previously developed autopilot flight controls. The rudder and elevator are controlled by separate servo motors located in front of the structural hydrogen tank. These are connected to their respective control surfaces via graphite control rods along the carbon fiber spine. As stated above, the empennage assembly was taken from a pre-constructed sailplane. It is bonded to the carbon fiber spine that extends from the fuselage structure using wood buttresses and epoxy. Plastic body filler was used to ensure a premium surface finish and smooth spine-empennage transition. 1.3 Fuselage Structure Due to the large frontal surface area of the hydrogen storage tank, much of the fuselage shape was dictated by this tank. Acting as a skin between the internal components and the environment, a thin layer of fiberglass was manufactured to enclose all components except the infrared sensors used by the autopilot telemetry. Due to the shape and size of the hydrogen storage tank, a cylindrical fuselage shape was used with conical shapes to transition from the nose to the tail. Using hose clamps and custom fixtures, the hydrogen tank is secured to a one inch diameter hollow carbon fiber tube. This serves as the spine of the plane providing structural support along the length from the front motor mount all the way to the empennage in the rear. As the main structure of the aircraft, everything stems form the carbon fiber spine. The wings, servo motors, electronics, propeller motor and hydrogen storage tank are attached to this spine via ASTM 6061 lightweight aluminum brackets that were manufactured using a computer numeric controlled (CNC) milling machine. All structural components were computationally tested against theory using finite element analysis. 1.4 Landing Gear This is the one aspect of the airplane that has caused a number of problems during the testing stages, though it will be replaced by a skid plate for the final 24 hr flight. The difficulties presented with this aircraft are its large size and weight, and the ground clearance needed for the large diameter propeller (20+ inches). The initial landing gear setup was a composite two-wheel tail-dragger configuration which, while lightweight, was structurally unstable and turned out to be too narrow. Following this a mono-wheel configuration with wing skids was employed, but proved to be too unstable for use on a multi-flight aircraft. The landing gear design has since moved to a traditional tricycle configuration with two wheels of a large
wheelbase behind the center of gravity and a single wheel directly behind the propeller with steering controlled by the rudder servo motor. Designed into this configuration is a lower angle of attack to increase the acceleration during initial take off. Angle of attack refers to the difference between the horizontal datum plane and the angle made by the wing in which zero lift is produced. While this reduces lift temporarily, it also reduces drag significantly allowing the plane to achieve a higher velocity in a shorter distance. Once the desired velocity is achieved, the elevator is moved quickly to induce high lift for take off. This tricycle landing gear configuration provides more stability and control while permitting lower induced drag. These advantages come with the minor cost of additional weight. 2 Autopilot System Integration For the hands-free control of this aircraft and optimal flight management we have integrated the open source Paparrazzi autopilot developed by Ecole Nationale de l Aviation Civile in France and used by a number of other research UAVs (USU-OSAM, USU Aggie Air Remote Sensing, UCSD, U of Arizona Autonomous Glider, Team UAV UALR). This flexible ARM7 based system uses IR (Infrared) Thermopiles for horizon sensing on the pitch and roll axes of the aircraft. For the flight pattern and altitude control of the aircraft, a small ublox LEA-5H GPS receiver is used. With the included transceiver system, waypoints and other commands can be given and performance data obtained from the aircraft throughout the flight. 2.1 IR Sensors The use of IR sensors for attitude (pitch, and roll) control is based on the principle that the ambient temperature IR signal from the ground and the sky are distinctly different. While terrain, and weather can have an impact on this form of sensing, it is remarkably robust, and all of our flight testing will be performed over virtually flat terrain. Yaw control is provided primarily by the GPS waypoint commands and any coordinated flight control schemes written in the controller. Figure 2 - Diagram of IR sensing 2.2 GPS Receiver The GPS Receiver is a combination of the u-blox chipset with Sarantel s SL1206 helical antenna to produce an incredibly sensitive 50 channel GPS receiver. Some of the advantages of this receiver is the 2 Hz update rate, low power, and small form factor. The Sarantel antenna also has its own filtering giving high immunity to RF interference. 2.3 Transceiver System Figure 3 - GPS receiver/antenna The transceivers used for communicating between the ground station and the aircraft are the Digi XBee Pro 900 RPSMA and allow a very reliable and simple communication. This low power, Figure 1 - Autopilot board Figure 4 - XBee transceiver
high data rate wireless module allows for up to 6 miles line of site communication and have been tested to work well with other wireless modules on the aircraft. 2.4 Processing and Servo Control The processing of sensor readings and outputting servo control is based on common PID control. The desired closed loop dynamics of flight are tuned by changing proportional, integral, and derivative gains in the autopilot software either permanently in the code or in flight using the ground station software. The critical core of the autopilot code has been tested formally using Lustre. 2.5 Graphical Interface The ground station interface for the autopilot runs in a linux environment. Currently our ground station consists of a laptop running Ubuntu linux with the Paparazzi Center software installed. When a flight is executed, a satellite image of the current aircraft location and flight plan is loaded. Here we are able to keep track of important aspects of the plane like battery voltage, GPS signal, altitude, location, and autopilot mode (manual, wing leveling, fully autonomous). The software also records the flight for future playback. 3 Battery Power System The battery power system in use is to readily and safely provide multiple flights for flight testing and data acquisition. This data will be used to determine the final setup of the fuel cell power management. The current heavy-duty power system in the aircraft uses 2, 5000 mah Lithium polymer batteries to provide power to a Hacker A60-18L motor through a Phoenix 110 speed controller. This setup is capable of delivering over 2kW of power. The previously attempted flight tests using Axi motors were thwarted by an overloaded speed controller, shorted motor coils, and broken magnets, thus the switch to the more durable system despite a 1lb weight penalty. 4 Fuel Cell System 4.1 PEM Fuel Cell The 33 - cell stack we will be using is developed specifically for this application by United Technologies Research Center. It is a 600 W nominal system at max power and operates at 200 W for cruise performance. Its weight is 1.68 kg, providing 357 W/kg at max power with a hydrogen utilization of 90%. See Figure 6 for characteristics. 4.2 Hydrogen Storage The hydrogen is stored in a 9L, 4.5 kg composite wound pressure vessel at 5500 Psi (MCS International). Pressure regulation is provided by three stages of regulators. The first regulator drops the pressure from 5500 Psi to Figure 5 - View of Graphical Interface
management controller were included due to the results of a DFEMA completed by UTRC engineers and our team. Figure 6 - Fuel cell performance characteristics 500 Psi. Second stage regulator brings the pressure from 500 psi to 50 psi, and finally from 50 to 1 psi. On the exhaust side of the fuel cell, an on-off purge valve is used to maintain the proper humidity, pressure and stoichiometric conditions inside the fuel cell. This is controlled by the power management system discussed later. 4.3 Air Supply The air supply for the fuel cell is provided by a Micronel U51DX 51mm High Performance Radial Blower. This fan is capable of a max flow of 16.7 CFM and max pressure of 4,900 Pa. This blower was chosen for its performance specifications, power usage, and weight. 4.4 Power Management Currently in development is the power management controller for the fuel cell system. This device, developed by our team, provides control for the air and fuel utilization by measuring current and adjusting the air supply blower and the hydrogen purge rate accordingly. Also included on this board are sensors to determine the health of the fuel cell while in flight, a data logger to record these details during the flight and a telemetry system for sending the readings back to the ground. Many of the features of the power 4.5 Byproducts The byproducts of the fuel cell system are heat, water, hydrogen, and air. Cut into the nose of the aircraft are vents to provide air to the blower as well as to remove heat, and at the tail of the aircraft we have a vent for the escaping air, hydrogen, and water vapor. 5 Flight Testing Current flight testing is focused on achieving level flight for verifying the aircraft s general handling and stability characteristics. These experimental results will allow for tuning the autopilot controls and power consumption characteristics. Due to design iterations in the landing gear configuration and battery power consumption, these flights are scheduled for the first two weeks in May 2010. A minimum of two successful test flights will be needed; the first to determine the aircrafts characteristics and then set the controller for optimal power and control scheme efficiencies, and the second to operate at optimal conditions and record data. This will be used to perform accurate lab tests on the fuel cell system before installation of the fuel cell in the aircraft. 6 Future Proposed Work While we are currently working towards achieving the fuel cell long endurance flight, there are possibilities for future work with this aircraft. Gaseous hydrogen systems have a slightly higher specific power than existing boro-hydride systems [1], however cryogenic systems have roughly 10 times the power density. We are currently investigating
possibilities of creating an insulated tank system for use with cryogenic hydrogen, and have spoken with some tank and specialized materials manufacturers about such an endeavor. Depending on funding developed and interest from future students and external parties, more testing will be possible to investigate different power schemes, and flight envelope limits. 7 Acknowledgements The team has greatly enjoyed working on this cutting edge project, gaining invaluable skills in a variety of engineering tasks, and providing a useful segway into graduate school and career work. Many thanks are due to Dr. Thomas Bradley, the pilot, Rich Schoonover, the team at United Technologies Center, and Dr. Azer Yalin, with the CSU Space Grant Program. References 1. Bradley, T.H., Moffitt, B.A., Fuller, T.F., Mavris, D.N., Parekh, D.E. "Comparison of Design Methods for Fuel-Cell-Powered Unmanned Aerial Vehicles," Journal of Aircraft, Volume 46, Number 6, 2009. 2. Bradley, T.H., Moffitt, B., Mavris, D., and Parekh, D.E., Development and Experimental Characterization of a Fuel Cell Powered Aircraft, Journal of Power Sources, Vol. 171, 2007, pp. 793-801. 3. Bradley, T.H., Moffitt, B.A., Mavris, D.N., Fuller, T.F., Parekh, D.E. "Hardware-in-the-Loop Testing of a Fuel Cell Aircraft Powerplant," Journal of Propulsion and Power 2009, Vol 25, No 6. 2009