VALIDATED MODELING AND SYNTHESIS OF MEDIUM-SCALE POLYMER ELECTROLYTE MEMBRANE FUEL CELL AIRCRAFT

Proceedings of FUELCELL2006: 4 th International ASME Conference on Fuel Cell Science, Engineering and Technology June 18-21 2006, Irvine, California FUELCELL2006-97233 VALIDATED MODELING AND SYNTHESIS OF MEDIUM-SCALE POLYMER ELECTROLYTE MEMBRANE FUEL CELL AIRCRAFT Thomas H. Bradley Georgia Institute of Technology Woodruff School of Mechanical Engineering 801 Ferst Drive, Atlanta, Georgia 30332-0405 Blake A. Moffitt, Dimitri Mavris Georgia Institute of Technology Guggenheim School of Aerospace Engineering, 270 Ferst Drive, Atlanta, Georgia 30332-0150 David E. Parekh Georgia Tech Research Institute 400 10 th St. N.W., Atlanta, Georgia 30332-0801 ABSTRACT This paper describes a methodology for design and optimization of a polymer electrolyte membrane (PEM) fuel cell unmanned aerial vehicle (UAV). The focus of this paper is the optimization of the fuel cell propulsion system and hydrogen storage system for a baseline aircraft. Physics-based models, and experimentally-derived sub-system performance data are used to characterize the performance of each configuration within a design space. The results of aircraft synthesis and performance modeling routines are used to create response surface equations where tradeoffs among component specifications can be explored. Significant tradeoffs between fuel cell performance, hydrogen storage and aircraft aerodynamic and propulsion system design are presented. Validation and test results from a proof-of-concept fuel cell UAV propulsion system are presented. Validated models of the fuel cell and aircraft systems are used to predict the performance of fuel cell UAVs at the scale of the baseline aircraft. INTRODUCTION Proposed aircraft applications of fuel cells include auxiliary power units for conventionally powered aircraft or primary powerplants for low-power or long-endurance aircraft [1]. Fuel cell powerplants for UAVs are particularly attractive because of the fuel cell s high energy density when compared to other rechargeable energy storage systems. Where advanced batteries can reach energy densities of 150Wh/kg at the module level [2], fuel cells can achieve >800Wh/kg at the system level [3]. This study is concerned with the design of fuel cell powerplants to serve as the primary method of propulsion in UAVs. A few researchers have proven the viability of small scale fuel cell powered UAVs by constructing demonstration aircraft [4, 5]. Ofoma and Wu, [6] and Soban and Upton [7] have performed conceptual design of larger-scale fuel cell aircraft. AeroVironment has designed and demonstrated a large-scale fuel cell UAV [8]. Despite these recent achievements, there exists a need for a documented and validated methodology for design and analysis of fuel cell UAVs. In order to begin developing tools and technologies for analysis of fuel cells as powerplants in aeronautical vehicles, a PEM fuel cell UAV demonstration project was started in 2005 as a collaboration between the Aerospace Systems Design Laboratory at the Georgia Institute of Technology Daniel Guggenheim School of Aerospace Engineering and the Georgia Tech Research Institute. The primary research objectives of this project are: the development of validated methodologies and tools for the fuel cell aircraft design, and the demonstration of a series of fuel cell UAVs. At present, a self-contained 500W fuel cell propulsion system has been designed, constructed and tested [9]. A full-scale baseline UAV shown in Figure 1 has also been designed, constructed and is undergoing battery-powered flight testing. The purpose of this study is to quantify important tradeoffs in propulsion system design for fuel cell aircraft and to define optimal fuel cell propulsion systems for a baseline UAV. For this study, the propulsion system is defined to include the PEM fuel cell, fuel cell balance of plant, compressed hydrogen storage system, electric motor system and propeller. A mathematical model of the fuel cell aircraft is constructed in the MATLAB programming environment where the fuel cell propulsion components can be scaled within a range of values. The scaling of the propulsion component models affects the 1 Copyright 2006 by ASME

performance, mass and geometry of the propulsion component models and of the aircraft model as a whole. By defining the performance of a wide variety of propulsion system design configurations, the effect of variation of the fuel cell propulsion system design on the aircraft performance can be described. Optimal configurations can be characterized on the basis of these results. Table 1. Range of input variables for this study Design Variable Description Range Units L/d H 2 Radius H 2 Pressure fc_technology D p Kv n Hydrogen Tank Length-todiameter Ratio Hydrogen Tank Radius Hydrogen Tank Pressure Fuel Cell Technology Indicator Propeller Diameter Propeller Pitch Motor Voltage Constant Number of Fuel Cells in Stack 1-3 - 2.5-8.5 (6.35-21.59) 2500-4500 (17.24-31.03) in (cm) psi (MPa) 1-6 - 20-24 (50.8-61.0) 18-24 (45.7-61.0) 773-2320 (81-242) in (cm) in (cm) rpm/v (rad/vs) 32-64 - Figure 1. General configuration for fuel cell UAV, dimensions refer to baseline aircraft SYSTEM MODELING FRAMEWORK Scalable mathematical models of the primary aircraft subsystems have been constructed including: hydrogen storage system, fuel cell, fuel cell balance of plant, electric motor, propeller, fuselage, and aerodynamics. Each of these models represents a contributing analysis (CA). The CAs are connected to form a large system performance simulation which is scalable within a limited design space. To calculate the performance of all of the aircraft within the design space, each input variable is discretized over the ranges specified in Table 1 and a full-factorial design of experiments (DOE) is executed. For each case in the DOE, fixed point iteration is used to bring all variables passed between the CAs into agreement. The response metrics are then calculated and saved. In some DOE cases, either the CAs could not be brought into agreement, or the aircraft was incapable of flight. These cases were not used for further analysis. For this study, the design space includes the baseline demonstration aircraft. The baseline aircraft is a low-wing monoplane shown in Figure 1. The baseline aircraft is designed to be a low-power medium-endurance aircraft capable of carrying useful payloads for sensing and reconnaissance missions. As a result of the scale of the baseline aircraft, the design space encloses larger powerplants and aircraft structures than previous fuel cell aircraft demonstrations. However, the design space for the aircraft components is limited so as to enclose constructable aircraft whose scale is suitable for an academic research project. All powerplant components models represent the range of performance, mass and geometry that is available as off the shelf components. For each convergent design configuration, hundreds of variables are output that describe the performance, mass and geometry of the aircraft and powerplant. Response Surface Methodology (RSM) is used to organize, quantify and visualize these results. These outputs are condensed into several figures of merit that can be more easily interpreted. Where the raw output variables are used in this analysis for validation and detailed analysis, the figures of merit are more useful for meaningful interpretation of results. The primary figures of merit are: specific power, specific energy, thrust margin ratio, and climb rate. These figures of merit are described and defined as follows. Specific power (W/kg) and specific energy (Wh/kg) are defined using the fuel cell system output power, including balance of plant losses. All fuel cell, balance of plant, and hydrogen system components are included in the weight. Thrust margin ratio (T r ) describes the ratio of excess thrust to required thrust at the cruise condition: max_thrust T r =. (1) cruise_thrust Any design configuration with a T r >1 is able to achieve steady-level flight, but T r >3 is recommended for robust performance in flight. A larger thrust margin ratio is indicative of higher performance, better flyability and larger payload capacity. Climb rate (CR) is another figure of merit that quatifies the performance, flyability and payload capacity of the aircraft. Climb rate is defined for small angles of attack as: max_thrust - cruise_thrust CR = cruise velocity. (2) aircraft weight 2 Copyright 2006 by ASME

CONTRIBUTING ANALYSES Each CA calculates the performance, weight and geometry of the subsystem under analysis. In each of the following sections, the performance calculation algorithm is described, as are the methods used to scale that algorithm. Figure 2. Comparison of modeled fuel cell technologies in 32-cell stacks Fuel Cell Stack Contributing Analysis The fuel cell system is modeled as a static, linearized voltage versus current curve. This simplification of the conventional fuel cell polarization curve is used because the on-design performance of the fuel cell is dictated by the linear region of the polarization curve. The current output of the fuel cell is constrained to be less than the current that produces maximum power. The mass of the fuel cell system is based on a mass breakdown of the BCS Fuel Cells 500W fuel cell stack (Bryan, Texas), a low-pressure (5psi (0.21bar) air at cathode, 3psi (0.34bar) hydrogen at anode), self-humidified, commercially available stack of graphite and stainless steel construction. The geometry of the fuel cell systems have the same active area, bipolar plate area, bipolar plate thickness, and endplate dimensions as the BCS fuel cell. The fuel cell stack mass, length, open circuit voltage, and internal resistance are scalable by the number of cells within the fuel cell. The fuel cell stack open circuit voltage, internal resistance, and maximum current density are scalable by the fuel cell technology scaling factor. A fuel cell technology factor of 1 represents the performance of the commercially available (in stack form) BCS fuel cell. A fuel cell technology factor of 6 represents the performance of state of the art commercially available membrane electrode assemblies (Series 58, W. L. Gore & Associates, Inc., Newark, Delaware). The polarization curves of the varying fuel cell technologies in 32- cell stacks is shown in Figure 2. Fuel cell technologies that have higher performance than the present, self-humidified, commercially available state of the art are not considered. Fuel Cell Balance of Plant Contributing Analysis The fuel cell balance of plant represents the air delivery, hydrogen delivery and regulation, water cooling and power management and distribution subsystems of the fuel cell. All stacks are assumed to be self-humidified. The electrical power consumption and mass of the fuel cell balance of plant are based on measurements of the prototype fuel cell system. The hydrogen consumption of the stack is scalable using extrapolation from the experimental data obtained with the prototype fuel cell system. The electrical power consumption, and mass of the balance of plant, are scalable by the number of fuel cells and the peak output current according to the air requirements of the modeled fuel cell stack: [10] n M O I 2 m W = 2.0 (3) O2 4F A cathode stoichiometry of 2.0 is assumed for all stacks. The geometry of the balance of plant is assumed to be constant. The scaling of the geometry of the fuel cell balance of plant is not considered. Hydrogen Storage Contributing Analysis Compressed gaseous hydrogen in carbon fiber/aluminum pressure vessels is the hydrogen storage technology used for this design investigation. The mass, length and diameter of the hydrogen storage tank are scalable by the hydrogen storage tank radius and hydrogen storage pressure. The mass of the tank is determined using the Tank Performance Factor (TPF) of James et. al, 1999 where: [11] 5 5 5 TPF = min( rtank 1.2 10 + 2.9 10, 9.0 10 ) Pb Vtank TPF = (4). W The variation of TPF with tank radius (r tank, in inches) is a linear fit to a database of commercially available carbon fiber/aluminum pressure vessels. Electric Motor Contributing Analysis The electric motor analysis uses a lumped parameter, equivalent circuit model of the motor, as shown in Figure 3. This model uses a no-load current (I 0 ) and motor internal resistance (R m ). The current draw from the fuel cell is I m and the speed of the motor is proportional to the voltage (V m -I m R m ) through the motor constant Kv. The motor controller losses are modeled as a linear resistance. The motor analysis is scalable by the torque constants, no-load current, maximum output torque, speed constants and motor mass. The range of motor constants within the design space is shown in Figure 3. 3 Copyright 2006 by ASME

Fuselage Contributing Analysis The weight and geometry of the fuel cell, balance of plant, hydrogen tank, electric motor and controller are the inputs to the fuselage analysis. The fuselage analysis wraps a minimumvolume cylindrical fuselage around these systems and uses a volume of revolution formula for streamlining the nose and tail sections of the fuselage, as suggested by Roskam, 2000 [15]. The fuselage mass consists of a skin mass and a structural mass. The mass of the skin is scaled by the fuselage surface area. The mass of the structure (including landing gear mass) is scaled by the fuel cell system mass and the hydrogen storage system mass. Both the skin mass and the structure mass are scaled based on coefficients determined by two prototype fuselages built for the baseline aircraft. The skin mass assumes a fiberglass material construction while the structure mass is based on a carbon fiber structure with aluminum landing gear. Wing spar structure is scaled with aircraft mass. Figure 3. Values of measured motor constants used in electric motor CA Propeller Contributing Analysis The propeller analysis is based on Goldstein s vortex theory of screw propellers using the Betz condition [12]. The baseline propeller geometry used in this analysis is derived from measurements of a typical high performance small-scale propeller (2-blade 22x20, Bolly Products, Elizabeth West, South Australia) [13]. To account for propellers of varying diameter and pitch, the baseline aerodynamic pitch distribution and the planform blade shape is appropriately scaled while assuming that the baseline airfoil distribution along the blade span remains constant. This assumption provides excellent results for estimating the performance of fixed pitch propellers of similar size and pitch distribution as the baseline propeller. For any given propeller diameter, pitch, and number of propeller blades, the propeller thrust and power coefficients are calculated as a function of the propeller advance ratio. This allows the torque and the thrust of a given propeller to be calculated as a function of propeller RPM and airspeed. The effect of fuselage blockage is modeled using the recommendations of Lowry [14]. Based on a propeller optimization analysis using a BCS 500W fuel cell paired with several commercially available electric motors and propellers, it was found that a small range of propeller diameters and pitches consistently provide the most thrust for a given fuel cell power over a large range of electric motors [9]. These propeller diameters and pitches were used as the input variables to the propeller CA. Aerodynamic Contributing Analysis The aerodynamic analysis is conducted using Wings2004, a potential flow analysis code developed by Utah State University [16]. The profile drag and section lift data is calculated based on low speed wind tunnel test data [17]. Fuselage parasitic drag as well as fuselage drag due to lift is calculated using methods specified in Roskam [15]. Wings2004 then calculates the lift and induced drag to develop a drag polar for the aircraft. The empennage is an inverted V-tail design using a NACA 0009 airfoil. Sizing of the empennage is based on maintaining a static margin (scaled by the wing chord) of 26.4 and an aircraft yawing moment coefficient of 0.04. The static margin and yawing moment coefficients are determined based on pilot preference and were verified as adequate through several remotely piloted tests of the baseline aircraft. Sizing of the tail is accomplished using an iterative method involving Wings2004. The wing, as modeled, is based on the baseline aircraft. This wing is made up of a Selig-Donovan 7062 airfoil. The planform consists of a straight inboard section as well as a tapered outboard section (taper ratio=0.67) with no quarter chord sweep or washout. The tapered section employs 10 degrees of dihedral. Although the wing design can be modified and scaled within the aerodynamic contributing analysis, the wing design is held fixed at S w =17ft 2 (1.6m 2 ) and AR=18.2. The wing design was fixed since the goal of the study is to analyze and optimize the propulsion system within a baseline aircraft design. Aircraft thrust, cruise airspeed and maximum climb rate at cruise airspeed are calculated at standard atmospheric conditions at the elevation of Atlanta, Georgia, USA. VALIDATION RESULTS Because of the high computational load required to analyze every permutation within the design space, each CA incorporates the simplifying assumptions described above. 4 Copyright 2006 by ASME

Validation of the results from the CAs ensures that the results of the system performance simulation are accurate. for the weight and geometric scaling models for designs varying from the baseline UAV were also analyzed and found to behave as expected. The most difficult CA to experimentally validate is the aerodynamic CA. The calculated drag polar for the baseline aircraft was used to estimate the power required for steady, level flight. This power required was then compared to flight test data of the baseline aircraft and found to be in agreement. Also, a vortex lattice analysis of the performance of the baseline aircraft was created and found to be in agreement with the results of the aerodynamic CA. Figure 4. Comparison of experimentally measured and modeled powertrain performance (32 fuel cell stack, motor 14, propeller 1, fuel cell technology 1) To validate the performance models within the fuel cell CA, the fuel cell balance of plant CA, the electric motor CA and the propeller CA, a series of bench-top tests were performed. These tests compared the measured performance (fuel cell voltage, fuel cell current, motor speed, static thrust) of the hardware against its modeled performance for a subset of the design space permutations. All permutations were tested for one fuel cell, one fuel cell balance of plant, two electric motors and two propellers. Sample results from one of these validations is shown in Figure 4. In all cases, the results show good correlation between the modeled and tested performance. The results validate the performance models within these CAs. To validate the weight and geometric scaling models within the hydrogen storage CA, the tanks that were designed by the hydrogen storage CA are compared to commercially available tanks (L-series carbon-wrap, Luxfer Gas Cylinders, Riverside, California). The result of this comparison for a subset of the designed tanks is shown in Figure 5. There is good agreement between the two data sets. For other CAs, the weight and geometric scaling models can only be validated by reconstructing the fuel cell powerplant and aircraft structure for the baseline UAV. The mass and geometry of all components of the baseline UAV can be accurately reconstructed using the CA s. In addition, the trends Figure 5. Comparison volume and weight of modeled and commercially available hydrogen storage tanks FUEL CELL POWERPLANT DESIGN TRADEOFFS Large Scale Analysis Analysis of the results of the system modeling is accomplished using Response Surface Methodology (RSM). RSM is a mathematical modeling technique designed to represent the relationship between a response and the variables that influence the response [18]. The influencing design variables used in this study are given in Table 1. The responses chosen are the system level powerplant and aircraft performance metrics summarized in Table 2. The responses are modeled using polynomial Response Surface Equations (RSE) of the form: 3 3 3 β i xi + β ij xi x j + y = β + β x + ε, (5) 0 i= 1 i< j i= 1 where y represents a response, x i and x j represent the design variables, β ij are regression coefficients, and ε represents an assumed normally distributed error term. ii 2 i 5 Copyright 2006 by ASME

Figure 6. Sample of response surface equation profiles Out the 47250 cases executed, 31142 cases, or combinations of different values of the design parameters, were successful. Using these successful cases, a linear least squares regression was used to calculate each of the β ij terms in Eq. (5). The results of the regression are given in Table 2. Note that over the ranges of the design variables considered, the R 2 adjusted values are very high and the root mean squared error terms are relatively low indicating that the RSE model suggested by Eq. (5) can accurately model the system responses. One major advantage of using RSEs to fit data throughout the design space is that RSEs lend themselves to fast computation times. As a result, the design tradeoffs present in any engineering system are easy to quantify and visualize. Figure 6 shows sample profiles of the response surface equations. Figure 6 is structured as a matrix of plots where each plot represents how the response metric on the y-axis varies as a function of the corresponding design parameter on the x-axis. Each curve is calculated assuming that all of the adjacent design variables are fixed. The current fixed values of each of the design variables are shown next to the x-axis of each column. The predicted response variable values and their relative error are shown on the y-axes of each row. These values correspond to the design variable values shown on x- axes. 6 Copyright 2006 by ASME

Table 2. Summary of response surface fit Response Metric Mean R 2 Adjusted Root Mean Squared Error Thrust Margin Ratio 2.5 0.983 0.139 Endurance, mins 489 0.971 47.4 Climb Rate, m/min 77.3 0.983 6.66 Cruise Airspeed, m/s 20.2 0.994 0.244 Aircraft Mass, kg 31.6 0.987 1.21 Drag Coefficient (CD) 0.0365 0.999 0.0001 Specific Energy, Wh/kg 159 0.988 11.9 Specific Power, W/kg 58.8 0.989 2.27 The response surface equation profiles in Figure 6 show the fuel cell airplane system designer what the relative effect of the design variables are on the system performance. For instance, analyzing the relationship between the speed constant of the electric motor (Kv) and the thrust margin ratio (T r ), at a value of roughly 200rpm/V, the Kv has a small effect on the T r. As Kv increases, the motor becomes less effective and the T r decreases. In another example, increasing the number of fuel cells increases the maximum power of the fuel cell stack and increases T r. Figure 6 shows that the coupling between the number of fuel cells and the T r weakens as the number of fuel cells increase. As the number of fuel cells increase, the aircraft mass, and cruise airspeed increases. All of these have the effect of increasing cruise thrust and reducing the effectiveness of increasing the number of fuel cells to increase thrust margin ratio. RSM allows for efficient visualization and analysis of the tradeoffs that are present in the design of complex systems. For the fuel cell aircraft it enables the designer to consider the effect of the fuel cell powerplant system on aircraft-level performance metrics. Hydrogen Storage Geometry and Drag Tradeoff As shown in the preceding analysis, the geometry of the hydrogen storage system (hydrogen tank radius and hydrogen tank length-to-diameter ratio) has a large effect on the aircraft performance metrics. To further analyze the effect of the hydrogen geometry, a small region of the design space is examined in further detail. An example aircraft is designed that achieves 330 minutes of endurance and has a hydrogen tank volume of 6.6 L. The hydrogen tank radius and hydrogen tank length-to-diameter ratio are varied to determine the effect on aircraft performance. The volume of the hydrogen tank and all other design variables are kept constant. Under steady-level flight, the available propulsive power is equal to the required propulsive power (P req ). The equation for required propulsive power is: 1 Preq = Drag V = ρ 2 Sw CD V V. (6) 2 At constant density, wing area (S w ), and design lift coefficient, there are two main components of the aircraft propulsive power, a component proportional to CD and a component proportional to V 3. The power component proportional to CD is due to the drag caused by the shape of the aircraft. The cruise airspeed (V) required to maintain steadylevel flight is directly proportional to the square root of the aircraft mass. Thus, the power component proportional to V 3 is caused by the mass of the aircraft. The total propulsive power required for cruise is equal to the product of these components. By normalizing the CD and V 3 terms in Eq. (5), the percent changes in propulsive power required due to shape and weight of the aircraft can be compared. The cruise airspeed (V), drag coefficient (CD), and total propulsive power for this aircraft with a spherical hydrogen tank radius are used to normalize the components of Eq.(5). Figure 7. Normalized total drag increase for hydrogen tanks Figure 7 shows the effect that varying the hydrogen tank length-to-diameter ratio has on the components of the aircraft propulsive power required for steady, level flight at cruise. Because the volume of the hydrogen tank remains constant, as the hydrogen tank length increases, the tank radius must decrease. At the vertical line labeled in Figure 7, the radius of the hydrogen tank is equal to the effective radius of the fuel cell. In this example, this transition occurs at a length-todiameter ratio of 2.7. For tank length-to-diameter ratios less than 2.7, the increasing tank radius increases the frontal area of the aircraft thereby increasing the normalized CD. For hydrogen tank radii 7 Copyright 2006 by ASME

less than the fuel cell equivalent radius, the increasing tank radius has no effect on the frontal area of the aircraft because the fuel cell is of larger diameter than the hydrogen tank. In this case increasing the length-to-diameter ratio increases the length and wetted area of the fuselage thereby increasing the normalized CD. Note that for the streamlined fuselage considered, length-to-diameter ratios affecting the frontal area of the fuselage have a much larger impact on CD as compared to length-to-diameter ratios that cause the fuselage to be lengthened. As shown in Figure 7, increasing the length-to-diameter ratio always increases the normalized V 3 component of the aircraft power. Since spherical tanks are lighter weight per unit volume, any deviation from a spherical tank will increase the tank mass and thus increase the cruise airspeed. In addition, as the hydrogen tank length-to-diameter ratio increases, the fuselage length must increases, adding further mass to the aircraft. For hydrogen tank length-to-diameter ratios less than 2.7, the increased fuselage mass necessary to accommodate the large frontal area of the hydrogen tank makes up for any mass lost due to a more spherical tank. At length-to-diameter ratios more than 2.7, increasing the fuselage length to accommodate the length of the hydrogen tank increases mass as in addition to the mass increase due to the increasing aspherical shape of the tank. This results in the increase in the normalized V 3 component slope for length-to-diameter ratios above 2.7. The normalized required propulsive power in Figure 7 is a product of the normalized CD and V 3 components. This normalized required power displays a shallow minimum at a length-to-diameter ratio slightly less than 2.7. The minimum represents a tradeoff between decreasing the CD component while increasing the V 3 component. Note however, that optimizing the length-to-diameter ratio overall can only decrease the propulsive power required by 3%. In addition to normalized power required, Figure 7 also shows normalized thrust margin ratio and climb rate values as a function of length-to-diameter ratio. Note that for this aircraft design, the climb rate optimizes with a nearly spherical tank while the optimum thrust margin ratio occurs very near the optimal propulsive power required length-to-diameter ratio. As a result, slightly different length-to-diameter ratios may be chosen for the hydrogen storage depending on which performance metrics are most important. The final aspect of Figure 7 to note is the scale of the variation in each of the normalized components. Although Figure 6 shows that the hydrogen storage tank design parameters have a profound effect on aircraft performance metrics, once a tank volume and pressure have been decided, the competing effects of tank radii and length-to-diameter have only a small effect on the changes in aircraft performance metrics. Thus, the overall aircraft performance will be primarily influenced by high-level design variable choices. Specific Energy and Specific Power Tradeoff One of the proposed benefits of fuel cell powerplants is that the specification of the energy content of the system is relatively independent of the power content of the system [19]. Hydrogen tanks, which dictate energy content, can be sized separately from the fuel cell, which dictates power content. Because of this simplification, fuel cell system design for aircraft has often been performed separately from design of the aircraft. In these cases, fuel cell system design focuses on maximizing energy density and the aircraft is designed to minimize the power required for flight [20]. In fact, the requirements of the aircraft force a tradeoff between energy and power specifications as simultaneous design of aircraft and fuel cell systems can show. To quantify the effects of this tradeoff, Figure 8 shows a Ragonne plot where the area inside of the contours contains all of the fuel cell powerplants designed for this study. By varying fuel cell technology, cell number, hydrogen storage pressure, and hydrogen storage tank radius, the power and energy density of the fuel cell powerplant can be varied within the limits of the contours. Of course, there are multiple designs that can have the same specific power and specific energy, Figure 8 shows only the design with the higher climb rate (CR). In summary, Figure 8 shows the maximum performance that is available from any aircraft configuration within the design space as a function of specific power and specific energy. Figure 8. Ragonne plot for all viable fuel cell powerplant systems within the design space, with contours of constant aircraft climb rate Steady, level flight is achieved at any point within the region where the CR is greater than 0. Figure 8 shows that within the design space of this study, there is a limit on powerplant specific energy that is obtainable at a given climb rate. So, although some of the fuel cell powerplants designed 8 Copyright 2006 by ASME

for this study can have very high specific energies, their high specific energy is unusable for aircraft application because of their lower specific power. For example, the aircraft designer might specify that the aircraft design must have a CR>150m/min. This specification provides a limit of 460Wh/kg as the maximum fuel cell powerplant specific energy available to the designer, although higher specific energy configurations are available within the design space. This specification also limits the specific power of the fuel cell system. No aircraft within the design space can achieve CR>150m/min with a fuel cell system specific power less than 73W/kg. By incorporating fuel cell system design with the design of the aircraft, the constraints imposed by the aircraft can considered in specification of the fuel cell design variables. Characteristics of High Performance Configurations In order to predict what the specifications of the fuel cell system would be for a medium-endurance aircraft at this scale, a set of high performance design configurations were chosen from all of the aircraft that were simulated. Configuration #1, #2, and #3 are the designs with the highest climb ratios for cruise endurances between 0-100 mins, 200-300 mins and 400-500 mins, respectively. in radius and increasing in length. The net effect of the differences between the high performance configurations is shown in Table 3. As the endurance required increases, both the specific energy and the mass of the aircraft increase as well. Because the fuel cell does not change between the high performance configurations, the specific power goes down with increasing endurance. Table 3. Design characteristics for high performance configurations Design Variable Cruise Endurance High Perf. Config. #1 High Perf. Config. #2 High Perf. Config. #3 35 mins 205 mins 404 mins Aircraft Mass 26.4kg 28.8kg 33.9kg Specific Energy 8.4 Wh/kg 56.8 Wh/kg 120 Wh/kg Specific Power 121 W/kg 110 W/kg 91 W/kg Fuel Cell Technology Number of Fuel Cells 6 6 6 64 64 64 Cruise Power 269W 244W 446W Figure 9. Fuselage specifications and packaging diagrams for high performance configurations Each high performance configuration uses the maximum number of fuel cells (64), and the maximum fuel cell technology (6). As suggested by the response surface equation profiles in Figure 6, there exists an optimal motor and propeller for each high performance configuration that maximizes thrust margin ratio. Figure 9 shows the packaging of the powerplant for the high performance configurations and the effect that the fuel cell components have on the fuselage geometry. As the hydrogen storage requirements of the airplane increase with increasing endurance, the hydrogen tank volume increases by increasing CONCLUSIONS A system design method for fuel cell aircraft has been developed that uses subsystem-level models as part of a large system performance simulation. A design space is described by choosing design variables and discretizing the variables over appropriate ranges. The system performance simulation then calculates the performance, weight and geometric configuration of the aircraft for a full factorial combination of all settings of the discretized design variables. The results of this analysis have been validated at a few accessible points in the design space using results from a proof-of-concept fuel cell aircraft and its powerplant. The results of this design method are presented for a limited design space at the scale of the proof-ofconcept fuel cell aircraft. The effects of the individual powerplant design variables were quantified through the use of Response Surface Methods. The design of the fuel cell powerplant affects the aircraft mass, geometry and performance and are found to have a large effect on the aircraft performance metrics. The design tradeoff between hydrogen tank geometry and aircraft drag is explored in detail. The optimal configuration is found to depend on the aircraft performance metric pursued. At the scale of the baseline aircraft, for a fixed hydrogen tank volume, the effect of hydrogen tank geometry is relatively minor. Aircraft performance is primarily determined by other performance metrics such as endurance or minimum climb rate. The tradeoff between specific power and specific energy specifications of the fuel cell is quantified within the 9 Copyright 2006 by ASME

considered design space. The results of this design tradeoff study can be generalized to show that aircraft performance requirements provide a limitation on the fuel cell system specific energy and specific power. This design study highlights a few high performance fuel cell aircraft configurations with varying design endurance. The geometry and performance of one of these configurations was compared to quantify the characteristics of viable fuel cell powered UAVs. NOMENCLATURE AR Wing aspect ratio β ij Response surface regression coefficients CA Contributing Analysis CD Aircraft total coefficient of drag F Faraday's constant, 96,485 C/mol ε Response surface error term I 0 Electric motor no-load current, A I m Fuel cell output current Kv Electric motor voltage constant, rpm/v M O2 Molecular weight of oxygen, kg/mol P b Hydrogen tank burst pressure, psi R cont Controller resistance, Ω R m Electric motor internal resistance, Ω r tank Hydrogen tank radius, in SL Standard Liters S w Planform wing area TPF Tank Performance Factor, in T r Thrust margin ratio T req Aircraft propulsive thrust required for cruise V Aircraft cruise airspeed V tank Hydrogen tank volume, in 3 V m Electric motor input voltage, V W Hydrogen tank weight, lbs W O2 Oxygen flow rate to fuel cell system, kg/sec x i, x j Response surface design variables y Response surface response ACKNOWLEDGMENTS This research was funded in part by the NASA University Research Engineering Technology Institute (URETI) grant to the Georgia Institute of Technology. The authors would like to thank the research engineers at the Georgia Tech Research Institute and the Aerospace Systems Design Laboratory who provided valuable expertise and guidance during the course of the project. REFERENCES [1] Liang, A. T. Emerging Fuel Cell Development at NASA for Aerospace Applications, 4 th Annual SECA Meeting, Seattle, Washington, April 15-16, 2003 [2] Anderman, M. Brief Assessment of Improvements in EV Battery Technology since the BTAP June 2000 Report, California Air Resources Board, 2003. [3] Burke, K. A., High Energy Density Regenerative Fuel Cell Systems for Terrestrial Applications, NASA/TM 1999-209429. [4] Scheppat, B. Betriebsanleitung für das brennstoffzellenbetriebene Modellflugzeug, Fachhochschule Wiesbaden, 2004. [5] Kellogg, J., Fuel Cells for Micro Air Vehicles, Joint Service Power Expo, Tampa, Florida, May 2-5, 2005. [6] Ofoma, U. C., Wu, C. C. Design of a Fuel Cell Powered UAV for Environmental Research, AIAA-2004-6384. [7] Soban, D. S. and Upton E. Design of a UAV to Optimize Use of Fuel Cell Propulsion Technology, AIAA 2005-7135. [8] AeroVironment Press Release, AeroVironment Flies World s First Liquid Hydrogen Powered UAV, June 28, 2005. [9] Moffitt, B. A., Bradley, T. H., Parekh, D., Mavris, D., Design and Performance Validation of a Fuel Cell Unmanned Aerial Vehicle. AIAA 2006-823. [10] EG&G Technical Services, Fuel Cell Handbook, 7 th Ed. U.S. Department of Energy, Morgantown, West Virginia, 2004. [11] James, B. D., Thomas, C. E., Lomax Jr., F. D., Onboard Compressed Hydrogen Storage, Canadian Hydrogen Association Meeting, Vancouver, February, 1999. [12] Goldstein, S., On the Vortex Theory of Screw Propellers, Proceedings of the Royal Society of London, Series A. Vol. 123, No. 792, pp. 440-495, 1929. [13] Kelly, Q. J., Validation and Implementation of Goldstein s Vortex Theory of Screw Propellers, Utah State University, 2002. [14] Lowry, J. T., Performance of Light Aircraft, AIAA Education Series, 1999. [15] Roskam, J., Airplane Design Part VI: Preliminary Calculation of Aerodynamic, Thrust and Power Characteristics, DAR Corporation, 2000. [16] Phillips, W. F., and Snyder, D. O., Modern Adaptation of Prandtl s Classic Lifting-Line Theory, Journal of Aircraft, Vol. 37, No. 4, pp. 662-670, July 2000. [17] Lyon, C. A., Broeren, A. P., Giguere, P., Gopalarathnam, A., Selig, M.S., Summary of Low-Speed Airfoil Data: Volume 3, SoarTech Publications, Virginia Beach, Virginia: 1997. [18] Wu, C. F. J., and Hamada, M., Experiments: Planning, Analysis, and Parameter Design Optimization, John Wiley and Sons, Inc. New York, 2000. [19] Wickenheiser, T. J., Sehra, A. K., Seng, G. T., Freeh, J. E., Berton, J. J. Emissionsless aircraft: requirements and challenges, AIAA 2003-2810. [20] Nam, T., Soban, D. S., Mavris, D. N., Power based sizing method for aircraft consuming unconventional energy, AIAA 2005-818. 10 Copyright 2006 by ASME