AIAC-2011-052 DESIGN, BUILD AND FLIGHT

6. ANKARA INTERNATIONAL AEROSPACE CONFERENCE 14-16 September 2011 - METU, Ankara TURKEY

AIAC-2011-052

DESIGN, BUILD AND FLIGHT TESTING OF A VTOL TAILSITTER UNMANNED AERIAL VEHICLE WITH HYBRID PROPULSION SYSTEM*

Miraç Aksugür1

Gökhan İnalhan2

Istanbul Technical University Istanbul, Turkey

Istanbul Technical University Istanbul, Turkey

ABSTRACT As a kind of vertical take off and landing capable unmanned vehicles, tailsitter UAVs with their combined VTOL and fixed-wing aircraft with full flight-speed regime capability provides a distinct alternative to rotary-wing and ducted fan UAVs. ITU tailsitter concept is tailored towards city and urban operations with possible autonomous recharging capability to allow 24 hour on demand reconnaissance and surveillance for traffic and law-enforcement. The development of manned tailsitter aircraft had begun as early as 1950’s. Such manned tailsitter aircraft were hard to control especially during landing phase, as the early tailsitter aircraft did not have any stability augmentation system to help the pilots during the critical landing phase. However, as unmanned systems developed, the distinct tailsitter concept is realized again by using recent autopilot technology. In ITU tailsitter, a folding propeller system is used for hovering, vertical take-off, vertical landing and low speed transition mode, whereas an electric ducted fan (EDF) system is used for level and high speed flight mode where the propeller folds onto the fuselage in order to reduce drag. Initial system performance analysis with candidate propulsion units indicate that up to 35m/s cruise speed and maximum 90 minutes of flight endurance can be achieved while carrying 1.2 kg payload – a distinct performance in comparison to the same class rotary-wing and OAV alternatives. This flight time includes 3 minutes of vertical take-off and landing phase. After being proven the new tailsitter concept with hybrid electric propulsion system with the help of prototyping and several flight tests, a new concept with several usage areas, such as reconnaissance and surveillance for traffic and law-enforcement, scientific research, defense industry, is going to born. INTRODUCTION Tailsitter UAVs combine vertical take off and landing (VTOL) operation and relatively high speed capabilities in single airframe and such concept provides manifest advantages over the other VTOL aircraft concepts including helicopters and organic air vehicles (OAVs). As a result of the increasing requisition on efficient and silent UAV concepts which require no regular ”runway” for urban-civilian applications, the design of ITU-BYU tailsitter concept is adopted for tailoring towards city and urban operations with possible autonomous recharging capability to allow 24 hour on demand reconnaissance and surveillance for various usage areas from tra c/law-enforcement to border patrolling. The design and development of manned tailsitter concepts had begun in the beginning of 1950s and many experimental aircraft are build and tested during the period between 1950 and 1960s. The pilots of such manned aircraft were in charge of the aircraft’s attitude control by only looking at the displays and sensing the behaviors of the aircraft in an upside position with unfamiliar control inputs comparing to the helicopter pilots. Moreover, such manned aircraft had control problems especially during landing and hover-to-cruise transition phase, because of having no stability augmentation system which helps test pilots to reduce their extremely-high-workload. Therefore, none of the experimental tailsitter concepts have proven the advantages over helicopters or fixed wing aircrafts. As a result of several accidents mainly due to the high workload over the test pilots, the development of the manned tailsitter projects were backed off after mid 1960s. However, with the * This work is supported under ITU BAP Project # 33364. 1 Graduate Student at the Faculty of Aeronautics and Astronautics , Email: [email protected] 2 Assoc. Prof. Dr. at the Faculty of Aeronautics and Astronautics and Director of CAL, Email: [email protected]

AIAC-201 11-052

Aksugü ür, İnalhan

help of th he advancess in both hard dware and so oftware base ed technologies [1], [2], [33], the distinc ct tailsitter concept wass realized and took place e among the other VTOL UAVs. Likew wise, there arre many VTOL co oncepts have e developed or still underr development. Now, OAV Vs are the m most popular ones that can be classified d under tailsitter conceptss. Besides, Allied A Aerosp pace’s iStar sseries Organ nic Air vehicles (OAV) and Honeywell’s H electric pow wered mini cla ass OAV are the foremosst examples that t have wid de range of application a area on the w world. On the other hand, although thee OAVs show w efficient static-low sp peed flight due to the shrroud and duc ct around the e propeller bllades, such ducted d fan UAV V concepts sttill have some e aerodynam mic problems s especially about a efficiennt forward fliight regime a and hover to cruise transition phase. In addition, OAVs O with th heir internal ccombustion engines e also sufffer from high levels of noise during op peration. The erefore, the noise n problem m makes the ese types of tailsittter UAVs unssuitable for ”s silent intellige ence” in urba an operational areas. Othher examples to the mini-mid d size tailsitte er concepts are a T-Wing a and Heliwing. T-Wing was s developed at the Unive ersity of Sydney a at 2000s [4],, and Heliwin ng was develloped by Boe eing. Howeve er, both of thhese similar concepts c are not ssize wise suittable for urba an applicatio ons because of their noisy internal co mbustion en ngines. Besides,, there are allso several micro m class ta ailsitter UAV concepts su uch as Brighaam Young Universitty’s tailsitter UAV [5] and d University o of Arizona’s coaxial c prope eller driven U UAV [6]. Neverthe eless, such micro m size UA AVs are not suitable for carrying c “use eful” payload and for serv vicing in severe w weather cond ditions. Two of o the discusssed tailsitterr concepts arre seen in Figgure 1 Since e 2001, related to o recently increasing technology on L Lithium base ed batteries, aircraft a desiggners have started s to think outt electric pow wered aircraftt concepts [7 7].

Figurre 1: a: T-win ng concept demonstrator d r from Univerrsity of Sydney [1], b: M Micro tailsitterr UAV Proto otype from University of Arizona A [6] Moreove er, advances in brushless s DC electricc motors have e accelerated the develoopment period of such kind of a aircraft. In ITU U Tailsitter, electric e propu ulsion system m choice as a major pre-ddesign selec ction, because e of electric powered p prop pulsion’s low w noise levels s and the unique capabilitty to autonom mously recharge e the units fro om base land ding stationss. However, for f an electric powered ve vehicle within n the mini-UAV V class, this unique capa ability calls fo or a trade-o between spe eed limited hhigh power prropeller configura ation and the e power limited high spee ed electric du ucted fan (ED DF) system. In this study y, for mini class UA AVs, design optimization o and intuitive e thrust-powe er-airspeed trrade-off apprroach which leads to a hybrid ”propeller-du ucted fan pro opulsion systtem” design that can ach hieve maximuum horizonta al flight d maximum range r for the ongoing ITU U Tailsitter Project is prov vided. time and Due to th he aim of designing an efficient tailsittter UAV, pro opulsion system has the hhighest priorrity among the design re equirements. Since electrric powered propulsion p sy ystem is alsoo advantageo ous to internal ccombustion engines e in te erms of main ntenance and d noise level; electric proppulsion syste em has been cho osen as one of the requirrements. Pro opeller driven n system sup pplies high thhrust to powe er ratio for VTOL L operations. However, th he thrust is rrapidly decre eased as the incoming airrspeed is inc creased. Hence, tthe performa ance is decre eased at high h speed and the system becomes b inssufficient for cruise. c On the o other hand, Electric E EDF system is ca apable of pro oducing the same s thrust w with higher th hrust to power grradient as the propeller system. s Thou ugh the powe er consumption is increassed, EDF system’s high thru ust is necesssary for long range cruise e operations. As a result, the hybriid propulsion n system, con nsisting of bo oth propulsio on systems, w was decided to be used in o order to design an “all flight regime” efficient aircraft which meets the requuirements. In n ITU Tailsitterr, a folding prropeller syste em located o on the nose of o the aircrafft, is used forr hovering, ve ertical take-o , vvertical landiing and low-s speed transittion mode, whereas w an EDF E system, which is pla aced between n the stabilize ers, is used for f level and high speed flight mode where w the prropeller folds s onto the fusellage in orderr to reduce drag when it i s turned o . In addition, to calculate thhe approxim mate empty w weight calcula ation, instead d of the classsic method of empty weig ght fraction, ””aircraft-base ed” weight m modeling and d optimization n study have e been condu ucted in orde er to see the most efficien nt design which is possible. Iniitial system performance p analysis with candidate propulsion uunits represent that 2 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-2011-052

Aksugür, İnalhan

up to 40m/s cruise speed and maximum 90 minutes of flight endurance can be achieved while carrying 1.5 kg of payload in 10kg of flying system with 3 minutes of vertical take-o and landing operation duration. - a distinct performance in comparison to the same class rotary-wing and OAV alternatives. In the proceeding sections, a trade off analysis is given and hybrid-dual propulsion approach with a qualitative analysis of the selected propulsion systems is described. This is followed by the design overview and the design and design optimization approach. After that, the control methodology and the results are denoted. PROPULSION SYSTEM As well as the aerodynamic design, selection of the right propulsion system/component is utmost determinative on the performance of an aircraft. Comparing the propulsion systems of both “heavy lifter” helicopter and “fast and agile” jet fighter, it is comprehended that relatively high diameter and low weight loading propeller blades are efficient for hovering when relatively small diameter blades having high induced velocity, are used as effective way to reach high speeds. First, we compare the propeller and EDF propulsion systems separately with both qualitative and quantitative approach. Then we describe the advantages of combined propulsion system that we call “hybrid” propulsion. Qualitative Comparison of Candidate COTS Propulsion Systems To select appropriate propeller for an aircraft, all the performance data of the candidate propellers should be carefully analyzed. Although there is a large number of available performance data about propellers [8], these propellers are mostly used on commercial or military manned aircraft. On the other hand, there is no sufficient and systematic cataloged propeller performance charts which are used on small scale UAVs, except for some test results [9]. In the analysis conducted, propeller performance is measured by plotting propeller coefficients against advance ratio (J). For tailsitter UAV application, propeller system is considered to be used as primary lift generating device during VTOL operations. Thus, to get the highest specific thrust (T/P) value, a propeller having largest diameter available and relatively low pitch value should be selected. Because, propellers having high pitch value are designed for high-speed applications and high percentage of propeller blades are stalled during zero speed (hovering) or low speed regimes (i.e. low speed vertical climb). To determine the specific propeller coefficients analytically, three important variables must be known; propeller chord distribution, airfoil twist distribution and airfoil data for each section of the blade. However, most of the commercially available hobby purpose propeller manufacturers do not provide such detailed information about their designs. For illustrative purposes, Graupner’s carbon folding, 20x12 size (20 inches of diameter and 12 inches of pitch) propeller is considered. Figure 2 shows the thrust and the power coefficients against the advance ratio for this propeller; a commonly used propeller for this class of UAVs. After T/P ratio vs. airspeed conversion, the propeller data can be illustrated as given in Figure 3. 0,08 0,07 0,06

Ct, Cp

0,05 Ct

0,04

Cp

0,03 0,02 0,01 0 0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

J (V/nD)

Figure 2 : Thrust (Ct) and power (Cp) coefficients vs. advance ratio, for Graupner 20x12 carbon folding propeller. These coefficients were found analytically by Dr. Martin Hepperle and the source data can be reached by his website [10].

3 Ankara International Aerospace Conference

AIAC-2011-052

Aksugür, İnalhan

Ducted Fan Propulsion System: There are many commercially available EDF units consisting of 3 to 7 blades regarding to their size and made from plastic or carbon fiber related to the working conditions. In addition, the commercially available EDF systems’ diameters can vary between 5 and 14.5 cm.

0,08 0,07 0,06

T /P

0,05 0,04 0,03 0,02 0,01 0 0

5

10

15

20

25

V (m/s)

Figure 3 : Specific thrust (T/P) vs. airspeed, for Graupner 20x12 carbon folder propeller. However, like small size propellers, hobby purpose EDF units also suffer from lack of any catalogued performance data. Moreover, they do not exhibit identical ducted fan behavior because of having wider gaps between shroud and blade tips than the full size precisely manufactured ducted fans. Many of the commercially available EDF systems are designed for high-speed applications, such as radio controlled model jets. In addition, at low speeds, for a given unit power input, EDFs produce less thrust than the propeller systems. Thus, EDF systems are suitable for relatively high cruise speed in comparison to the propellers. Hence, even though the T/P ratio of EDF systems are quite low at the static condition, second derivative of the T/P curve with respect to the airspeed is lower than the propellers’ T/P curve’s second derivative. Note that, there is only one unit with available wind tunnel test measurements, which is officially published on the manufacturer’s website [8]. This is Schübeler brand’s DS51 type EDF unit. According to the measurements, T/P vs. airspeed graphic for DS51 EDF system is shown in Figure 4. 0,02 0,018 0,016 0,014

T/P

0,012 0,01 0,008 0,006 0,004 0,002 0 0

5

10

15

20

25

30

35

40

45

50

V

Figure 4: Specific thrust (T/P) vs. airspeed graph of Schübeler DS51 EDF unit, based on wind tunnel measurements, graphed by using the manufacturer’s own data Because of the higher efflux velocity, specific thrust loss of EDF system in dynamic conditions is lower than the propeller system’s loss, which in turn is advantageous for EDF systems in high-speed conditions. However, although the EDF unit can produce as much thrust as a propeller does, the power consumption at those thrust levels are much higher than the propeller system because of the lower T/P ratio. Therefore, EDF usage provides the ideal solution within the high speed flight while 4 Ankara International Aerospace Conference

AIAC-2011-052

Aksugür, İnalhan

propeller system usage provides the ideal solution within the VTOL and low speed flight. This is illustrated in Figure 5. 0,08 0,07 0,06 0,05 T/P

Graupner 20X12 0,04

Schübeler DS51 EDF

0,03 0,02 0,01 0 0

5

10

15

20

25

30

35

40

45

50

V (m/s)

Figure 5 : Comparison of change of specific thrust values with airspeed. Propeller is Graupner 20x12; EDF system is Schübeler DS51. For 20x12 propeller and Schübeler DS51 EDF combination, an active region of each propulsion system’s usage is obvious. For VTOL (nearly static condition where V is zero) and transition up to 22.5 m/s velocity, propeller propulsion system can be considered to be active. After that, for cruising and high speed flying, propeller stops and fold onto fuselage and EDF system is activated to maintain desired airspeed above 22.5 m/s. Combining the benefits; Hybrid Propulsion Approach In this hybrid propulsion system, as discussed before, for both VTOL operations and hover to cruise transition phase, propeller system is used because of its energy efficiency. During cruising at high speed demand, EDF system is advantageous than propeller system. For the ITU tailsitter aircraft, the initial propeller selection is a RASA 28x12.5 size propeller and the EDF system is Schübeler DS51. Although there is wind tunnel test data for DS51 EDF system, the selected propeller unit has a measurement data only for static condition where incoming airflow is zero. Thus, some assumptions are made based on the performance values of the propeller propulsion systems to start the initial design process. The assumptions and the performance results for the selected propeller system take shape after inspecting the similar folding propellers’ geometry, After making a comparison between RF brand 20x13, 20x12 and 23x12 propellers, it was seen that the chord distribution is almost directly proportional to size scaling. So, the chord length distribution for 28.5x12 RASA brand propeller is determined by using the similar chord distribution. After that, the twist distribution from the root of the propeller to the tip of the propeller is estimated in the light of the Graupner 20x12 and RASA 23x12 propellers’ shape data. Because, the pitch values of both 20x12, 23x12 and 28.5x12 propellers are identical. Thus, both chord length and twist distribution of the selected propeller were entered to the propeller performance calculation software [11]. According to the results of the software, it is seen that the thrust and power coefficients nearly match with the static test results [12]. T/P curve based on the software can be seen in Figure 6.

5 Ankara International Aerospace Conference

AIAC-201 11-052

Aksugü ür, İnalhan

28.5x12 0,1 0,09 0,08

T/P (N/W)

0,07 0,06 0,05 0,04 0,03 0,02 0,01 0 0,00

2,00

4,00

6,00

8,00

10,00

12,00

14,00

1 16,00

18,00

20 0,00

V (m/s)

Figure e 6 : Estimatted specific tthrust changing for RASA A 28.5x12 pro ropeller As a result, the final T/P graphic for selected propulsion system s can be b seen in Figgure 7. Note e that, the initia al condition iss given for a constant thru ust value. He ence, at zero o speed, bothh of the syste ems are assigned d the same th hrust value.

Figure 7 : T/P vs. airspeed a com mparison of the t selected propulsion ssystems Qualitative Comparisson of Candidate COTS P Propulsion Systems; S sca aled specific tthrust concept: Specific thrust, whicch is defined d as thrust to power ra atio, is a pre eferential meethod to define and compare e the efficiency values of propulsion ssystems. Mo ore, since the specific thhrust value is sa function of advance ratio r that is also a a functio on of angula ar and airspeed, three dim mensional co omplex surface g geometry an nalysis should d be conduccted in orderr to see and quantize thee performanc ce of both of the propulsion systems. However, H witth the help off the scaled specific s thrusst method, which w are structure ed during the e design pha ase of ITU T Tailsitter UA AV, the thre ee dimensionnal specific thrust t determin nation proble em has sca aled down to two dimen nsional prob blem. To doo that, x axiis represen nts incoming g air velocity y, which is sccaled from ze ero airspeed d to the maxi mum desired d airspeed d, while y axxis represen nts specific tthrust values s as seen in Figure 8. T The calculattion method o of scaled spe ecific thrust for f both EDF F and propeller propulsion n systems iis seen in Figures 9 and 10 0. Note that, for the selected “prope eller based” propulsion systems, thruust and powe er coefficients can be written as a quadratic function of advance rattio. Thereforre, the letterrs a1, a2, a3 and a2, b2, c2 in Figure F 9 and Figure 10, are the co oefficients off the quadraatic equation ns of EDF and propeller systems s resp pectively.

6 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 8 : Scaled specific thrust ccomparison of the selec cted propellerr and EDF

F Figure 9 : Sccaled specific thrust calcculation meth hodology for EDF propuulsion system m. In Figurre 11, there are two re egions, whicch are valid for our con nsideration. The first reg gion is zero sp peed, which is simulate ed as hoverring maneuv ver; the seco ond region is from the black b vertical lline, which in ntersects the stall speed (16 m/s) on the x axis s, to the dessired maximu um airspeed d, which is 35 5 meters per second. Th he pink area between “ve ery-low” speeed (0-5m/s) and a stall speed iss intentionallly left blank k. This is beccause the sim mulation of exact e and opttimized trans sition maneuvvers hasn’t be een complete ed yet. To g ive more infformation, propulsion sys ystem only ov vercomes parasite and induce ed drag from m the stall sp peed to the maximum airspeed. a Hoowever, durin ng the transition maneuverss, aircraft’s weight is a added to drrag compone ent, which ppropulsion sy ystem must ovvercome. The erefore, the accurate a tran nsition phase e airspeed is s investigatedd after calcu ulating the optim mized transittion maneuve ers. As seen from Figure 11, each of the propulsioon systems with w any given airrspeed condition results in different re evolution perr minute and d advance raatio values. Thereforre, the comp parison betwe een the prop pulsion syste ems can be made outrighht in Figure e 11. Thus, th he EDF sysstem is abou ut 64 times more efficie ent than the propeller syystem for the whole flight reg gime from sta all to maximu um airspeed .

7 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

F Figure 10: Sccaled specifiic thrust calcculation meth hodology forr EDF propuulsion system m. Moreove er, for static case c analysis s, specific th hrust for both propeller an nd EDF systeems are calc culated as 0.105 52 and 0.012 22. Hence, de epending on the static ca ase calculatio ons (for hoveer and low sp peed climb) it is seen that the propellerr system is a about 9 times s more efficie ent than the E EDF system. However, due to ang gular velocity y restriction, tthe EDF system cannot produce p adeqquate thrust as the propellerr system.

Figure 11: Scaled sp pecific thrus st comparison n of both ED DF and prope eller propulsioon systems in hover and hig ght speed co ondition. As a result, the breakkthrough adv vantage of th he hybrid pro opulsion apprroach is seenn as an evide ent in Figure 1 11. O OPTIMIZATIO ON The de eveloped dessign philosophy hinges on n obtaining the maximu um possible ppayload capa acity while achieving both high thrust to weight ra atio for VTOL, maneuverability andd low energy y demand per unit operation time (i.e. lo ow power de emand for enhanced e endurance). e T To do this, an a aircraft, w which has a relatively high h cruise sp peeds with vertical takeoff and landinng capabilities, has been delineated. In addition, res strictions com ming through city operatio onal environm ment were re eflected via area a and volume limitations before startiing the desig gn. In this sec ction, the opttimization de esign methodo ology for the ITU Tailsitter UAV is sum mmarized. As s indicated in n Figure 12, the design methodo ology approa ach consists of o three main n phases.

8 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure e 12: Genera al design me ethodology off ITU Tailsittter UAV These p phases are inputs, basic c calculationss and optimiz zation. Input part includess two sub-pa arts. The first sub--part is called d componen nt based consstant weight inputs, which includes thhe weights off the components that are e fixed for the e design proccess. The se econd sub-part is designn inputs including wingspa an, wing load ding, maximum takeoff we eight and the e horizontal tail arm. In baasic calculations part, drag coe efficient, emp pty weight an nd battery we eights are ca alculated. Op ptimization paart is used to o get most desirable desig gn within our criteria and constraints. Inputs nent Based Constant C Weight Inputs: In this part, non-variable n weight inputts including Compon electroniic and powerr related equ uipments are summarized d. Table 1: Specifications and weights of the se lected COTS S componentts C Component

Servvo Rece eiver

Specificcations

# of Piece P Us sed

T Total Weightt (gr)

10 03

215

Channel, PCM M 10 C Modulatio on

214

45

7.4V Lith hium-Polyme er

12 20

84

Ph hoenix HV85 5 Brushless s ESC C

12 21

230

7.4 V Volts to 6V V Convverter Circuitt

210

20

Brushless motor and d propellerr

20 2

557

Brrushless mo otor and EDF un nit

53 5

450

Va arious Size fro rom 12 AWG G to 24 AWG G

11

50

JR-D58611 Digital Serv vo

Rece eiver Batteryy Moto or Driver Uniit Volta age Regulato or Prop peller Unit EDF F Unit Cables

9 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-2011-052

Aksugür, İnalhan

The components contributing this category are servos, a radio receiver, a receiver battery, electronic brushless motor drivers, a voltage regulator, cables and the brushless motors of the propulsion systems. Note that, the sum of the weight of the component based weight inputs are kept constant for each design for whole optimization process. All the components are commercial off-the-shelf (COTS) equipments and seen in Table 1. Design Inputs: Design inputs consist of four variables; wingspan, wing loading, horizontal tail arm and aircraft’s maximum takeoff weight. Upper and lower bounds of these variables will be described and shown in optimization subsection. Basic Calculations Empty Weight Calculation: During the design process of a manned aircraft, empty weight fraction value is generally obtained using the historical data of the same size class aircraft. Nevertheless, it is hard to find any sufficient data to construct empty weight prediction. Consequently, in ITU Tailsitter design, a quasi exact weight prediction method has been developed to calculate the “more accurate” empty weight value in terms of aircraft size and total weight. Empty weight of the aircraft is composed of four main weight components as seen in Equation 1 and these are described in detail in the following subchapters;

(1) Fuselage Skin Weight Modeling: Before starting the calculations, composite fabrication method had been selected. As mentioned previously, during cruise flight, the folding propeller folds onto the fuselage to reduce the drag and then the EDF system is activated. Moreover, in "cruise to hover" transition phase, EDF system is shut down and the propeller system is reactivated in order to perform a power efficient landing. It is important to notice that, unfolding process of the propeller requires symmetrical fuselage shape to prevent the propeller from destructing the fuselage skin. Therefore, in the light of this prediction, a symmetrical airfoil, NACA 642-015, was selected to shape the fuselage geometry. After that, the relation between fuselage surface area and fuselage weight was derived. The measurements on the selected fuselage airfoil show that the surface area is about 0.303 m2, for 1 meter of fuselage length. Therefore, the function for the fuselage weight can be written as given in Equation 2 in terms of surface area and composite skin's surface weight density. .

∙

∙

(2)

It should be noted that, is the fuselage length in meters and ρskincomposite is the composite skin's surface weight density in N/m2. However, fuselage length in Equation 2 is not a sufficient parameter for the optimization problem since fuselage length depends on horizontal tail arm ( ) and root chord ) which are variables of the optimization problem. Hence, fuselage length is of the wing ( expressed in terms of these variables as given in Equation 3 (3) The variable lnose, which is determined as a constant value of 0.4 meters, represents the distance, which is assigned by considering folding propeller's clearance, between the leading edge of the wing root and nose of the fuselage. Therefore, the equation of fuselage weight is derived and seen in Equation 4. Moreover, to make a geometrical sense, The variables,

lHT

and

croot , are

explained in Figure 2 .

W fuselage  0.303  (0.342  lHT 

Cwingroot 4

)  composite

10 Ankara International Aerospace Conference

(4)

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 4.1 4 : Repres sentation of the varriables in geometric g c sense Wing We eight Modelin ng: While obtaining wing weight functtion; wing airrfoil, taper raatio, wing loading, wingspan are the ma ain paramete ers. Hence, th he calculatio ons have bee en started witth wetted surrface area of tthe wing. Thu us, for one meter m of chorrd length, the e wetted surrface area iss calculated as a 2.06633b m2, where b is wingspan. After tha at, the wing skin s weight ca an be expresssed as given in Equation n 4.5:

W wingskiin  2 . 06613  C meaan wing  b   composite

(5)

It should d be noted th hat, the produ uction metho od and the material m that is s going to bee used for bo oth fuselage e and wing are same. Mo oreover, the w weight of win ng spar, whic ch is made off hollow carb bon tube, must be obtained to derive the entire wing we eight. Since the deflection at the tip oof the wing is s a crucial parametter for the wing stiffness, the tip defle ction is restrricted as 5 % of the wingsspan upon having 2.5g of loading. Acco ording to the previous exxperiences, th his loading condition cann be modeled d by applying g the force, which w is equa al to half weig ght of the airrcraft to the wing w tip whenn the fuselag ge is stationary. As known n from the mechanics of m materials lec ctures, 2.5g load case forr the half win ng can be simulate ed as the tip deflection d for cantilever b beam that is loaded with concentratedd force from the free edge. Th he deflection n is then exprressed as givven in Equattion 6:

 Wo   b     3 F l 2 2   b  0.05      12  E  I 12  E  I

3

(6)

For the e equation give en above, E is the modullus of elasticity of spar material and I represents moment m of inertia a for circular wing spar se ection. The m moment of inertia can the en be calculaated in order to find the innerr diameter off the spar, which is a holllow carbon tu ube. Note tha at, the outer diameter of the spar is restriccted as given n in Equation 7:

 outherspaar

diameter

t  c tip     0 . 003 c

(7)

Where ctip represen nts the wing tip t chord, t c representts thickness ratio of wingg airfoil and th he numerica al value reprresents the th hickness of sskin composite material. After A that, thhe moment off inertia can be ccalculated ass given in Equ uation 8. 11 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-2011-052

Aksugür, İnalhan

3

I

 4



   4

outherspar diameter

 4

innerspard iameter

 Wo   b     2  2   12  E  (b  0.05)

(8)

More, the inner diameter of the spar can be expressed as seen in Equation 9.

   Wo   b      4 2  2 4     outherspardiameter      12  E  (b  0.05)     3

innerspardiameter

1 4

(9)

Finally, the weight of the carbon spar can be calculated in Equation 10 and given as follows:

Wspar 

 4





 outherspardiameter   innerspardiameter   b  spar 2

For the equation given above,

2

 spar

(10)

is the density of the carbon material per unit volume. Summing

skin and spar weights, weight function for the wing is derived as given in Equation 11 given below.

Wwing 

 4







 outherspar diameter   innerspard iameter   b  spar  2.06613  C meanwing  b   composite 2

2

 (11)

Stabilizer Weight Modeling: The total weight of the stabilizers can be expressed as the summation of horizontal and vertical tail weights. Manufacturing method applied for stabilizers is strictly different from the method applied for fuselage and wing. In this manufacturing technique, stabilizers are going to be made of high-density foam covered with carbon fiber. Therefore, wetted area and inner volume of the stabilizers are used to derive the weight function, given in Equation 12:

W HT  S HTwetted   skincompos ite  v HT   foam

(12)

In Equation 4.11, S HTwetted represents the wetted area of horizontal stabilizer in m2,

 skincompos ite

2

represents the density of stabilizers’ composite covering material in N/m , vHT represents the inner volume of horizontal stabilizer in m3, and

 foam represents the density of the foam that fills the

3

horizontal tail in N/m . As first step, horizontal stabilizer area can be written in terms of wing area, mean aerodynamic chord, tail arm and volume coefficient as seen in Equation 13, given below:

S HT 

V HT  c wing  S wing

(13)

l HT

In the equation given above, V HT is horizontal tail volume coefficient, 3wing mean aerodynamic chord of the wing, S wing is wing area and ‘HT is horizontal tail arm which is equal to the distance between quarter mean aerodynamic chords of wing and horizontal stabilizer respectively. In ITU Tailsitter airplane, inverted V-tail configuration is used with 25 degree of anhedral on both left and right horizontal stabilizer parts. Therefore, related to the anhedral angle, the actual area of horizontal stabilizer can be calculated by using Equation 14 and the figure representation is seen in Figure 14:

12 Ankara International Aerospace Conference

AIAC-201 11-052

Aksugü ür, İnalhan

Fiigure 13: Re presentation n of the Varia ables in Geom metric Sense e

S HTactual 

V HT  c wingg  S wing



l HTT

1 cos 25 

(14)

After the e actual horizzontal tail are ea is determi ned, wetted area can be found with rrespect to the e tail airfoil. Ass indicated before, b NACA A0014 is sele ected for horrizontal and vertical v stabi lizers. There efore, after insp pecting the geometric g pro operties of th he selected airfoil, a it is fou und that the wetted area is 2.133 m2 per u unit planform area, which can easily b be found as given g in Equa ation 15:

S HTwetted  2.133 

V HHT  c wing  S wing w l HT



1 cos25 

(15)

After dettermination of o wetted are ea for horizon ntal stabilizerr, the weight of the compposite skin co overing material can be found as given in n Equation 16 6:

WHTskin  2.133 

VHTT  c wing  S winng l HT



1   skincomposite cos25

(16)

Next, the e weight funcction for “foam core” musst be found in n order to derive weight ffunction for horizontal h tail. Acco ording to geo ometry, the airfoil a side are ea coefficien nt for one me eter of chord is determine ed as 0.000951. Hence, the volume of horizontal ta ail is expressed as given in Equation 117:

 HT 

V HT  S wing  C mean wing l HT



1  (0.0000951) Coss ( 25)

(17)

In order to determine e the weight of foam core e, the volume e is multiplied d by the foam m density as given in Equation n 18:

W foamcore 

V HT  S wing  C mean wing l HTT



1  (0.0000951 )   foam Co os ( 25)

(18)

As a result, the total weight of the e horizontal sstabilizer is derived d by su umming bothh foam core and a skin quation 19: composiite weights as seen in Eq

VHT  C meanwing  S wing   1   skiincomposite  WHT  2.133   l HT cos25   VHT  S wing  C mmeanwing  1   (0.00095 1)   foam   l HT Cos(25 2 )   13 Ankara A Interna ational Aerospa ace Conferenc ce

(19)

AIAC-2011-052

Aksugür, İnalhan

Similarly, weight function for vertical tail can be derived by using the same steps and then obtained as Equation 20:

 VVT  S wing  bwing WVT   lVT   VVT  S wing  bwing   lVT 

  VHT  S wing  C meanwing    tan(25)  (2.113)   skincomposite  l HT  

(20)

  VHT  S wing  C meanwing    tan(25)  (0.000951)   foam  l HT  

Finally, entire weight function for stabilizers is derived by summing horizontal tail and vertical tail weight functions. By plugging Equation 19 and 20 in Equation 21, weight function for stabilizers is obtained as in Equation 22:

Wstabilizer  WHT  WVT

(21)

 V HT  S wing  C mean wing 1 W stabilizer     ( 2 . 113 )   skincompos ( 20 ) l Cos HT   V HT  S wing  C mean wing 1    ( 0 . 000951 )   l HT Cos ( 20 ) 

foam

ite

  

  

  V VT  S wing  C mean wing    lVT  

 V HT  S wing  C mean wing   tan( 20 )  ( 2 . 113 )   skincompos  l HT 

  V VT  S wing  C mean wing    lVT  

 V HT  S wing  C mean wing   tan( 20 )  ( 0 . 000951 )    l HT 

foam

ite

  

  

(22)

Structural Weight Modeling: Structural weight consists of the weight of spar box, bulkheads, longerons and glue. Spar box weight is assumed a constant, while glue weight can be approximated as 10 % of bulkheads, longerons and spar box weights. Moreover, for building fuselage frame, it is planned to locate four longerons along the fuselage and one bulkhead per each 0.15 meters of fuselage length. The total structural weight is then expressed as given in Equation 23:

 l fuselage  Wstructural  1.1    Wbulkhead  4  l fuselage  Wlongeron  Wcarrythrou gh   0.15  Where

Wbulkhead is the assigned constant average weight of a bulkhead, Wlongeron

(23)

is the weight of a

carbon longeron per unit length and Wcarrythrou gh is the assigned constant spar carrythrough weight. After determination of each component composing the equation, previously given as Equation 1, the empty weight can be written as summation of each component. Battery Weight Modeling: In this section, we first discuss about propeller propulsion system and then going into the EDF propulsion system.The propeller propulsion system is responsible for VTOL operations including hovering, low speed vertical climb and descent. To simplify the calculations, the total VTOL operation duration is set as 3 minutes with only hover mode. This is because, even though the propeller consumes more energy than hovering mode during vertical climb; the energy consumption level reduced below the level during hovering while vertical descent maneuver. For hover mode, the thrust produced by the propeller is equal to the summation of weight of the aircraft and the drag force created by the propeller’s induced velocity, which is calculated by using helicopter theory [14]. The battery weight determination logic for the propeller system is shown in Figure 14 systematically. The battery weight of the propeller propulsion system is a function of the VTOL operation duration, propeller’s characteristics, airplane’s geometry creating drag force and the takeoff weight of the aircraft. Note that, in order to calculate the “worst case” and simplify the 14 Ankara International Aerospace Conference

AIAC-201 11-052

Aksugü ür, İnalhan

calculatio ons, the para asite drag co oefficient for w whole aircraft is used, instead of the parasite drag coefficient that is con nstructed by the “induced d velocity wetted” part of the aircraft.

Figure e 14: Battery Weight Calcculation Step ps for Propeller Propulsionn System The sele ected EDF unit u is mainly dedicated to o cruise flight. For this rea ason, thrust generated by the EDF unitt is equal to the t drag forc ce on the airccraft. Therefo ore, the batte ery weight caalculation methodo ology is consstructed on cruise flight a nd seen in Figure F 15

15 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure e 15: Battery Weight Calcculation Step ps for EDF Prropulsion Sysstem Drag Co oefficient Mo odeling: The drag force of an aircra aft can be written w as thee summation n of both parasite (zero lift) and a induced drags. In ITU Tails sitter design n, induced drag coefficient is equalize ed to the parrasite drag coefficient c so o as to fly y at “minim mum thrust” level. Acc cording to drag co oefficient asssumption made, parasitte drag coe efficient is the t only vaariable that must be formulate ed for each design durin ng optimizatio on process. To do that, the parasitte drag coeff fficient of each o of the com mponents is s calculated d using the empirical “componeent buildup” method describe ed by Raym mer [15]. Acc cording to Ra aymer component buildu up method iis used to calculate drag coefficient for sub-sonic flight and ccounts on bo oth flat platte skin fricttion and co omponent form factor (pressurre drag due e to the visccous separation). Relate ed to compoonent buildup p method, the drag coefficient of o each comp ponent can be described in Equation n 24;

C Dp 

C

fc

 FFc  Qc  S wet v S reff

 C Dmisc  C D LP

(24)

In Equattion 24, Cf de enotes flat-plate skin fricttion coefficient, FF repres sents form faactors, Q indicates interference factor, Swet is wetted d area of the selected com mponent and d Sref is the reeference win ng area. Where th he subscript "c" indicates s that those vvalues are diifferent for ea ach componeent. In the fo ollowing descriptiions, the equ uations used for each com mponent of the aircraft arre shown. Drag Force Modeling g for Aerodyn namic Surfacces: The parasite drag co oefficient for the aerodynamic surfacess, CDaero , com mposed of three discrete parts; wing, horizontal sttabilizer and vertical stab bilizer, and figurred out in Eq quation 25:

16 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-2011-052

C Daero 

1 S ref

Aksugür, İnalhan

 C f w  FFw  Qw  S wetw  C f HT  FFHT  QHT  S wetHT      C f  FFVT  QVT  S wet  VT VT  

(25)

Flat-plate skin friction varies depending on the type of flow, laminar or turbulent, over the surfaces. For wing and stabilizers, turbulent flow assumption has been made and flat-plate skin friction coefficient is written as Equation 26

Cf 

0.455

log 10 Re 2.58  1  0.144  M 2 0.65

(26)

Where "Re" represents Reynolds number, which is taken 500.000 for wings and 300.000 for stabilizers, "M" indicates mach number, which is selected 0.1 as constant for each design during optimization. To continue, form factor (FF) equation for wing and tails is written in Equation 27

  FF  1    

 t 0 .6    4  c   100   t    1.34  M 0.18  cos  0.28   m x c      c  max 



(27)



In Equation 24, the term "(x/c)max" is the chord wise location of the airfoil maximum thickness point, which is 0.3 for the selected airfoils of wing and stabilizers;

m refers to the sweep of the maximum

thickness line, which are 0 and 23 degrees for wing and stabilizers used in ITU Tailsitter, respectively. The interference factor, "Q", is chosen as 1.1 for both wing and stabilizers. Drag Force Modeling for Fuselage: To calculate the drag coefficient for fuselage, the steps must be followed are similar to the steps followed in wing and stabilizer calculations. However, there is a change on form factor estimation. The form factor for fuselage or smooth canopy can be calculated using Equation 28

 60 f   FF  1  3  400   f

(28)

Where;

f 

l l  d 4    Amax  

(29)

In Equation 29, l is the length and Amax is the maximum frontal area of the fuselage. Optimization Defining the Optimization Problem: During the preliminary design, it has been observed that an aircraft, which carries much more payload and flies longer, would be more challenging for the design. Therefore, optimization process is focused on maximizing payload weight and cruise duration. Since it is not possible to optimize all the design parameters, the parameters with crucial effect are selected as primal variables of the optimization problem. Maximum Takeoff Weight (W0 ), Wing Loading (W/S ), 17 Ankara International Aerospace Conference

AIAC-2011-052

Aksugür, İnalhan

Wing Span (b), Horizontal Tail Arm (lHT ) and Fan Battery Weight (Wfanbattery) are determined as primal variables. The boundary constraints of those variables are going to be discussed in the 'Formulation Section'. In addition, the other design constraints for the optimization problem are going to be discussed in the same section as follows. Consequently, the optimization problem can be classified as multiobjective, multidisciplinary, constrained and continuous. It is a multiobjective problem because the objective is having a maximum payload capacity with maximum cruise duration. It is a multidisciplinary optimization problem because it consists of aerodynamics, propulsion, structure and design. It is a constrained optimization since it includes both boundary and design constraints which are going to be discussed later. It is a continuous problem since the variables are free to change within the side constraints. As it was stated before, the objectives of the optimization for the ITU Tailsitter UAV can be listed as follows; 

Maximization of payload weight

 Maximization of cruise duration  Assuming the Maximum Takeoff Weight (MTOW) as a constant; the objectives, given above, are in contradiction with each other. Since maximization demand on payload weight gives the minimum cruise duration while maximization of cruise duration, which also means maximization of the battery weight, minimizes the payload weight. In the light of such relations between the given objectives, the aim is to find the most suitable configuration by the optimization variables satisfying the design constraints. Requirements, Variables and Constraints of the Optimization Problem: Before beginning the design process, design requirements were determined as seen in Table 2. After that, MTOW, wing loading, wingspan, horizontal tail arm and fan battery weight, which are listed in Table 3, are determined to be primal variables. The constraints of the primal variables are then considered as given in Table due to the listed reasons. Table 2: Operational Requirements for the Optimization Problem Variable Values Minimum Range Minimum Operation Duration Opertation Altitude Maximum Airspeed Maximum Operation Condition Wind Maximum VTOL Operation Area Minimum Payload Weight

20 km 30 minutes 1 km 50 m/s 15 m/s 2m x 2m 0.8 kg

Table 3: Primal Variables and Side Constraints for the Optimization Problem Primal Lower Upper Bound Explanation Variable Bound W0 30N 100N W0/S 100 N/m2 200 N/m2 The limits are based on the similar type UAVs b 1m 2m -

lHT

0.6m

1.5m

Structural limits during vertical landing phase. Wfanbattery 3N 30N The boundaries are set by the previous experiences In addition to the side constraints, some operational and aerodynamic restrictions are also applied to the optimization problem to obtain the desired design. At the beginning, stall and cruise speeds are assumed crucial parameters for operation capability. Stall speed is limited up to 20 m/s where cruise speed is limited up to 50 m/s. Moreover, in order to prevent the aircraft stall due to the gust effects during cruise flight or landing approach, the cruise speed must be at least 5 m/s more than the stall speed. Second, although cruise duration and payload weight are being maximized as a result of the optimization algorithm, there are lower limits which come from the design requirements. In Table 2, it 18 Ankara International Aerospace Conference

AIAC-2011-052

Aksugür, İnalhan

was stated that payload weight must be equal to or more than 0.8 kg, where the cruise duration must be at least 30 minutes. Next, in order to have an efficient, easily controllable and non-stubby design, the fuselage length is determined to be less than the wingspan. Moreover, aspect ratio's minimum value is set as 4. Consequently, all design constraints can be classi¯ed into three groups as operational capability, design requirements and geometrical limits. The determined design constraints of the optimization problem are then summarized in Table 4. Table 4: Design Constraints for the Optimization Problem Variables

Values

/ / /

Operation Capability Design Requirements Geometrical Limits

Mathematical Formulation of the Problem and Objective Function: In ITU Tailsitter design process, the main aim is to design an airplane which is configurable with different weight of payloads. To see the performance of the aircraft with different type (weight) of payloads, two objective functions have been defined; payload weight and cruise time. Therefore, in this multi-objective optimization problem, maximizing both of our objective functions is the main purpose. Before explaining the objective functions, some descriptions should be made on the way that is following. Aircraft's maximum takeoff weight(W0) can be expressed in Equation 30: (30) Where Winputs is described in "Component Based ConstantWeight Inputs" section and can be rewritten in Equation 4.31. (31) In Equation 31; Wservo; Wrx; Wrxbatt; WESC; Wreg; Wcable; Wpropunit; WEDFunit denotes the weights of servos, receiver, electronic speed controller of the brushless motors, voltage regulator, cables, propeller propulsion unit and EDF unit respectively. In the light of the given formulations, one of the objective functions can be expressed in a compact form in Equation 32. Note that,Winputs;Wpropbattery; Wfanbattery and Wempty were described in the previous chapters as functions or invariant values. (32) The second objective function is the cruise time for EDF propulsion system. As seen in Figure 14, the general process to determine the battery weight for the EDF system was summarized, but not formulated. Therefore, cruise time formulation, which is seen in Equation 33, is obtained with the help of the battery weight equation. ∙ ∙

∙

∙

∙

∙

∙

(33) ∙ ∙

∙

∙

∙

∙ ∙

∙ ∙

∙

∙ ∙ ∙

Optimization Methodology After the optimization problem is discussed in detail, optimization process becomes ready to be carried through. Instead of developing a new code, the commercially available software is preferred for the multidisciplinary design optimization of ITU Tailsitter UAV and so Optimization process is decided 19 Ankara International Aerospace Conference

AIAC-201 11-052

Aksugü ür, İnalhan

to be performed by using u Esteco ModeFRON NTIER 3.2[16 6] and Micros softExcel com mmercial sofftware, which arre run simulta aneously. Th herefore, Estteco ModeFR RONTIER so oftware is seleected as the e main optimiza ation driver to ool and "Micrrosoft Excel" is selected for f the analysis tool. In orrder to solve e optimiza ation problem m, mathematical formulatiions, which are a already given g in the pprevious chapters, are writte en into the Microsoft M Exc cel and a calcculation Exce el Sheet is obtained. Afteer that, Estec co ModeFR RONTIER is connected c to o Excel Shee et in order to perform the calculations . Next, the optimiza ation flowchart is started to t be built byy adding five input nodes, which are aalso the prev viously considerred optimizattion variables s (W0, Wing Loading, Fa an Battery Weight, Horizoontal Tail Arm m and Wing Sp pan). The side constraints s of optimiza ation variable es are then applied to theese input nod des and seven ou utput nodes are added to o the flowcha art. Aspect Ratio, R Cruise Speed, Stall Speed, Spe eed Differencce and Fuselage Length output node s are used in n order to ap pply design co constraints which are previoussly given in Table T 3. More e, Payload W Weight and Cruise Time output o nodes are the resu ults of objective e functions. In addition, minimum m misssion requirem ments of the optimizationn problem, which w are given in Table 3, are also applied d to these ou utput nodes. The last outp put node, Em mpty Weight, is connecte ed in order to o monitor the e empty weig ght of the airc craft. After alll of the inputt and output nodes are adde ed into the ModeFRONTI M IER software e, Scheduler Node is add ded in order tto determine Design of Experriments (DOE E) properties s, and optimizzation algoritthm. Full Fac ctorial is prefferred for DO OE and Multi Ob bjective Gene etic Algorithm m (MOGA) is preferred fo or the optimiz zation algorithhm. Finally, the t flowcharrt showing the optimizatio on methodolo ogy, which is s given in Fig gure 16, is buuilt up for the e optimiza ation problem m. The results s for the varia ables strongly depend on n the optimizzation methodology. Thus, the e methodolo ogy, which is being used w with the optim mization driv ver, is very im mportant to obtain o the optimum m design. More, the Desig gn of Experim ments (DOE)) node and optimization aalgorithm are e considerred carefully before starting the optim mization proce ess.

Figure 16: F Flowchart for the optimization problem m

20 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

OPTIMIZ ZATION PRO OCESS After the e optimization n problem is considered iin detail, and d the optimization methoddology is discussed with grea at care, optim mization proc cess become e ready to be e carried out. Previously ddiscussed ob bjective functionss, side consttraints and de esign variablles are applie ed to the Estteco modeFR RONTIER an nd Microsofft Excel softw wares as previously conssidered. Notin ng that, Full Factorial DO OE node and MOGAII algorith hm are used d for optimiza ation processs. When the flowchart, f wh hich is previoously given in n Figure4.1, is ra an the total number of 13550 designs is obtained. The total de esign points aare consistin ng of feasible designs, unffeasible designs and erro ors. Feasible designs are e the design ppoints with all a the constrain nts are satisffied, while un nfeasible dessigns consistt of the desig gn points withh at least one e unsatisfied constrain nt. The errors s are the dessigns with ind definite desig gn points. Thee distribution n of feasible, unfeasible design d points s and errors is shown witthin the chartt given as Figgure 17. It ca an be stated th hat, the optim mization meth hodology is ccorrectly sele ected according to the disstribution of obtained o design p points. The design summary is given iin Table 5 be elow and the en all of the ddesign points s obtained d for the optim mization prob blem are sho own in Figure e 18 as follow ws:

F Figure 17 : Design D summ mary chart off the optimiza ation problem m

Table 5 : Design sum mmary of the optimization n problem Design ns Numb ber Total Design D 1355 50 Feasib ble Design 1121 11 Unfeas sible Design 2337 2 Errors

Figure 18 8: All design p e optimization problem points for the

21 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

As previo ously stated, the design points with a at least one broken b consttraint are the unfeasible design d points. F For the optim mization probllem 3 of the constraints are a strictly sa atisfied, whilee the other 4 of the constrain nts are not satisfied at 23 337 design p points. The nu umber of bro oken constra ints is listed in Table 6. For so ome of the de esign points more than o one design co onstraint is broken. b Thereefore, the tottal number of broken co onstraints is more m than th he number off unfeasible design d pointss. Aspect ratio constrain nt is the mosst broken con nstraint while e stall speed,, stall speed--cruise speedd relation and cruise speed co onstraints are never brok ken by the de esign points. The distribution of the brroken constrraints is given in Figure 19 ass pie chart. Table 6 : Broken con nstraints at unfeasible de sign points Constraintss Wingspan--Fuselage Le ength b  l fuselage Relation Stall Speed d V  30m / s

Broken C Constraints 981 0

stall

Stall Speed d-Cruise Spe eed Relation eed Cruise Spe

Vcruise Vstall  3m / s

0

Vcruise  50m / s

0

Aspect Ratio

AR  4

me Cruise Tim

tcruise  30min m

380

Payload Weight W

Wpayload  8N

731

1473

Fig gure 19 : Chart of broken n constraints at unfeasible design poi nts For the o optimization process, fea asible design points mustt be evaluate ed in detail. T The feasible design d points ob btained for th he multiobjec ctive-multidissciplinary des sign optimiza ation of unmaanned tailsittter aircraft iss shown in Figure F 20.

22 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

F Figure 20 : Feasible F desi gn points forr the optimiza ation problem m As seen from the figu ure given above, there arre 11211 fully distributed feasible dessign points obtained as a resu ult of optimizzation. Due to o having a m multiobjective e optimization n problem, thhe design point both having m maximized pa ayload weigh ht and cruise e time must be b determined. Thereforee, Pareto Cha art is used in o order to dete ermine the se et of optimal solutions. In Figure 21, Pareto P Chartt which consiists of 3021 opttimum design points for objective o fun nctions is give en.

Figure 21 : Pareto chart for opttimum design n points After all the pareto design points are examine ed in detail, following f figu ures are obtaained with the e help of modeFR RONTIER. Th he figures sh how the frequ uencies of the design varriables and oobjective func ctions. It can be sstated accord ding to the fig gures that, frrequency distributions of maximum taakeoff weightt and horizonta al tail arm arre normal dis stributions. A Additionally, th he distributio ons of objectiive functions s, payload weight and cruise c time, are a also norm mal distributions. Howeve er, the distribbutions of win ng loading, wing span and a fan battery weight are e different fro om the normal distributionn. According g to Figure 23, nearly 90 % of the parreto designs have wing lo oading of mo ore than 190 N/m2. Next, Accordin ng to Figure 5.8, 5 nearly all of the pare eto designs have h wing sp pan of 2 m whhich is also the t upper limit of w wing span constraint. The erefore, it can n be stated th hat, wing spa an constraintt limits the pa areto designs. According to Figure 26, nearly all off the pareto designs d have e battery weigght of 15 N which w is also upp per limit of ba attery weight constraint. T Therefore, similar to wing g span constrraint, it can be b stated that batte ery weight co onstraint limiits the pareto o designs. 23 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 22: Frequency F histogram forr maximum ta akeoff weight of pareto deesign points

Figurre 23: Freque ency histogra am for wing loading of pa areto design points

Figu ure 24: Frequ uency histog gram for wing gspan of pare eto design pooints

24 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 25: 2 Frequenc cy histogram for horizonta al tail arm of pareto desiggn points

Figure 26: 2 Frequenc cy histogram for fan batte ery weight of pareto desiggn points

Figure e 27: Frequecy histogram m for payload d weight of pa areto design points

25 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 28: Frequency histogrram for cruise time of parreto design ppoints Considering the Pare eto Chart giv ven in Figure 5.5; the des sign points with better (maaximized) va alues of both obje ective functio ons, is pointe ed as shown in Figure 29 9. Due to the severe variaation of desig gn points over the selected design set, the optimum de esign point is considered within this deesign set. By y the way the desig gn points havving optimum m points onlyy for one obje ective functio on is eliminatted. The optiimum design p points for the selected design set, sho own in Figure e 29, are liste ed in Table 7 given as follows:

Figure 29: 2 Selected design set from f the pare eto chart As previo ously stated, the severe variation of tthe pareto ch hart should consist c of opttimum design n point. Thereforre the design n point, with ID: I 2478 sho own in Figure e 30, at the corner c of the chart is deciided to be the re eal design po oint for the multiobjective m e-multidisciplinary design optimizationn of unmanne ed tailsitter aircraft. The comparison n of the initiall design and the optimum m design is suummarized in Table 8. As a rresult of the optimization, initial MTOW o W is increase ed from 56.2 N to 178 N, which is nea arly 3 times the e initial value e. Moreover, maximum p ayload capacity is increa ased from 133.5 N to 92.39 9 N, which is nearly 7 times of the initial value and d therefore it is a good op ptimized valuue. Additiona ally, fan battery w e help of this weight is incrreased to 15 N and by the s increment cruise c time iss changed fro om 40 min to 74 4.37 min whiich is also an nother good optimization result.

26 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Table e 7 : Design points of dominant desig gn set Design ID

W0

242 594

W /S

b

lHT

Wfanbatteery

Wpayload ad

tcruise

(m)

(N)

(N)

(min)

2 2

1.50 1.48

15.0 15.0

94.022 93.988

73.65 73.66

200

2

1.50

15.0

88.522

76.50

180

200

2

1.50

14.4

94.622

70.71

1768

180

200

2

1.50

14.8

94.222

72.67

2159

180

200

2

1.41

15.0

93.822

73.67

2478

178

200

2

1.38

15.0

92.399

74.37

3349

177

200

2

1.50

15.0

91.966

74.70

3428

173

200

2

1.44

15.0

89.133

76.15

3613

180

200

2

1.45

14.4

94.522

70.72

4221

180

200

2

1.23

15.0

93.11

73.68

4698

175

199

2

1.25

15.0

89.722

75.44

4898

180

200

2

1.49

15.0

94.000

73.66

6219

175

200

2

1.50

15.0

90.599

75.41

6746

180

200

2

1.40

15.0

93.800

73.68

7089

180

200

2

1.46

15.0

93.944

73.66

7159

180

200

2

1.46

14.4

94.544

70.72

7222

173

200

2

1.50

15.0

89.21

76.13

7708

180

199

2

1.50

14.6

94.244

71.68

2 (N/m ( )

(m m)

180 180

200 200

1642

172

1653

(N)

Figure 30: Selected de esign point as optimum design point

27 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Tab ble 8: Compa arison of initiial design an nd selected design d (ID:24478) Parame eters Initial Des sign Seelected Deesign W0 (N) 56.2 178 Wing Loading (N/m2) 200 200 Wingsp 1.3 2 pan (m) Horizon ntal Tail Arm(m) 0.78 1.38 15 FanBatttery Weight((N) 4.76 Payload d Weight (N)) Cruise Time (min) Aspect Ratio ge Length (m m) Fuselag Cruise Speed (m/s)) Stall Sp peed (m/s) Empty Weight W (N) 2 Wing Area A (m )

13.5 40 6 1.2 22.1 16.81 15.73 0.28

92.39 74.37 4.49 1.86 30.87 16.81 36.73 0.89

ROTOTYPIN PR NG Because e of their high h strength to weight ratio and good fa atigue charac cteristics, com mposite materials are wide ely being use ed in aerospa ace industry. In ITU tailsittter airplane, carbon-kevllar hybrid fiber cloth is used. The density of the cloth used u is 68 grr/m2. Moreov ver, to ensure the structuural integrity and a having h high strength to weight ratio compositte shell, sand dwich shell system is useed. To give a thickness to the com mposite shell, aramid hone eycomb material having 1.5mm of thiickness and 44 gr/m2 of surfacce density, iss used betwe een the uppe r and lower layers l of the composite sshell. As men ntioned in design n section and d during design calculatio ons, aircraft weight w mode eling is consttructed on the prototyping method. In practice, there t are num merous composite produ uction methodds; pultrusion, resin transfer molding, vaccuum assiste ed resin transsfer molding, filament win nding and haand lay-up. These T methodss have wide variety v of applications fro om toys to full scale civil aviation a or m military airplan nes. his method, fibres are be Pultrusio on is consiste ent molding process. p In th eing combineed with therm mosetting . With the he resin; wh hich cures un nder proper temperature. t elp of this me ethod, profile or plate sha aped composiite materials can be easilly made. The e principle off pultrusion method m is illuustrated in Figure 31 below;

Fig gure 31 : Illusstration of pu ultrusion method As anoth her method; resin transfe er molding (R RTM) is a type of close mold process. This is beca ause the reinforce ement materiial (fibers) is placed betw ween two matching mold surfaces; onne is male an nd the other is ffemale. Afterr placing the material, the e mold couplle is closed and a thermoseetting resin is s injected via the injecttion port into o the mold. T The injection process continues until thhe resin com mes up from the vent port. The RTM metthod is illustrrated in Figurre 32.

28 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure F 32 : Illlustration of RTM method um assisted resin r transferr method (VA ARTM) is wid dely used in today’s moddern and com mplex In vacuu structure es like turbine e blades, boats, cars and d many otherr vehicles an nd constructioons. In VART TM method, in general, female f mold is used with vacuuming equipment, which w works as male mold as in RTM me ethod. This method m is verry useful for ccomplex sha apes and thic ck compositee parts. Filamentt winding me ethod is gene erally used in n the fabricattion of cylindirical compossite parts; lik ke circular o or rectangula ar cross secttional beamss, composite tanks and pipes. In this m method, fiber roving is wet wiith the resin and than wra apped onto a rotating ma andrel with th he specified aangles. Afterr completiing the wrapp ping process s, the resin iss cured and the t part is removed from the mandrell. Compossite hand lay--up method, which is deccided to be used in ITU Tailsitter aircrraft, is a relattively cheap prroduction tecchnique. For additional acccuracy and strength, va acuum techniique is comb bined with the lay-up metho od. According to the fab brication method, producttion process has been created as bellow: A A- Draw a 3D 3 CAD mod del B B- Prepare drawing for CNC machin ne C C- CNC ma achining D D- Apply mo old releasing g agent on to o the mold an nd let it dry fo or 20 minutess E E- Place ca arbon-kevlar fiber sheet o onto the mold d F F- Wet the fiber with epoxy resin G G- Put plasttic film and blanked b onto the wet fiber H H- Cover th he mold and apply a vacuum m In step A A, 3D cad mo odeling of the e aircraft is o obtained by using u CATIA software. Beecause, CAT TIA is a valuable e 3D modeling and mecha anical analyzze and simulation software, which is ppreferred by many area from kittchen access ospace. Afte product a sories to aero er modeling the designedd aircraft in co omputer environm ment, which is i seen in Fig gure 33, the drawing is set up for the CNC machinne in step B..

29 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 33 3 : 3D CAD illustration of ITU Tailsitter aircraft Because e, CNC mach hine accepts a generalize ed form of da ata, called G code. Step C is CNC ma achining stage.

ogress Figurre 34 : Stabillizer mold miilling is in pro m reads the codes and mills the e material, which is placeed into the wo orking In this sttage, CNC machine area. Ma achining proccess is seen in Figure 34 4;

30 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figurre 35 : Mach hined wing an nd fuselage molds After gettting the mold ds being machined by ussing CNC ma achine, it is time to apply wet lay-up method. m In this m method, epoxyy or polyeste er resins are impregnated d by hand intto fibers, whiich are in the e woven form. Ro ollers or brushes usually accomplish a tthis. After tha at, wetted lam minates are lleft to cure under u high tem mperature and d low-pressu ure condition . To give dettail, some liquid mold releeasing agentt is applied o on the mold with w the help p of sponge i n step A. Bru ush or roller shouldn’t bee used for we etting, because e surface stru ucture of thatt materials arre not suitab ble for such application. a

Figu ure 36 : Illusstration of we et lay-up method After 20 minutes of drying d time, in n step C, carrbon-kevlar fibers f are pla aced onto thee mold and wet w with epoxy re esin by rollerss or brushes. Note that, tthe curing tim me of the epo oxy resin useed in the production is about 24 hours. 31 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 37 7 : Molds to be b polished However, the resin gets g thicker after a 10 minu utes from stirrring the compunds. In steep D, plastic film having h holes is applied onto the wet w fibers to take out exc cess epoxy frrom the bottoom layer.

Fig gure 38 : Cuttting fibers fo or lay-up metthod Meanwh hile, some bla ankets are la aid out onto tthe plastic film m to widen th he pressure created by the vacuum pump. After that, in step E, mold stru ucture is plac ced into the plastic p cover and vacuum m is applied. A basic wet lay-up method is illustratted in Figure e 36. Finished d and polisheed wing mold ds is seen in F Figure 37.

32 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figurre 39 : Wing mold and co omposite wing shell Preperattion for comp posite lay-up method is sseen in Figure 38. More, the mold annd products are a seen in Figure e 39, 40 and 41. After the e final assem mbly, the ITU Tailsitter airplane can bee seen in Fig gure 42;

Figure 40 : W Wing structu ure and molds

33 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure F 41: Fu uselage struc cture and mo old

Figure 4 2 : ITU Tails sitter UAV

34 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

FL LIGHT TEST TS During th he design process, empirical method s are widely used to dete ermine the aeerodynamic characte eristics of the e airplane. So o as to valida ate the desig gn and see if it meets ourr requirements, some flight tessts are condu ucted. Instead of using th he autopilot system, s manu ual control syystem with gyro g assists iss preferred in n the first flig ght test. Therrefore, by sensing the pitch, roll and yyaw rates, th he gyros help giving correction ns during hover flight. Th he gyros to be used in the e flight test w were the gyro os, which are commonly used in radio conttrolled hobbyy helicopters to hold their headings coonstant. More eover, the radio o system, wh hich is used to t control all of the contro ol surfaces, as a well as thee motor conttroller, is Futaba b brand, 10CAP model com mmercial hob bby radio sys stem, operating 2.4 GHzz frequency. Electronic speed con ntrol unit is used to drive the brushles ss motor to help rotating tthe propellerr located at the fro ont of the airp plane.

ure 43 : Rece eiver and gyrro wiring diag gram Figu Servo m motor take sig gnals and pow wer from the e receiver and actuates th he control suurfaces. Therrefore, to control th e high torque he ailerons, elevators and rudder, five e servo moto ors are used.. Wiring diag gram of the comp ponents is se een in Figure e 43. After m making the ele ectrical conncections andd ensuring th he structura al stifness of the airframe e, the aircraftt is taken to the t flight field d to see the hhoverin perfo ormance. Hovering g is one of th he flight phas ses in which the aircraft holds h its vertical position and hangs on o its propellerr as seen in Figure 44;

e 44: ITU Taiilsitter is read dy for the flig ght test Figure After settting up the airplane, a the flight sequen nce is started d by holding the wing tipss by two stud dents, as seen in F Figure 45.

35 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 45: B Beginning of the flight test However, as seein in n Figure 46, the t person lo ocated at the e left side of the t airplane,, was late to release the airpla ane.

Figure e 46: ITU Taiilsitter is read dy for the flig ght test As a result, instead of o climbing ve ertically, the aircraft couldn’t maintain n its position and started d low speed hiigh angle of attack flight, which is nott suitable for its nature.

36 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-201 11-052

Aksugü ür, İnalhan

Figure 48 : IITU Tailsitterr is taking offf Thereforre, after 10 seconds from m taking off, th he aircraft crrashed becau use of the coontrollability problems p due to th he disturbancce given in th he take off ph hase. In addition to the disturbance dduring take-o off, lack of authority on the co ontrol surface es is affected d by the flyin ng weight of the t aircraft. B Because, the e flying weight o of the aircraft is determine ed as 10kg. H However, it was w only 7 kg g during the flight tests. Weight W of the airpla ane affects the propeller’’s RPM. The rfore, the low wer RPM me eans lower aiirspeed behin nd the propellerr. In hover mode, m the forc ces generate ed by the con ntrol surfaces s is highly reelated to the airspeed a coming ffrom the prop peller. As a result, r the low wer speed co oming from the propeller makes the control c surfacess generate lower forces to o control the aircraft’s behavior. CON NCLUSION A AND RECOM MMENDATIO ONS In this w work, the desiign optimizattion study of a tailsitter aiircraft with a revolutionary ry hybrid/dua al propulsio on system ha as been desc cribed. The o obtained results in this th hesis are bassed on analytical calculatio ons on the propeller p prop pulsion syste em, experime ental data on n the EDF proopulsion system and the desig gn inputs wh hich are in clo ose relation w with both the e design cons straints and ddesign criterria. Initial sysstem perform mance analys sis with cand didate propullsion units indicate that uup to 27.5 m//s cruise speed an nd maximum m 2 hours of flight f endura nce, includin ng 3 minutes of vertical taake-off and la anding duration, can be achieved while carrying c a 1 kg payload and a 90 km off range – a m marvelous ance in comp parison to the e same classs rotary-wing g and OAV alternatives. performa Prototyp ping of ITU Ta ailsitter is completed and d flight tests were w conduc cted. Accordiing to the flig ght tests, even if th he airplane crashed, c controllability off the airplane e under hover or low speeed flight regim mes, has been pro oven. In order to see and validate v the cruise c perform mance, more e flight tests can be madee.

37 Ankara A Interna ational Aerospa ace Conferenc ce

AIAC-2011-052

Aksugür, İnalhan

References [1] R. Beard, D. Kingston, M. Quigley, D. Snyder, R. Christiansen, W. Johnson, T. McLain, and M. Goodrich, Autonomous vehicle technologies for small fixed wing UAVs, AIAA Journal of Aerospace, Computing, Information, and Communication, 2(1):92–108, 2005. [2] N.B. Knoebel, Adaptive Quaternion Control of a Miniature Tailsitter UAV, PhD thesis, Brigham Young University, 2007. [3] S. Ates, I. Bayezit and G. Inalhan, Design and HIL Integration of a UAV Microavionics System in a Manned—Unmanned Joint Airspace Flight Network Simulator, Journal of Intelligent and Robotic Systems, 54(1-3):359–386, 2009. [4] R.H. Stone, The T-wing tail-sitter research UAV, In International Powered Lift Conference, Williamsburg, Virginia, 2002. [5] N.B.K.S.R. Osborne, D.O. Snyder, T.W.M.L.R.W. Beard, and A.M. Eldredge, Preliminary modeling, control, and trajectory design for miniature autonomous tailsitters, AIAA Guidance, Navigation, and Control Conference and Exhibit, 2006. [6] S. Shkarayev, J.M. Moschetta, and B. Bataille, Aerodynamic design of micro air vehicles for vertical flight, Journal of Aircraft , 45(5):1715–1724, 2008. [7] Torres G. E. Mueller, T. J. and D. W. Srull, Introduction to the Design of Fixed-Wing Micro Aerial Vehicles , chapter 5, page 391. AIAA, 2006[3] [8] Url-1 , retrieved 10 January 2008 [9] Merchant, M. P., Propeller Performance Measurement for Low Reynolds Number Unmanned Aerial Vehicle Applications, Wichita State University, Kansas, Wichita, 2005 [10] Url-2 , retrieved on March 2010 [11] Url-3 , retrieved on January 2010 [12] Url-4 , retrieved on January 2010 [13] Darrol Stinton, The Design of The Aeroplane, chapter: Reciprocating Engines, pages 308–310. BSP Professional Books, 1993. [14] R.H. Stone, The T-wing tail-sitter research UAV, In International Powered Lift Conference, Williamsburg, Virginia, 2002. [15] Daniel P. Raymer, Aircraft Design: A Conceptual Approach, chapter Aerodynamics, pages 280– 288. American Institute of Aeronautics and Astronautics, 2nd edition, 1992. [16] Url-5 [17] ModeFRONTIER 3.2, Help Menu [18] Arora, Jasbir S., Introduction To Optimum Design, Elsevier Academic Press, 2004, Second Edition, London [19] Singiresu S. Rao, Engineering Optimization, Printed in the United States of America, 1996 by John Wiley & Sons, Inc., Wiley Eastern Limited, Publishers, and New Age International Publishers, Ltd. [20] Tomas Melin, A Vortex Lattice MATLAB Implementation for Linear aerodynamic Wing Applications, Master Thesis, Royal Institute of Technology (KTH), December 2000. [21] Multhopp, H., , Aerodynamics of Fuselage, NACA-TM 1036, 1942

38 Ankara International Aerospace Conference