Christian Rodriguez Final Paper

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

A study of the Saab J35 Draken

Christian Rodriguez Royal Institute of Technology, 100 44, Stockholm, Sweden March 10, 2011
Abstract In this paper, an analysis is presented for the Saab J35 Draken. In order to make a valid model, empirical data from previous experiment are used. The paper is divided into three parts, performance analysis, unsteady aerodynamics and nonlinear simulation model. Calculations were done using dierential equation for the ight path and numerical methods for achieving the angles of attack. Wind tunnel experiments were done to acquire the derivatives Clp and Cl . To analyze the stability of the J35 eigenvalues were calculated and plotted on a Root locus plot.

Nomenclature
0 T D W V g b h m Sref Thrust installation angle [ ] Flight path angle [ ] Angle of attack [ ] angle of attack correction[ ] Engine Thrust [N ] Drag [N ] Weight [N ] Ground speed [m/s] Gravitational acceleration [g/m2 ] Fuel burned rate [kg/s] Altitude [km] Total mass [kg] Wing reference area [m2 ] nz M CL CD CD0 CL u x Density [kg/m3 ] Load factor [] Mach number [] Lift induced coecient [] Lift coecient Drag coecient [] Zero Drag coecient [] Derivative dCL /d [-] Control Variable State variables

Introduction
The Saab J35 Draken is a results of several years of research [1]. Supersonic ight was new in the early 60s and not much was known about ying these circumstances. The J35 was requested as a new interceptor that would follow up and replace the Saab 29 Tunnan. A new design approach was used using delta wings instead of swept wings. Further design studies led to a double delta wing conguration. The inner thicker delta wing had benets of having room for fuel and landing gear. The outer thinner delta wing gave additional lift for low speed ight but still kept low drag penalty at high speed ight. This type of design was the rst of its kind. To understand this unique design a study of the J35 Draken is performed. The paper is divided into three parts, performance analysis, unsteady aerodynamics and nonlinear simulation model.

C.Rodriguez

Flight Mechanics

VT-11

Part I - Performance analysis


1 Project Brief
The rst analysis is the static performance. Here the excess thrust (Tex ) and Specic excess power (SEP )are processed. To get a visual interpretation, the results are presented in a Tex vs M -graph and SEP-graph. for a pilot a SEP-graph is an important tool to understand the ight performance of the aircraft. The ights are done in steady level ight at full thrust, for altitudes 0 to 16 km and from Mach 0 to 2. The second task is to nd the minimum time to climb from V = 100 m/s, h = 0.1km to M 1.5, h 11km. This is controlled by regulating (t), for nding in this case a ight path angle (t) that minimizes th time to climb. The requirements are to start with a ground speed of V = 100 m/s at 0.1 km and nd an optimal trajectory to an altitude of 11 km at 1.5 Mach. In addition nding the trajectory giving the shortest time to the maximum Mach number is carried out. Initial conditions are the same as for the previous part in task two.

1.1

Equations of motion

Given that the Draken reaches altitudes below 16 km and never reaches a Mach number larger then two, its more convenient to derive all equations with a at earth point of view. For this model the aircraft can be assumed a point mass. From Fig.1 a force equilibrium can be derived using the equation of motion:

Figure 1: Forces acting on the J35 from a body xed reference.

mV = T cos ( + T ) D mg sin mV = T sin ( + T ) + L mg cos h = V sin m = b

(1) (2) (3) (4)

Forces perpendicular to the ight path in Eq.(2) is assumed equal to zero. This is based on the approximation that the changes in are considerably small and can be assumed equal to zero. Dotted variables are time derivative of respective states.

C.Rodriguez

Flight Mechanics

VT-11

1.2

Lift and drag models


1 D = CD V 2 Sref 2 1 L = CL V 2 Sref 2

In order to calculate the lift L and drag D the equation below are used: (5) (6)

As for the drag and lift coecients:


2 CD = CD0 + Sref CL

(7) (8)

CL = CL ( 0 )

CL , and 0 are design characteristics of the delta wing. A very important aspect is that these values change over Mach number. As a consequence the drag parameter is both dependent on altitude h and Mach number M .

1.3
1.3.1

Method
Static performance

Tex is dened as the dierence in forces parallell to the velocity vector as seen in Eq.(1). For a steady level ight the dierence in thrust has to be equal to zero. The angle of attack can be solved from Eq.(2). However this is a nonlinear algebraic equation and has to be solved with numerical methods. This can be done by using Bisection method. In this graph there should be a curve that represent each altitude from 0 to 16 km. In task one results are done with full thrust and afterburners at a 30% fuel level, full thrust in ISA 0 conditions and a load factor of nz = 1 at a ight path angle equal to = 0 This can be carried out by using contour in MATLAB. For this, we need the specic excess power which is dened as: SEP = 1.3.2 Dynamic performance V Tex mg (9)

For the dynamic problem we want to solve: x = f (x, u) (10)

To nd the state variables (x), the dierential equation from Section 1.1 can be solved. Rewriting this to an equation system we get: T D V m cos ( + T ) m g sin x= h = (11) V sin b m This can be found numerical by using ode45 in MATLAB, this method is based on Runge-Kutta method. As for choosing (c), there is no simple way of nding this variable. Instead, two ight trajectories are computed with dierent ight strategies. The rst test starts out with full tank level and will be ying through the sound barrier at a low altitude, (approximately 6 km). As for the second test a slightly lower fuel level is used and this time the aircraft will y at a higher altitude (approximately 12 km) before ying trough the sound barrier. Using the results from previous task we can eventually nd the trajectory giving the maximum Mach number. 3

C.Rodriguez

Flight Mechanics

VT-11

1.4

Results

Results from the static performance are seen in Fig.2:

Figure 2: Excess thrust vs Mach number to the left, SEP-graph to the right.

Results from the dynamic performance are seen in Fig.3, Fig.4 and Tab.1.

M=1.77 h=12km

M=1.5 h=11km

data1 data2 data3

0.4

0.6

0.8

1.2

1.4

1.6

1.8

Figure 3: SEP-graph with corresponding ight trajectories.

C.Rodriguez

Flight Mechanics

VT-11

Velocity vs time
600 Flight Velocity [m/s] 500 400 300 200 100 0 Height [km] 10 15

Altitude vs time

100

200 300 Time [s]

400

500

100

200 300 Time [s]

400

500

Fuel level vs time


1 0.8 0 0.6 0.4 0.2 100 [o] 50

Flight path angle vs time

Fuel level []

50

100

200 300 Time [s]

400

500

100

200 300 Time [s]

400

500

Figure 4: State- and control- variables vs time. Corresponding color of curves are the same as in previous gure.

Data 1 Data 2 Data 3

Initial fuel [%] 70 100 100

End fuel [%] 32 50 40

climb time [s] 275 315 500

h [km] 11 11 12

M 1.5 1.5 1.77

Table 1: Fuel mass and climb time results.

1.5

Discussion

Flying at regions with high SEP values give a high rate of climb, the three trajectories have this in common as shown Fig.3. Looking at Fig.4, the trajectories have dierent state and control attributes. By climbing and then diving (rapid change of ight path angle to negative value) the velocity increases signicantly, see green curve in Fig.4. This reduces both climb time and fuel consumption. Problems occurred when nding the trajectory for reaching the maximum M . It was more of a question of nding a path that gave a fuel level > 30%. 1.5.1 Errors and inaccuires

Changes in mass and ight path angle over time will aect the shape of the SEP graph. However these changes are small enough to assume that they signicant aect in our results. Looking at Fig.4 there is discontinuity of gamma angle. This means that at a certain time there is a innite change in which as a consequence gives a innite large size in nz . This is of course not realistic but this is something we have to accept in our model. When trying to nd the trajectory giving maximum M there seems to be a error in the MATLAB program. The SEP-graph was computed for a fuel level of 30 %.If the Draken starts with an fuel level larger then 30% the SEP graph will change shape, as mentioned earlier. In theory, this means that the Draken can never reach this point with a fuel level above 30 %. However this seemed to work with my program, this shows a few bugs in the MATLAB-script. 5

C.Rodriguez

Flight Mechanics

VT-11

Part II - Unsteady aerodynamics


Additional Symbols in part II Cl Cl Clp L kexp q J p n Rolling-moment coecient [] Derivative dCl /d [] Damping in roll derivative [] Sideslip angle [ ] Bank angle [ ] Dihedral angle [ ] Rolling moment [N m] Torsional stiness [N/m] Dynamic pressure [P a] Mass moment of inertia [kgm2 ] Angular velocity around x-axis [kgm2 ] Eigenvalues [] Damping (real part of ) [] Frequency (imaginary part of ) [rad/s]

Project brief

In this part of the project the purpose is to nd the unsteady aerodynamic derivative Clp . This derivative is known as the damping-in-roll derivative [2]. This can be achieved by analyzing the dynamics of roll through wind tunnel testings. The experimental model consist of a down scale model of the J35 that is mounted on a exible sting. By twisting and locking it to place, the sting becomes a mechanism for triggering roll oscillations. Worth mentioning is that the scale isnt an exact model of the J35, more about this in the discussion. For the second task the aerodynamic derivative Cl is to be determine. Unlike Clp , Cl is a function of . Compared from previous part the J35 will now be considered a rigid body.

Figure 5: Lateral schematic of rolled airplane.

C.Rodriguez

Flight Mechanics

VT-11

2.1

Method

The approach for nding these derivatives can be done by looking at a forced vibrating system governed by the dierential equation: m + cx + kx = L x (12)

Looking at our problem we want to gure out the rotation rather then the x-direction. Rewriting the equation gives us: J + C + k = L (13) Where each term represent inertial force (J ), damping force (C ) and elastic force (k). The damping is relative small and can be assumed zero. Clp and Cl are expressed in in Eq.(13) implicitly. However to nd these derivatives the moment of inertia (J) has to be calculated rst by doing a Ground Vibration test. Setting Eq.(13) equal to zero meaning that there are no aerodynamic forces acting on the model. Rewriting the equation and using complex analysis the general solution can be set as = et , this gives the nal expression to: k k = J = 2 (14) J k is equal to 98.3 N/m Once J is found, the derivatives can be calculated. Looking at gure Fig.5, the roll moment and roll rate can be expressed as: 2 = L = Jp p= The roll moment can also be written as: L = Cl bqS Where, pb + Cl 2V Inserting Eq. (15) (16) (17) and (18) in Eq. (13) gives us: ][ ] [ ] [ 0 1 = Cl bqSkexp Clp b2 qS p p J 2V J Cl = Clp (18) (17) (15)

(16)

(19)

Since the model is linear, eigenvalues can be calculated. By setting equal to zero the derivative Cl is dropped. Clp can be found by setting an arbitrary value for it as a function of V that ts the eigenvalues from the experimental data. This yields the eigenvalues: 2 Clp q 2 S 2 b4 k Cl qSb Clp qSb2 (20) 2J 2 4V J 16V J J Once Clp is found, one can vary and and apply the same method to Cl .

C.Rodriguez

Flight Mechanics

VT-11

2.2

Results
Damping vs Velocity
0 Fitted curve =0o Damping [Nm] 1 2 3 4 Experimental data =0o Fitted curve =5o Experimental data =5o

10

15

20 Velocity [m/s]

25

30

35

40

Angular velocity vs Velocity


47.5

47

46.5

10

15

20 Velocity [m/s]

25

30

35

40

Figure 6: Angular and damping vs velocity.

The value for J was calculated to around 0.0448. For the derivatives, the results which correspond to these curves are Clp = 0.250 and Cl = 0.205.

2.3

Discussion

The results say a lot about Clp . It has a close relation to the damping n as seen in Fig.6. A large value on Clp gives a high damping on the contrary, the derivative does not aect frequency. Now if we increase the angle of attack, signicant changes occur in but the damping remains unchanged. Conrming this in Eq.(20) Clp only aects the real part (the imaginary part can considered small enough to assume its equal to zero) and Cl aects the imaginary part. Another approach for nding Cl is by having a constant velocity and changing the . This was also done and gave almost the same results. 2.3.1 Errors and inaccuracy

Cl expression is only valid for small changes in and . Experimental data can deviate depending on the accuracy from measurement from wind tunnel data. Comparing the diagram [3] for Cl with the one calculated, its clear that there is a large dierence. One factor for this is that the model used in the wind tunnel had some geometric dierences with the real scale model, however the biggest factor for this error that the model was missing the n. This clearly has a huge impact on Clb eta .

C.Rodriguez

Flight Mechanics

VT-11

Part III - Nonlinear Simulation model


Additional Symbols in part III V u w e p xcg xr cr True airspeed [m/s] Velocity x-component in the body frame reference [m/s] Velocity z-component in the body frame reference [m/s] Elevator deection [o ] Thrust output [%] Center of gravity [m] Attitude [o ] State variables Control variables

Project brief

For the nal part of the project the stability and control abilities are to be inspected. As in the previous section, the J35 is considered a rigid body. Only the longitudinal degrees of freedom are considered in this project. Simulations are going to be done and to keep realistic ight conditions some constrains have to be dened. These can be seen in Tab.2: Parameter V Fuel level nz e Minimum value 10% 4 25o 50o Maximum value 1350km/s 90% 7 10o 140o

Table 2: Flight constrains.

Another constrain is that the largest forward center of gravity position is xcg = 10m. Flight is done in level ight meaning that = . The nal part can be divided into three task; trim conditions, linear stability and nonlinear simulations. These are explained in depth in the next section.

3.1
3.1.1

Method
Trim conditions

The condition for trim is that the aircraft is in longitudinal equilibrium. This requites that the pitching moment is equal to zero and all external forces are in equilibrium. In this problem the state and control variables which are of interest are xr = (u, w, qp , ) and cr = (e , p ). Using the general equation of motion in Etkin [2] the states can be calculated. However, looking at these equations, one can see that they are a system of non linear equations. For simplication we assume the model is linear, this is a valid approximation considering that in most cases the aerodynamic eects are nearly linear functions. Using Newton Raphson method (Eq.(21)) the trim values can be found: xn+1 = xn J1 f (xn ) (21)

where J is the Jacobian matrix, this is computed by using nite dierent method. Once the trim conditions are accomplished, investigations can be made on how the position of the xcg aects the elevator trim settings. The goal is to nd the maximum rearward position of this quantity. The approach here is to nd when maximum deection occurs on the elevators (see, Fig.2).

C.Rodriguez

Flight Mechanics

VT-11

3.1.2

Linear stability

In the second section the objective is to nd the eigenvalues for the longitudinal degrees of freedom. The eigenvalues can be acquired by linearizing the nonlinear model. This is visualized by performing a Root locus plot. The purpose of this is to see how the eigenvalues vary when changing altitude and Mach number. In addition, the short period and phugoid are identied by using the eigenvalues. 3.1.3 Nonlinear simulation - The Looping maneuver

The nal part in this report is about the Drakens performance setting up a nonlinear simulation. The goal is to compute a loop using a combination of the thrust (p ) and elevator setting e which is change over time. To get valid results the constraints mentioned in Tab.2 has to be taken to account. Initial condition in this simulation are 1km, fuel level 500kg and a true airspeed of 700km/h. Additional investigation on the aircrafts maneuverability is also done. When in the pull up maneuver theres a sudden change in elevator deection (e ). The ratio seen in Eq.(22) is a measurement of how sensitive the aircraft is to maneuvers. This ratio is known as elevator per g: e nz 1 (22)

This quantity can be found by doing a simulation where you change the elevator settings over time.

3.2

Results

Set of altitudes used in Section 7 were 0, 5 and 10 km with a fuel mass of 500 kg. Alt [km] 0 5 10 max xcg [m] 10.93 10.53 10.57
e nz 1

0.034 0.07 0.119

Table 3: Maximum postion of xcg and values showing maneuverability.

10

C.Rodriguez

Flight Mechanics

VT-11

Alfa trim vs mach


30 [o] 20 10 0 0.3 0 km 5 km 10 km

0.35

0.4

0.45

0.5 Mach

0.55

0.6

0.65

0.7

Elevator angle trim vs mach


0 e [o]

10 0.3

0.35

0.4

0.45

0.5 Mach

0.55

0.6

0.65

0.7

Thrust level trim vs mach


1 p [%]

0.5

0 0.3

0.35

0.4

0.45

0.5 Mach

0.55

0.6

0.65

0.7

Figure 7: Trim values for respective parameters, thrust, elevator and alfa.

Following results show the linear stability of the aircraft for dierent h and M :
Root locus plot
4 3 2 Imaginary (A) 1 0 1 2 3 4 2 Possible instability Increasing Mach

1.5

0.5 Real (A)

0.5

Figure 8: Root locus plot, where red 0km, green 5km, blue 10km.

Mach 0.7 0.7 0.3 0.3

Name Short term Phuigoid Short term Phugoid

Period [s] 2.20 107 11.1 54

thalf [s] 0.67 141 1.41 478

Nhalf [Cycles] 0.31 1.32 0,13 8.85

Table 4: Dierent modes for the lowest and highest mach number at 5 km altitude.

11

C.Rodriguez

Flight Mechanics

VT-11

Results from loop simulation shown in Fig.9 and Fig.10:


Range vs Altitude
Altitude [m] 2500 2000 1500 1000 0 500 1000 1500 Range [m] 2000 True Airspeed [m/s] 3000 200 150 100 50

True Airspeed vs Time

10

20 30 Time [s]

40

Alfa vs Time
40 6 nz [] 20 [ ]
o

Load factor vs Time

4 2 0

0 20 2

10

20 30 Time [s]

40

10

20 30 Time [s]

40

Figure 9: State variables vs time.

Elevator setting vs Time


0 e [] 2 4 6

10

15

20 25 Time [s]

30

35

40

45

Thrust vs Time
1 p []

0.5

10

15

20 25 Time [s]

30

35

40

45

Figure 10: Control variables vs time.

3.3
3.3.1

Discussion
Trim values

The results shown in Fig.7 are between Mach between 0.3 and 0.7, values below 0.3 gave unstable trim points. The p and increases in magnitude when increasing altitude on the contrary for e . The reason for this is the low density in higher altitudes. To keep trim conditions these relations has to be valid. If cruising at 10 km, ight is only possible for M > 0.54 due to maximum throttle conguration reached. As the altitude increases the maneuverability decreases as shown in Fig.3. The reason for this is that the maneuverability is depended on the aerodynamic forces which are decreased as the density is decreased. 3.3.2 Root locus

Looking at the results in Fig.8 one can see that ying in high Mach numbers, gives a high (absolute) eigenvalues. This means that when the J35 ies in higher Mach number it will be settle faster to steady state ight. This seems logical, if ying in a higher altitude the air is thinner and to

12

C.Rodriguez

Flight Mechanics

VT-11

keep a steady state ight a higher Mach number is needed. As shown in Fig.8 there is a possible instability. This occurs when ying at 10 km with a speed of Mach 0.3. This is what one intuitive would think, seeing as how the air is thinner at those altitudes. To provide the same lift necessary to sustain ight, the aircraft has to y at high enough speed. This is also conrmed in Tab.4 showing the dierent modes at 5 km altitude. 3.3.3 Loop simulation

The loop in Fig.9 was computed by having an high trust output into and around the loop. To get a load factor below the limit the aircraft had to y in a large enough radius, in our case it was around 750 m. After around 4 seconds the loop is initiate a rapid change in the deection and maximum thrust is initiated going in to the loop d, as seen in Fig.10 . To maintain the arc of the loop and maintaining V , the thrust is kept at max throttle. The elevator on the other hand is slightly varied through the loop. When coming out of the loop the elevator is increased to get a nose up pitch moment. This is desired to get back to a steady level ight.

Conclusions
Static and Dynamic performance By reducing mass and ying through the sound barrier in a higher altitude and increase the velocity by diving will reduce the minimum time to climb. Flying at areas with maximum SEP will decrease the climb time. Derivative Clp and Cl The derivatives Clp and Cl has strong relations with the damping and frequency. The conclusions are n = n(Clp , V ) and = w(Cl , V ). Deviation between the computed results and the diagram [3] can be aected by the scale model which had no n. Stability and Control Increase in Mach number leads to a more stable ight state (large eigenvalues). Flying in higher altitudes pushes the eigenvalues closer to instability ( real part n > 0). To maintain the limits of the load factor when doing a loop maneuver, the J35 has to have large enough loop radius.

References
[1] Military Factory Specications Saab J35 Draken, web site http://www.militaryfactory.com/aircraft/detail.asp?aircraft id=121 Last retrieved 2011-03-10 [2] Bernard Etkin, Loyd Du Reid. Dynamics of Flight Stability and Control. John While and Sons, inc. 3rd International Edition, 1995. ISBN 978-0-471-03418-6 [3] KTH Flyg Cl diagrams, web site http://fdl8.flyg.kth.se/j35/section 2.21/diagram268.jpg Last retrieved 2011-03-10

13

You might also like