16.
333: Lecture # 6 Aircraft Longitudinal Dynamics
Typical aircraft openloop motions Longitudinal modes Impact of actuators Linear Algebra in Action!
Fall 2004
16.333 51
Longitudinal Dynamics
Recall: X denotes the force in the Xdirection, and similarly for Y and Z, then (as on 413) X ,... Xu u 0
Longitudinal equations (see 413) can be rewritten as: mu = Xuu + Xw w mg cos 0 + X c
m(w qU0) = Zuu + Zw w + Zw w + Zq q mg sin 0 + Z c Iyy q = Muu + Mw w + Mw w + Mq q + M c
There is no roll/yaw motion, so q = .
Control commands X c, Z c, and M c have not yet been specied.
Fall 2004
16.333 52
Rewrite in state space form as mu (m Zw )w Mw w + Iyy q 0 mg cos 0 Xu Xw Zu Zw Zq + mU0 mg sin 0 = Mq 0
Mu Mw 0 0 1 0 u X w
Z c +
q
M c 0
c
m 0 0 m Zw
0 Mw 0 0 Xu Xw Z Zw = u
Mu Mw 0 0
0 0 Iyy 0
u 0
0 w
0
q
1 u X w
Z c +
q
M c 0
c
0 mg cos 0 Zq + mU0 mg sin 0 Mq 0 1 0
E X = AX + c
descriptor state space form
X = E 1(AX + ) = AX + c c
Fall 2004
16.333 53
Write out in state space form:
Xu m Xw m Zw m Zw 0 Zq + mU0 m Zw g cos 0
Zu m Zw A = 1 Iyy [Mu + Zu]
0
1 1 Iyy [Mw + Zw ] Iyy [Mq + (Zq + mU0 )]
mg sin 0 m Zw 1 Iyy mg sin 0
0
Mw m Zw
Note: slight savings if we dened symbols to embed the mass/inertia Xu = Xu/m, Zu = Zu/m, and Mq = Mq /Iyy then A matrix collapses to:
Xu Xw Zw 1 Zw 0 Z q + U0 1 Zw g cos 0
Z 1 uZ w A= M + Z u u
M w + Zw Mq + (Zq + U0 ) 0 1
g sin 0 1 Zw g sin 0
Mw 1 Zw
Check the notation that is being used very carefully
To gure out the c vector, we have to say a little more about how the control inputs are applied to the system.
Fall 2004
16.333 54
Longitudinal Actuators
Primary actuators in longitudinal direction are the elevators and thrust. Clearly the thrusters/elevators play a key role in dening the steadystate/equilibrium ight condition Now interested in determining how they also inuence the aircraft motion about this equilibrium condition deect elevator u(t), w(t), q(t), . . .
Recall that we dened X c as the perturbation in the total force in the X direction as a result of the actuator commands Force change due to an actuator deection from trim Expand these aerodynamic terms using same perturbation approach X c = Xe e + Xp p e is the deection of the elevator from trim (down positive) p change in thrust Xe and Xp are the control stability derivatives
Fall 2004
16.333 55
Now we have that
X Xe Xp c
Z Zp 1
Z c =
E = E 1
e
M c
Me Mp 0 0 0
e = Bu
p
For the longitudinal case
Zp Ze m Zw m Zw B
=
1 1 Iyy [Me + Ze ] Iyy Mp + Zp
0 0 Typical values for the B747 Xe = 16.54 Ze = 1.58 106 Me = 5.2 107 Aircraft response y = G(s)u X = AX + Bu G(s) = C(sI A)1B y = CX We now have the means to modify the dynamics of the system, but rst lets gure out what e and p really do. Xp = 0.3mg = 849528 Zp 0 Mp 0
Xe m
Xp m
Fall 2004
16.333 56
Longitudinal Response
Final response to a step input u = u/s, y = G(s)u, use the FVT u lim y(t) = lim s G(s) t s 0 s u lim y(t) = G(0) = (CA1B) u
t
Initial response to a step input, use the IVT u u y(0+) = lim s G(s) = lim G(s) s s s For your system, G(s) = C(sI A)1B + D, but D 0, so
s
lim G(s) 0
Note: there is NO immediate change in the output of the motion variables in response to an elevator input y(0+) = 0 Consider the rate of change of these variables y(0+) y = C X = CAX + CBu (t) and normally have that CB = 0. Repeat process above1 to show + (0 that y ) = CBu, and since C I, y +) = Bu (0 Looks good. Now compare with numerical values computed in Mat lab. Plot u, , and ight path angle = (since 0 = 0 = 0 see picture on 48)
1 Note
that C(sI A)1 B + D = D +
CB s
CA1 B s2
+ ...
Fall 2004
16.333 57
Elevator (1 elevator down stick forward) See very rapid response that decays quickly (mostly in the rst 10 seconds of the response) Also see a very lightly damped long period response (mostly u, some , and very little ). Settles in >600 secs Predicted steady state values from code: 14.1429 m/s u (speeds up) 0.0185 rad (slight reduction in AOA) 0.0000 rad/s q 0.0161 rad 0.0024 rad Predictions appear to agree well with the numerical results. Primary result is a slightly lower angle of attack and a higher speed Predicted initial rates of the output values from code: 0.0001 0.0233 1.1569 0.0000 0.0233 m/s2 rad/s rad/s2 rad/s rad/s u q
All outputs are zero at t = 0+, but see rapid changes in and q. Changes in u and (also a function of ) are much more gradual not as easy to see this aspect of the prediction Initial impact Change in and q (pitches aircraft) Long term impact Change in u (determines speed at new equilib rium condition)
30 0.005 0.05 0.01 (rad) 0.015 0.02 0.05 0.025 0.03 0 time 200 400 600 0 0.1 0
Step response to 1 deg elevator perturbation 0 0.1
Fall 2004
25
20
15
10
0 400 600
200
time
200
time
400
600
30 0.005 0.01 (rad) 0.015 0.02 0.025 0.03
0.02 0 0.02 0 10 30 40 0.04 0.06 0.08 0.1 20 time
25
20
15
10
Figure 1: Step Response to 1 deg elevator perturbation B747 at M=0.8
30 40 0 10 20 time 30 40
16.333 58
10
20 time
Fall 2004
16.333 59
Thrust (1/6 input) Motion now dominated by the lightly damped long period response Short period motion barely noticeable at beginning. Predicted steady state values from code: 0 m/s u 0 rad 0 rad/s q 0.05 rad 0.05 rad Predictions appear to agree well with the simulations. Primary result now climbing with a ight path angle of 0.05 rad at the same speed we were going before. Predicted initial rates of the output values from code: 2.9430 0 0 0 0 m/s2 rad/s rad/s2 rad/s rad/s u q
Changes to are very small, and response initially at. Increase power, and the aircraft initially speeds up Initial impact Change in u (accelerates aircraft) Long term impact Change in (determines climb rate)
15 0.015 0.01 0.005 0.06 (rad) 0 0.005 0.01 0.02 0.015 400 time 600 0.02 0 200 400 600 0 0 200 400 time 600 0.04 0.08
Step response to 1/6 thrust perturbation 0.02 0.1
Fall 2004
10
10
15
200
time
20
x 10
0.09 0.08
6 15
0.07 0.06
5 10 5
4 (rad)
0.05 0.04 0.03
2 0
0.02 0.01
Figure 2: Step Response to 1/6 thrust perturbation B747 at M=0.8
30 40 5 0 10 20 time 30 40 0 0 10 20 time 30 40
16.333 510
10
20 time
Fall 2004
16.333 511
Frequency Domain Response
Plot and inspect transfer functions from e and p to u, w, and See following pages
From elevator: Huge response at the phugoid mode for both u and (very lightly damped) Short period mode less pronounced Response falls o very rapidly Response to w shows a pole/zero cancelation (almost) of the phugoid mode. So the magnitude level is essentially constant out to the frequency of the short period mode
Why would we expect that?
From thrust: Phugoid peaks present, but short period mode is very weak (not in u, low in , w). entirely consistent with the step response. Thrust controls speed (initially), so we would expect to see a large response associated with the phugoid mode (speed variations are a key component of this mode)
10
3 3
Fall 2004
Transfer function from elevator to flight variables 4 4 10 10
10 10 |
de
10
2 2
10
10 |G |
de
ude
|G
|G
1
10
10
0 0
10
10 10
1 1
10
0 0
10 2 10 10
2
10 10 Freq (rad/sec)
10 10 10 Freq (rad/sec)
1 0
10 10
2
10 10 Freq (rad/sec)
0 50 100
de
0 50 100
de
50
100 150 200 250 300 350
0
arg Gude
250
arg G
200
arg G
2 1 0
150
150 200 250 300 350
Figure 3: TFs from elevator to ight variables B747 at M=0.8
10 10 10 Freq (rad/sec) 10
2
300
350
16.333 512
10
10 10 Freq (rad/sec)
10 10 Freq (rad/sec)
10
2 2
Transfer function from thrust to flight variables 3 3 10 10
Fall 2004
10
1 1
10
10
10 |
dp
10
10 |G |
dp
|Gudp|
10
1 1
|G
0
10
10
10
2 2
10
10
10
0
10
10 10 Freq (rad/sec)
10 10
2
10 10 Freq (rad/sec)
10 10
2
10 10 Freq (rad/sec)
0 150 100 50
dp dp
150
50 100 150 arg G
2 1 0
100
50
arg Gudp
50 100 150
0
arg G
0 50
200 250 300 350
Figure 4: TFs from thrust to ight variables B747 at M=0.8
10 10 10 Freq (rad/sec) 10
2
100
150
16.333 513
10
10 10 Freq (rad/sec)
10 10 Freq (rad/sec)
Fall 2004
16.333 514
Summary:
To increase equilibrium climb rate, add power.
To increase equilibrium speed, increase e (move elevator further down).
Transient (initial) eects are the opposite
and tend to be more consistent with what you would intuitively expect to occur
Fall 2004
16.333 515
Modal Behavior
Analyze model of vehicle dynamics to quantify the responses seen. Homogeneous dynamics of the form X = AX, so the response is X(t) = eAtX(0) a matrix exponential. To simplify the investigation of the system response, nd the modes of the system using the eigenvalues and eigenvectors is an eigenvalue of A if det(I A) = 0 which is true i there exists a nonzero v (eigenvector) for which (I A)v = 0 Av = v
If A (n n), typically get n eigenvalues/eigenvectors Avi = ivi Assuming that eigenvectors are linearly independent, can form 1 0 ... A v1 vn = v1 vn 0 n AT = T T 1AT = , A = T T 1
1 Given that eAt = I + At + 2! (At)2 + . . ., and that A = T T 1, then it is easy to show that n At t 1 X(t) = e X(0) = T e T X(0) = vieiti i=1
State solution is a linear combination of system modes vieit eit determines nature of the time response vi gives extent to which each state participates in that mode i determines extent to which initial condition excites the mode
Fall 2004
16.333 516
The total behavior of the system can be found from the system modes Consider numerical example of B747
0.0069 0.0139 0 9.8100
0 0.0905 0.3149 235.8928 A =
0
0.0004 0.0034 0.4282 0 0 1.0000 0 which gives two sets of complex eigenvalues = 0.3717 0.8869 i, = 0.962, = 0.387, short period
= 0.0033 0.0672 i, = 0.067, = 0.049, Phugoid long period Result is consistent with step response heavily damped fast response, and a lightly damped slow one. To understand eigenvectors, must do some normalization (scales each element appropriately so that we can compare relative sizes) u = u/U0, = w/U0, q = q/(2U0/c) Then divide through so that 1 u q Short Period 0.0156 + 0.0244 i 1.0202 + 0.3553 i 0.0066 + 0.0156 i 1.0000 Phugoid 0.0254 + 0.6165 i 0.0045 + 0.0356 i 0.0001 + 0.0012 i 1.0000
Short Period primarily and = w in the same phase. The u and q response is very small. Phugoid primarily and u, and lags by about 90. The and q response is very small. Dominant behavior agrees with time step responses note how initial conditions were formed.
90 2 60 120 30 150 30 0.5 60 1
90
Fall 2004
120 1
150
180
180
210 300 270 240 270 0.067282
330 210
330 300
240
0.96166
1 u w q 0.5
u w q
Perturbation States u,w,q
Perturbation States u,w,q
Figure 5: Mode Response B747 at M=0.8
0 0.5 10 15 1 0 100 200 300 time (sec) 400
0.5
16.333 517
0.5
time (sec)
500
600
Fall 2004
16.333 519
Relative motion between aircraft and an observer ying at a constant speed U0t
Motion of perturbed aircraft with respect to an unperturbed one Note phasing of the forward velocity xe with respect to altitude ze aircraft faster than observer at the bottom, slower at the top The aircraft speeds up and slows down leads and lags the ob server.
Consistent with ight path? Consistent with Lanchesters approximation on 41?
Fall 2004
16.333 520
Summary
Two primary longitudinal modes: phugoid and shortperiod Have versions from the full model but can develop good approx imations that help identify the aerodynamic features that deter mine the mode frequencies and damping Impact of the various actuators claried: Short timescale Long timescale
Fall 2004
16.333 521
Matrix Diagonalization
Suppose A is diagonizable with independent eigenvectors V = [v1, . . . , vn] use similarity transformations to diagonalize dynamics matrix x = Ax xd = Adxd 1 ... = Ad V 1AV = = n Corresponds to change of state from x to xd = V 1x
System response given by eAt, look at power series expansion At = V tV 1 (At)2 = (V tV 1)V tV 1 = V 2t2 V 1 (At)n = V ntnV 1 1 eAt = I + At + (At)2 + . . . 2 1 = V I + + 2t2 + . . . V 1 2 t e1 ... V 1 = V etV 1 = V ent
Fall 2004
16.333 522
Taking Laplace transform, (sI A)1 = V
1 s1
...
1 sn
1 V
n Ri = s i i=1
T where the residue Ri = viwi , and we dene
. V = v1 . . . vn , V 1 = . . T wn
T w1
Note that the wi are the left eigenvectors of A associated with the right eigenvectors vi AV = V
T w1
... n
... n V 1
V 1A = 1 ...
. . A = . T w n
T T where wi A = iwi
. . . T n wn
T w1
Fall 2004
16.333 523
So, if x = Ax, the time domain solution is given by x(t) = x(t) =
n i=1 n i=1 T eitviwi x(0)
dyad
T [wi x(0)]eitvi
The part of the solution vieit is called a mode of a system solution is a weighted sum of the system modes weights depend on the components of x(0) along wi
Can now give dynamics interpretation of left and right eigenvectors:
T Avi = ivi , wiA = iwi , wi vj = ij
so if x(0) = vi, then
x(t) =
n
i=1 i t T (wi x(0))eitvi
= e vi so right eigenvectors are initial conditions that result in relatively simple motions x(t).
With no external inputs, if initial condition only disturbs one mode, then the response consists of only that mode for all time.
Fall 2004
16.333 524
If A has complex conjugate eigenvalues, the process is similar but a little more complicated. Consider a 2x2 case with A having eigenvalues a bi and associated eigenvectors e1, e2, with e2 = e1. Then a + bi 0 1 A = e1 e2 e1 e2 0 a bi a + bi 0 1 = e1 e1 T DT 1 e1 e1 0 a bi
Now use the transformation matrix 1 i 1 1 M = 0.5 M 1 = 1 i i i Then it follows that A = T DT 1 = (T M )(M 1DM )(M 1T 1) = (T M )(M 1DM )(T M )1 which has the nice structure: 1 a b A = Re(e1) Im(e1) Re(e1) Im(e1) b a where all the matrices are real.
With complex roots, the diagonalization is to a block diagonal form.
Fall 2004
16.333 525
For this case we have that at eAt = Re(e1) Im(e1) e
1 cos(bt) sin(bt) Re(e1) Im(e1) sin(bt) cos(bt)
Note that Re(e1) Im(e1)
is the matrix that inverts Re(e1) Im(e1) 1 1 0 Re(e1) Im(e1) Re(e1) Im(e1) = 0 1
So for an initial condition to excite just this mode, can pick x(0) = [Re(e1)], or x(0) = [Im(e1)] or a linear combination.
Example x(0) = [Re(e1)] x(t) = eAtx(0) = Re(e1) Im(e1) eat cos(bt) sin(bt) sin(bt) cos(bt)
1 [Re(e1)] Re(e1) Im(e1) at cos(bt) sin(bt) 1 = Re(e1) Im(e1) e sin(bt) cos(bt) 0 cos(bt) = eat Re(e1) Im(e1) sin(bt) = eat (Re(e1) cos(bt) Im(e1) sin(bt)) which would ensure that only this mode is excited in the response
Fall 2004
16.333 526
Example: Spring Mass System
Classic example: spring mass system consider simple case rst: mi = 1, and ki = 1
k5
M1 k1 k2
M3 k3
M2 k4
Z1
Z3
Z2
x = A =
z1 z2 z3 z1 z2 z3 0 I M = diag(mi) M 1K 0
k1 + k2 + k5 k5 k2 K = k5 k3 + k4 + k5 k3 k2 k3 k2 + k3 Eigenvalues and eigenvectors of the undamped system 1 = 0.77i 2 = 1.85i 3 = 2.00i v1 1.00 1.00 1.41 0.77i 0.77i 1.08i v2 1.00 1.00 1.41 1.85i 1.85i 2.61i v3 1.00 1.00 0.00 2.00i 2.00i 0.00
Fall 2004
16.333 527
Initial conditions to excite just the three modes: xi(0) = 1Re(vi) + 2Im(v1) Simulation using 1 = 1, 2 = 0 Visualization important for correct physical interpretation Mode 1 1 = 0.77i M1
-
j R
M3
-
M2
-
Lowest frequency mode, all masses move in same direction Middle mass has higher amplitude motions z3, motions all in phase
0.6 V1 z1 z2 z3
0.4
0.2
displacement
0.2
0.4
0.6
0.8
5 time
10
Fall 2004
16.333 528
Mode 2 2 = 1.85i M1
-
M3
M2
-
Middle frequency mode has middle mass moving in opposition to two end masses Again middle mass has higher amplitude motions z3
0.4 V3 z1 z2 z3
0.3
0.2
0.1 displacement
0.1
0.2
0.3
0.4
5 time
10
Fall 2004
16.333 529
Mode 3 3 = 2.00i M1
-
M3
M2
0 Highest frequency mode, has middle mass stationary, and other two masses in opposition
0.4 V2 z1 z2 z3
0.3
0.2
0.1 displacement
0.1
0.2
0.3
0.4
5 time
10
Eigenvectors with that correspond with more constrained motion of the system are associated with higher frequency eigenvalues
