MAE 318, Fall 2020: Exam-3 (Final Exam)

Your name: __________________________________________

General Instructions:

1. There are 7 questions in the exam. The total score of the exam is 120.
2. Justify/explain/show work for your results so that we can distinguish an informed answer from a guess/cheating.
Always show the steps leading to your final answer. Just the final answer is not enough to get the credit.
3. You can have three letter sized paper with any formulae/information on it. With things written on front and back of each 3
papers. It should be handwritten, it should not include direct copies of notes/HWs. 2 of these are sheets from the past exams.
4. Some formulae are provided below!

Step response for 1st order system with no zeros

2.2 4 1
Pole: −𝜎, Rise time: 𝑇𝑟 = , Settling time: 𝑇𝑠 = Time constant: 𝜏 =
𝜎 𝜎 𝜎
Step response for 2nd order system with no zeros
𝑓(𝜁) 𝜋
Poles:−𝜎 ± 𝑖𝜔𝑑 = −𝜁𝜔𝑛 ± 𝑖𝜔𝑛 √1 − 𝜁 2 Rise time: 𝑇𝑟 = Peak time: 𝑇𝑝 = ,
𝜔𝑛 𝜔𝑑
4 2
Settling time: 𝑇𝑠 = (time to reach with 2% of steady state), Maximum overshoot: 𝑀𝑝 = 𝑒 −𝜁𝜋 /√1−𝜁
𝜎

Error Analysis
1
For the following control structure: (called unity feedback system), steady state error = lim 𝑠 𝑅
s→0 1+𝐺𝐾
1 1
Steady state error for a step input = =
1+lim 𝐺𝐾 1+𝐾𝑝
s→0
1 1
Steady state error for a ramp input = =
lim 𝑠𝐺𝐾 𝐾𝑣
s→0
Root locus
𝐺𝐾
Root locus of a transfer function GK is the plot of the poles of as the gain on 𝐾 goes from 0 to ∞.
1+𝐺𝐾
Root locus of a transfer function G is the plot of close-loop poles of “G with a proportional controller”.
As the gain of K goes from zero to infinity, the close-loop poles travel from the poles of GK (open-loop poles) and go toward the zeros of
GK (open-loop zeros). Those poles who cannot find a zero, go to infinity along asymptotes. Asymptotes have angles based on how many
they are. All asymptotes intersect at a point on the real axis.
And a positive gain root locus will pass through all points on the real axis, which have odd number of OL poles + OL zeros to their right.

Bode plots and Nyquist plots

Bode plot of G(s) is the plot of the decibel magnitude and angle of complex number 𝐺(𝑗𝜔) as a function of log10 𝜔 for 0 < 𝜔 < ∞.
Nyquist plot of G(s) is the plot of the complex number 𝐺(𝑗𝜔) as a function of 𝜔 for −∞ < 𝜔 < ∞.
1
Bode magnitude is 20 log10 |𝐺(𝑗 𝜔)| dB. Note: 0 = −∞ dB, = −20 dB, 1 = 0 dB , 10 = 20 dB, ∞ = ∞ dB.
10
Bode plots tell you about steady state amplitude amplification factor and phase change when you give a sine input of frequency 𝜔.
If the input to G(s) is 𝐴𝑖 sin(𝜔𝑡 + 𝜙𝑖 ) , then the steady state output / particular solution is: 𝐴𝑖 |𝐺(𝑗𝜔)| sin(𝜔𝑡 + 𝜙𝑖 + ∠𝐺(𝑗𝜔)) .
Bode plot of G(s) is the Bode plots of individual dc gain, poles, and zeros, added together.

Common controllers

1. Proportional controller: 𝐾 = 𝑘𝑑𝑐 , (this is the first one to be tried)

𝑠
𝑠 ( +1)
𝑧
2. Proportional derivative (PD): 𝐾 = 𝑘𝑑𝑐 ( + 1) , Lead: 𝐾 = 𝑘𝑑𝑐 𝑠 (with 𝑧 < 𝑝) (these help with transient response)
𝑧 ( +1)
𝑝
𝑠 𝑠
( +1) ( +1)
𝑧 𝑧
3. Proportional integral (PI): 𝐾 = 𝑘𝑑𝑐 Lag: 𝐾 = 𝑘𝑑𝑐 𝑠 (with 𝑝 < 𝑧) (these help with steady state errors)
𝑠 ( +1)
𝑝
𝑠 𝑠 𝑠 𝑠
( +1)(𝑧 +1 ) ( +1) (𝑧 +1)
𝑧1 2 𝑧1 2
4. PID: 𝐾 = 𝑘𝑑𝑐 Lag-Lead : 𝐾 = 𝑘𝑑𝑐 𝑠 𝑠 (they help with both)
𝑠 ( +1) ( +1)
𝑝1 𝑝2
5. Any combination of poles, zeros, and gain. You can possibly cancel the plant poles / zeros too (if they are stable).
Dc gain of a controller: 𝑘𝑑𝑐 = lim 𝐾(𝑠) (if lim 𝐾(𝑠) = ∞, then by ‘sisotool convention’, remove the s factors to measure 𝑘𝑑𝑐 )
𝑠→0 𝑠→0

Nyquist Stability Criterion

Original: number of unstable closed loop poles (i.e. poles of GK/(1 + GK) ) = number of unstable open loop poles (i.e. poles of GK)
+ number of clockwise encirclements of -1 by Nyquist of GK.
For stable GK (i.e. open loop stable): number of unstable closed loop poles = number of clockwise encirclements of -1 by Nyquist of GK.
Gain and Phase margin tell us how far are you from encircling -1. They should be positive for no encirclements.
1. Total 20 points
a. Consider a differential equation: 2𝑦̈ + 7𝑦̇ + 3𝑦 = 𝑢 − 𝑢̇ . Is the differential equation stable? (2 pts)

poles: -0.5 -3 Stable!

b. Write the expression for (as function of time) and plot the solution 𝑦(𝑡) of 𝑦̈ + 4𝑦̇ + 4𝑦 = 2𝑢, if 𝑢(𝑡) = 6
for 𝑡 > 0? Assume zero ICs. (4 pts)

poles: -2, -2 1pt

PS: 3 1pt
graph: 1pt
y(t) = e(-2t) (a + bt) + 3 1pt

PS = Particular Solution

c. Here is the plot of the solution of some 2nd order differential equation for a constant input, what can you
say about its poles: are they real, imaginary, complex, negative, positive, or zero?
Explain briefly! (2 pts)

1 pole = 0 1pt
1 poles negative real 1pt
Explanation: → PS is a line , → approaches PS 1 pt
 Linear Time-Invariant
d. To qualitatively plot the solution of a stable LTI system for a given input, the three steps are (2 pts)
i. First find the _______poles__________________________________________.

ii. Then find and draw the ____PS_____________________________________.

initial conditions and goes towards the

iii. The solution graph starts from ________IC______________
________PS__________________________.

e. Write the general symbolic expression (no numbers) for the particular solution of 𝑦̇ + 2𝑦 = 𝑢(𝑡), if
i. 𝑢(𝑡) = 𝑘 is a step : (2 pts)

k2

ii. 𝑢(𝑡) = 𝑎𝑡 + 𝑏 is a ramp : (2 pts)

ct+d

iii. 𝑢(𝑡) = 𝐴 sin 𝑏𝑡 is a sinusoid : (2 pts)

Bsin(ct+d)

100(𝑠+1)
f. Draw the unit step response of transfer function 𝐺 = (4 pts)
(𝑠+1.1)(𝑠2 +4𝑠+5)(𝑠+10)

Approximation: G = 10/(s^2 + 4s + 5) 1 pt
PS = 1.818 1pt
Dom poles: -2+/- 1i -- decaying oscillations. 2pt
 1
2. Consider a simple controller architecture, with 𝐺 = 𝑠2
and 𝐾 = 𝑘(𝑠 + 4) where 𝑘 > 0 is the (controller) gain.
(𝑠+4) 𝐾 𝐺
The rlocus of 𝑠2 is shown below. (Total 12 points)
For your own good, make sure you understand that this rlocus has two
poles at origin, and one zero at -4.

a. What are the open loop poles? (2 pts)

0,0
Gain 8

b. What are the closed loop poles when the gain is 0? (2 pts)

0,0 Gain 16

c. What are the closed loop zeros? (2 pts)

-4

d. For what values of the gain is the closed loop system stable? (2 pts)

>0

e. If 𝑅(𝑠) is unit step, will the steady state error 𝑅 − 𝐶 be zero or not? Explain (2 pts)

yes , due to integrator

f. If 𝐷(𝑠) is unit step, will the steady state error 𝑅 − 𝐶 be zero or not? Explain (2 pts)

no, it needs 1/s in K

3. Total 18 points
a. Find the equivalent TF from R to C in terms of 𝐺1 , 𝐺2 . (5 pts)

1
1 𝐺1
(𝐺2 + ) ∗ 1+𝐺1
𝐺1 C(s)

𝐺1𝐺2+1
1+𝐺1

everyone who tried reduction methods got it wrong.. except 2…

b. Find the TF from voltage 𝑣𝑖 to the output 𝑣0 in terms of 𝑅 and 𝐶. (5 pts)

1
𝐶𝑠 1
=
1 𝑅𝐶𝑠 + 1
𝑅+
𝐶𝑠

1pt VR = IR
1pt VC = I/Cs
1pt KVL: vi -IR – 1/Cs = 0
1pt Vo = I/Cs
1pt solve to get vo TF
c. Give a reason why prefilters may be useful. And a problem they may cause. (2 pts)
i. Reason:

ii. Problem:

d. Complete the table below: (6 pts)

Controller K(s) Pole-zero diagram of K(s) 1 pro of this controller 1 con of this controller

P 0.4 points 0.4 points 0.4 points

PD

PI

PID

Lead
4. Total 10 points
a. Draw the Bode plot of GK = (s+2)/(s+20). Make sure to label the x and y axes with numbers. (4 pts)

start: -20dB , 0 deg 1pt

end: 0dB, 0 deg 1pt
up at zero 1pt
down at pole 1pt

105
b. Here is the Nyquist plot of 𝐺𝐾 = (𝑠+2)(𝑠+3)(𝑠+4) .The Nyquist plot intersects negative real axis at −0.5
Explain using Nyquist stability criterion if the closed loop system GK/(1+GK) stable or not? (2 pts)

Nyquist Diagram
ucLP = uOLP + encirc 1pt 3

=0+0=0 2
= stable 1pt
1
Imaginary Axis

-1

-2

-3
-1 0 1 2 3 4 5
Real Axis

c. If you give an input 𝑢 = 2 sin 𝑡 to the plant whose Bode plot is shown on the side, what will be the steady
state output? (3 pts)

gain = -20dB = 0.1 0.5 pt

phase = - 5 deg 0.5 pt
2*0.1* sin (t – 5*pi/180) 2pt

d. What is the dc gain of the TF whose Bode

plot if shown in part c. (1 pt) 0.1
5. (Total 20 points) For the system shown, the plant has a D(s)
(𝑠+4)
transfer function 𝐺 = , Design a controller 𝐾(𝑠), so that
𝑠
R(s) + U(s) + Y(s)
if you give a step reference command 𝑟(𝑡), then 𝐾(𝑠) 𝐺(𝑠)
1) the output 𝑦(𝑡) reaches the value of step 𝑟(𝑡) exactly, -
2) the settling time is less than 1s,
3) even if there is a step disturbance 𝐷(𝑠) the error 𝑟 − 𝑦 is zero in steady state.
4) make sure that 𝒖(𝒕) is finite.
5) any overshoot is fine!

a. What do you need in 𝐾(𝑠) to make sure 𝑢(𝑡) remains finite for a step 𝑟(𝑡) ? Explain. (3 pts)

prefilter or a proper controller

b. What do you need in 𝐾(𝑠) to make sure 𝑦(𝑡) reaches the value of step 𝑟(𝑡) exactly? Explain. (3 pts)

nothing G has 1/s

c. What do you need in 𝐾(𝑠) to make sure 𝑦(𝑡) reaches the value of step 𝑟(𝑡) exactly despite the presence of
a step disturbance 𝐷(𝑠)? Explain. (3 pts) Hint: The answer is not ‘nothing’!

K needs 1/s

d. Show the steps and reasoning you followed to get your final controller. (11 pts)
Show the rootlocus of your final GK, with the appropriate region for settling time marked in it.
Write down your final controller K(s). I need numerical values of the gain, poles, and zeros of the controller.

Proper settling region < -4 → 2 pt

rlocus: 4pt
CLTF: 2pt
K(s): 3pt (need K = gain/s, gain > 8 to get poles to the left of -4)
6. (Total 20 points) To change the angle 𝜃 of a satellite, you need to apply torque 𝑇 to its flywheel. The relationship
between the torque 𝑇 and the satellite angle is: 𝑇 = 𝐽𝜃̈, where inertia 𝐽 = 0.1 kg m2. Hence the TF from torque to
Θ 1
the angle is = . The angle, 𝜃 is measured by a gyroscope to produce the measured angle 𝜃𝑚 . Gyroscope TF
𝑇 0.1 𝑠2
Θm 10
is 𝐻 = = . A controller 𝐾(𝑠) reads the difference in the desired angle 𝜃𝑑 and the measured angle 𝜃𝑚
Θ (𝑠+10)
from the gyroscope and provides the voltage signal 𝑉 to the flywheel motor. The transfer function from the motor
𝑇 40
voltage 𝑉 to the flywheel torque 𝑇 is 𝑉
= 𝑠+40.
If 𝜃𝑑 is a ramp reference command of unit slope, design a proper controller 𝐾(𝑠) and if needed a pre-filter 𝐹(𝑠) so
that the output 𝑦(𝑡) will follow the ramp with no steady state error and a settling time less than 10 seconds.

a. Draw the feedback block diagram of this control system. Fill in all the TFs in the blocks (5pts)

-----o ------[ K ]----[ 40/(s+40) ]-----[ 10/s^2 ] -------

| |
|____[10/(s+10) ]___________________|

b. Explain what you need to meet the ramp error spec. (you’ll have to do some serious algebra!) (6pts)

For no ramp error I need 1/s^2 in GK , G already has it! --- 3pt
also I need a pre-filter 10/(s+10) ---- 3pt

c. Shade in the s-plane the region for the closed loop poles, so that settling time specification is met. (2 pts)

poles < 0.4

d. Show the steps and reasoning you followed to get your final controller. (7 pts)
Show the root locus of your final GK.
Write down a general expression of your final controller K(s). I DO NOT need the actual numerical values of
the gain, poles, and zeros of the controller. I just need a qualitative general expression of the controller,
that will meet all the specifications as evidenced by the rlocus.

show rlocus of GKH not GK 2pt

plan 3pt
cancel -10, -40
put zero at -1 to pull to the left
put 3 poles far to the left
rlocus ------------- 2pt
7. We shall attempt to make a very simplistic model of a (rear-wheel drive) electric car run by a dc motor.
The armature resistance of the motor is 𝑅, the armature inductance is about 0, the motor torque and back emf
constants are both 𝐾, and the rotational inertia of the motor’s armature is 0. The total mass of the car is 𝑀 and the
car moves in the air with linear air drag 𝑐𝑣.
When you apply the voltage 𝑉 to the motor, it provides a torque 𝑇 and speed 𝜔 to the input of the transmission.
The transmission then delivers the torque to the rear wheels. The gear ratio of the (very light & frictionless)
transmission is 𝐺. Assume the wheels too are very light and have radius 𝑟, and do not slip on the road

Find the transfer function from the motor voltage 𝑽 to the car speed 𝒗. (Total 20 points)
The points shall be awarded for
1) drawing the FBD of the whole car,
2) a separate FBD of the rear car wheels, and
3) a separate FBD for the front car wheels,
4) writing all the needed equations for electrical circuit, and all the FBDs, etc., and
5) combining them to get the final TF.

Some hints: Speed of the car, assuming no slipping of the wheel, is 𝑣 = 𝑟𝜔𝑤 , where 𝜔𝑤 is the angular velocity of
the wheels. In going from the motor to the wheels through the transmission, while the torque increases by a factor
of 𝐺 (assuming no friction and massless gears), the angular velocity decreases by the same factor 𝐺.

KVL U-IR - Vb = 0 3pt

Bemf Vb = Kw 1pt
Motor torque Tm = KI 1pt

gear wheel torque T = G*Tm 1pt

gear Motor speed w = G*ww 1pt
Wheel speed ww = v/r 1pt

FBD of car frictions + Drag 2pt

equations: F = Msv + cv 3pt

FBD of rear wheel torque and friction 1pt

equation: F = T/r 1pt

FBD of front wheel + eq : fr=0 1pt

Combine and eliminate var: 2pt

Final TF: 2pt

U – ((Ms+c)vr/GK)R – K (Gv/r) = 0
if: GK/r = N
Then rewrite: U – ( (Ms+c)R/N - N )v = 0
TF = 1/ ( (Ms+c)R/N + N) = N / ( R(Ms+c) + N^2)