Control Theory - 035188

Project 1

Barak Henen – 941170565

Rafael Gruber Bernstein – 941170375

Question 1

To create the block diagram of the overall system, we first need to find and linearize the
equations of motion in the system, to do this we use the following standard DC motor
relations:
𝑉𝑒
𝑉𝑒 = 𝑅𝐼 + 𝐿𝑖; (𝐿 = 0) → 𝑖 =
𝑅
𝑉𝑒 = 𝑢 − 𝑘𝑏 𝜔𝑚

𝑇𝑚 = 𝑘𝑚 𝑖

Knowing the 𝑉𝑎 (Armature Voltage) is generated with a static gain amplifier (𝑘𝑎 ) and
combining the equations, we find:
𝑘𝑎 𝑉𝑖𝑛 − 𝑘𝑏 𝜔𝑚
𝑇𝑚 = 𝑘𝑚
𝑅
Also given, the ratio of the gear box (𝑁𝑔 ), so that the angular velocity of the motor shaft is:

𝜔𝑚 = 𝑁𝑔 𝜙̇ and 𝑇 = 𝑁𝑔 𝑇𝑚

Therefore:

𝑘𝑎 𝑣𝑖𝑛 − 𝑘𝑏 𝑁𝑔 𝜙̇
𝑇 = 𝑁𝑔 𝑘𝑚
𝑅
The same system can be found by defining new coordinate systems, find the equations of
motion around the centre of mass, finding the equations for kinetic and potential energy
and then linearizing the system.

Therefore, the block-diagram of the system is as follows:

Question 2

Let us start off by defining each component of the block diagram as parameters 𝐺1 to 𝐺7 as
defined on the diagram above, in red.

By manipulating the way the system is described, we can easily see the following transfer
functions:

𝐺1 𝐺2 𝐺3 𝐺4 𝐺5
𝑃𝜃𝑢 (𝑠) =
1 + 𝐺2 𝐺3 𝐺7 + 𝐺3 𝐺4 𝐺5 𝐺6

𝑃𝜃𝑢 (𝑠) 𝐺1 𝐺2 𝐺3 𝐺4
𝑃∅𝑢 (𝑠) == =
𝐺5 1 + 𝐺2 𝐺3 𝐺7 + 𝐺3 𝐺4 𝐺5 𝐺6

𝑃∅𝑢 (𝑠)
𝑃∅𝜃 = = 𝐺5
𝑃𝜃𝑢 (𝑠)
We also required to derive 𝑃∅𝑢1 (𝑠)
𝐺1 𝐺2 𝐺3 𝐺4
𝑃∅𝑢1 (𝑠) = lim (𝑃∅𝑢 (𝑠)) →
𝜁→∞ 1 + 𝐺2 𝐺3 𝐺7

Question 3

The following figures were calculated through MATLAB:

Part 3
Question 4

We are required to design a minimal- order unity feedback controller 𝐶∅ (𝑠) in the setup
shown in in Fig.2 (a) of the question paper, such that:

• The steady state response/ errors to step d(t) and 𝑟𝜃 (𝑡) vansish,
• The crossover frequency of the loop 𝜔𝑐 ≈ 50 𝑟𝑎𝑑/𝑠𝑒𝑐
 • The phase margin 𝜇𝑝ℎ ≈ 45°,
• The magnitude of the complementary sensitivity frequency response 𝑇∅ (𝑗𝜔) ≤ 3𝑑𝐵
• The magnitude of the disturbance sensitivity frequency response 𝑇∅,𝑑 (𝑗𝜔) ≥
−20𝑑𝐵
Most likely, the controller that we will design will either be a PI or PID controller as the
requirements are too much for a proportional controller to handle alone.

We will start off by find the proportional aspect of the controller (𝐶𝑝 ) to move our crossover
frequency from 5.44 rad/sec to 50 rad/sec, as requested. To do this we need a proportional
controller that we found by finding the magnitude of 𝑃∅𝑢 (𝑖𝜔) which was 0.0240. Therefore,
𝟏
𝑪𝒑 = |𝑷(𝒊𝝎𝒄)| = 𝟒𝟏. 𝟔𝟔𝟕

To get rid of the disturbance factor of the system we need to create a lead-lag controller
10𝑠+𝜔𝑐
where 𝑙𝑎𝑔 = 𝜔 . However, as 𝛽 → ∞ the lag controller will be of the form:
10𝑠+ 𝛽𝑐

10𝑠 + (5.44 + 41.67) 10𝑠 + 47.107

𝑙𝑎𝑔 = =
10𝑠 10𝑠
In order to create the lead aspect of the controller we will use the following standard
formula:

√𝛼𝑠 + 𝜔𝑚
𝑙𝑒𝑎𝑑 =
𝑠 + √𝛼𝜔𝑚
Where:
1 + sin (𝜙) 1 + sin (47.107)
𝛼= = = 6.4802
1 − sin (𝜙) 1 − sin (47.107)

Therefore, our lead controller is:

√6.4802 𝑠 + 47.107
𝑙𝑒𝑎𝑑 =
𝑠 + 119.9163

As we can see the phase margin is close to 45

However, after adding the whole controller to our system:

Therefore, our controller 𝐶(𝑠) is:

10𝑠 + 47.107 √6.4802 𝑠 + 47.107

𝐶(𝑠) = 𝐶𝑝 ∗ 𝑙𝑎𝑔 ∗ 𝑙𝑒𝑎𝑑 = 41.667 ∗ ∗
10𝑠 𝑠 + 119.9163
1601𝑠 3 + 2.462 ∗ 10−4 𝑠 2 + 9.246 ∗ 10−4 𝑠
𝐶(𝑠) =
10𝑠 + 1199
Now we need to check the complementary sensitivity of the system respectively, to do this
we plotted each of their bode diagrams:

The following time responses are also plotted for d-u and d-theta, respectively below, as
requested:
Question 5

Let us find the transfer function of the system described in the question i.e. from current
𝑖(𝑡) to angle 𝜙(𝑡) for 𝜉 → ∞, 𝜃 → 0, 𝐴 = 0
𝜙 𝑁𝑔 𝑘𝑚 𝐵 𝑁𝑔𝑘𝑚
𝑃𝑖𝜙 = = =
𝑖 𝑠 (𝐽𝑏 + 𝑚𝑟 2 )𝑠 2 + 𝑓𝑠

However, we need system in the time domain to create our equations of motion:

(𝐽𝑏 + 𝑚𝑟 2 )𝜙̈(𝑡) + 𝑓𝜙̇(𝑡) = 𝑁𝑔 𝑘𝑚 𝑖(𝑡)

If we set:

𝐽𝑏 + 𝑚𝑟 2 𝑁𝑔 𝑘𝑚
𝜏= 𝑎𝑛𝑑 𝑘 =
𝑓 𝑓

We can simplify the equation and find

𝜏𝜙̈(𝑡) + 𝜙̇(𝑡) = 𝑘𝑖(𝑡)

This allows us to plug in our boundary conditions of:

𝜙(0) = 0 𝑎𝑛𝑑 𝜙̇(0) = 0 𝑎𝑛𝑑 𝜙(𝑡𝑓 ) = 𝜋 𝑎𝑛𝑑 𝜙̇(𝑡𝑓 ) = 0

Applied to the system:

𝑘𝑎 ((𝐽𝑏 + 𝑚𝑟 2 )𝑠 + 𝑓)
𝑃𝑢−𝑖 (𝑠) =
𝑁𝑔2 𝑘𝑚
2
𝑅 (((𝐽𝑏 + 𝑚𝑟 2 )𝑠 + 𝑓) + 𝑅 )

Now let’s set up and find the switches using the bang bang controller format, for our plant
with only 2 poles, from there the control signal can be found:

𝑡𝑠𝑤 ∈ (0, 𝑡𝑓 )

for

𝑖(𝑡) = 𝑖_ max (1(𝑡) − 2 ∗ 1(𝑡 − 𝑡𝑠𝑤 ) + 1(𝑡 − 𝑡𝑓 ))

And after applying Laplace:

1 − 2𝑒 −𝑠𝑡𝑠𝑤 + 𝑒 −𝑠𝑡𝑓
𝑖(𝑠) = 𝑖𝑚𝑎𝑥 ( )
𝑠

Therefore, using the formulas we learned in class:

|𝜃𝑓 −𝜃0 |
|𝜃𝑓 − 𝜃0 | √
𝑡𝑠𝑤 = + 𝜏 ∗ ln (1 + 1 − 𝑒 𝜏𝑘𝑖𝑚𝑎𝑥 ) = 0.7794
𝑘 ∗ 𝑖𝑚𝑎𝑥

|𝜃𝑓 −𝜃0 |
|𝜃𝑓 − 𝜃0 | √
𝑡𝑓 = + 2 ∗ 𝜏 ∗ ln (1 + 1 − 𝑒 𝜏𝑘𝑖𝑚𝑎𝑥 ) = 1.2731
𝑘 ∗ 𝑖𝑚𝑎𝑥
After substituting back into 𝑖(𝑠), and as we saw in the step repose of 𝑢 → 𝑑 we have already
found the optimal 𝑖(𝑡), to find the optimal voltage 𝑢(𝑡), we derive the transfer function:
𝑖 𝑘𝑎
𝑃𝑢𝑡 = =
𝑢 𝑅 + 𝑁𝑔2 𝑘𝑚 𝑘𝑏 𝐵

Therefore,

𝑢 𝑅 + 𝑁𝑔2 𝑘𝑚 𝑘𝑏 𝐵
𝑢(𝑠) = 𝑖(𝑠) ∗ = 𝑖(𝑠) ∗
𝑖 𝑘𝑎
Therefore, for interest, the optimal current 𝑖(𝑠) is graphed below:

Therefore, the optimal voltage 𝑢(𝑡) is graphed below:

Now let us find the optimal angle 𝜙(𝑡) and the 𝑡𝑓 that it reaches

𝑁𝑔𝑘𝑚
𝜙(𝑠) = 𝑖(𝑠) ∗ 𝑃𝑖𝜙 (𝑠) = 𝑖(𝑠) ∗
(𝐽𝑏 + 𝑚𝑟 2 )𝑠 2 + 𝑓𝑠

Using inverse Laplace, the optimal angle 𝜙 in the time domain is graphed below:
Where 𝜙 = 𝜋 is reached in 𝑡𝑓 = 1.2731 [𝑠𝑒𝑐]

Question 6

In order to stimulate the response of 𝑢(𝑡), 𝜙(𝑡)𝑎𝑛𝑑 𝜃(𝑡), we must first find the transfer
functions 𝑃𝑢→𝑟𝜙 , 𝑃𝜙→𝑟𝜙 and 𝑃𝜃→𝑟𝜙 :

𝑢(𝑠) 𝐶𝜃
𝑃𝑢→𝑟𝜙 = =
𝑟𝜙 (𝑠) 𝐶𝜃 𝑃𝜙𝑢 + 1

𝜙(𝑠) 𝐶𝜃 𝑃𝜙𝑢
𝑃𝜙→𝑟𝜙 = =
𝑟𝜙 (𝑠) 1 + 𝐶𝜃 𝑃𝜙𝑢

𝜃(𝑠) 𝐶𝜃 𝑃𝜙𝑢
𝑃𝜃→𝑟𝜙 = = 𝑃𝜃𝜙 (𝑠)𝐶𝜃 𝑃𝜙𝑟 (𝑠) = 𝐴
𝑟𝜃 (𝑠) 1 + 𝐶𝜙 𝑃𝜙𝑢

We can further derive the frequency responses of 𝑢, 𝜙 and 𝜃 and then convert them to time
domain to achieve the following figures:

𝑢(𝑠) = 𝑟𝜙 (𝑠)𝑃𝑢−𝑟𝜙 ; 𝜙(𝑠) = 𝑟𝜙 (𝑠)𝑃𝜙−𝑟𝜙 ; 𝜃(𝑠) = 𝑟𝜙 (𝑠)𝑃𝜃−𝑟𝜙

Part 4
Question 7

Considering the following figure below, describing a control setup of mind positive feedback,
we will design a minimal order feedback controller 𝐶𝜃 (𝑠) such that:

• The pole excess of 𝐶𝜃 (𝑠) is at least 2,

• |𝐶𝜃 (𝑗𝑤)| ≤ 1 for all values of 𝜔
• The response to 𝜙(𝑡) = 𝟙(𝑡)of the closed-loop system in the figure below verifies
|𝜃(𝑡)| ≤ 1° for all 𝑡 ≥ 2 sec

To construct a controller with a minimum pole excess of 2, we selected two LPF controllers
and their poles were selected according to their crossing frequency, and their bandwidth.
Using the Nichols diagram of the plant, P, and we want to move away from the critical
points.

Now we look at the Nichols plot of the negative of the plant. because we would like to place
Nichols between the 2 critical points. We want the critical phase as not to affect the
amplification at the resonant frequency that is close to it:
We will use the critical point frequency to use the low frequency filter to stay away from the
1
critical points so that our initial controller will be: 𝐶 = (𝑠+4.27)2 and we will get the following
Nichols plot of the loop:

Step response of 𝜃
Question 8

From Fig 2(c) and question 8, we can infer that:

𝑇𝜙 (𝑠)𝑃𝜃𝜙
𝑃𝜃𝑟 =
1 + 𝑇𝜙 (𝑠)𝑃𝜃𝜙 𝐶𝜙

Where:
𝑃𝜙𝑢 (𝑠)𝐶𝜙
𝑇𝜙 =
1 + 𝑃𝜙𝑢 𝐶𝜙
Unfortunately we did not manage to successfully combine the controllers and couldn’t
provide the right step responses and Nichols chart.