Kuis 1 (13/11/2015) = part 1
Kuis 2 (24/11/2015) = part 1 + part 2
Kuis 3 (04/12/2015) = part 1 + part 2 + part 3
part 1
: Fundamental of Control Theory (Basic Understanding)
Brian Videos
a1 - Why Learn Control Theory
a2 - Control Systems Lectures - Closed Loop Control
a3 - Control Systems Lectures - Time and Frequency Domain
a4 - Control Systems Lectures - LTI Systems
a5 - Modeling Physical Systems, An Overview
a6 - System Identification Methods
a7 - Introduction to System Stability and Control
a8 - Stability of Closed Loop Control Systems
part 2
: Root locus ( Concept and Hand Drawing)
Brian Videos
b1 - The Root Locus Method Introduction
b2 - Sketching Root Locus Part 1
b3 - Sketching Root Locus Part 2
b4- Root Locus Plot- Common Questions and Answers
: PID ( Gain estimation/adjustment)
Brian Videos
c1 - PID Control - A brief introduction
c2 - Simple Examples of PID Control
c3 - Standard HW Problem #1_ PID and Root Locus
part 3
Final Score = best of part 1 + best of part 2 + best of part 3 + UAS
Try sketching root locus plot for following open-loop TF
PID controller
PID: proportional,
integral and derivative
Most commonly found in
industry
>90% controller in
industry is PID controller
Simple to use, can
perform various control
tasks
The effect of Kp, Ki, and Kd
Effects of Ki with others held constant
Effects of Kp with others held constant
Kp is a dominant quantity that normally has
some nonzero value
Ki helps eliminate steady state error but too
high value could introduce overshoot and
oscillation.
Kd could help the response to reach steady
state faster but could amplify high frequency
noise, and could affect stability if set too high.
Effects of Kd with others held constant
6
PID summary
Effects of increasing a parameter independently
Parameter
Rise time
Overshoot
Steady-state error
Kp
Decrease
Increase
Decrease
Ki
Decrease
Increase
Eliminate
Kd
Minor change
Decrease
No effect in theory
PID tuning with xcos/Scilab
Call xcos from the console
Construct the following block diagram
Simulation > setup >final integration time: 10
Setting CSCOPE:
- Refresh period: 10
- Ymin and Ymax accordingly
PID tuning with xcos/Scilab (2)
3
2.5
1.5
0.5
-0.5
-1
0
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Can you manually tune Kp, Ki, and Kd to get the
optimum shape?
- Not easy, without proper experience
- One way to aid the tuning: Ziegler-Nichols
Frequency Domain (ZNFD) method
Ziegler-Nichols method
Turn off both Ki, and Kd. So now the PID is left only
with the proportional gain Kp. We adjust Kp up to the
point that the closed-loop system starts to oscillate. At
this point, the output will swing in a constant sinusoid
motion, not growing and not dying out. Write this
value down on a paper as Ku.
Measure the period of oscillation. Write it down as Tu.
Use Table below to tune the controller parameters
Controller form
Kp
Ki
Kd
o.5Ku
PI
o.4Ku
o.5Ku/Tu
PID
o.6Ku
1.2Ku/Tu
0.075KuTu
10
Ziegler-Nichols method (2)
Obtained Ku = 333
which gives: ->
Tu= 1 sec
From the table:
Kp = 200
Ki = 400
Kd = 42
With Kp = 200; Ki = 400; Kd = 42
11
Can still adjust Ki to decrease overshoot
Industrial PID Controller
A box, not an algorithm
Auto-tuning functionality:
pre-tune
self-tune
Manual/cascade mode
switch
Bumpless transfer between
different modes, setpoint
ramp
Loop alarms
Networked or serial port
12
PID and root locus: sample problem
A diesel engine is controlled by a PID controller. It has following block
diagram. Determine the range of A for which the system is stable
Characteristic equation:
BUT not in a typical format of root locus equation -> need to rearrange
13
PID and root locus: sample problem
An open loop TF with 3 zeros
s = 0, -20.5 9.5j
Asymptotes:
r=3 -> 3 asymptotes
intersection: -13.7
Breakaway points?
Can calculate stability requirement of A
with Routh-Hurwitz
14
PID and Routh Hurwitz sample problem
Use the Zeigler Nichols method to tune a PID-controller for a
cruise control system applied in a road vehicle. Assume the
road vehicle has a unity feedback with TF as follows. And
assume that when the system oscillates with Kp only, the
period is 1 second.
Controller form
Kp
Ki
Kd
o.5Ku
PI
o.4Ku
o.5Ku/Tu
PID
o.6Ku
1.2Ku/Tu
0.075KuTu
15
PID and Routh Hurwitz sample problem
Since we need to first tune Kp so that the system oscillates, Ki
and Kd = 0
With Routh-Hurwitz:
Row 1:
82
Row 2:
10
10+40K
Row 3:
80-8K
Row 4:
10+40K
Hence we get that Ku=10 will get
the system to start oscillates
Then finally we can
get Kp, Ki, and Kd
from the table
16
Transient response analysis
Time response specs
Maximum
overshoot
ymax(t) - yss
Settling time
Time for y(t) to
settle within 5%
of yss
Rise time
Delay time
17
Routh-Hurwitz
Why we need it? because most of the time,
finding the roots of a TF is not an easy task
Purpose: to evaluate the stability of a system
by determining the location of the roots (of a
given transfer function)
It does NOT explicitly compute the roots
This method is strictly for LTI (linear timeinvariant) system with a polynomial
denominator (without sin, cos, exponential,
etc)
More useful for design purpose
18
Routh-Hurwitz (2)
Given a TF with characteristic equation/denominator
We can construct a table:
The number of sign changes in the
first column of Routh array =
the number of roots in the right
half-plane
19
Routh-Hurwitz (3)
Example #1:
TF:
Denominator: 5s3 - 10s2 + 7s + 20
Routh-Hurwitz table:
Row 1:
Row 2:
-10
20
Row 3:
17
Row 4:
20
Change of signs: 2 times
2 poles on RHP
Unstable
Evaluate the stability of characteristic
equation s4+10s3+35s2+50s+24 with
Routh-Hurwitz method
20
Routh-Hurwitz (4)
Example #2: Routh-Hurwitz for parameter design
Characteristic equation: s3+3Ks2+(K+2)s+4
Determine range of K for which the system is stable
Row 1:
K+2
Row 2:
3K
Row 3:
[3K(K+2)-4]/3K
Row 4:
For the system to be stable, the first column has
to be all positive:
3K > 0
3K(K+2)-4 > 0
21
Routh-Hurwitz (5)
Try for yourself!
Example #3: find the range of K that stabilizes
the system
Answer: 0 < K < 35
22
Sample problem: finding stable K with Routh-Hurwitz
Let's take another example:
Plot open-loop poles (and zeros)
Intersection of asymptote:
With r=3 -> 3 lines go to infinity
23
Sample problem: finding stable K with Routh-Hurwitz (2)
Compute breakaway point
For each s, check the positivity of K
24
Sample problem: finding stable K with
Routh-Hurwitz (3)
Evans root locus
25
asymptotic directions
open loop poles
20
15
10
Imaginary axis
-5
-10
-15
-20
-25
-30
-25
-20
-15
-10
Real axis
-5
10
15
25
Sample problem: finding stable K with Routh-Hurwitz (4)
Row 1:
Row 2:
Row 3:
(30-K)/6
Row 4:
Stability condition:
0<K<30
Can you calculate the s value when K is 30?
26
Trick of scilab to determine K
Click any spot on the root
locus and the K value will be
determined
27
With complex poles
For poles/zeros at real axis, angle of departure must be
0o or 180o
28
Additional rule: angle of departure
Evans root locus
asymptotic directions
open loop poles
open loop zeroes
Imaginary axis
0.5
-0.5
-1
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
1.2
1.4
Real axis
29
Recap
-
L(s)
With Scilab:
evans(L)
plzr(L) -> for K=0 (no K or
open loop TF)
plzr(KL/1+KL) -> for K>0
Can you plot root locus for
following system with Scilab?
characteristic equation: used for
most root locus calculation
L(s)
F(s)
30
