Lecture 9

PID Control 1/3

Shanying Zhu
Department of Automation, SJTU
March, 2017
▪ Control systems
Problems with open loop systems
▪ They fly “blind”

▪ Cannot respond to disturbances

▪ Cannot adjust to different plants

Open loop CNC ▪ Models may be difficult or impossible to
derive
VS

▪ Add feedback so controller knows

results of actions.

▪ But, how to utilize these information

to design a desirable controller?

▪ PID controller! Closed-loop CNC

Outline

1 Introduction

2 Analog PID Controller

3 Digital PID Controller

4 PID Controller Tuning

5 Summary
A Brief History of PID Control
▪ 1890’s, PID (Proportional - Integral – Derivative) Control,
originally developed in the form of motor governors, which
were manually adjusted

▪ 1922, the first theory of PID Control was published by Nicolas

Minorsky, who was working for the US Navy

▪ 1940’s, the first papers regarding PID tuning appeared

▪ there are several hundred different rules for tuning PID
controllers (See Dwyer, 2009) Nicolas Minorsky
(1885-1970)
▪ Nowadays, 97% of regulatory controllers utilize PID feedback a Russian American
control theory
▪ based on a survey of over eleven thousand controllers in the mathematician,
refining, chemicals and pulp and paper industries (see engineer and applied
scientist
Desborough and Miller, 2002).
PID Control
Pros:
▪ Process independent
▪ The best controller where the specifics of the process can not be modeled
▪ Leads to a “reasonable” solution when tuned for most situations
▪ Inexpensive: Most of the modern controllers are PID
▪ Can be tuned without a great amount of experience required

Cons:
▪ Not optimal for the problems
▪ Can be unstable unless tuned properly
▪ Not dependent on the process
▪ Hunting (oscillation about an operating point)
▪ Derivative noise amplification
Ways to Implement PID Control
Analog PID:
▪ Receives a measured process variable y(t) using an electronic controller;
▪ Compares this value with that of a desired setpoint signal;
▪ Calculates an error value e(t) as the difference between the setpoint
signal and process variable in a PID control circuit;
▪ The correction signal u(t) is then sent to the actuator to apply a
correction.

setpoint w + e
PID Controller
u y process variable
Plant
(output)
-
Digital PID:
▪ Computer/Microcontroller aided;
▪ The computer registers the process variable y(t) via an AD converter, and
produces a numerical value y(k);
▪ Calculates an error value e(k) as the difference between the setpoint signal
and process variable in a discrete-time PID control circuit;
▪ The correction signal u(k) is then sent to the DA converter producing u(t),
followed by the actuator to apply a correction.

process variable
setpoint w + e Discrete-time u y
DAC Plant (output)
PID Controller
-

ADC
Today’s Lecture- PID/1

1 Introduction

2 Analog PID Controller

3 Digital PID Controller

4 PID Controller Tuning

5 Summary
Today’s Lecture- PID/1

2 Analog PID Controller

 PID control overview
 P control
 PI control
 PD control
 Simulation results
 Summary
Analog PID Controller
▪ Block diagram of a PID controller
setpoint control error control law output
P Kp
+
w(t) + e(t) + u(t) y(t)
I K i/s Plant
- +
D K ds

▪ Textbook form

: integration/reset time
: derivative time

P I D
P Control
▪ Proportional control (P): accounts for present values of the error

— control signal
e(t)
— proportional gain （
1

0 t
— error signal t0

▪ In the Laplace domain u(t)

K pe
0 t
t0
▪ Pros&Cons Step response for P control
▪ Rapid response to track the error signal
▪ Steady-state error
▪ Prone to be unstable for large Kp

▪ Proportional control is always present, either by itself, or allied with derivative

and/or integral control
▪ Integral control (I): accounts for past values of the error

e(t)

1
— control signal
0 t
t0
— integral gain

— error signal u(t)

)
（t
▪ In the Laplace domain Kp
）
0 t
t0
Ti

Step response for I control

▪ Pros&Cons
▪ Eliminates the steady-state error that occurs with pure P control
▪ Prone to cause the present value to overshoot the setpoint (responds to accumulated
errors from the past)
PI Control
▪ The I control action is rarely used by itself, but is coupled with proportional (P)
action for PI controller.

▪ Proportional-Integral control (PI): a combination of P and I control

▪ In the Laplace domain

▪ Derivative control (D): accounts for possible future trends of the error
e(t)

1
— control signal
0 t
— derivative gain t0
— error signal
u(t)
▪ In the Laplace domain
0 t
t0
Step response for D control
▪ Pros&Cons
▪ Predicts system behavior and thus improves settling time/transient response and
stability of the system
▪ Helps reduce overshoot, but amplifies noise (derivative kick)
▪ Seldom used in practice, 80% of the employed PID controllers have the D part
switched-off (see Ang et al., 2005)
PD Control
▪ Proportional-Derivative control (PD): a combination of P and I control

▪ Take Td as a step size, then

▪ D control action is able to predict system behavior and thus improving settling
time/transient response.
Effect of P-I-D Control
▪ We will examine effect of PID control on a canonical 2nd order system to
gain insight.
W + E U Y
Controller (C) Plant (P)
Setpoint: unit step signal
-
Sensor

Plant: 2nd order system

Transfer function:

Controller: P-I-D

Error signal:
P Control
▪ Effect on steady-state performance
Steady-state error for a unit step reference

▪ Final-value theorem

y(t)

Steady-state error • Steady-state error always occurs;

1 • Larger Kp makes steady state error goes to zero
t
PI Control
▪ Effect on steady-state performance
Steady-state value for a unit step reference
▪ Final-value theorem

▪ Steady-state error is zero for a step reference, even for small Ki (just takes
longer to reach steady state).
y(t)

1
t
PD Control
▪ Effect on steady-state performance
Steady-state error for a unit step reference
▪ Final-value theorem

▪ Steady-state error remains the same as the steady-state error with

pure P control. Indeed, D control does not track error, only the rate of change of it.
▪ No significant value added by including the D control with respect to the steady-
state performance (transient performance probably differs).
PID Control
▪ Proportional integral derivative control (PID): a combination of P, I and D
control

0 t
t0
▪ Effect on steady-state performance
y

Kp

Therefore, steady-state error is zero for a step Kp

0 t
reference. It can be used to control the response t0 Ti

characteristics better than the other types of Step response for PID control

controllers, e.g., P, PI, PD. Nevertheless, more

complex to tune the parameters.
Simulation using MATLAB
▪ A canonical 2nd order system ▪ MATLAB code

Setpoint: unit step signal %plant

clc; clear all; close all;
Plt = tf(1,[10,7,1]); %transfer function
Plant: 2nd order system %PID control:
sys = feedback(C*Plt,1); %feedback
connection
step(sys); %unit step response
Controller: PID
 Kp=10

Kp=5
setpoint

Kp=1

Kp=0.1

%P control  Kp increases, the response speed

Plt = tf(1,[10,7,1]); of the system increases, the
Kp = [0.1,1,5,10]; overshoot of the closed-loop
for k = 1:4
system increases, and the steady-
sys = feedback(Kp(k)*Plt,1);
step(sys),hold on
state error decreases.
end  Kp large enough, the closed-loop
system becomes unstable
 Ti=2

Ti=6

setpoint

Ti=12
Ti=24

%PI control:
Plt = tf(1,[10,7,1]);
 No steady-state error in the step
Kp = 2; Ti = [2,6,12,24]; response
for m = 1:4  Ti increases, the overshoot tends
Cpi = tf([Kp,Kp/Ti(m)],[1,0]); to be smaller, but the speed of
sys = feedback(Cpi*Plt,1); response tends to be slower.
step(sys); hold on;
end
 Td=0

Td=0.4 setpoint

Td=1
Td=4

%PD control:
Plt = tf(1,[10,7,1]);  Td increases, the response has a
Kp = 10; Td = [0,0.4,1,4]; smaller overshoot with a slightly
for m = 1:4
slower rise time but similar settling
Cpd = tf([Kp*Td(m),Kp],[0,1]);
sys = feedback(Cpd*Plt,1); time
step(sys); hold on;
end
▪ Some intuition about effects of the terms:
▪ Increasing Kp: Same amount of error generates a proportionally larger
amount of control, makes system faster, but overshoot more (less stable)
▪ Increasing Ki: Control effort builds as error is integrated over time, helps
reduce steady state error, but can be slow to respond
▪ Increasing Kd: Allows controller to anticipate an increase in error, adds
damping to the system (reduces overshoot), can amplify noise

• These guidelines do not hold for

all situations.
• For systems that are not
canonical first or second order,
need to use trial and error.
Practical Modifications of PID controllers
▪ Textbook form

P I D

▪ Seldom used in practice because of a few problems arise leading to poor

practical performance.

▪ Modifications:
▪ P part: setpoint weighting
▪ I part: anti-windup
▪ D part: setpoint weighting and limited gain
Summary
▪ The controller performs the PID mathematical functions on the error and
applies their sum to a process.
▪ We can build a PID controller that works well in practice in most situations
without knowing control theory.

Math Function Effect on Control System

P Typically the main drive in a control loop, Kp reduces a
Proportional large part of the overall error.
Reduces the final error in a system. Summing even a small
I
error over time produces a drive signal large enough to
Integral move the system toward a smaller error.

Counteracts the Kp and Ki terms when the output

D
changes quickly. This helps reduce overshoot and ringing.
Derivative No effect on final error.
References
▪ Genke Yang, and Jianying Xie, Micro-Computer Control Technology,
4th ed. Changsha: National Defense Industry Press, 2016 (in Chinese).
▪ K. J. Åström, and T. Hägglund, PID Controllers: Theory, Design, and
Tuning, 2nd ed. Research Triangle Park, NC: Instrument Society of
America, 1995.
▪ A. O’Dwyer, Handbook of PI and PID Controller Tuning Rules, 3rd
ed. London: Imperial College Press, 2009.
▪ L. Desborough, and R. Miller, Increasing customer value of industrial
control performance monitoring—Honeywell’s experience. AIChE
Symposium Series, vol. 326, pp.158-186, 2002.
▪ K. H. Ang, G. Chong, and Y. Li, PID control system analysis, design,
and technology, IEEE Transactions on Control Systems Technology,
vol. 13, no. 4, pp.559-576, 2005.
 Thanks for your attention!

E-mail: [email protected]

Wechat ID: S14528707