20/03/2020

Dynamic behavior of first-order and

second-order processes

DINAMICA E CONTROLLO DEI PROCESSI CHIMICI

A.A. 2019-2020
Corso di Laurea Magistrale in Ingegneria di Processo e dei Materiali

ANDREA MIO, PhD

MolBNL @ DIA-UniTS; www.molbnl.it
[email protected] - 0405583440

80

THE INPUTS TO A DYNAMIC SYSTEM

▰ To design a process control system, the dynamic system (i.e., the process)
response to the input needs to be known INPUT

▻ i.e., how the system responds to any input variable change that can
affect the output (specifically, the target variable to control)
▻ The TF is the tool allowing to evaluate the effect that different inputs
transfer on the output
MANIPULATED DISTURBANCES
▻ The adoption of TFs allows the generalization of the description of
the dynamic behavior of a process

Output input 1 input 2

𝑋 𝑠 = 𝐺 𝑠 𝑋 𝑠 + 𝐺 (𝑠)𝑋 ′(𝑠)

81
© Andrea Mio, University of Trieste

81

1
 20/03/2020

ANALYSIS OF THE DYNAMICS OF A SYSTEM

▰ In order to study the study dynamics we will follow the following procedure:
1. We perturb the system applying a variation (forcing function) on each input (𝑢 ) individually, keeping any existent
control loop open
 In case of multiple input, they are changed alternatively one at a time

2. We collect the observation on the dynamic response of the output y

3. We extract general information about the steady-state and dynamic behavior of the system
4. We exploit this information to design the controller, i.e. to determine the closed-loop response of the system

𝑢 y

PROCESS
t t
© Andrea Mio, University of Trieste

82

LIST OF PERTURBATIONS

REAL EXAMPLES FORCING FUNCTION LAPLACE TRANFORM

𝑢

STEP – S(t) M A valve whose opening 0 𝑡<0 𝑀

is changed in a very
𝑢 𝑡 = 𝑈 𝑠 =
𝑀 𝑡≥0 𝑠
t short time interval

𝑢
A valve whose 0 𝑡<0
RAMP – R(t) opening is changed 𝑢 𝑡 = 𝑎
𝑎𝑡 𝑡 ≥ 0 𝑈 𝑠 =
at a constant rate 𝑠
t

𝑢 A valve whose opening 0 𝑡<0 ℎ

RECTANGULAR is istantaneously 𝑢 𝑡 = ℎ 0≤𝑡<𝑡 𝑈 𝑠 =
𝑠
(1 − 𝑒 )
h 0 𝑡≥𝑡
PULSE increased for a time
t interval and set back to
tw initial value at the end
© Andrea Mio, University of Trieste

83

2
 20/03/2020

LIST OF PERTURBATIONS /cont’d

𝑢 REAL EXAMPLES FORCING FUNCTION LAPLACE TRANFORM

1
A
SINUSOIDAL – sin(t) 0 Daily variation of 𝑢 𝑡 =
0 𝑡<0 𝐴𝜔
𝐴 sin 𝜔𝑡 𝑡 ≥ 0 𝑈 𝑠 =
-1 cooling water 𝑠 +𝜔
t temperaure
T

𝑢 Difficult to obtain; it 0 𝑡<0

∞ 𝑢 𝑡 = ∞ 0≤𝑡<𝑡
can be approximated
UNIT PULSE – UP(t) by a tracer injection
0 𝑡≥𝑡 𝑈 𝑠 =1
for RTD studies
t With 𝑡 → 0

𝑢
RANDOM Usually characterized
M
with statistical term
0
(mean, st.dev.)
-M t
© Andrea Mio, University of Trieste

84

FIRST-ORDER SYSTEMS

▰ They are constituted by a 1°-order differential equation with 𝑎 ≠ 0: Notice: the speed of
𝑎 𝑑𝑦 𝑡 𝑏 𝑑𝑦(𝑡) variation of the output
+𝑦 𝑡 = 𝑢 𝑡 → 𝜏 + 𝑦 𝑡 = 𝐾𝑢(𝑡) depends on the current
𝑎 𝑑𝑡 𝑎 𝑑𝑡 output value itself!
Called: first-order lag (in deviation variable terms)

𝑌(𝑠) 𝐾
𝐺 𝑠 = = ▻ K: steady-stete gain of the system [output u.m./input u.m.]
𝑈(𝑠) 𝜏𝑠 + 1
▻ 𝜏 > 0: time constant of the system [time]

▰ Assume that u(t) is at steady-state and at time t=0 a perturbation occurs

▰ We need to take into account two different types of perturbations
𝑀
▻ Step input change: 𝑈 𝑠 =
𝑠
𝐴𝜔
▻ Sinusoidal input change: 𝑈 𝑠 =
𝑠 +𝜔 85
© Andrea Mio, University of Trieste

85

3
 20/03/2020

0 𝑡<0
STEP RESPONSE OF FO SYSTEMS 𝑢 𝑡 =
𝑀 𝑡≥0

𝐾 𝑀 𝐾𝑀 𝑦 =0
𝐺 𝑠 = ; 𝑌 𝑠 =𝐺 𝑠 𝑈 𝑠 ; 𝑈 𝑠 = → 𝑌 𝑠 =
𝜏𝑠 + 1 𝑠 𝑠(𝜏𝑠 + 1) 𝑦 = 𝐾𝑀

Upon inverse transform: 𝑦 𝑡 = 𝐾𝑀 1 − 𝑒

▻ The process starts responding to the perturbation instantaneously

▻ The speed of change y is maximum at t=0, then it progressively decreases to 0

▻ A new steady-state is attained only as 𝑡 → ∞

▻ It only takes a time equal to 4-5 time constants for the transient to die out

▻ 𝜏 provides an indication on the speed of response: ↑ 𝜏 → slow response

▻ K indicates how far the new steady state value of y is from the original one:
↑ K → ↑ amplification of input variation onto the output 86
© Andrea Mio, University of Trieste

86

0 𝑡<0
RESPONSE OF FO SYSTEMS TO A SINUSOIDAL 𝑢 𝑡 =
𝐴 sin 𝜔𝑡 𝑡 ≥ 0

𝐾 𝐴𝜔 𝐾𝐴𝜔
𝐺 𝑠 = ; 𝑌 𝑠 =𝐺 𝑠 𝑈 𝑠 ; 𝑈 𝑠 = → 𝑌 𝑠 =
𝜏𝑠 + 1 𝑠 +𝜔 (𝜏𝑠 + 1)(𝑠 + 𝜔 )
𝐾𝐴𝜔𝜏 𝐾𝐴
Upon inverse transform: 𝑦 𝑡 = 𝑒 + sin(𝜔𝑡 + 𝜙) 𝜙 = tan −𝜔𝑡 =phase angle [rad]
𝜔 𝜏 +1 𝜔 𝜏 +1
▻ An initial transient exists, which dies out after a few 𝜏

▻ Then the response is purely sinusoidal (periodic), as the input perturbation

▻ The amplitude of the response sinusoid is different from that (A) of the forcing function;
It depends on:
 The process: K, 𝜏
 The forcing function, namely its angular frequency 𝜔
(hence on frequency f or period 𝑇 = 2𝜋⁄𝜔 = 1⁄𝑓 )
▻ The system response lags behind the input
 The phase shift 𝜙 depends on the dynamic characteristics of the system (𝜏) and of
those of the forcing function (𝜔) 87
© Andrea Mio, University of Trieste

87

4
 20/03/2020

EFFECT OF FREQUENCY

𝑢 𝑡 = 𝐴 sin(𝜔𝑡)
𝐾𝐴𝜔𝜏 𝐾𝐴
𝑦 𝑡 = 𝑒 + sin(𝜔𝑡 + 𝜙) 𝜙 = tan −𝜔𝑡 =phase angle [rad]
𝜔 𝜏 +1 𝜔 𝜏 +1

▰ What’s the influence of the input frequency f (hence its angular frequency ω) on the response?
▰ The ratio between the response amplitude and the input amplitude is given by:
 ↑ 𝑖𝑛. 𝜔 →↓ 𝑜𝑢𝑡. 𝑎𝑚𝑝𝑙
𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝐾  ↑↑↑ 𝑖𝑛. 𝜔 → 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 flattens, i.e. the input
Amplitude ratio 𝐴𝑅 = = is filtered by the process
𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 𝑖𝑛𝑝𝑢𝑡 𝜔 𝜏 +1  The same holds true if the process has a ↑↑↑ 𝜏

▰ The phase shift (lag) 𝜙 between the sine waves is maximum (−𝜋⁄2𝑟𝑎𝑑 ) when the input frequency is very large
▰ The input frequency can be denoted as «large» of «smalle» depending on the dynamic characteristic of the process (𝜏)
What matters is the product 𝜔𝜏
88
© Andrea Mio, University of Trieste

88

EXAMPLES

▰ 4° ed., Example 5.1, p. 71; 3° ed, 5.1 p.78

𝑘𝑔
𝑇; 𝑤 [ ]
ℎ𝑟

𝑘𝑔
𝑇; 𝑤[ ]
ℎ𝑟

89
© Andrea Mio, University of Trieste

89