Scribd
Upload
26 views

Process Control - Chapter 2 - Continued

This document discusses methods for linearizing nonlinear models and identifying parameters of linear models from process data. It introduces Taylor series expansion to linearize nonlinear models around a steady state. It also presents various methods to estimate parameters of first-order and first-order plus time delay models from step response experimental data, including determining gain, time constant and time delay. Skogestad's 'half rule' is introduced for approximating higher order systems with lower order models.

Uploaded by

Huy Nguyễn Hoàng
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
26 views

Process Control - Chapter 2 - Continued

This document discusses methods for linearizing nonlinear models and identifying parameters of linear models from process data. It introduces Taylor series expansion to linearize nonlinear models around a steady state. It also presents various methods to estimate parameters of first-order and first-order plus time delay models from step response experimental data, including determining gain, time constant and time delay. Skogestad's 'half rule' is introduced for approximating higher order systems with lower order models.

Uploaded by

Huy Nguyễn Hoàng
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

2.

4 Linearization of nonlinear models


 (𝑥𝑠 , 𝑢𝑠 ) are the steady-state of the system, we have:
x  f ( xs , us )  0 (2.50)
 The equation describing the system is differentiable at
the steady state

A linear approximation of this equation can be obtained


by using a Taylor series expansion and truncating after
the first-order terms
Given
 x  xs  x
 (2.51)
u  us  u
2.4 Linearization of nonlinear models

x  x  f ( xs  x, us  u )
f f
 f ( xs , us )  x  u (2.52)
x xs ,us u xs ,us

Example 4:
Fi ,ρ
Determine the mathematic
model of the system, given
the outlet flow rate
h
Fo ,ρ Fo (h)  C h
(2.53)
2.4 Linearization of nonlinear models
Similarly to the example 1, we have the differential equation:
dh
A  Fi  Fo (h) (2.54)
dt
Because of the nonlinearity of the right term, we could
not apply the Laplace transforms. Then using Taylor
expansion at the steady-state

Fo (h)  C h  Fo  Fo (hs )  Fo (hs )[h  hs ]


1 C
 Fo (h)  C hs  [h  hs ] (2.55)
2 hs
2.4 Linearization of nonlinear models
Replace eq. (2.55) into (2.54)
dh 1 C
A  Fi  C hs  [h  hs ] (2.56)
dt 2 hs
At the steady-state, we have:
Fis  C h s (2.57)
From Eq. (2.56) & (2.57)
dh 1 C
A  Fis  Fi  [h  hs ]
dt 2 hs
dh 1 C
A  [h  hs ]  Fi  Fis (2.58)
dt 2 hs
2.4 Linearization of nonlinear models
Given H  h  hs , F  Fi  Fis
Eq. (2.58) is rewritten
d (h  hs ) 1 C
A  [h  hs ]  Fi  Fis
dt 2 hs
dH 1 C
A  H F (2.59)
dt 2 hs

Apply the Laplace transforms to eq. (2.59)


1 C 1 C
AsH ( s)  H ( s)  F ( s)  [ As  ]H ( s )  F ( s )
2 hs 2 hs
2.4 Linearization of nonlinear models
The transfer function of linearized model
2 hs
H (s) 1 C K
    (2.60)
F ( s) As  1 C 2 A hs  s 1
2 hs s 1
C
Where, 2 hs
K
C

2 A hs

C
2.5 Empirical models from process data
 Open-loop identification

Step input Process


Process output

 Closed-loop identification

Controller Process
Input Process
output

In this course, only open-loop identification is introduced.


Profound knowledge of system identification, students read
other related books
2.5.1 First-order models
The transfer function of the first-order system

u y

Y( s) K
G (s)   (2.61)
U( s)  s  1

 Step 1: Determine K

Y
K  (2.62)
U
2.5.1 First-order models
u,y

Y2
A
ΔY=Y2 – Y1
0.632ΔY

Y1

U1
ΔU=U2 – U1
U1

t1 τ t2
2.5.1 First-order models
 Step 2: Determine 𝝉
- Determine the point A at value 𝑌𝐴 = 0.632∆𝑌 on the
graph
- Determine the time corresponding with A on the time
axis, 𝑡2
- 𝝉 = 𝒕𝟐 − 𝒕𝟏 (2.63)
Example: Figure shows the response of the temperature
T in a continuous stirred-tank reactor to a step change in
feed flow rate w from 120 to 125 kg/min. Find an
approximate first-order model for the process and these
operating conditions
2.5.1 First-order models

Feed

Products
Coolant

- ∆𝑈 = ∆𝑤 = 125 − 120 = 5 (kg/min)


- ∆𝑌 = 𝑇 ∞ − 𝑇 0 = 160 − 140 = 20 (oC)
2.5.1 First-order models

Y 20 0
C
K  4
U 5 kg min
The temperature at the point of 0.632∆𝑌
T  140  0.632(20)  152.6 (oC )
Based on the graph, 𝑡2 = 5 (min)
Finally, τ = 𝑡2 − 𝑡1 = 5
The transfer function of the system
Y ( s) 4
G( s )  
U ( s ) 5s  1
2.5.2 First-order plus time delay models (FOPTD)
The transfer function of the FOPTD
e  s
G( s )  K . (2.64)
 s 1
- The process gain K is found similarly to the previous
method.
- A tangent is drawn at the point of inflection of the
step response; the intersection of this line and the
time axis is the time delay, 𝜽
- The tangent is extended to intersect the steady-state
line, the point of intersection corresponds to time:
t    (2.65)
2.5.2 First-order plus time delay models (FOPTD)
u,y

Y2

ΔY=Y2 – Y1
0.632ΔY

Y1
Inflection
point
U1
ΔU=U2 – U1
U1

t
t1 t2 t3
θ τ
2.5.2 First-order plus time delay models (FOPTD)
The major disadvantage of the above method is that it is
difficult to find the point of inflection.
The two points method was proposed
 From the step response curve we determine the times 𝑡1 ,
𝑡2 that correspond to the 28.3% and 63.2% the steady
state value of the response, respectively.
 Determine the parameters by equations
   1.5(t2  t1 ) (2.66)

  t2  
Another two points can be used: 35.3% and 85.3%, then
  0.67(t2  t1 ) (2.67)

  1.3t1  0.29t2
2.5.2 First-order plus time delay models (FOPTD)
u,y

Y2

ΔY=Y2 – Y1
0.632ΔY

Y1 0.283ΔY

U2
ΔU=U2 – U1
U1

t
t1 t2
2.5.3 Approximation for higher order systems

 Taylor series expansion

 s  2 s 2  3s3
e  1 s    ...
2! 3!
 e  s  1   s
(2.68)
or e s  1   s
 The 1/1 Padé approximation


1 s
e  s  G1 ( s )  2
(2.69)

1 s
2
2.5.3 Approximation for higher order systems

 Skogestad’s “Half Rule” (2003)


- Determine the largest neglected time constant
- One-half of its value is added to the existing time delay
- Other half is added to smallest retained time constant
- Time constants that are smaller than the largest
neglected time constant are approximated as time delay
using (2.68)
Example: Consider a transfer function
K (0.1s  1)
G (s) 
(5s  1)(3s  1)(0.5s  1)
2.5.3 Approximation for higher order systems

 Skogestad’s “Half Rule” (2003)


- Determine the largest neglected time constant
- One-half of its value is added to the existing time delay
- Other half is added to smallest retained time constant
- Time constants that are smaller than the largest
neglected time constant are approximated as time delay
using (2.68)
Example 1: Consider a transfer function
K (0.1s  1)
G (s) 
(5s  1)(3s  1)(0.5s  1)
2.5.3 Approximation for higher order systems

Derive an approximate first-order plus time delay model


Ke  s
G (s) 
 s 1
Using two methods:
a) The Taylor series expansion
b) Skogestad’s half rule
Homework: Using Matlab to compare the response of the
actual model and approximation models

Example 2: Consider a transfer function


2.5.3 Approximation for higher order systems
K (1  s) e  s
G (s) 
(12s  1)(3s  1)(0.2s  1)(0.05s  1)
Using Skogestad’s method to derive two approximate models:
a) A first-order plus time delay model
b) A second-order plus time delay model in the form:
Ke  s
G (s) 
( 1s  1)( 2 s  1)

Homework: Using Matlab to compare the response of the


actual model and approximation models

You might also like