5-1 CHBE320 Process Dynamics and Control Korea University CHBE320 LECTURE V LAPLACE TRANSFORM AND TRANSFER FUNCTION Professor Dae Ryook Yang Fall 2019 Dept. of Chemical and Biological Engineering Korea University
5-1CHBE320 Process Dynamics and Control Korea University
CHBE320 LECTURE VLAPLACE TRANSFORM AND
TRANSFER FUNCTION
Professor Dae Ryook Yang
Fall 2019Dept. of Chemical and Biological Engineering
Korea University
5-2CHBE320 Process Dynamics and Control Korea University
PROCESS
Sensor
ActuatorController+-
Lectures IV to VII
Road Map of the Lecture V
• Laplace Transform and Transfer functions– Definition of Laplace transform– Properties of Laplace transform– Inverse Laplace transform– Definition of transfer function– How to get the transfer functions– Properties of transfer function
5-3CHBE320 Process Dynamics and Control Korea University
SOLUTION OF LINEAR ODE
• 1st-order linear ODE– Integrating factor:
• High-order linear ODE with constant coeffs.– Modes: roots of characteristic equation
– Depending on the roots, modes are• Distinct roots:• Double roots:• Imaginary roots:
• Many other techniques for different cases
For 𝑑𝑥𝑑𝑡 𝑎 𝑡 𝑥 𝑓 𝑥 , I.F. exp 𝑎 𝑡 𝑑𝑡
𝑥𝑒 𝑓 𝑡 𝑒 𝑥 𝑡 𝑓 𝑡 𝑒 𝑑𝑡 𝐶 𝑒
For 𝑎 𝑥 𝑎 𝑥 𝑎 𝑥 𝑓 𝑡 ,
𝑎 𝑝 𝑎 𝑝 𝑎 𝑎 𝑝 𝑝 𝑝 𝑝 0
𝑒 , 𝑒
𝑒 cos 𝛽 𝑡, 𝑒 sin 𝛽 𝑡
𝑒 , 𝑡𝑒
Solution is a linear combination of modes and the coefficients are
decided by the initial conditions.
5-4CHBE320 Process Dynamics and Control Korea University
LAPLACE TRANSFORM FOR LINEAR ODE AND PDE
• Laplace Transform – Not in time domain, rather in frequency domain– Derivatives and integral become some operators.– ODE is converted into algebraic equation– PDE is converted into ODE in spatial coordinate– Need inverse transform to recover time-domain solution
ODE or PDEu(t) y(t)
Transfer FunctionU(s) Y(s)
(Algebraic calculation)
(D.E. calculation)
𝔏 𝔏 𝔏 𝔏𝔏 𝔏
5-5CHBE320 Process Dynamics and Control Korea University
DEFINITION OF LAPLACE TRANSFORM
• Definition
– F(s) is called Laplace transform of f(t).– f(t) must be piecewise continuous.– F(s) contains no information on f(t) for t < 0.– The past information on f(t) (for t < 0) is irrelevant.– The s is a complex variable called “Laplace transform variable”
• Inverse Laplace transform
– and are linear.𝔏 𝔏 𝔏 𝑎𝑓 𝑡 𝑏𝑓 𝑡 𝑎𝐹 𝑠 𝑏𝐹 𝑠
5-6CHBE320 Process Dynamics and Control Korea University
LAPLACE TRANSFORM OF FUNCTIONS
• Constant function, a
• Step function, S(t)
• Exponential function, e-bt
𝔏 𝑎 𝑎𝑒 𝑑𝑡𝑎𝑠 𝑒 0
𝑎𝑠
𝑎𝑠
t
f(t)
a
t
f(t)
1
𝑓 𝑡 S 𝑡 1 for 𝑡 00 for 𝑡 0
𝔏 S 𝑡 𝑒 𝑑𝑡1𝑠 𝑒 0
1𝑠
1𝑠
𝔏 𝑒 𝑒 𝑒 𝑑𝑡1
𝑠 𝑏 𝑒1
𝑠 𝑏t
f(t)
1b>0
b<0
5-7CHBE320 Process Dynamics and Control Korea University
• Trigonometric functions– Euler’s Identity:
• Rectangular pulse, P(t)
𝑒 ≜ cos 𝜔 𝑡 𝑗 sin 𝜔 𝑡
cos 𝜔 𝑡12 𝑒 𝑒 sin 𝜔 𝑡
12𝑗 𝑒 𝑒
𝔏 sin 𝜔 𝑡 𝔏12𝑗 𝑒 𝔏
12𝑗 𝑒
12𝑗
1𝑠 𝑗𝜔
1𝑠 𝑗𝜔
𝜔𝑠 𝜔
𝔏 cos 𝜔 𝑡 𝔏12 𝑒 𝔏
12 𝑒
12
1𝑠 𝑗𝜔
1𝑠 𝑗𝜔
𝑠𝑠 𝜔
𝑓 𝑡 P 𝑡0 for 𝑡 𝑡
ℎ for 𝑡 𝑡 00 for 𝑡 0
𝔏 P 𝑡 ℎ𝑒 𝑑𝑡ℎ𝑠 𝑒
ℎ𝑠 1 𝑒
t
f(t)
h
tw
t
sin(t)1
2𝜋𝜔
5-8CHBE320 Process Dynamics and Control Korea University
• Impulse function,
• Ramp function, t
• Refer the Table 3.1 (Seborg et al.) for other functions
𝑓 𝑡 𝛿 𝑡 lim→
0 for 𝑡 𝑡 1/𝑡 for 𝑡 𝑡 0
0 for 𝑡 0
𝔏 𝛿 𝑡 lim→
1𝑡 𝑒 𝑑𝑡 lim
→
1𝑡 𝑠 1 𝑒 1
L′Hospital′s rule: lim→
𝑓 𝑡𝑔 𝑡 lim
→
𝑓 𝑡𝑔 𝑡
t
f(t)1/tw
tw
t
f(t)
1
1
𝛿 𝑡
𝔏 𝑡 𝑡𝑒 𝑑𝑡
𝑡𝑠 𝑒
𝑒𝑠 𝑑𝑡
1𝑠 𝑒 𝑑𝑡
1𝑠
Integration by part: 𝑓′ ⋅ 𝑔𝑑𝑡 𝑓 ⋅ 𝑔 𝑓 ⋅ 𝑔 𝑑𝑡
5-9CHBE320 Process Dynamics and Control Korea University
5-10CHBE320 Process Dynamics and Control Korea University
5-11CHBE320 Process Dynamics and Control Korea University
PROPERTIES OF LAPLACE TRANSFORM
• Differentiation𝔏
𝑑𝑓𝑑𝑡 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 𝑡 𝑒 𝑓 ⋅ 𝑠 𝑒 𝑑𝑡 by 𝑖. 𝑏. 𝑝.
𝑠 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 0 𝑠𝐹 𝑠 𝑓 0
𝔏𝑑 𝑓𝑑𝑡 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 𝑡 𝑒 𝑓 ⋅ 𝑠 𝑒 𝑑𝑡 𝑠 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 0
𝑠 𝑠𝐹 𝑠 𝑓 0 𝑓 0 𝑠 𝐹 𝑠 𝑠𝑓 0 𝑓 0
𝔏𝑑 𝑓𝑑𝑡 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 𝑡 𝑒 𝑓 ⋅ 𝑠 𝑒 𝑑𝑡
𝑠 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 0 𝑠 𝔏𝑑 𝑓𝑑𝑡 𝑓 0
𝑠 𝐹 𝑠 𝑠 𝑓 0 ⋯ 𝑠𝑓 0 𝑓 0
⋮
5-12CHBE320 Process Dynamics and Control Korea University
• If f (0) = f ’(0) = f ”(0) =…= f (n-1)(0) = 0,– Initial condition effects are vanished.– It is very convenient to use deviation
variables so that all the effects ofinitial condition vanish.
• Transforms of linear differential equations.
𝔏𝑑𝑓𝑑𝑡 𝑠𝐹 𝑠
𝔏𝑑 𝑓𝑑𝑡 𝑠 𝐹 𝑠
⋮
𝔏𝑑 𝑓𝑑𝑡 𝑠 𝐹 𝑠
𝑦 𝑡 𝔏
𝑌 𝑠 , 𝑢 𝑡 𝔏
𝑈 𝑠𝑑𝑦 𝑡
𝑑𝑡 𝔏
𝑠𝑌 𝑠 if 𝑦 0 0
𝜏𝑑𝑦 𝑡
𝑑𝑡 𝑦 𝑡 𝐾𝑢 𝑡 𝑦 0 0 𝔏
𝜏𝑠 1 𝑌 𝑠 𝐾𝑈 𝑠
𝜕𝑇𝜕𝑡 𝑣
𝜕𝑇𝜕𝑧
1𝜏 𝑇 𝑇
𝔏 𝜏 𝑣
𝜕𝑇 𝑠𝜕𝑧 𝜏 𝑠 1 𝑇 𝑠 𝑇 𝑠
5-13CHBE320 Process Dynamics and Control Korea University
• Integration
• Time delay (Translation in time)
• Derivative of Laplace transform
𝔏 𝑓 𝜉 𝑑𝜉 𝑓 𝜉 𝑑𝜉 𝑒 𝑑𝑡
𝑒𝑠 𝑓 𝜉 𝑑𝜉
1𝑠 𝑓 ⋅ 𝑒 𝑑𝑡
𝐹 𝑠𝑠 by 𝑖. 𝑏. 𝑝.
0
Leibniz rule: 𝑑𝑑𝑡 𝑓 𝜏 𝑑𝜏 𝑓 𝑏 𝑡
𝑑𝑏 𝑡𝑑𝑡 𝑓 𝑎 𝑡
𝑑𝑎 𝑡𝑑𝑡
𝑓 𝑡
𝑓 𝑡 𝜃 S 𝑡 𝜃
𝔏 𝑓 𝑡 𝜃 S 𝑡 𝜃 𝑓 𝑡 𝜃 𝑒 𝑑𝑡 𝑓 𝜏 𝑒 𝑑𝜏 let 𝜏 𝑡 𝜃
𝑒 𝑓 𝜏 𝑒 𝑑𝜏 𝑒 𝐹 𝑠
t
f(t)
𝜃
𝑑𝐹 𝑠𝑑𝑠
𝑑𝑑𝑠 𝑓 ⋅ 𝑒 𝑑𝑡 𝑓 ⋅
𝑑𝑑𝑠 𝑒 𝑑𝑡 𝑡 ⋅ 𝑓 𝑒 𝑑𝑡 𝔏 𝑡 ⋅ 𝑓 𝑡
5-14CHBE320 Process Dynamics and Control Korea University
• Final value theorem– From the LT of differentiation, as s approaches to zero
– Limitation: has to exist. If it diverges or oscillates, this theorem is not valid.
• Initial value theorem– From the LT of differentiation, as s approaches to infinity
𝑑𝑓𝑑𝑡 𝑑𝑡 𝑓 ∞ 𝑓 0 lim
→𝑠 𝐹 𝑠 𝑓 0 ⇒ 𝑓 ∞ lim
→𝑠 𝐹 𝑠
lim→
𝑑𝑓𝑑𝑡 ⋅ 𝑒 𝑑𝑡 lim
→𝑠𝐹 𝑠 𝑓 0
lim→
𝑑𝑓𝑑𝑡 𝑒 𝑑𝑡 0 lim
→𝑠 𝐹 𝑠 𝑓 0 ⇒ 𝑓 0 lim
→𝑠 𝐹 𝑠
𝑓 ∞
lim→
𝑑𝑓𝑑𝑡 ⋅ 𝑒 𝑑𝑡
𝑑𝑓𝑑𝑡 ⋅ lim
→𝑒 𝑑𝑡 lim
→𝑠𝐹 𝑠 𝑓 0
5-15CHBE320 Process Dynamics and Control Korea University
EXAMPLE ON LAPLACE TRANSFORM (1)
•
•
– Using the initial and final value theorems
– But the final value theorem is not valid because
t
f(t)
3
2 6
𝑓 𝑡
1.5𝑡 for 0 𝑡 23 for 2 𝑡 60 for 6 𝑡 0 for 𝑡 0
𝑓 𝑡 1.5𝑡 S 𝑡 1.5 𝑡 2 S 𝑡 2 3 S 𝑡 6
∴ 𝐹 𝑠 𝔏 𝑓 𝑡1.5𝑠 1 𝑒
3𝑠 𝑒
For 𝐹 𝑠2
𝑠 5 , find 𝑓 0 and 𝑓 ∞ .
𝑓 0 lim→
𝑠 𝐹 𝑠 lim→
2𝑠𝑠 5 2 𝑓 ∞ lim
→𝑠 𝐹 𝑠 lim
→
2𝑠𝑠 5 0
lim→
𝑓 𝑡 lim→
2 𝑒∞
5-16CHBE320 Process Dynamics and Control Korea University
EXAMPLE ON LAPLACE TRANSFORM (2)
• What is the final value of the following system?
– Actually, cannot be defined due to sin t term.
• Find the Laplace transform for ?
𝑥 𝑥 𝑥 sin 𝑡 ; 𝑥 0 𝑥 0 0
⇒ 𝑠 𝑋 𝑠 𝑠𝑋 𝑠 𝑋1
𝑠 1 ⇒ x s1
𝑠 1 𝑠 𝑠 1
𝑥 ∞ lim→
𝑠𝑠 1 𝑠 𝑠 1 0
𝑥 ∞
From 𝑑𝐹 𝑠
𝑑𝑠 𝔏 𝑡 ⋅ 𝑓 𝑡
𝔏 𝑡 ⋅ sin 𝜔 𝑡𝑑
𝑑𝑠𝜔
𝑠 𝜔2𝜔𝑠
𝑠 𝜔
5-17CHBE320 Process Dynamics and Control Korea University
INVERSE LAPLACE TRANSFORM
• Used to recover the solution in time domain
– From the table– By partial fraction expansion– By inversion using contour integral
• Partial fraction expansion– After the partial fraction expansion, it requires to know some
simple formula of inverse Laplace transform such as
𝑓 𝑡 𝔏 𝐹 𝑠1
2𝜋𝑗 𝑒 𝐹 𝑠 𝑑𝑠
1𝜏𝑠 1 ,
𝑠𝑠 𝑏 𝜔 ,
𝑛 1 !𝑠 ,
𝑒𝜏 𝑠 2𝜁𝜏𝑠 1 , etc.
5-18CHBE320 Process Dynamics and Control Korea University
PARTIAL FRACTION EXPANSION
• Case I: All pi’s are distinct and real– By a root-finding technique, find all roots (time-consuming)– Find the coefficients for each fraction
• Comparison of the coefficients after multiplying the denominator• Replace some values for s and solve linear algebraic equation• Use of Heaviside expansion
– Multiply both side by a factor, (s+pi), and replace s with –pi.
– Inverse LT:
𝐹 𝑠𝑁 𝑠𝐷 𝑠
𝑁 𝑠𝑠 𝑝 ⋯ 𝑠 𝑝
𝛼𝑠 𝑝 ⋯
𝛼𝑠 𝑝
𝛼 𝑠 𝑝𝑁 𝑠𝐷 𝑠
𝑓 𝑡 𝛼 𝑒 𝛼 𝑒 ⋯ 𝛼 𝑒
5-19CHBE320 Process Dynamics and Control Korea University
• Case II: Some roots are repeated
– Each repeated factors have to be separated first.– Same methods as Case I can be applied.– Heaviside expansion for repeated factors
– Inverse LT
𝐹 𝑠𝑁 𝑠𝐷 𝑠
𝑁 𝑠𝑠 𝑝
𝑏 𝑠 ⋯ 𝑏𝑠 𝑝
𝛼𝑠 𝑝 ⋯
𝛼𝑠 𝑝
𝛼1𝑖!
𝑑𝑑𝑠
𝑁 𝑠𝐷 𝑠 𝑠 𝑝 𝑖 0, ⋯ , 𝑟 1
𝑓 𝑡 𝛼 𝑒 𝛼 𝑡𝑒 ⋯𝛼
𝑟 1 ! 𝑡 𝑒
5-20CHBE320 Process Dynamics and Control Korea University
• Case III: Some roots are complex
– Each repeated factors have to be separated first.– Then,
– Inverse LT
𝐹 𝑠𝑁 𝑠𝐷 𝑠
𝑐 𝑠 𝑐𝑠 𝑑 𝑠 𝑑
𝛼 𝑠 𝑏 𝛽 𝜔𝑠 𝑏 𝜔
𝛼 𝑠 𝑏 𝛽 𝜔𝑠 𝑏 𝜔 𝛼
𝑠 𝑏𝑠 𝑏 𝜔 𝛽
𝜔𝑠 𝑏 𝜔
𝑓 𝑡 𝛼 𝑒 cos 𝜔 𝑡 𝛽 𝑒 sin 𝜔 𝑡
where 𝑏 𝑑 /2, 𝜔 𝑑 𝑑 /4
𝛼 𝑐 , 𝛽 𝑐 𝛼 𝑏 /𝜔
5-21CHBE320 Process Dynamics and Control Korea University
EXAMPLES ON INVERSE LAPLACE TRANSFORM
•
– Multiply each factor and insert the zero value
𝐹 𝑠𝑠 5
𝑠 𝑠 1 𝑠 2 𝑠 3𝐴𝑠
𝐵𝑠 1
𝐶𝑠 2
𝐷𝑠 3 distinct
𝑠 5𝑠 1 𝑠 2 𝑠 3 𝐴 𝑠
𝐵𝑠 1 𝑠
𝐶𝑠 2 𝑠
𝐷𝑠 3 ⇒ 𝐴 5/6
𝑠 5𝑠 𝑠 2 𝑠 3
𝐴 𝑠 1𝑠 𝐵
𝐶 𝑠 1𝑠 2
𝐷 𝑠 1𝑠 3 ⇒ 𝐵 2
𝑠 5𝑠 𝑠 1 𝑠 3
𝐴 𝑠 2𝑠
𝐵 𝑠 2𝑠 1 𝐶
𝐷 𝑠 2𝑠 3 ⇒ 𝐶 3/2
𝑠 5𝑠 𝑠 1 𝑠 2
𝐴 𝑠 3𝑠
𝐵 𝑠 3𝑠 1
𝐶 𝑠 3𝑠 2 𝐷 ⇒ 𝐷 1/3
∴ 𝑓 𝑡 𝔏 𝐹 𝑠56 2𝑒
32 𝑒
13 𝑒
5-22CHBE320 Process Dynamics and Control Korea University
•
– Use of Heaviside expansion
𝐹 𝑠1
𝑠 1 𝑠 2𝐴𝑠 𝐵𝑠 𝐶
𝑠 1𝐷
𝑠 2 repeated
𝛼1𝑖!
𝑑𝑑𝑠
𝑁 𝑠𝐷 𝑠 𝑠 𝑝 𝑖 0, ⋯ , 𝑟 1
1 𝐴𝑠 𝐵𝑠 𝐶 𝑠 2 𝐷 𝑠 1𝐴 𝐷 𝑠 2𝐴 𝐵 3𝐷 𝑠 2𝐵 𝐶 3𝐷 𝑠 2𝐶 𝐷
∴ 𝐴 𝐷, 2𝐴 𝐵 3𝐷 0, 2𝐵 𝐶 3𝐷 0, 2𝐶 𝐷 1
⇒ 𝐴 1, 𝐵 1, 𝐶 1, 𝐷 1
𝑠 𝑠 1𝑠 1
𝛼𝑠 1
𝛼𝑠 1
𝛼𝑠 1
𝑖 1 : 𝛼11!
𝑑𝑑𝑠 𝑠 𝑠 1 1
𝑖 2 : 𝛼12!
𝑑𝑑𝑠 𝑠 𝑠 1 1
𝑖 0 : 𝛼 𝑠 𝑠 1 1
∴ 𝑓 𝑡 𝔏 𝐹 𝑠 𝑒 𝑡𝑒12 𝑡 𝑒 𝑒
5-23CHBE320 Process Dynamics and Control Korea University
• 𝐹 𝑠𝑠 1
𝑠 𝑠 4𝑠 5𝐴 𝑠 2 𝐵
𝑠 2 1𝐶𝑠 𝐷
𝑠 complex
𝑠 1 𝐴 𝑠 2 𝑠 𝐵𝑠 𝐶𝑠 𝐷 𝑠 4𝑠 5𝐴 𝐶 𝑠 2𝐴 𝐵 4𝐶 𝐷 𝑠 5𝐶 4𝐷 𝑠 5𝐷
∴ 𝐴 𝐶, 2𝐴 𝐵 4𝐶 𝐷 0, 5𝐶 4𝐷 1, 5𝐷 1
⇒ 𝐴 1/25, 𝐵 7/25, 𝐶 1/25, 𝐷 1/5
𝐴 𝑠 2 𝐵𝑠 2 1
125
𝑠 2𝑠 2 1
725
𝐵𝑠 2 1
𝐶𝑠 𝐷𝑠
125
1𝑠
15
1𝑠
∴ 𝑓 𝑡 𝔏 𝐹 𝑠1
25 𝑒 cos 𝑡7
25 𝑒 sin 𝑡1
2515 𝑡
5-24CHBE320 Process Dynamics and Control Korea University
• 𝐹 𝑠1 𝑒
4𝑠 1 3𝑠 1𝐴
4𝑠 1𝐵
3𝑠 1 1 𝑒 Time delay
𝐴 1/ 3𝑠 1/
4, 𝐵 1/ 4𝑠 1/
3
∴ 𝑓 𝑡 𝔏 𝐹 𝑠 𝔏4
4𝑠 13
3𝑠 1 𝔏4𝑒4𝑠 1
3𝑒3𝑠 1
𝑒 / 𝑒 / 𝑒 / 𝑒 / S 𝑡 2
5-25CHBE320 Process Dynamics and Control Korea University
SOLVING ODE BY LAPLACE TRANSFORM
• Procedure1. Given linear ODE with initial condition,2. Take Laplace transform and solve for output3. Inverse Laplace transform
• Example: Solve for 5𝑑𝑦𝑑𝑡 4𝑦 2; 𝑦 0 1
𝔏 5𝑑𝑦𝑑𝑡 𝔏 4𝑦 𝔏 2 ⇒ 5 𝑠𝑌 𝑠 𝑦 0 4𝑌 𝑠
2𝑠
5𝑠 4 𝑌 𝑠2𝑠 5 ⇒ 𝑌 𝑠
5𝑠 2𝑠 5𝑠 4
∴ 𝑦 𝑡 𝔏 𝑌 𝑠 𝔏0.5𝑠
2.55𝑠 4 0.5 0.5𝑒 .
5-26CHBE320 Process Dynamics and Control Korea University
TRANSFER FUNCTION (1)
• Definition– An algebraic expression for the dynamic relation between the
input and output of the process model
• How to find transfer function1. Find the equilibrium point2. If the system is nonlinear, then linearize around equil. point3. Introduce deviation variables4. Take Laplace transform and solve for output5. Do the Inverse Laplace transform and recover the original
variables from deviation variables
TransferFunction, G(s)5
𝑑𝑦𝑑𝑡 4𝑦 𝑢; 𝑦 0 1
Let 𝑦 𝑦 1 and 𝑢 𝑢 4
5𝑠 4 𝑌 𝑠 𝑈 𝑠 ⇒ 𝑌 𝑠𝑈 𝑠
15𝑠 4
0.251.25𝑠 1 𝐺 𝑠
𝑈 𝑠 𝑌 𝑠
5-27CHBE320 Process Dynamics and Control Korea University
TRANSFER FUNCTION (2)
• Benefits– Once TF is known, the output response to various given inputs
can be obtained easily.
– Interconnected system can be analyzed easily. • By block diagram algebra
– Easy to analyze the qualitative behavior of a process, such as stability, speed of response, oscillation, etc.
• By inspecting “Poles” and “Zeros”• Poles: all s’s satisfying D(s)=0• Zeros: all s’s satisfying N(s)=0
𝑦 𝑡 𝔏 𝑌 𝑠 𝔏 𝐺 𝑠 𝑈 𝑠 𝔏 𝐺 𝑠 𝔏 𝑈 𝑠
G1 G2
G3
+-
X Y 𝑌 𝑠𝑋 𝑠
𝐺1 𝑠 𝐺2 𝑠1 𝐺1 𝑠 𝐺2 𝑠 𝐺3 𝑠
5-28CHBE320 Process Dynamics and Control Korea University
TRANSFER FUNCTION (3)
• Steady-state Gain: The ratio between ultimate changes in input and output
– For a unit step change in input, the gain is the change in output– Gain may not be definable: for example, integrating processes
and processes with sustaining oscillation in output– From the final value theorem, unit step change in input with
zero initial condition gives
– The transfer function itself is an impulse response of the system
Gain 𝐾ΔouputΔinput
𝑦 ∞ 𝑦 0𝑢 ∞ 𝑢 0
𝐾𝑦 ∞
1 lim→
𝑠 𝑌 𝑠 lim→
𝑠 𝐺 𝑠1𝑠 lim
→𝐺 𝑠
𝑌 𝑠 𝐺 𝑠 𝑈 𝑠 𝐺 𝑠 𝔏 𝛿 𝔱 𝐺 𝑠
5-29CHBE320 Process Dynamics and Control Korea University
EXAMPLE
• Horizontal cylindrical storage tank (Ex4.7)
– Equilibrium point:(if , can be any value in .)
– Linearization:
L q
qi
Rh
wi
𝑑𝑚𝑑𝑡 𝜌
𝑑𝑉𝑑𝑡 𝜌𝑞 𝜌𝑞
𝑉 ℎ 𝐿𝑤 ℎ 𝑑ℎ ⇒𝑑𝑉𝑑𝑡 𝐿𝑤 ℎ
𝑑ℎ𝑑𝑡
𝑤 𝐿𝑑ℎ𝑑𝑡 𝑞 𝑞 ⇒
𝑑ℎ𝑑𝑡
12𝐿 𝐷 ℎ ℎ
𝑞 𝑞 Nonlinear ODE
𝑤 ℎ /2 𝑅 𝑅 ℎ 2𝑅 ℎ ℎ
0 𝑞 𝑞 / 2𝐿 𝐷 ℎ ℎ𝑞 , 𝑞, ℎ𝑞 𝑞 ℎ 0 ℎ 𝐷
𝑑ℎ𝑑𝑡 𝑓 ℎ, 𝑞 , 𝑞
𝜕𝑓𝜕ℎ , ̄ , ̄
ℎ ℎ𝜕𝑓𝜕𝑞 , ̄ , ̄
𝑞 𝑞𝜕𝑓𝜕𝑞 , ̄ , ̄
𝑞 𝑞
5-30CHBE320 Process Dynamics and Control Korea University
𝑠𝐻 𝑠 𝑘𝑄 𝑠 𝑘𝑄 𝑠
𝜕𝑓𝜕ℎ , ̄ , ̄
𝑞 𝑞𝜕
𝜕ℎ1
2𝐿 𝐷 ℎ ℎ0 ∵ 𝑞 𝑞
𝜕𝑓𝜕𝑞 , ̄ , ̄
1
2𝐿 𝐷 ℎ ℎ,
𝜕𝑓𝜕𝑞 , ̄ , ̄
1
2𝐿 𝐷 ℎ ℎ
Let this term be k
• Transfer function between (integrating)
• Transfer function between (integrating)
• If is near 0 or D, k becomes very large and is around D/2, kbecomes minimum.
The model could be quite different depending on the operating condition used for the linearization.
The best suitable range for the linearization in this case is around D/2. (less change in gain)
Linearized model would be valid in very narrow range near 0.
𝐻 𝑠 and 𝑄 𝑠 : 𝑘𝑠
𝐻 𝑠 and 𝑄 𝑠 : 𝑘𝑠
ℎ ℎ
5-31CHBE320 Process Dynamics and Control Korea University
PROPERTIES OF TRANSFER FUNCTION
• Additive property
• Multiplicative property
• Physical realizability – In a transfer function, the order of numerator(m) is greater
than that of denominator(n): called “physically unrealizable”– The order of derivative for the input is higher than that of
output. (requires future input values for current output)
G1(s)
G2(s)
X1(s)
X2(s)
Y(s)
++
Y1(s)
Y2(s)
𝑌 𝑠 𝑌 𝑠 𝑌 𝑠𝐺 𝑠 𝑋 𝑠 𝐺 𝑠 𝑋 𝑠
G1(s) G2(s)X1(s) X2(s) X3(s)
𝑋 𝑠 𝐺 𝑠 𝑋 𝑠𝐺 𝑠 𝐺 𝑠 𝑋 𝑠
5-32CHBE320 Process Dynamics and Control Korea University
EXAMPLES ON TWO TANK SYSTEM
• Two tanks in series (Ex3.7)– No reaction
– Initial condition: c1(0)= c2(0)=1 kg mol/m3 (Use deviation var.)– Parameters: V1/q=2 min., V2/q=1.5 min.– Transfer functions
𝑉𝑑𝑐𝑑𝑡 𝑞𝑐 𝑞𝑐
𝑉𝑑𝑐𝑑𝑡 𝑞𝑐 𝑞𝑐
𝐶 𝑠𝐶 𝑠
1𝑉 /𝑞 𝑠 1
𝐶 𝑠𝐶 𝑠
1𝑉 /𝑞 𝑠 1
𝐶 𝑠𝐶 𝑠
𝐶 𝑠𝐶 𝑠
𝐶 𝑠𝐶 𝑠
1𝑉 /𝑞 𝑠 1 𝑉 /𝑞 𝑠 1
Ci
C1
C2
V1
V2
q
5-33CHBE320 Process Dynamics and Control Korea University
• Pulse input
• Equivalent impulse input
• Pulse response vs. Impulse response
t
5
0.25
𝑐
𝐶 𝑠5𝑠 1 𝑒 .
𝐶 𝑠 𝔏 5 0.25 𝛿 𝑡 1.25
𝐶 𝑠1
2𝑠 1 𝐶 𝑠5
𝑠 2𝑠 1 1 𝑒 .
5𝑠
102𝑠 1 1 𝑒 .
⇒ 𝑐 𝑡 5 1 𝑒 /
5 1 𝑒 . / S 𝑡 0.25
𝐶 𝑠1
2𝑠 1 𝐶 𝑠1.25
2𝑠 1
⇒ 𝑐 0.625𝑒 /
5-34CHBE320 Process Dynamics and Control Korea University
𝐶 𝑠1
2𝑠 1 1.5𝑠 1 𝐶 𝑠
1.252𝑠 1 1.5𝑠 1
52𝑠 1
3.751.5𝑠 1
⇒ 𝑐 2.5𝑒 / 2.5𝑒 / .
𝐶 𝑠1
2𝑠 1 1.5𝑠 1 𝐶 𝑠5
𝑠 2𝑠 1 1.5𝑠 1 1 𝑒 .
5𝑠
402𝑠 1
22.51.5𝑠 1 1 𝑒 .
⇒ 𝑐 𝑡 5 20𝑒 / 15𝑒 / .
5 20𝑒 . / 15𝑒 . / . S 𝑡 0.25