Sunteți pe pagina 1din 28

The School of Chemical Engineering

Process Control & Instrumentation


with special thanks to Manta Controls
Last Lecture SV
r(t) e(t) c(t)
+ Controller
-

Control systems are everywhere! PV ym(t)


Three basic control strategies Controller mechanism
Feedback, Inferential, Feedforward
Need to know objective
Directly measure the output that defines the objective wherever
possible; this is a primary measurement
Connect controller input to the most effective MV
Define the system and assumptions carefully
Dynamic systems; stability vs. instability
Hardware and software requirements
Good process control is now essential!

University of Adelaide 2
Inlet Fi, Ti Outlet F, T

Velocity v
Volume V

Force F T
Steam Inlet
h Fst, Tst

Drag cv2 Q

Condensate

Topic 2: Developing Open-Loop Transfer Functions

1st Principles Model Development


Linearisation and Deviation Form
Process Plant Models
Heat exchangers
Continuous stirred tank reactors (CSTRs)
isothermal and variable holdup, biological growth, gas-phase
Batch reactors with mass transfer
Flash systems
single- and multi-component vapourisers
Distillation columns
ideal binary phase, multi-component and non-ideal
Mass transfer
stripping, absorbing
Minerals processing operations

University of Adelaide 4
Model Types
Qualitative
Design, preliminary economic analysis

Algebraic equations
Chemical reactions, vapour-liquid equilibria, steady-state
balances, equations of state, heat and mass transfer with very
fast dynamics
Ordinary differential equations
Dynamic balances

Often a model will require a combination of algebraic and


ordinary differential equations

University of Adelaide 5
Heated Stirred Tank
Problem Statement
The level and effluent temperature of a heated stirred tank
are to be held constant. Develop relationships between
these outputs and the inputs affecting the process given
the outflow rate can be controlled.
Represent the process
Inlet F , Ti Outlet F, T
i

with a block diagram


Volume V Fi Ti Tst
T
Steam Inlet MV PV*
h Fst, Tst
F h
Q
Fst Process T
Condensate

University of Adelaide 6
Heated Stirred Tank
State assumptions
Constant density
Constant heat capacity
Perfectly mixed
Constant cross-sectional area
In the real world, verify your assumptions!

Determine which conservation equation(s) to use


Think about what is accumulating
No chemical reaction and material balance required
Mass balance
Energy balance as temperature is involved

University of Adelaide 7
Heated Stirred Tank
Write out conservation equations

accumulation = in out + generation

dm dV
Mass: m i m Vi V
dt dt
where V Ah

dh
A Fi F
dt
Check physical sense and dimensional homogeneity

University of Adelaide 8
Heated Stirred Tank
Write out conservation equations

accumulation = in out + generation

d H
Energy: C p Fi Ti Tref C p F T Tref Q
dt

where H C pV T Tref ; Q UAst (Tst T )


d
C pVT C p FiTi C p FT UAst Tst T
dt
if we choose Tref = 0

University of Adelaide 9
Heated Stirred Tank
d
C pVT C p FiTi C p FT UAst Tst T
dt

A hT FiTi FT Tst T
d UAst
dt C p
Check physical sense and dimensional homogeneity

Using the chain rule


dh
TA Ah
dT
FiTi FT
UAst
Tst T
dt dt C p

University of Adelaide 10
Heated Stirred Tank

T Fi F Ah Tst T
dT UAst
FiTi FT
dt C p

Fi Ti T Tst T
dT UAst
Ah
dt C p
What can the mass and energy balances tell us about
control strategy?
h f ( Fi , F ) T f ( Fi , Ti , Tst ,U )

, Cp, A, Ast are system parameters and t is the


independent variable

University of Adelaide 11
Heated Stirred Tank
Fi Ti T Tst T
dh dT UAst
A Fi F Ah
dt dt C p
Can we solve these differential equations?
Need to know whether any other dependencies exist first
The energy balance will be non-linear, but we may be able to
assume some variables remain at steady-state
No general solution; normally numerical e.g. Runge-Kutta
Need to convert to a linear approximation for analytical solution

dy
1st order linear: cy r (t ) y = output
dt
c = constant
r(t) = forcing/input function
University of Adelaide 12
Linearisation

What is meant by tangent


linearisation? f(x)

Replacing a non-linear function


by the equation of the tangent f(x0)
f(x)
through the point (x0, f(x0))
Equivalent to a first order Taylor
series expansion about (x0, f(x0))
Why linearise? x
x0
So we can solve analytically
The linear solution should be a reasonable approximation near
the point (x0, f(x0))
All significant control theory has been derived for linear systems
Allows generalisations to be made about the result

University of Adelaide 13
Linearisation
The tangent approximation y f(x) through (x0, y0) is:

df
y mx c where m
dx ( x0 , y0 )

df
and y0 mx0 c c y0 x0
dx ( x0 , y0 )

y y0
df
x x0
dx ( x0 , y0 )
What point should be chosen for the linearisation?
University of Adelaide 14
Linearisation
Steady-state is the logical choice

y yss
df
x xss
dx ( xss , yss )

Use the subscript ss to indicate steady-state


Different notation is used elsewhere

If the function depends on n independent variables:

f f f
y yss x1 x1,ss x2 x2,ss ... x xn, ss
x1 x2 xn
n
ss ss ss

University of Adelaide 15
Linearisation
Say outflow is from the base of the tank, then flow
through a control valve is likely: Fi

F k h
dh
A Fi F Fi k h A h
dt
k
Linearise F: F Fss h hss
2 hss F

dh k
A Fi Fss h hss
dt 2 hss

University of Adelaide 16
Linearisation
Fi Ti T Tst T
dT UAst
Energy balance: Ah
dt C p
All terms are non-linear if we cannot assume any inputs
remain at steady-state. Linearise:

e.g. FiTi Fi ,ssTi ,ss Ti ,ss Fi Fi ,ss Fi ,ss Ti Ti ,ss

Substitute linearised terms back into the energy balance


to derive the linear 1st order differential equation
Exercise for you- well come back to this next lecture
Re-check the linearised equations for physical sense and
dimensional homogeneity
Is there a more convenient method?
University of Adelaide 17
Deviation Form
In control systems, we are concerned with how much
variables deviate from steady-state
Define deviation variables:
y ' y yss
Re-write mass balance:

dh k dh k
A Fi Fss h hss A Fi Fss
h'
dt 2 hss dt 2 hss
At steady-state we can write:
=0
dh k
A Fi ,ss Fss h'
dt ss 2 hss
University of Adelaide
Deviation Form
Subtract the steady-state balance from the dynamic
balance so that all inputs can be written in deviation
form:
h'Fi ,ss Fss
dh dh k
A Fi Fss
dt dt ss 2 hss
dh' k
A Fi ' h'
dt 2 hss
dh' k Fi '
i.e. h'
dt 2 A hss A
r(t)
c
Now we have a linear, 1st order ordinary differential
equation in terms of deviations
University of Adelaide
Deviation Form
To remove a step, we can re-write the generalised
linearisation formula in terms of deviation variables:

f f f
y yss x1 x1,ss x2 x2,ss ... x xn, ss
x1 x2 xn
n
ss ss ss

f f f
y y ss x1 ' x2 '... xn '
x1 ss x2 ss xn ss
Reminds us that we do not need to expand these brackets
e.g. FiTi Fi ,ssTi ,ss Ti ,ss Fi ' Fi ,ssTi '
Not as many terms means less chance of errors

University of Adelaide
Degrees of Freedom
Number of unknowns minus the number of equations
Underspecified: No. of unknowns > No. of equations
Overspecified: No. of unknowns < No. of equations
Exactly specified: No of unknowns = No. of equations
Two equations in the heated stirred tank example
If all system parameters are known, there are seven
unknowns
Each control objective reduces the degrees of freedom by
one
Therefore with two control objectives, we need to know
three variables
e.g. Ti = Ti,ss, Fi = Fi,ss and Tst = Tst,ss

University of Adelaide 21
Moving Car
Problem Statement
A car weighing 1000 kg is propelled by the force originating
from the engine of F = 250N. It has an effective drag
coefficient of 0.4 Ns2/m2. Determine the linearised
relationship between speed, engine force and time.
Represent the process
Velocity v with a block diagram
cv2
Force F

MV PV*
Drag cv2
F v
Process

University of Adelaide 22
Moving Car
State assumptions
Constant air density
Horizontal road
No other factors to consider
In the real world, verify your assumptions!

Determine which conservation equation(s) to use


Think about what is accumulating
Forces are involved
Momentum balance (Newtons 2nd Law; F = ma)

University of Adelaide 23
Moving Car
Write out conservation equation

accumulation = in out + generation

dp dv
Momentum: m Fleft Fright (left = positive)
dt dt
where Fleft F and Fright cv 2

dv
m F cv 2
dt
Check physical sense and dimensional homogeneity

University of Adelaide 24
Moving Car
Linearise non-linear term:
cv 2 cvss 2cvss v'
2

Substitute back into differential equation:


dv

m F cvss 2cvss v'
dt
2

At steady-state:
2
dv F 250 Nm
Fss cvss 0 vss 25 m/s
2 ss
m 2
dt ss c 0.4 Ns
And subtracting from dynamic equation:
dv'
dv'
m F '2cvss v' F '0.02v'
dt dt
University of Adelaide 25
Moving Car
Now, say the car starts at rest and the engine force never
deviates from its steady value:

dv'
0.02v' and v'(0) = -25 m/s
dt
Can solve analytically easily in this case:

v' (t ) 25e 0.02t v(t ) 251 e 0.02t

University of Adelaide 26
Moving Car

Linearised
solution
Non-linear (true)
solution

University of Adelaide 27
Next Lecture
Go through a simple and convenient transformation we
can use to help us solve linearised ordinary differential
equations
Expresses relationships in terms of a different independent
variable

Show the utility of block diagrams when working with


transformed equations

University of Adelaide 28

S-ar putea să vă placă și