Chapter 18 Batch Process Control

Appendix A
394

The reaction rate is first order with a rate constant (in min -I) in the liquid (where it reacts), keeping the pressure low.
However, if the reaction slows or the gas feed is greater than
k = 2.4 X lOIS e-20 ,OOOIT (Tin OR).
can be absorbed, the pressure will start to rise. The pressure
rise can be compensated by an increase in liquid feed, but this
For the batch case, linearize the model around T = T.
may cause the cooling capacity to be exceeded. Describe a so-
18.10 A batch reactor converts component A into B, which lution to this problem using overrides (see Chapter 15).
(I) in turn decomposes into C:
k1
A-'>B-'>C
k2 (1
18.12 Fogler2 describes a safety accident in which a batch
reactor was used to produce nitroanaline from ammo-
1 nia and o-nitro chlorobenzene. On the day of the acci-
Laplace Transforllls
dent, the feed composition was changed from the
where kl = klOe-EIIRT and k2 = k20e-E2IRT. normal operating value. Using the material/energy balances
The concentrations of A and B are denoted by XI and X2, re- and data provided by Fogler, show that the maximum cooling In Chapter 2 we developed a number of mathematical where F(s) is the symbol for the Laplace transform, s is
spectively. The reactor model is rate will not be sufficient to prevent a temperature runaway
models that describe the dynamic operation of se- a complex independent variable, f(t) is some function
under conditions of the new feed composition. Use a simula-
dXI tor to solve the model equations. lected processes. Solving such models-that is, finding of time to be transformed, and ;E is an operator,
- = - kloXle-EIIRT defined by the integral. The function f(t) must satisfy
dt
18.13 Consider the batch reactor system simulated by Aziz the output variables as functions of time for some
mild conditions that include being piecewise continu-
dt -
dX2 _ k
lOxle
-E1IRT
-
k20x2e-E21RT (I) et al. 3 The two reactions, A + B -'> C and A + C -'> D, change in the input variable(s)-requires either ana-
ous for 0 < t < 00 (Churchill, 1971); this requirement
are carried out in a jacketed batch reactor, where C is lytical or numerical integration of the differential
the desired product and D is a waste product. The ma- almost always holds for functions that are useful in
equations. Sometimes considerable effort is involved
Thus, the ultimate values of XI and X2 depend on the reactor nipulated variable is the temperature of the coolant in the process modeling and control. When the integration is
temperature as a function of time. For in obtaining the solutions. One important class of performed, the transform becomes a function of the
cooling jacket. There are two inequality constraints: input
models includes systems described by linear ordinary Laplace transform variable s, The inverse Laplace
klO = 1.335 X 1010 min-I, k20 = 1.149 X 10
17
min- l bounds on the coolant temperature and an upper limit on the
maximum reactor temperature. Using the model parameters differential equations (ODEs). Such linear systems transform (;E-1) operates on the function F(s) and
EI = 75,000 Jig mol, E2 = 125,000 J/gmol specified by Aziz et aI., evaluate the following control strate- represent the starting point for many analysis tech- converts it to f(t). Notice that F(s) contains no infor-
R = 8.31 J/(gmol K) XIO = 0.7 mol/L, X20 = 0 gies for a set-point change from 20°C to 92 0c. niques in process control. mation about f(t) for t < O. Hence, f(t) = ;E-1[F(s)} is
(a) PID controller In this Appendix, we introduce a mathematical tool, not defined for t < 0 (Schiff, 1999).
Find the constant temperature that maximizes the amount of (b) Batch unit the Laplace transform, which can significantly reduce the One of the important properties of the Laplace trans-
B, for 0 t 8 min. Next allow the temperature to change as form and the inverse Laplace transform is that they are
(c) Batch unit with preload effort required to solve and analyze linear differential
a cubic function of time linear operators; a linear operator satisfies the superpo-
(d) Dual-mode controller equation models. A major benefit is that this transfor-
T(t) = ao + alt + a2? + a3t3 mation converts ordinary differential equations to alge- sition principle:
Find the values of ao, ai, a2, a3 that maximize X2 by integrating braic equations, which can simplify the mathematical ?:F(ax(t) + by(t» = a?:F(x(t» + b?:F(y(t» (A-2)
the model and using a suitable optimization method. 2 Elements of Chemical Reaction Engineering, 4th ed., Prentice Hall, manipulations required to obtain a solution or perform
Upper Saddle River, NJ, 2005, Chapter 9. where ?:F denotes a particular operation to be performed,
I)
18.11 Suppose a batch reactor such as the one in Fig. 18.12
has a gas ingredient added to the liquid feed. As long as
the reaction is proceeding normally, the gas is absorbed
3N. Aziz, M. A. Hussain, and I. M. Mujtaba, Performance of
Different Types of Controllers in Tracking Optimal Temperature
Profiles in Batch Reactors, Comput. Chem. Eng, 24, 1069 (2000).
an analysis.
First, we define the Laplace transform and show how
it can be used to derive the Laplace transforms of sim-
such as differentiation or integration with respect to
time. If?:F ==;E, then Eq. A-2 becomes
ple functions. Then we show that linear ODEs can be ;E(ax(t) + by(t» = aXes) + b Yes) (A-3)
solved using Laplace transforms, along with a technique
Therefore, the Laplace transform of a sum of functions
called partial fraction expansion. Some important gen-
x(t) and yet) is the sum of the individual Laplace trans-
eral properties of Laplace transforms are presented, forms Xes) and Yes); in addition, multiplicative con-
and we illustrate the use of these techniques with a series stants can be factored out of the operator, as shown
of examples. in (A-3).
In this book we are more concerned with operational
aspects of Laplace transforms-that is, using them to
A.1 THE LAPLACE TRANSFORM OF
obtain solutions or the properties of solutions of linear
REPRESENTATIVE FUNCTIONS
differential equations. For more details on mathematical
The Laplace transform of a function f(t) is defined as aspects of the Laplace transform, the texts by Churchill

F(s) = ;E[f(t)] = 1
00

f(t)e- st dt (A-I)
(1971) and Dyke (1999) are recommended.
Before we consider solution techniques, the appli-
cation of Eq. A-I should be discussed. The Laplace

A-l
A-2 Appendix A Laplace Transforms A.1 The Laplace Transform of Representative Functions A-3

transform can be derived easily for most simple func- is taken to be the moment at which the input changes Table A.I Laplace Transforms for Various Time-Domain Functions a
tions, as shown below. in magnitude. In many process modeling applications,
f(t) F(s)
functions are defined so that they are zero at initial
Constant Function. For f(t) = a (a constant), time-that is,f(O) = O. In these cases, (A-9) simplifies 1. Set) (unit impulse) 1
to ::E(dfldt) = spes).
1
1
00 2. Set) (unit step)
The Laplace transform for higher-order derivatives
ae- SI dt = e- sl [
S
::E(a) = (A-4)
can be found using Eq. A-9. To derive ::E[.f"(t)], we 1
3. t (ramp)
define a new variable (<p = dfldt) such that s2
(n - I)!
= ::E( = s<p(s) - <p(0) (A-lO) Sll
1
Step Function. The unit step function, defined as <p(s) = spes) - f(O) (A-ll) 5 . e -bt
s + b
.!.e- liT 1
< 0 Substituting into Eq. A-I0 6.
S(t) = (A-S) T TS + 1
2: 0 2f
tll-Ie- bt
::E ( ddt 2) = s[sP(s) - f(O)] - <p(0) (A-12) 1
is an important input that is used frequently in process 7. (n _ I)! (n > 0)
(S + b)ll
dynamics and control. The Laplace transform of the = s2p(s) - sf(O) - 1'(0) (A-13) 8. 1 /ll-Ie-th 1
unit step function is the same as that obtained for the
where 1'(0) denotes the value of dfldt at t = O. The
Tll(n - I)! (TS + 1)11
constant above when a = 1:
Laplace transform for derivatives higher than second 1 1
1 9. (e- b2t - e- blt )
::E[S(t)] = - (A-6) order can be found by the same procedure. An nth- bi - b2
s order derivative, when transformed, yields a series of 1
10. - - - (e- tiTl - e- th2 )
If the step magnitude is a, the Laplace transform is als. (n + 1) terms: TI - T2 hs + 1)(T2s + 1)
The step function incorporates the idea of initial time, S + b3
zero time, or time zero for the function, which refers to = sI1P(s) - sl1-1f(0) - sl1-2f(1)(0) - ... 11. b3 - bi
e-blt + b - b
-,-3_---=-2 -b 2t
b2 - bi b i - b2 e (S + bl)(s + b2 )
the time at which Set) changes from 0 to 1. To avoid
any ambiguity concerning the value of the step function - sf(n-2)(0) - f(n-l)(o) (A-14) 12. -1 _
TI _
- _
T3 e-thl 1 T2_
+ __ - _
T3 e-th2 T3S + 1

at t = 0 (it is discontinuous), we will consider Set = 0) TI TI - T2 T2 T2 - TI (TIS + 1)(T2S + 1)

where fiCO) is the ith derivative evaluated at t = O. If
to be the value of the function approached from the n = 2, Eq. A-13 is obtained. 1
positive side, t = 0+.
w
Exponential Functions. The Laplace transform of an 14. sin wt
Derivatives. The transform of a first derivative of f is exponential function is important because exponential S2 + w2
important because such derivatives appear in dynamic functions appear in the solution to all linear differential S
15. cos wt
models: equations. For a negative exponential, e- bl , with b > 0, s2 + w2

::E(dfldt) = 1 00

(dfldt)e- SI dt (A-7) ::E(e- bt ) = 1 00

e-b1e-S1dt =
roo e-(b+s)tdt (A-IS)
Jo 16. sin(wt + <p)
w cos <P + s sin <p
s2 + w2
Integrating by parts, w
17. e- bt sin wt

::E(dfldt) = 1 00

f(t)e-Sls dt+ f(t)e- SI [ (A-8)

= _1_[ -e -(b+s)t] I 00
b +s 0
1
s + b
(A-16)
18. e-btcoswt
} b, w real

The Laplace transform for b < 0 is unbounded if s < b;

= s::E(f(t)) - f(O) = spes) - f(O) (A-9) therefore, the real part of s must be restricted to be
19.
larger than -b for the integral to be finite. This condi- T 1-1;2
where pes) is the Laplace transform of f(t). Generally, tion is satisfied for all problems we consider in this book.
the point at which we start keeping time for a solution (0 :s; Isld)
is arbitrary. Model solutions are most easily obtained Trigonometric Functions. In modeling processes atid 1
20. 1 + ---(Tle- tiTl - T2e-th2)
T2 - TI
assuming that time starts (i.e., t = 0) at the moment in studying control systems, there are many other
the process model is first perturbed. For example, if important time functions, such as the trigonometric (TI #- T2)
the process initially is assumed to be at steady state functions, cos wt and sin wt, where w is the frequency in
and an input undergoes a unit step change, zero time radians per unit time. The Laplace transform of cos wt ( continued)
 ,2

A·4 Appendix A Laplace Transforms A2 Solution of Differential Equations by Laplace Transform Techniques A.5

Table A.1 (Continued) as given in Table Al. The unit impulse function may
f(t) F(s) h 1---_ _ _--.
also be interpreted as the time derivative of the unit
step function Set). The response of a process to a unit
1 1 impulse is called its impulse response, which is illus-
21. 1 + tV]
trated in Example A. 7.
A physical example of an impulse function is the
f(t)
tV = tan- I , ,(0 :s; 1'1<1) rapid injection of dye or tracer into a fluid stream,
where f(t) corresponds to the concentration or the flow
1 rate of the tracer. This type of signal is sometimes used
22. 1 - + ,
tIT)
vI - ,2 sin tIT)]
in process testing, for example, to obtain the residence
(O:S; 1,1 < 1) time distribution of a piece of equipment, as illustrated
°oL------L-------
tw in Section A5.
23. 1 + T3 - TI e-tiTl + T3 - T2 e-th2
Time. t
TI - T2 T2 - TI
Figure A.1 The rectangular pulse function.
(TI "* T2) A.2 SOLUTION OF DIFFERENTIAL
EQUATIONS BY LAPLACE
24. df sF(s) - f(O)
dt The Laplace transform of the rectangular pulse can be TRANSFORM TECHNIQUES
derived by evaluating the integral (A-I) between t = 0
d"f In the previous section we developed the techniques
s"F(s) - s"-If(O) - s"-2f(1)(0) - ... and t = tIV because f(t) is zero everywhere else:

1
25. dt" required to obtain the Laplace transform of each term

26. f(t - to)S(t - to)

- Sf(11-2)(0) - f(Il-I)(O)
e-tosF(s)
F(s) = 100

f(t)e- st dt =
1w
he-sl dt (A-21)
in a linear ordinary differential equation. Table Al
lists important functions of time, including derivatives,
and. their Laplace transform equivalents. Because the
aNote thatf(t) and F(s) are defined for t 2: 0 only. Laplace transform converts any function f(t) to F(s)
F(s) = - -h e- sl 1111 = -h (1 - e-II\,s) (A-22) and the inverse Laplace transform converts F(s) back
s 0 s
to f(t) , the table provides an organized way to carry out
Note that an exponential term in F(s) results. For a unit these transformations.
or sin wt can be calculated using integration by parts. Table A.l lists some important Laplace transform rectangular pulse, h = 1itIV and the area under the pulse The procedure used to solve a differential equation
An alternative method is to use the Euler identityl pairs that occur in the solution of linear differential is unity. is quite simple. First Laplace transform both sides of
equations. For a more extensive list of transforms, see the differential equation, substituting values for the
ejwI + e- jwI
cos wt = 2 ,j v==-t (A-17) Dyke (1999). Impulse Function. A limiting case of the unit rectangu- initial conditions in the derivative transforms. Re-
Note that in all the transform cases derived above, lar pulse is the impulse or Dirac delta function, which has arrange the resulting algebraic equation, and solve for
and to apply (A-I). Because the Laplace transform of a F(s) is a ratio of polynomials in s, that is, a rational the symbol oCt). This function is obtained when tw 0 the transform of the dependent (output) variable. Fi-
sum of two functions is the sum of the Laplace transforms, form. There are some important cases when nonpoly- while keeping the area under the pulse equal to unity. A nally, find the inverse of the transformed output vari-
nomial (nonrational) forms occur. One such case is pulse of infinite height and infinitesimal width results. able. The solution method is illustrated by means of
;£(cos wt) = !;£(e jwt ) + !;£(e -jwI) (A-18) discussed next. Mathematically, this can be accomplished by substitut- several examples.
Using Eq. A-16 gives ing h = 11tw into (A-22); the Laplace transform of oCt) is
The Rectangular Pulse Function. An illustration of
;£( cos wt) = 1.(_1_._ + _1_._) ;£(o(t)) = lim (1 - e-II"s) (A-23) EXAMPLEA.l
2 s - JW S + JW the rectangular pulse is shown in Fig. A.l. The pulse ->0 tws
111
has height h and width two This type of signal might be
= 1.(s + jw + s - jW) = -,:-_s_-:- (A-19) used to depict the opening and closing of a valve regu- Equation A-23 is an indeterminate form that can be
Solve the differential equation,
2 s2 + w2 evaluated by application of L'Hospital's rule (also dy
lating flow into a tank. The flow rate would be held at h 5 - + 4y = 2 yeO) = 1 (A-26)
Note that the use of imaginary variables above was for a duration of tw units of time. The area under the spelled L'Hopital), which involves taking derivatives of dt
merely a device to avoid integration by parts; imagi- curve in Fig. A.l could be interpreted as the amount of both numerator and denominator with respect to tw:
using Laplace transforms.
nary numbers do not appear in the final result. To find material delivered to the tank (= htw)' Mathematically, se-tl"s
;£(sin wt), we can use a similar approach. the functionf(t) is defined as ;£(o(t)) = lim - - = 1 (A-24)
111->0 S SOLUTION
If the impulse magnitude (i.e., area twh) is a constant a First, take the Laplace transform of both sides of Eq. A-26:
(A-20) rather than unity, then
lThe symbol j, rather than i, is traditionally used for
control engineering literature.
v=I in the
;£(ao(t)) =a (A-25)
:£(5 d: + 4Y ) = :£(2) (A-27)
A.6 Appendix A Laplace Transforms A3 Partial Fraction Expansion A·7

Using the principle of superposition, each term can be trans- A.3 PARTIAL FRACTION EXPANSION terms (s + bD and then set s = -bi, which causes all
EXAMPLEA.2 terms except one to be multiplied by zero. Multiplying
formed individually: The high-order denominator polynomial in a Laplace
Solve the ordinary differential equation Eq. A-42 by s + 1 and then letting s = -1 gives
transform solution arises from the differential equation
d:) + = (A-28) terms (its characteristic polynomial) plus terms con-
(A-38) tributed by the inputs. The factors of the characteristic
al = S + 51 4
s + 4 s=-l 3
polynomial correspond to the roots of the characteristic
= = 5(sY(s) -1) = 5sY(s) - 5 Similarly, after multiplying by (s + 4) and letting
with initial conditions yeO) = y' (0) = y"(O) = O. polynomial set equal to zero. The input factors may be
(A-29) quite simple. Once the factors are obtained, the s = -4, the expansion gives
= = 4Y(s) (A-30) Laplace transform is then expanded into partial frac- 1
SOLUTION tions. As an example, consider a2 = s + 51
= 2 (A-31) s +1 s=-4 3
s Take Laplace transforms, term by term, using Table A1: Yes) _ s +5 (A-41) As seen above, the coefficients can be found by sim-
- s2 + 5s + 4
Substitute the individual terms: ple calculations.
The denominator can be factored into a product of For a more general transform, where the factors
5sY(s) - 5 + 4Y(s) = -
2
(A-32) first-order terms, (s + 1)(s + 4). This transform can be are real and distinct (no complex or repeated factors
s expanded into the sum of two partial fractions: appear), the following expansion formula can be used:
Rearrange (A-32) and factor out Yes): s +5 al
2
(s + 1)(s + 4) = s +1+s+4 (A-42) Yes) = N(s) = n N(s) = '£ (A-46)
Y(s)(5s + 4) = 5 +- (A-33) D(s) IT (s + bi) i=l s + i
s d:) = l1sY(s) where al and a2 are unspecified coefficients that must 1=1
satisfy Eq. A-42. The expansion in (A-42) indicates where D(s), an nth-order polynomial, is the denomina-
or = 6Y(s) that the original denominator polynomial has been fac- tor of the transform. D(s) is the characteristic polyno-
5s + 2 tored into a product of first-order terms. In general, for miaL The numerator N(s) has a maximum order of n - 1.
(A-34) = 1
Yes) = s(5s + 4) every partial fraction expansion, there will be a unique The ith coefficient can be calculated using the Heavi-
s
set of ai that satisfy the equation. side expansion
Take the inverse Laplace transform of both sides of Eq. A-34: Rearranging and factoring Yes), we obtain There are several methods for calculating the appro-

-1 [Yes)] _- -1[ s(5s5s ++ 24) ] (A-35) Y(s)(s3 + 6s2 + l1s + 6) =

1
(A-39)
priate values of al and a2 in (A-42):
a· = (s
I
N(s) I
+ b)--
I D(s) s=-b
(A-47)
s Method 1. Multiply both sides of (A-42) by (s + 1)(s + 4): I

The inverse Laplace transform of the right side of (A-35) can 1 Alternatively, an expansion for real, distinct factors
Yes) = - - -2- - - - (A-40) s + 5 = aleS + 4) + a2(s + 1) (A-43)
may be written as
be found by using Table AI. First, divide the numerator and s(s3 + 6s + l1s + 6)
denominator by 5 to put all factors in the s + b form corre- Equating coefficients of each power of s gives
sponding to the table entries: To invert (A-40) to find yet), we must find a similar expression Yes) = N'(s) = N...c./(s....:..)_
__
(A-48)
Sl: al + a2 = 1 (A-44a)
in Table AI. Unfortunately, no formula in the table has a .IT (TiS + 1)
yet) =
-1( s(sS++0.40.8) ) (A-36) fourth-order polynomial in the denominator. This example will
be continued later, after we develop the techniques necessary to
so: 4al + a2 = 5 (A-44b)
D'(s)
z=l
Using Method 3, calculate the coefficients by
Because entry 11
in the table, (s + b 3 )/[(s + b 1 )(s + b 2 )],
generalize the solution method in Section A3. Solving for al and a2 simultaneously yields al =
l

- -I
a2 = -3"'
matches (A-36) with b 1 = 0.8, b 2 = 0, and b 3 = 0.4, the a·
I
= (T'S + I )N'(s) (A-49)
solution can be written immediately: I Z D'(s) s=-l.
Method 2. Because Eq. A-42 must be valid for all values 'it'

In general, a transform expression may not exactly

yet) = 0.5 + 0.5e- O.8t (A-37) of s, we can specify two values of s and solve for the Note that several entries in Table A.l have the TS + 1
match any of the entries in Table A.1. This problem al-
two constants: format.
Note that in solving (A-26), both the forcing function (the ways arises for higher-order differential equations,
We now can use the Heaviside expansion to com-
constant 2 on the right side) and the initial condition have because the order of the denominator polynomial (A-45a)
plete the solution of Example A.2.
been incorporated easily and directly. As for any differential (characteristic polynomial) of the transform is equal to
equation solution, (A-37) should be checked to make sure it the order of the original differential equation, and no s = - 3: - = - + a2 (A-45b)
satisfies the initial condition and the original differential table entries are higher than third order in the denomi- EXAMPLE A.2 (Continued)
equation for t 2: O. nator. It is simply not practical to expand the numtfer Solving, al = t a2 = First factor the denominator of Eq. A-40 into a product of
of entries in the table ad infinitum. Instead, we use a
Method 3. The fastest and most popular method is first-order terms (n = 4 in Eq. A-46). Simple factors, as in this
procedure based on elementary transform building
case, rarely occur in actual applications.
Next we apply the Laplace transform solution to a blocks. This procedure, called partial fraction expan- called the Heaviside expansion. In this method multiply
higher-order differential equation. sion, is presented in the next section. both sides of the equation by one of the denominator s(s3 + 6s2 + l1s + 6) = s(s + 1)(s + 2)(s + 3) (A-50)
A-8 Appendix A Laplace Transforms A.3 Partial Fraction Expansion A-9

This result determines the four terms that will appear in the Time _ _ Laplace value of s. MUltiplying (A-54) by (s + 2)2 and letting s = -2 from which
domain E( domain ---'>->-
partial fraction expansion-namely, yields
1
Y(s) = ----:------:------,-----,-
s(s + 1)(s + 2)(s + 3) Step 1 0'2 = s : l\s=_2 = (A-55) i = 0: a2 =s+
s
1\ s=-2 1
2
(A-64a)
ODE Take Laplace
(A-51) Initial transform
s s+1 s+2 s+3 conditions (Table 3.1) Multiplying (A-54) by s and letting s = 0 yields
The Heaviside expansion method gives a I = 1/6, 0'2 = -1/2, s +1 \ 1
0'3 = 1/2, 0'4 = -1/6.
After the transform has been expanded into a sum of first- Step 2
0'3 = s2 + 4s + 4 s=O = "4 (A-56)
= -;
-1\ s=-2
1
4
order terms, invert each term individually using Table A.l: Solve for Substituting the value s = -1 in (A-54) gives s=-2
----
yes) = N(s) (A-57)
yet) = :£-1 [yes)] D(s)
0= 0'1 + 0'2 - 0'3 (A-64b)
1
=:£_1(1/6 0'1 = -"4 (A-58)
Complex Factors
s s+1 s+2 s+3
An alternative approach to find 0'1 is to use differentiation of An important case occurs when the factored character-
= .!.:£-1(.!.) _ .!.:£-1(_1 ) Step 3
Factor D(s),
the transform. Equation A-54 is multiplied by s(s + 2)2, istic polynomial yields terms of the form
6 s 2 s+1
perform partial
fraction expansion
s+ 1 = al(s + 2)s + a2s + a3(s + 2)2 (A-59) CIS + Co
+ .!.:£-1(_1 ) _ .!.:£-1(_1 )
2 s+2 6 s+3 Then (A-59) is differentiated twice with respect to s,
111 1 e- 3t t where
+ e- 2t
= - _ -e- t _ _ (A-52)
6 2 2 6
Solution
Step 4
Take inverse dr
-<do
Equation A-52 is thus the solution yet) to the differential equa- yet) Laplace transform
Note that differentiation in this case is tantamount to equat- 4
tion (A-38). The a/s are simply the coefficients of the solution. (Table 3.1)
ing powers of s, as demonstrated earlier. Here the denominator cannot be written as the product
Equation A-52 also satisfies the three initial conditions of the Figure A.2 The general procedure for solving an ordinary
differential equation. The reader should verify the result. of two real factors, which can be determined by using
differential equation using Laplace transforms.
the quadratic formula.
The differentiation illustrated above can For example, consider the transform
A.3.1 General Procedure for Solving Differential be used as the basis of a morEl general method to evalu-
s +2
Equations in addition to the other factors. Repeated factors arise ate the coefficients of repeated factors. If the denomi- (A-65)
Yes) = s2 +s+1
infrequently in process models of real systems, mainly nator polynomial D(s) contains the repeated factor
We now state a general procedure to solve ordinary (s + by, first form the quantity
for a process that consists of a series of identical units To invert (A-65) to the time domain, We complete the
differential equations using Laplace transforms. The
or stages. square of the first two terms in the denominator:
procedure consists of four steps, as shown in Fig. A2.
Q(s) = N(s) (s + by = (s + by-1 a1 + (s + by-2 a2 + ...
Note that solution for the differential equation in- D(s) s +2 (s + 0.5) + 1.5

( 0)2
Yes) - - - - - - : ; : - - - -
volves use of Laplace transforms as an intermediate step.
Step 3 can be bypassed if the transform found in Step 2 EXAMPLEA.3
+ a r + (s + b n other partial fractions] (A-61) - (s + 0.5)2 - 0.25 + 1
(s + 0.5)2 + 2
matches an entry in Table AI. In order to factor D(s) in Setting s = -b will generate a r directly. Differentiat-
Step 3, software such as MATLAB, Mathematica, or For ing Q(s) with respect to s and letting s = -b generates (A-66)
Mathcad can be utilized (Chapra and Canale, 2010). a r -1' Successive differentiation a total of r - 1 times Dividing the numerator of Yes) into two terms,
Yes) = s + 1 = + 0'2 + 0'3 (A-54)
In Step 3, other types of situations can occur. Both re- s(s2 + 4s + 4) s + 2 (s + 2)2 S will generate all ai, i = 1,2, ... , r from which we obtain
peated factors and complex factors require modifications the general expression s + 0.5 1.5
of the partial fraction expansion procedure. evaluate the unknown coefficients ai'
= 1:.. d(i)Q.(s) \
Yes) =
(s + 0.5)2 +
(0)2 +2
3
(s + 0.5)2 +
(0)2
2
u. i = 0, ... , r - 1 (A-62)
Repeated Factors
SOLUTION
I dsl l
) s=-b
(A-67)
If a term s + b occurs r times in the denominator, r
To find 0'1 in (A-54), the Heaviside rule cannot be used for mul- For i = 0 in (A-62), O! is defined to be 1 and the zeroth
terms must be included in the expansion that incorpo- /1' To determine yet), we invert each term separately. Note
tiplication by (s + 2), because s = -2 causes the second term derivative of Q(s) is defined to be simply Q(s) itself.
rate the s + b factor
on the right side to be unbounded, rather than 0 as desired. We Returning to the problem in Example A3, . s+b -b
that in Table A1, ". 2 2 transforms to e I
0'1 a2 therefore employ the Heaviside expansion method for the (s+b) +w
Yes) = - - + + " .+ . + ...
s+b (s+b)2 (s+b)' other two coefficients (0'2 and 0'3) that can be evaluated nor- Q(s) =s+ 1 (A-63) coswt, while w 2 transforms to e- bt sinwt.
(A-53) mally and then solve for 0'1 by arbitrarily selecting some other s (s+b) 2 +w
A-lO Appendix A Laplace Transforms
A.4 Other Laplace Transform Properties A-ll

Therefore the corresponding time-domain solution is Equate coefficients of like powers of s: Taking the limit as s 0 and assuming that dy/dt is
continuous and that sY(s) has a limit for all Re(s) 2:: 0, EXAMPLEA.6
_ V3 _ . V3 s3: a1 + as = 0 (A-7Sa)
yet) = e 0.51 cos t + V3 e 0.51 sm- t (A-68) 00
2 2 s2: 4a1 + a2 + a6 = 0 (A-7Sb) A process is described by a third-order ODE:
If the denominator is factored into a pair of complex
terms (complex conjugates) in the partial fraction
sl: Sal + 4a2 = 1 (A-7Sc) J
o
t
Y
dd dt = lim [sY(s)] - yeO).
s->O
(A-80) d3y d 2y dy
- + 6 - + 11- + 6y = 4u
dr3 d? dt
(A-84)
equation, we can alternatively express the transform as so: Sa2 = 1 (A-7Sd) Integrating the left side and simplifying yields
with all initial conditions on y, dy/dt, and di/d? equal to
(Xl + j!31 (X2 + j!32 Solving simultaneously gives al = 0.04, a2 = 0.2, as = 0.04, lim yet) = s---::".
lim [sY(s)] (A-8l) zero. Show that for a step change in II of 2 units, the steady-
Yes) = + (A-69) a6 = -0.36. The inverse Laplace transform of Yes) is
f--'j. 00 (X)

state result in the time domain is the same as applying the

s + b + jw s + b - jw If yet) is unbounded for t 00, Eq. A-8l gives final value theorem.
Appearance of these complex factors implies oscillatory yet) = ::£_1(0.04) + ::£_1(0.2) + ::£_I(-0.04S - 0.36) erroneous results. For example, if Yes) = l/(s - 5), Eq.
behavior in the time domain. Terms of the form e- bl s s2 s2+4s+S A-8l predicts y(oo) = O. Note that Eq. A-79, which is SOLUTION
sinwt and e-bl coswt arise after combining the inverse (A-76) the basis of (A-79), requires that lim yet 00) exists. In
transforms e-(b+jw)1 and e-(b-jw)l. Dealing with complex Before using Table A.l, the denominator term (s2 + 4s + S) this case, yet) = eSI, which is unbounded for t 00. If II= 2 the steady-state result for y can be found by setting

factors is more tedious than analyzing real factors. must be converted to the standard form by completing the However, Eq. A-79 does not apply here, because s Yes) = all derivatives to zero and substituting for u. Therefore
A partial fraction form that avoids complex algebra is square to (s + 2)2 + 12; the numerator is -0.04(s + 9). In order s/(s - 5) does not have a limit for some real value of 6y = 8 or y = 4/3 (A-8S)
to match the expressions in Table A.l, the argument of the last s 2:: 0, in particular, for s = 5.
The transform of (A-84) is
(A-70) term in (A-76) must be written as
A.4.2 Initial Value Theorem (s3 + 6s2 + 11s + 6)Y(s) = 8/s (A-86)
-0.04s - 0.36 -0.04(s + 2) -0.28
-----= +
Using Table A.l, the corresponding expression for (s + 2)2 + 1 (s + 2? + 1 (s + 2)2 + 1 Analogous to the final value theorem, the initial value Yes) = 8(A-87)
yet) is (A-77)
theorem can be stated as s4 + 6s3 + 11s2 + 6s
This procedure yields the following time-domain expression: lim yet) = lim [sY(s)] One of the benefits of the final value theorem is that we do
(A-82)
(->O s-> 00 not have to solve for the analytical solution of yet). Instead,
y(t) = 0.04 + 0.2t - 0.04e- 2t cos t - 0.28e- 2t sin t
The proof of this theorem is similar to the development simply apply Eq. A-81 to the transform Yes) as follows:
However, the coefficients al and a2 must be found by in (A-78) through (A-8l). It also requires that yet) is 8 4
solving simultaneous equations, rather than by the Heav- It is clear from this example that the Laplace transform continuous. The proof is left to the -reader as an exercise. lim sY(s) = lim 8 - (A-88)
s ...... o s ...... o s3 + 6s2 + 11s + 6 6 3
iside expansion, as shown as follows in Example AA. solution for complex or repeated roots can be quite
cumbersome for transforms of ODEs higher than This is the same answer as obtained in Eq. A-8S. The time-
second order. In this case, using numerical simulation domain solution obtained from a partial fraction expansion is
EXAMPLEA.5
EXAMPLEA.4 techniques may be more efficient to obtain a solution, y = 4/3 2e- t + 2e- 2t 2/3e- 3t (A-89)
as discussed in Chapters 4 and 5. Apply the initial and final value theorems to the transform
Find the inverse Laplace transform of As t --'? 00, only the first term remains, which is the same
derived in Example A.l:
result as in Eq. A-90 (using the final value theorem).
Yes) = s +1
A.4 OTHER LAPLACE TRANSFORM Yes) = Ss + 2
(A-72) PROPERTIES
s2(s2 + 4s + S) s(Ss + 4)
A.4.3 Transform of an Integral
In this section, we consider several Laplace transform SOLUTION
SOLUTION properties that are useful in process dynamics and Occasionally, it is necessary to find the Laplace transform
control. Initial Value of a function that is integrated with respect to time. By
The roots of the denominator term (s2 + 4s + S) are applying the definition (Eq. A-I) and integrating by parts,
imaginary (s + 2 + j, s + 2 - j), so we know the solution A.4.1 Final Value Theorem . . Ss+2

:e{ l' f(t') dt' } f {l'

will involve oscillatory terms (sin, cos). The partial fraction yeO) = hm [sY(s)] = hm - - = 1 (A-83a)
s ...... 00 Ss + 4
}e-" dt
s ...... 00
form for (A-72) that avoids using complex factors or roots is The asymptotic value of yet) for large values of time
Final Value f(t') dt'
y(oo) can be found from (A-78), providing that lim [sY(s)]
s+1 a1 ass + a6 exists for all Re(s) 2:: 0:
Y( s) = = - + + ----''-------''- s->O
(A-90)
s2(s2 + 4s + S) S s2 s2 + 4s + S y(oo) = lim [sY(s)] = lim Ss + 2 = O.S (A-83b)
lim yet) = lim [sY(s)] (A-78) s ...... o s ...... o Ss + 4
(A-73) {->oo s->O

Multiply both sides of Eq. A-73 by s2(s2 + 4s + S) and

collect terms:
Equation A-78 can be proved using the relation for t'he
Laplace transform of a derivative (Eq. A-9):
The initial value of 1 corresponds to the initial condition given
in Eq. A-26. The final value of O.S agrees with the time-domain
solution in Eq. A-37. Both theorems are useful for checking
[e-" l' f(t') dt'r
00
s+1 = (a1 + as)s3 + (4a1 + a2 + (6)s2
+ (Sal + 4(2)S + Sa2 (A-74) Jo
dy
- e- si dt
dt
= sY(s) - yeO) (A-79)
mathematical errors that may occur in obtaining Laplace trans-
form solutions. +- 11
s 0
00

e-Slf(t) dt (A-9l)
 A.S A Transient Response Example A-13
A-12 Appendix A Laplace Transforms

SOLUTION and
The first term in (A-93) yields 0 when evaluated at both
the upper and lower limits, as long as f(t*) possesses Equation A-lOO can be split into two terms: t 2: 2: yet) = e- tl4 e- tl3 + [e-(t-2)/4 - e-(t-2)/3]

a transform (is bounded). The integral in the second ret) = e- ti4 (1 + e2/4 ) - e- tl3 (1 + e2/3 )
term is simply the definition of the Laplace transform Yes) = Y1(s) + Y2 (s) (A-99)
= 2.6487e- tI4 - 2.9477e- tI3 (A-IDS)
of f(t). Hence, 1 e- 2 ,

:e{ l
= (4s + 1)(3s + 1) + (4s + 1)(3s + 1) (A-lOO) Note that (A-104) and (A-IDS) give equivalent results for
o t= 2, because in this case, yet) is continuous at t = 2.
f(t-') dt' } ; F(.,) (A-92) (a) The inverse transform of Yj(s) can be obtained directly from
Table A.l:
Note that Laplace transformation of an integral function
(A-lOl) A.S A TRANSIENT RESPONSE EXAMPLE
of time leads to division of the transformed function by s.
We have already seen in (A-9) that transformation of Because Y 2 (s) = e- 2s y j (s), its inverse transform can be writ- In Chapter 3 we will develop a standardized approach
time derivatives leads to an inverse relation - that is, ten immediately by replacing t by (t 2) in (A-lOl), and then for using Laplace transforms to calculate transient re-
multiplication of the transform by s. multiplying by the shifted step function: sponses. That approach will unify the way process mod-
els are manipulated after transforming them, and it will
A.4.4 Time Delay (Translation in Time) Y2(t) = [e-(1-2)/4 - e-(1-2)/3]S(t - 2) (A-102)
o further simplify the way initial conditions and inputs
(b)
Functions that exhibit time delay play an important (forcing functions) are handled. However, we already
Thus, the complete inverse transform is
role in process modeling and control. Time delays com- Figure A.3 A time function with and without time delay. have the tools to analyze an example of a transient re-
monly occur as a result of the transport time required (a) Original function (no delay); (b) function with delay to· yet) = e -1/4 - e -1/3 + [e -(1-2)14 - e -(t-2)/3]S(t - 2) sponse situation in some detail. Example A.7 illustrates
for a fluid to flow through piping. Consider the stirred- (A-l03) many features of Laplace transform methods in investi-
tank heating system example presented in Chapter 2. gating the dynamic characteristics of a physical process.
Suppose one thermocouple is located at the outflow Because (t - to) is now the artificial variable of integra- Equation A-103 can be numerically evaluated without dif-
point of the stirred tank, and a second thermocouple is tion, it can be replaced by t*. ficulty for particular values of t, noting that the term in
immersed in the fluid a short distance (L= 10 m) down-
stream. The heating system is off initially, and at time
zero it is turned on. If there is no fluid mixing in the
;£(f(t)) = e- sto 100

f(t*)e- st* dt* (A-95)

brackets is multiplied by 0 (the value of the unit step func-
tion) for t < 2, and by 1 for t 2: 2. An equivalent and sim-
pler method is to evaluate the contributions from the
EXAMPLEA.7
The Ideal Gas Company has two fixed-volume, stirred-tank
reactors connected in series as shown in Fig. A.4. The three
bracketed time functions only when the time arguments are
pipe (the fluid is in plug flow) and if no heat losses yielding the Real Translation Theorem: nonnegative. An alternative way of writing Eq. A-lOS is as IGC engineers who are responsible for reactor operations-
occur from the pipe, the shapes of the two temperature two equations, each one applicable over a particular inter- Kim Ng, Casey Gain, and Tim Delay-are concerned about
responses should be identical. However, the second Pd(S) = ;£(f(t - to)S(t - to)) = e-stop(s) (A-96) the adequacy of mixing in the two tanks and want to run a
val of time:
sensor response will be translated in time; that is, it will tracer test on the system to determine whether dead zones
In inverting a transform that contains an e- sto ele- o :S t < 2: yet) = e- 1/4 - e- 1/3 (A-104) and/or channeling exist in the reactors.
exhibit a time delay. If the fluid velocity is 1 mis, the
ment (time-delay term), the following procedure will
time delay (to = Llv) is 10 s. If we denote f(t) as the
easily yield results and also avoid the pitfalls of dealing
transient temperature response at the first sensor and
with translated (shifted) time arguments. Starting with
fd(t) as the temperature response at the second sensor,
the Laplace transform
Fig. A.3 shows how they are related. The function fd = 0
for t < to. Therefore, fd and f are related by Yes) = e-stop(s) (A-97)
fd(t) = f(t - to)S(t - to) (A-93)
1. Invert pes) in the usual manner; that is, perform q
Note thatfd is the functionf(t) delayed by to time units. partial fraction expansion, and so forth, to find f(t).
The unit step function Set - to) is included to denote 2. Find yet) = f(t - to)S(t - to) by replacing the ar-
explicitly thatfJCt) = 0 for all values of t < to· If ;£(f(t)) = gument t, wherever it appears in f(t), by (t - to); q
pes), then then multiply the entire function by the shifted
;£(fd(t)) = ;£(f(t - to)S(t - to)) unit step function, Set - to)·

= loof(t - to)S(t - to)e- st dt

Stage 1

= to f(t - to)(O)e- st dt + 100

f(t - to)e- st dt
EXAMPLEA.6
Find the inverse transform of

= 1
to
00

f(t - to)e-s(t-tole-sto d(t - to) (A-94) Y(S)------

1 + e- 2s
(4s + 1)(3s + 1)
(A-98) Figure A.4 Two-stage stirred-tank reactor system.
Stage 2
 -
A-14 Appendix A Laplace Transforms A5 A Transient Response Example A-IS

Table A.2 Two-Stage Stirred-Tank Reactor Process and where c is the reactant concentration of component A. Be- Substituting (A-114) into (A-l13), we have discussed above, we obtain the following inverse transform:
Operating Data cause the reaction term can be neglected in this example
(k = 0), the stages are merely continuous-flow mixers. Two ma- d 2 6.5 cf(t) = e- t12 + 6(1 - e- tI2 ) - 5[1 - e-(t-O.25)!2JS(t - 0.25)
Volume of Stage 1 = 4m 3 I(S) = s(4s + 2) + 4s + 2 (A-115)
terial balance equations are required to model the two stages:
Volume of Stage 2 = 3 m3 (A-121)
Total flow rate q = 2 m3/min dC1 Equation A-115 does not correspond exactly to any entries in
4dt- + 2C1 = 2c·I (A-106) Note that there are two solutions; for t < 0.25 min (or tw) the
Nominal feed reactant concentration (Ci) = 1 kg mol/m 3 Table AI. However, putting the denominator in 'TS + 1 form rightmost term, including the time delay, is zero in the time
yields solution. Thus, for
Their idea is to operate the reactors at a temperature low dC2
3 ---:it + 2C2 = 2CI (A-107) B( ) _ . 1 3.25
enough that reaction will not occur, and to apply a rectangu- (A-116) t < 0.25 min: q(t) = e- t12 + 6(1 - e- tI2 ) = 6 - 5e- t12
CI S - s(2s + 1) + 2s + 1
lar pulse in the reactant concentration to the first stage for If the system initially is at steady state, all concentrations are (A-122)
test purposes. In this way, available instrumentation on the which can be directly inverted using the table, yielding
equal to the feed concentration: t 2: 0.25 min: q(t) = e- tl2 + 6(1 - e- tI2 )
second-stage outflow line can be used without modification to
measure reactant (tracer) concentration. (A-lOS) c£(t) = 1 - e- tl2 + 1.625e- t12 = 1 + 0.625e- tI2 (A-117) - 5(1 - e-(t-O.25)/2)
Before performing the test, the engineers would like to have = 1 - 5e- t12 + 5e-tI2e+O.25/2
a good idea of the results that should be expected if perfect The rectangular pulse response is obtained in the same way.
(a) The pulse input is described by
mixing actually is accomplished in the reactors. A rectangular The transform of the input pulse (A-109) is given by (A-22), = 1 + 0.6657e- tI2 (A-123)
pulse input for the change in reactant concentration will be t<O so that Plots of (A-117), (A-122), and (A-123) are shown in Fig. A6.
5(1 - e- O•25s )
used with the restriction that the resulting output concentra-
tion changes must be large enough to be measured precisely.
,r {: o ::5 t < 0.25 min
t 2: 0.25 min
(A-109) p
Ci (s) =
1
s
+ ---'------'--
s
(A-11S)
Note that the rectangular pulse response approximates the
impulse response fairly well for t > 0.25 min. Obviously, the
The process data and operating conditions required to approximation cannot be very good before t = 0.25 min, be-
model the reactor tracer test are given in Table A2. Figure A.5 A convenient way to interpret (A-109) is as a constant input Substituting (A-11S) into (A-l13) and solving for Cf(s) yields cause the full effect of the rectangular pulse is not felt until
shows the proposed pulse change of 0.25 min duration that can of 1 added to a rectangular pulse of height = 5 kg mol/m3: that time, while the full effect of the hypothetical impulse be-
4 12 10e- O.25s
be made while maintaining the total reactor input flow rate p
C (s) - +--- (A-119) gins immediately at t = O.
constant. As part of the theoretical solution, Kim, Casey, and cf = 6 for 0::5 t < 0.25 min (A-110) I - 4s + 2 s(4s + 2) s(4s + 2)
(c) For the impulse response of Stage 2, Laplace transform
Tim would like to know how closely the rectangular pulse re- The magnitude of an impulse input that is equivalent to the Again, we have to put (A-121) into a form suitable for (A-107), using C2(0) = 1:
sponse can be approximated by the system response to an im- time-varying portion of (A-110) is simply the integral of the inversion:
pulse of equivalent magnitude. Based on all of these rectangular pulse: (A-124)
considerations, they need to obtain the following information: 2 6 5e- O.25s
kg mol . kg mol, min Ci(s) = 2s + 1 + s (2s + 1) (A-120)
(a) The magnitude of an impulse input equivalent to the rec- M = 5- - X 0.25 mill = 1.25 3 s(2s + 1) Rearrange to obtain C2 (s):
3
m m
tangular pulse of Fig. A5. Before inverting (A-120), note that the term containing e- 0.25s
(b) The impulse and pulse responses of the reactant concen- Therefore, the equivalent impulse input is (A-125)
will involve a translation in time. Utilizing the procedure
tration leaving the first stage.
c?(t) = 1 + 1.25o(t) (A-111)
(c) The impulse and pulse responses of the reactant concen-
tration leaving the second stage. Although the units of M have little physical meaning, the 1.75,---,-----,----,-----,----,
product
SOLUTION m3 kg mol, min - - - Impulse input

The reactor model for a single-stage CSTR was given

qM = 2 -. X 1.25 0 = 2.5 kg mol - - RectangUlar pulse input
III mill
Eq. 2-66 as
can be interpreted as the amount of additional reactant fed
dc into the reactor as either the rectangular pulse or the impulse. 1.50
Vdi = q(Ci - c) - Vkc
(b) The impulse response of Stage 1 is obtained by Laplace
transforming (A-106), using CI(O) = 1:
6
4sC I (s) - 4(1) + 2CI(s) = 2Ci(s) (A-112)
1.25
By rearranging (A-112), we obtain CI(s):
4 2
C1(s) = 4s + 2 + 4s + 2 Ci(s) (A-l13)
1 1
I The transform of the impulse input in feed concentration in 1.00 L -_ _..l.--_ _--L-_ _ _____I
o 0.25 (A-111) is o 2 4 6 8 10
Time (min) Time (min)
1
Figure A.S Proposed input pulse in reactant concentration. C?(s) = - + 1.25 (A-114)
s Figure A.6 Reactor Stage 1 response.
A-16 Appendix A Laplace Transforms Exercises A-17

For the input to (A-127), substitute the Laplace transform of 1.3,---,----,----,----,------, The use of Laplace transforms can be extended to of differential equation models into transfer function
the output from Stage I-namely, (A-116): obtain solutions for models consisting of simultaneous models, covered in the next chapter, represents an im-
- - - Impulse input

+
differential equations. However, before addressing such portant simplification in the methodology, one that can
- - Rectangular pulse input
extensions, we introduce the concept of input-output be exploited extensively in process modeling and control
CO(s) _ _3_ _2_[ 1 (A-126) models described by transfer functions. The conversion system design.
2 - 3s + 2 3s + 2 s(2s + 1) 2s + 1
1.2

which can be rearranged to REFERENCES

1.5 1 Churchill, R. V., Operational Mathematics, 3d ed., McGraw-Hili, Dyke, P. R. G., An Introduction to Laplace Transforms and Fourier
q(s) = loSs + 1 + s(1.5s + 1)(2s + 1) New York, 1971. Series, Springer-Verlag, New York, 1999.
Chapra, S. c., and R. P. Canale, Numerical Methods for Engineers, Schiff, 1. L., The Laplace Transform: Theory and Application,
3.25
+------- (A-127) 6th ed., McGraw-Hill, New York, 2010. Springer, New York, 1999.
(l.5s + 1)(2s + 1)

Because each term in (A-127) appears as an entry in Table AI, EXERCISES

partial fraction expansion is not required:
A.1 The differential equation (dynamic) model for a chemi- (b) Construct this function as the sum of simpler time ele-

= e- tl l. 5 + [1 + J:... (1.5e- tl l. 5 -
0.5
2e- tI2
)]
Time (min)

Figure A.7 Reactor Stage 2 response.

cal process is as follows:
d 2y dy
ments, some perhaps translated in time, whose transforms
can be found directly from Table A.1.
- + 5- + 3y = 2u(t) (c) Find U(s).
dF dt
3.25 [-tI2 (d) What is the area under the pulse?
+-- e -e -tl1.5] use the simpler impulse response solution to compare with real
0.5 where u(t) is the single input function of time. y(O) and dy/dt (0)
data obtained when the reactor is forced by a rectangular A.4 Derive Laplace transforms of the input signals shown in
are both zero.
= 1 - 2.5e- tl l. S + 2.5e- tI2 (A-128) pulse. The maximum expected value of C2(t) is approximately Figs. EA4a and EA.4b by summing component functions
1.25 kg mollm3. This value should be compared with the nomi- What are the functions of the time (e.g., e-tiT ) in the solution found in Table AI.
3 to the ODE for output y(t) for each of the following cases?
For the rectangular pulse response of Stage 2, substitute the nal concentration before and after the test (C2 = 1.0 kg mOI/m )
Laplace transform of the appropriate stage output, to determine if the instrumentation is precise enough to record (a) u(t) = be- 2t
Eq. A-120, into Eq. A-125 to obtain the change in concentration. If the change is too small, then the (b) u(t) = ct
pulse amplitude, pulse width, or both must be increased. 51------..
p 1.5 2 Because this system is linear, multiplying the pulse magni- band c are constants.
C2 (s) = 1.5s + 1 + (l.5s + 1)(2s + 1) tude (h) by a factor of four would yield a maximum concen- Note: You do not have to find y(t) in these cases. Just deter-
O25s tration of reactant in the second stage of about 2.0 (the mine the functions of time that will appear in y(t).
+ 6 _ 5e- . (A-129)
difference between initial and maximum concentration will ((t)
s(1.5s + 1)(2s + 1) s(1.5s + 1)(2s + 1) A.2 Solve the ODE
be four times as large). On the other hand, the solutions ob-
tained above strictly apply only for tw = 0.25 min. Hence, the d 4y d 3y d 2y dy
Again, the rightmost term in (A-129) must be excluded from
the inverted result or included, depending on whether or not
effect of a fourfold increase in tw can be predicted only by re-
solving the model response for tw = 1 min. Qualitatively, we
-
dt 4
+ 16- + 86-----:? + 176- + 105y
df d, dt
= 1
1 -------,1
t < 0.25 min. The calculation of the inverse transform of know that the maximum value of C2 will increase as tw in- using partial fraction expansion. Note you need to calculate 00 2 6
(A-129) gives creases. Because the impulse response model is a reasonably the roots of a fourth-order polynomial in s. All initial condi- t (min)
good approximation with tw = 0.25 min, we expect that small tions on y and its derivatives are zero.
t < 0.25: cf(t) = 6 + I5e- tl l. 5 - 20e- t12 (A-130) Figure EA.4a
changes in the pulse width will yield an approximately pro- A.3 Figure EA3 shows a pulse function.
tI2
t 2: 0.25: cf(t) = 1 - 2.7204e- tl l. 5 + 2.663e- (A-131) portional effect on the maximum concentration change. This (a) From details in the drawing, calculate the pulse width, two
argument is based on a proportional increase in the approxi-
Plots of Eqs. A-128, A-130, and A-131 are shown in Fig. A7. mately equivalent impulse input. A quantitative verification
The rectangular pulse response is virtually indistinguishable using numerical simulation is left as an exercise. h
from the impulse response. Hence, Kim, Casey, and Tim can Slope = a

((t)
Slope =-a
u(t)

SUMMARY
In this chapter we have considered the application of concentrate on the important properties of the Laplace
Laplace transform techniques to solve linear differen- transform and its inverse, and to point out the tech-
niques that make manipulation of transforms easier tw t
tial equations. Although this material may be a re-
view for some readers, an attempt has been made to and less prone to error. Figure EA.3 Triangular pulse function. Figure EA.4b
 Exercises A-19
A-IS Appendix A Laplace Transforms

A.9 For each of the following functions Xes), what can you What are the functions of time (e.g., e -I) in the solution for has initial conditions yeO) = 1, y' (0) = 2. Find Y(s) and, with-
A.S The start-up procedure for a batch reactor includes a each output yet) for the following cases? (Optional: find the out finding yet), determine what functions of time will appear
say about x(t) (0 :S t:S (0) without solving for x(t)? In other
heating step where the reactor temperature is gradually solutions for y(t).) in the solution.
words, what are x(O) and x(cx;)? Is x(t) converging, or diverg-
heated to the nominal operating temperature of 75°C. The (a) u(t) = as(t) unit step function
desired temperature profile T(t) is shown in Fig. EA.5. What ing? Is x(t) smooth, or oscillatory? s +1
(b) lI(t) = be-liT
(b) If Yes) = 2 ' find yet).
is T(s)?
77=7] "'72 s(s + 4s + 8)
6(s + 2)
(c) u(t) = ee- 1iT 7 = 7] '" 72 A.17 A stirred-tank blending system initially is full of water
(a) Xes) = (s2 + 9s + 20)(s + 4)
(d) u(t) = d sin wt 7] '" 72 and is being fed pure water at a constant flow rate, q. At a
10s2 - 3 A.13 Find the complete time-domain solutions for the fol- particular time, an operator shuts off the pure water flow and
(b) Xes) = -(s-2-6-s-+-1-0-)(-s-+-2-) lowing differential equations using Laplace transforms: adds caustic solution at the same volumetric flow rate q but
with concentration Ci. If the liquid volume V is constant, the
16s + 5 d 3x dx(O) dynamic model for this process is
(c) Xes) = s2 + 9
(a) - + 4x = el with x(O) = 0, - - = 0,
df dt
de
75 A.I0 For each of the following cases, determine what func- d2x(0) V- + qe = qc·
--=0 dt I

tions of time, e.g., sin 3t, e- 81 , will appear in yet). (Note that dr with e(O) = O.
you do not have to find yet)!) Which yet) are oscillatory? dx
Which exhibit a constant value of yet) for large values of t? (b) dt - 12x = sin 3t x(O) = 0 What is the concentration response of the reactor effluent
stream, e(t)? Sketch it as a function of time.
o 30 2 d2x dx dx(O)
t (min) (i) Yes) = s(s2 + 4s) (c) - + 6- + 25x = e- I x(O) = 0, - - = 0 Data: V = 2 m 3; q = 0.4 m3/min; c;= 50 kg/m3
Figure EA.S 2 (d) A process is described by two differential equations: A.IS For the dynamic system
(ii) Yes) = s(s2 + 4s + 3)
dy] dy
A.6 Using partial fraction expansion where required, find 2
iit + Y2 = Xl 2 dt = -y + 5u
x(t) for (iii) yes) = sCi + 4s + 4) dY2
s(s + 1) iit - 2y] + 3Y2 = 2X2 y and II are deviation variables-yin degrees, u in flow rate units.
(a) Xes) = (s + 2)(s + 3)(s + 4) 2 (a) u is changed from 0.0 to 2.0 at t = O. Sketch the response
(iv) yes) = s(? + 4s + 8) If Xl = e- I and X2 = 0, what can you say about the form of the and show the value of Yss. How long does it take for y to reach
s +1
solution for y]? For Y2? within 0.1 degree of the final steady state?
(b) Xes) = (s + 2)(s + 3)(s2 + 4) 2(s + 1)
A.14 The dynamic model between an output variable y and (b) If u is changed from 0.0 to 4.0 at t = 0, how long does it
s+4 (v) yes) = s(s2 + 4)
(c) Xes) = (s + 1)2 an input variable u can be expressed by take to cross the same steady state that was determined in
A.ll Which solutions of the following equations will exhibit d 2y(t) dy(t) duet - 2) part (a)? What is the new steady state?
(d) Xes) = s2 +
1
s +1 convergent behavior? Which are oscillatory? 7 + 3 dt + yet) = 4 dt - u(t - 2) (c) Suppose that after step (a) that the new temperature is
maintained at 10 degrees for a long time. Then, at t = t1, U is
_ s + 1 e- O.5s (a) Will this system exhibit an oscillatory response after an returned to zero. What is the new steady-state value of y?
(e) Xes) - s(s + 2)(s + 3) arbitrary change in u? Use Laplace transformation to show how to obtain the analy-
A.7 Expand each of the following s-domain functions into (b) What is the steady-state gain? tical solution to the above ODE for this case. (Hint: select a
partial fractions: (c) For a step change in u of magnitude 1.5, what is yet)? new time, t = 0, where yeO) = 10.)
6(s + 1) A.IS Find the solution of A.19 Will the solution to the ODE that follows reach a
(a) Yes) = i(s + 1) d 3x . steady state? Will it oscillate?
(c) df + x = sm t dx
- + 4x = f(t)
12(s + 2) dt d 2x dx
d2x -+-=4
(b) Yes) = s(s2 + 9)

(c) Yes) =
+ 2)(s + 3)
(s
(s + 4)(s + 5)(s + 6)
dx
(d) dr + dt = 4

Note: All of the above differential equations have one com-

wh", fit) G t
t < 0
o :S t < 1/h
1/h
dr dt
Show appropriate calculations using partial fraction expan-
sion and Laplace transforms.
1 mon factor in their characteristic equations. x(O) = 0
A.20 Three stirred-tanks in series are used in a reactor train
(d) Yes) = [(s + I? + If(s + 2) A.12 The differential equation model for a particular chemi- (see Fig. EA.20). The flow rate into the system of some inert
Plot the solution for values of h = 1, 10, 100, and the limiting
cal process has been found by testing to be as follows: species is maintained constant while tracer test are conducted.
A.S (a) For the integro-differential equation solution (h ---> (0) from t = 0 to t = 2. Put all plots on the same
graph. Assuming that mixing in each tank is perfect and volumes are
1
constant:
X+3X+2x=21 e-r d7 A.16 (a) The differential equation
(a) Derive model expressions for the concentration of tracer
d 2y dy leaving each tank, ei is the concentration of tracer entering
find x(t). Note that x = dx/dt, etc. where 7] and 72 are constant parameters and u(t) is the input - + 6- + 9y = cos t
dr dt the first tank.
(b) What is the value of x(t) as t ---> oo? function of time.
A-20 Appendix A Laplace Transforms

q
Appendix B
Ci

Digital Process Control SysteDls:

Hardware and Software
Process control implemented by computers has un- widespread application of computer control systems.
dergone extensive changes in both concepts and Digital control systems have largely replaced tradi-
equipment during the past 50 years. The feasibility of tional analog instrument panels, allowing computers
digital computer control in the chemical process in- to control process equipment while monitoring process
Figure EA.20 dustries was first investigated in the mid-1950s. Dur- conditions. Technological advancements, such as
ing that period, studies were performed to identify VLSI (very large-scale integrated) circuitry, object-
chemical processes that were suitable for process oriented programming techniques, and distributed
(b) If Ci has been constant and equal to zero for a long time Wi monitoring and control by computers. These efforts configurations have improved system reliability and
and an operator suddenly injects a large amount of tracer ma-
culminated in several successful applications, the first maintainability while reducing manufacturing and im-
terial in the inlet to tank 1, what will be the form of C3(t) (i.e.,
what kind of time functions will be involved) if ones being a Texaco refinery and a Monsanto chemi- plementation cost. This cost reduction has allowed
cal plant (both on the Gulf Coast) using mainframe small-scale applications in new areas, for example,
1. VI = V 2 = V3
computers. The first commercial systems were slow in microprocessors in single-loop controllers and smart
2. VI -:P V 2 -:P V 3.
execution and massive in size compared with the com- instruments (Herb, 1999). Programmable logic con-
(c) If the amount of tracer injected is unknown, is it possi- puters available today. They also had very limited ca- trollers have also gained a strong foothold in the
ble to back-calculate the amount from experimental data? ] pacity. For example, a typical first-generation process process industries.
How?
v control computer had 32K RAM and disk storage of Increased demand for digital control systems cre-
A.21 A stirred-tank reactor is operated with a feed mixture
containing reactant A at a mass concentration CAi . The feed 1MB. ated a new industry, consisting of systems engineering
W
flow rate is Wi, as shown in Fig. EA.2l. Under certain condi- The functionalities of these early control systems were and service organizations. Manufacturing companies
tions the system operates according to the model limited by capabilities of the existing computers rather moved toward enterprise-wide computer networks by
Figure EA.21 than the process characteristics. These limitations, cou- interfacing process control computers with business
pled with inadequate operator training and an unfriendly computer networks. These networks permit all com-
model in a satisfactory form for Laplace transform opera- user interface, led to designs that were difficult to operate, puters to use the same databases in planning and
tions? Why or why not? maintain, and expand. In addition, many systems had scheduling (see Chapter 19), and they also allow access
(b) For the case where the feed flow rate has been steady at customized specifications, making them extremely ex- to operator station information from locations outside
(a) For cases where the feed flow rate and feed concentra- Wi for some time, determine how CA changes with time if a pensive. Although valuable experience was gained in sys- the plant.
tion may vary and the volume is not fixed, simplify the model step change in CAi is made from CAl to CA2. List all assump- tems design and implementation, the lack of financial In the following sections, we provide an overview of
to one or more equations that do not contain product deriva- tions necessary to solve the problem using Laplace transform
tives. The density may be assumed to be constant. Is the techniques. success hindered the infusion of digital system applica- the hardware and software used for process control.
tions into the process industries until about 1970, when The distributed control system configuration is de-
inexpensive microprocessors became available commer- scribed first, followed by data acquisition for different
cially (Liptak, 2005). signal types. Digital hardware is then considered, and
During the past 40 years, developments in micro- concluding with a description of control system soft-
electronics and software technologies have led to the ware organization and architeetures.

A-21