Process Systems Steady-State Modeling and Design

Christos T. Maravelias
Department of Chemical and Biological Engineering
University of WisconsinMadison
Madison, WI 53706
September 1, 2014
c 2014 by Christos T. Maravelias
Copyright

Process Flow Diagram

Our goal is to develop a process for the production of chemical C from raw materials A and
B. Reaction 1 is carried out in reactor RC.
A + B - C

(1)

Due to thermodynamic limitations (that we will study in CBE 311), the conversion of A and
B into C is not complete; it reaches equilibrium, a state where raw materials A and B, and
product C coexist at a specific ratio Keq . Hence, the effluent of the reactor (stream S3 ) is
fed into flash tank FT to separate the heavy C from the more volatile A and B. Product C
is removed in the liquid stream S5 , whereas a fraction R of the vapor stream S4 is recycled

[S]
S7
S6
S4

S1

[M]

S2

S3

[FT]

[RC]
S5

Figure 1: Process flow diagram for the production of C from A and B.

CBE 255

Process Systems Steady-State Modeling and Design

2014

back to the process (stream S7 ). The input stream S2 of the reactor is obtained by mixing
streams S7 and S1 . A diagram of the process is shown in Figure 1.

Material Balances

To rigorously study this system, we need to have a tool that allows us to calculate the
molar flows of the three chemicals in the seven process streams. In other words, we need
to develop a model that describes this system. This model is based on the material balance equations for the three components. If Njk is the molar flow (kmol/hr) of chemical
j {A, B, C} in stream k {S1 , S2 , S3 , S4 , S5 , S6 , S7 }, then we can express the following
equations.
Mixer (M). Streams S1 and S7 are mixed to give stream S2
S

Nj 2 = Nj 1 + Nj 7

j {A, B, C}

(2)

Reactor (RC). If is the extent of the reaction (i.e. the production of C in kmol/hr), then
the molar flows in the effluent stream S3 are given by the following equations
S

Nj 3 = Nj 2 + j

j {A, B, C}

(3)

in which j is the stoichiometric coefficient of component j in Reaction 1; i.e. A = 1,

B = 1, and C = 1. If QS3 is the volumetric flow (m3 /hr) of stream S3 , then equilibrium
is described by
S

Keq =

cC3
S

cA3 cB3

NC3 /QS3
S

(NA3 /QS3 )(NB 3 /QS3 )

Keq = QS3

NC3
S

(4)

NA3 NB 3

If we assume an ideal mixture (another topic covered in CBE 311) and that the densities of
A, B and C are practically the same and equal to , then the volumetric flow of S3 can be
calculated by
S
S
S
MA NA3 + MB NB 3 + MC NC3
(5)
Q S3 =

in which Mj is the molar weight of component j. Exercise 8 considers the case in which the
mixture density is not constant.
Flash tank (FT). Flash tanks are designed to carry out specified separation tasks. In this
case, our goal is to recover C in the liquid stream S5 and the two (volatile) raw materials,
A and B, in the vapor stream S4 . If j is the fraction of the flow of chemical j that is

CBE 255

Process Systems Steady-State Modeling and Design

2014

recovered in the vapor stream (split fraction), then the molar flows in streams S4 and S5
can be calculated as follows:
S

Nj 4 = j Nj 3
S
Nj 5

= (1

j A, B, C

S
j )Nj 3

j A, B, C

(6)
(7)

Splitter (S). A fraction R of stream S4 is recycled back to the process (stream S7 ), while
the remaining stream is purged (stream S6 ). Note that the relative amounts of compounds
A, B and C in streams S6 and S7 are identical and equal to the relative amounts in stream
S4 . Hence, the molar flows in S6 and S7 are:
S

Nj 6 = (1 R)Nj 4
S
Nj 7

S
RNj 4

j A, B, C

j A, B, C

(8)
(9)

Equations 29 fully describe the process in Figure 1. But how can we solve these equations,
i.e. find values of variables that simultaneously satisfy them?

Newtons Method

In this section, we will study how we can find the root of a nonlinear equation f (x) = 0,
i.e. find the value of x at which function f becomes zero. We consider equation:
f (x) = x 3 2x 2 + 3x 6 = 0

(10)

Let x 0 = 1 be our initial guess. The value of f at x 0 = 1 is f (1) = 4 0. Our initial guess
was incorrect. To make a new guess, we have to first determine whether the root is larger
or smaller than 1. We check the derivative of f at x = 1: f 0 (x) = 3x 4x + 3 which gives
f 0 (1) = 2. Since f (1) = 4 and f increases at x = 1, the root must be larger than 1. But
how much larger? If we assume that f increases at a rate equal to f 0 (1) (and for now this
is a reasonable assumption), then the root x should be given by:
x = x0

f (x 0 )
f 0 (x 0 )

(11)

which is derived from the definition of the slope between points (x 0 , f (x 0 )) and (x , f (x ))
f 0 (x 0 ) =

f (x ) f (x 0 )
x x0

with f (x ) = 0, because x is supposed to be the root of f . Unfortunately, nonlinear

functions do not increase at a constant rate. Thus, the x calculated in Equation 11 is not
going to be the root of f (check: x = 3 and f (x = 3) = 12). However, not everything
is lost: the new x calculated in Equation 11 can be our next estimate x 1 . This is the main

CBE 255

Process Systems Steady-State Modeling and Design

Iteration k
0
1
2

xk

f (x k )

f 0 (x k )

1.0000
3.0000
2.3333

-4.0000
12.0000
2.8148

2.0000
18.0000
10.0000

2014

Table 1: Iterations of Newtons method.

idea behind Newtons method, an iterative scheme for the solution of nonlinear equations:
given a point (estimate) x k at the current iteration k, we calculate our next point using:
x k+1 = x k

f (x k )
f 0 (x k )

(12)

The ratio f (x k )/f 0 (x k ) in the right-hand side of this equation is the step we have to take
to move towards the root. The numerator tells us how big the gap is, and the denominator
tells us how fast we can close this gap. The step p k = f (x k )/f 0 (x k ) is called the Newton
step. In the Newtons algorithm we take successive steps until f (x k ) becomes (almost)
zero. The steps of the Newtons algorithm are as follows:
0. Set k = 0; choose tolerance ; guess x 0 .
1. Calculate step p k+1 = f (x k )/f 0 (x k ); new estimate x k+1 = x k + p k+1 .
2. If f (x k+1 ) <  stop: x k+1 is a root of f (x) = 0; otherwise, set k = k + 1 and go to 1.
The first two iterations of Newtons method for Equation 10 are shown in Figure 2 and
Table 1.

Solution of Systems of Nonlinear Equations

In many engineering applications we have to solve systems of nonlinear equations. Such

applications include nonlinear least squares, numerical solution of ordinary differential
equations, and steady state calculations of chemical processes as the one we examined in
Sections 1 and 2. A system of nonlinear equations in n unknowns can be written as follows
f1 (x1 , x2 , . . . , xn ) = 0
f2 (x1 , x2 , . . . , xn ) = 0

fn (x1 , x2 , . . . , xn ) = 0
or using vector notation
f (x) = 0

CBE 255

Process Systems Steady-State Modeling and Design

2014

15

f (x 1 )=12

10

f (x 2 )=2.82
x 0 =1

x 2 =2.33

t
x 1 =3

f (x 0 )=4

-5

-10
0

0.5

1.5

2.5

x
Figure 2: Two iterations of Newtons method for f (x) = x 3 2x 2 + 3x 6 = 0.
h
iT
in which f = f1 f2 fn is a vector of nonlinear functions in n variables (x1 , x2 , . . . , xn ).
To solve a system of equations, we can generalize the idea of Newtons method: given an esiT
h
k
at the current iteration k, we want to find a new estimate
timate x k = x1k x2k xn
x k+1 that is closer to the solution x . In this case, the change of multivariable function f
will depend on the partial derivatives of f with respect to variables (x1 , x2 , . . . , xn ). Thus,
the general formula for the Newton step for systems of equations is:
x k+1 = x k J(x k )1 f (x k )
in which J(x k )1 is the inverse of the Jacobian matrix

f1
f1

x1 x2
f2 f2
x1 x2
J(x)
..
..
..
.
.
.

fn
x1

fn
x2

(13)

f1
xn
f2
xn

..
.

fn
xn

Note the analogy between the scalar f (x k ) in Equation 12 and the vector f (x k ) in Equation 13, and the inverse of the derivative 1/f 0 (x k ) in Equation 12 and the inverse of the
5

CBE 255

Process Systems Steady-State Modeling and Design

2014

matrix of the partial derivatives J(x k )1 in Equation 13.

If we replace the step p k = x k+1 x k into Equation 13, we obtain
p k+1 = J(x k )1 f (x k )
or
J(x k )p k+1 = f (x k )
in which J(x k ) is a known (n n) square matrix, f (x k ) is a known n-vector and p k+1 is the
vector of n unknowns. In other words, to find the Newton step we have to solve a system
of linear equations. Hence, the Newtons algorithm for systems of nonlinear equations is
as follows
0. Set k = 0; choose tolerance ; guess x 0 .
1. Calculate step p k+1 by solving J(x k )p k+1 = f (x k ) ; calculate new point x k+1 =
x k + p k+1 .
2. If fi (x k+1 ) <  for all i, stop: x k+1 is the solution; otherwise, set k = k + 1 and go to
1.
Example 1: System of two equations
We want to solve the following system of two equations
f1 (x1 , x2 ) = 2x12 + x22 5 = 0
f2 (x1 , x2 ) = x1 + 2x2 3 = 0
h
starting at x 0 = 1.5

1.0

iT

(14)

and using  = 104 .

Solution
h
iT
h
iT
At x 0 = 1.5 1.0 , we calculate f (x 0 ) = 0.5 0.5 , which means that x 0 is not a
solution. First, we determine the Jacobian
"
#
4x1 2x2
J(x) =
1
2
Iteration k = 0.
#
"
#" #
"
#
2
6 2
p11
0.5
Step 1. At x we calculate J(x ) =
and solving
=
, we
2
1 2
0.5
p21
h
iT
h
iT
determine the first step p 1 = 0 -0.25 and the next point x 1 = 1.5 0.75 .
h
iT
h
iT
Step 2. At x 1 = 1.5 0.75 , we calculate f (x 1 ) = 0.0625 0 , i.e. x 1 is not a
solution.
"

6
1

CBE 255

Process Systems Steady-State Modeling and Design

x1k

x2k

f1 (x k )

f2 (x k )

p1k+1

p2k+1

1.5000
1.5000
1.4881
1.4889

1.0000
0.7500
0.7559
0.7560

0.5000
0.0625
3.2 104
6.7 106

0.5000
0.0000
0.0000
0.0000

0.0000
0.0119
6.1 104
0.1 106

0.2500
0.0595
3.1 105
0.0000

Iteration k
0
1
2
3

2014

Table 2: Iterations of Newtons method.

Iteration k = 1.
"

#
6 1.5
Step 1. At x we calculate J(x ) =
, and solving the corresponding system
1
2
h
iT
h
iT
of linear equations, we get p 2 = 0.11905 0.00595 and x 2 = 1.4881 0.75595 .
h
iT
h
iT
Step 2. We calculate f (x 2 ) = 0.00032 0 0 0 . So we have to continue.
1

h
The solution of this problem is x = 1.4889
if the tolerance is  = 10

0.7560

iT

, and is found after four iterations

(see Table 2).

Process Design Example

Equipped with Newtons method, we can now try to find the flows in the process in Figure 1.
We are given the following data
1. The molecular weights of the three compounds are: MA = 78, MB = 70, and MC = 148.
The common density can be assumed to be = 1000 kg/m3 .
2. The equilibrium constant Keq at the specified temperature and pressure is equal to
4.45 (m3 /kmol).
3. The separation in the flash tank is sharp: product C is fully recovered in the liquid
stream (i.e. C = 0.0), and raw materials are fully recovered in the vapor stream (i.e.
A = 1.0 and B = 1.0).
As you will see later, the process in Figure 1 corresponds to many different process designs.
Here we study a special case, where we are given the following inputs and specifications:
1. There is no C in stream S1 . The flows of raw materials A and B are 10 and 12 kmol/hr,
respectively.
2. We are also told that we should find a design that produces 9.5 kmol/hr of C.
Before using the computer to solve the flowsheet, it is helpful to first enumerate the
unknowns and equations to decide if the problem is well posed.
7

CBE 255

Process Systems Steady-State Modeling and Design

2014

Unknowns. We have the molar flows Nj k for j = {A, B, C}, k = 1, . . . , 7 giving 21 unknown
molar flows. In expressing the unit balances we introduced two new unknowns, the reaction
extent and recycle ratio, R, giving a total of 23 unknowns.
Equations. The mixer balance gives 3 equations, the reactor gives 4, the flash tank gives
6, and the splitter gives 6 for a total of 19 equations. The problem statement also defines
S
S
the feed flows (Nj 1 , j {A, B, C}) for 3 more equations and the production rate of C (NC5 )
1
for one more equation giving 23 equations. The problem should be well posed and ready
to be solved.

Exercises

Exercise 1: Code your own Newton method

Write your own Matlab code to find the root of Equation 10, f1 (x) = x 3 2x 2 + 3x 6 = 0
using Newtons method. Try different initial points. Do you find the same root? Plot f1 (x)
for x [2, 3].
Exercise 2: f2 (x) = x 3 3x 2 = 0
Modify your code for Exercise 1 to solve f2 (x) = x 3 3x 2 = 0. Try different initial points
again. Do you now find different roots? How many?
Exercise 3: f3 (x) = x 3 2x 2 x + 2 = 0
Repeat Exercise 2 for solving f3 (x) = x 3 2x 2 x+2 = 0. Try different starting points. How
many roots do you find? What is going on? Plot functions f2 (x) and f3 (x) for x [2, 3].
Exercise 4: Matlabs fsolve function
Use Matlabs fsolve to solve the system in Equation 14. Try different initial points.
Exercise 5: Solve the flowsheet
Develop a model in Matlab for the process in Figure 1. Solve it using fsolve.
Exercise 6: Production rate versus recycle ratio and best economic operation
Consider the flowsheet of Figure 1.
(a) Vary the recycle ratio, R, and plot the production rate of C and QS3 versus R for
0 R < 1. What production rate is achieved without recycle? What is the largest
production rate that can be achieved with this process?
1 We could also choose to include QS3 as an unknown and count Equation 5 as an equation to obtain 24
unknowns and 24 equations. This choice would not change the tentative conclusion that the problem is well
posed and ready to be solved.

CBE 255

Process Systems Steady-State Modeling and Design

2014

(b) Examine the QS3 versus R plot. Discuss the competing factors in selecting R for
the best economics of this process. What R do you expect to be near the economic
optimum for this process?
Exercise 7: Process flowsheet with nonsharp separation
(a) Repeat Exercise 6 with the separation factors changed to
A = 0.8

B = 0.9

C = 0.2

Assume the production rate of C is the molar flow in S5 only and you cannot recover
the C in S6 . Does production rate go up or down compared to the sharp separation
case considered in Exercise 6? How much C is lost in S6 ?
(b) How has the economic situation changed with respect to choosing the recycle ratio
for this case compared to the sharp separation case considered in Exercise 6? What
value of R do you expect to be near the economic optimum for this process?
Exercise 8: Flowsheet example with ideal mixture equation of state and nonconstant
mass density
Lets consider the case in which the mixture density varies with composition. The simplest
equation of state to use in modeling this dependence comes from treating the mixture as
a so-called ideal mixture.2 In an ideal mixture, the volume of a mixture of components is
the sum of the volumes occupied by the pure components comprising the mixture; for the
(A,B,C) mixture we have
nA VA0 + nB VB0 + nC VC0 = V

ideal mixture, volumes of components add

in which Vj0 is the specific volume (vol/mol) of pure component j, which can also be expressed in terms of the mass density and molecular weight
Vj0 =

Mj
j0

Dividing the ideal mixture equation by V and using the mass density in place of specific
volume gives
X Mj
cj 0 = 1
j
j
We can multiply this equation by the volumetric flowrate to obtain a relation between volumetric flowrate and molar flows for all streams in the process
QSk =

X
j

2 You

Nj k

Mj

ideal mixture

j0

will learn more general equations of state in the CBE 311 course.

(15)

CBE 255

Process Systems Steady-State Modeling and Design

2014

Notice if the pure component densities are all equal (j0 = 0 ), then we can factor that term
out of the sum and we have the relation that we used previously to calculate the volumetric
flowrate out of the reactor
P
Sk
j Nj M j
Sk
constant density assumption
(16)
Q =
0
If we want to know the mass density of each stream, we can calculate it from the molar
P S
S
concentration and molecular weights by Sk = j cj k Mj . If we have found the Nj k , then we
can compute Sk using this equation and the previous equation for QSk
P
Sk
X S
j Nj M j
k
Sk
=
cj Mj =
QSk
j
P
Sk
j Nj Mj
Sk
= P
Sk Mj
j Nj 0
j

(a) Compute the production rate of C as a function of recycle ratio 0 R < 1 using the
constant density assumption with
0 = 1000 kg/m3
and the other parameters given in Exercise 4. Plot production rate of C versus R for
this case. Note that this program uses Equation 16 for the reactor volumetric flowrate
QS3 .
(b) Replace Equation 16 in your program with Equation 15 and resolve the problem using
the ideal mixture equation of state with pure component densities given by
h
i h
i
0
A
B0 C0 = 800 900 1100 kg/m3
Plot production rate of C versus R for this case.
(c) Compute and plot the fractional error in the C production rate versus recycle ratio.
Is the constant density assumption a reasonable assumption for this problem?
(d) For R = 0.6, compute QSk and Sk for all seven streams i = 1, . . . 7. Print out and
clearly label these values. What are their values for the constant density case?
Exercise 9: A second reaction
Consider again the process flow diagram of Module 4 shown in Figure 3, but consider the
case of two reactions
-* C
A + B )-

K1

(17)

-* D
C + B )-

K2

(18)

10

CBE 255

Process Systems Steady-State Modeling and Design

2014

in which K1 and K2 are the equilibrium constants for the two reactions. Both reactors are
assumed to be at equilibrium and you may assume constant density for the mixture. The
parameter values are
h
i
h
i
M = 78 70 148 218 kg/kmol
= 0.8 0.9 0.15 0.1
h
i
h
i
Nf = 10 18 0 0 kmol/hr
K = 2 0.2 m3 /kmol
= 1000 kg/m3
We wish to manufacture product C, and reaction 2 is regarded as an unwanted side
reaction that converts our desired product into an undesired side product D. Define the
conversions of A and B to be
S

xA =

xB =

NA1

NA1 (NA6 + NA5 )

NB 1 (NB 6 + NB 5 )
S

NB 1

Define the yields of C and D to be

S

yC =

NC5
S

yD =

NC5 + ND5

ND5
S

NC5 + ND5

(a) Solve for the steady state as a function of recycle ratio in the range 0 R 1. Make
S
S
three plots: NC5 and ND5 (production rates) versus R; yC and yD versus R; and xA
and xB versus R.
(b) Given your plots, answer the following three questions
1. If you wish to maximize the yield of C, what recycle ratio do you use?
2. If you wish to maximize the conversion of A, what recycle ratio do you use?
3. If you wish to maximize production rate of C, what recycle ratio do you use?

[S]

S7

S4

S1

[M]

S2

S3

S6

[FT]

[RC 1]
-* C
A + B )-

K1

-* D
C + B )-

K2

S5

Figure 3: Process flow diagram for the production of C with two reactions.
11

CBE 255

Process Systems Steady-State Modeling and Design

[S]

S7

S8

S6
S4

[RC 2]
-* C
A + B )-

S1

[M]

2014

S2

[FT]

S3
[RC 1]
-* C
A + B )-

S5
Figure 4: Process flow diagram for the production of C with an additional reactor.
(c) These three objectives all seem like reasonable plant objectives. Discuss why these
objectives lead to different operations. Which operation would you choose? Justify
your answer.
Exercise 10: A single reaction but a second reactor
Consider now a single reaction as in Module 4, but a second reactor shown in Figure 4. The
parameter values are the same as the sharp separation case considered in Module 4
h
i
h
i
M = 78 70 148 kg/kmol
= 1.0 1.0 0
h
i
Nf = 10 12 0 kmol/hr
K = 4.45 m3 /kmol
= 1000 kg/m3
(a) Add the second reactor to your flowsheet model and solve for the steady state over
S8
the range of recycle ratios 0 R 1. You should add unknowns: 2 , QS8 , and NA,B,C
,
and appropriate reactor equations to include the second reactor in the model.
S

Plot production rate of C, which is now NC5 +NC8 , versus R in one figure. Plot flowrates
S
QS3 versus NC5 (not R) in a second figure for this two-reactor case.
(b) Change your set of equations so that 2 = 0 in the second reactor is the solution,
and solve the flowsheet again. Since you are not allowing any reaction in the second
reactor, N S8 should be be equal to N S6 and you should get the same results as in
Exercise 6 of Module 4. This is the one-reactor case.
S

Plot production rate of C, NC5 , versus R in one figure, and flowrate QS3 versus NC5
(not R) in a second figure for this single reactor case.
(c) Comparing these two cases, if you want to produce C at rate 9.995 kmol/hr, what
recycle ratio do you use with a single reactor. What recycle ratio do you use with two
reactors. Discuss the advantages and disadvantages of these two different designs.
12