Chapter 14 - Process Optimization

Department of Chemical Engineering

West Virginia University

Copyright - R. Turton and J.

Shaeiwitz, 2008 1
 Outline

▪ Base Case
▪ Decision Variables
▪ Topological vs. Parametric Optimization
▪ Presentation of Optimization Results

Copyright - R. Turton and J.

Shaeiwitz, 2008 2
 The Base Case
For any optimization, we need a starting point – this is our base case
• The base case may be any technically feasible process
• The further the base case is from optimum, the more optimization
must be done
• An objective function (OF) based on operating and capital investment,
e.g., EAOC or NPV should be chosen with the aim of minimizing or
maximizing the function – other OFs are possible, e.g., maximize
production of chemical B from the plant, etc.
• A Pareto analysis (80-20 rule) is often helpful to focus attention on
what should be looked at first
• Overall material balance tells us how efficiently raw materials are
being used – even though these costs are high, it may not be possible
to reduce them significantly.
Copyright - R. Turton and J.
Shaeiwitz, 2008 3
 Key Decision Variables
Even simple processes have tens if not hundreds of potential decision
variables. Therefore, it is important to identify which of these has a
significant effect on the OF.

• A Pareto analysis will identify the major costs but how those costs are
affected by changes in operating variables needs some preliminary
investigation.
• Usually, the single-pass conversion in the reactor is an important
decision variable since recycle rates are influenced directly.
• It is easy to mask the effect of conversion using different variables
such as reactor temperature and pressure.
• Interpreting results in terms of conversion is usually more straight
forward than T and P – but T and P may be easier to set in the
simulation.
Copyright - R. Turton and J.
Shaeiwitz, 2008 4
 Key Decision Variables
• If unreacted material leaves the process then decision
variables that reduce these streams by better recovery in the
separations sections (and higher conversion) should be
considered.
• High utility costs (steam and electricity) warrant decision
variables that focus on recovering heat and work. Also, there
is the whole concept of process-wide heat integration that is
covered in Chapter 15

Copyright - R. Turton and J.

Shaeiwitz, 2008 5
 Key Decision Variables
Costs associated with the production of a new facility to
produce acrylic acid via the partial oxidation of propylene
Contribution to Cost of Manufacturing, (million $/yr)

Largest
20 improvements
may not be
associated with
the largest costs

3
C3H 6  O2  C3H 4O2  H 2O
2
10 acrylic acid
5
C3H 6  O2  C2 H 4O2  H 2O  CO2
2
acetic acid
9
C3H 6  O2  3H 2O  3CO2
2

0
Electricity

Water
DI Water
Propylene

Labor
lps (feed)

FCI
lps (utility)
Cooling

Copyright - R. Turton and J.

Shaeiwitz, 2008 6
c
 Costs associated with the production of a new facility to
produce acrylic acid via the partial oxidation of propylene
Change Selectivity (Reactor Temperature)
Recover lost propylene in Off Gas
Use New Catalyst (reduce Steam to Reactor) Largest savings is
Optimize Pressure to Recduce Compression Costs associated with
Add Waste Heat Boiler to Molten Salt Loop generating and selling
Contribution to Cost of Manufacturing, (million $/yr)

steam from WHB

20

10

0
Electricity

cooling Water
DI Water
Propylene

Labor
lps (feed)

FCI
lps (utility)

Copyright - R. Turton and J.

Shaeiwitz, 2008 7
 Topological vs. Parametric Optimization

Topological Optimization involves changes in the arrangement (topology) of

process equipment. Questions that should be addressed include:
• Can unwanted by-products be eliminated?
• Can equipment be eliminated or rearranged?
• Can alternative separation methods or reactor configurations be employed?
• Can heat integration be improved (Chapter 15)?
Often, topological changes lead to large improvements in the OF and are
considered early on in the design phase.

Many of the more common topological changes were considered in Chapter 2

– Structure and synthesis of process flow diagrams

Copyright - R. Turton and J.

Shaeiwitz, 2008 8
 Topological vs. Parametric Optimization

Parametric Optimization involves the manipulation of

process variables such as single-pass conversion, reflux
ratio, product purity and yield, operating pressure, etc.
Parametric optimizations may lead to changes in the
topology of the process. For example, by increasing the
single-pass conversion, it may be possible to eliminate a
separation unit and the recycle of unused reactant.

Copyright - R. Turton and J.

Shaeiwitz, 2008 9
 Parametric Optimization – 1 variable
Parametric Optimization – Case Study No .1
Optimum reflux in a distillation column
For a fixed feed, operating pressure, and product yields and purities – the
size (diameter and height) of a distillation tower is fixed based on the reflux
ratio, R (assuming % flood, tray design, etc. are fixed)
As R increases
Reboiler duty, size, and utility costs increase Formulate costs in
Condenser duty, size, and utility costs increase terms of EAOC or NPV
and optimize
Tower Diameter increases
Number of stages + tower height decreases
EAOC

Copyright - R. Turton and J. R R

min
Shaeiwitz, 2008 10
 Parametric Optimization – 1 variable
Parametric Optimization – Case Study No .2
Consider the compression of feed air into a process that produces maleic
anhydride.
The air (10 kg/s) enters the process at atmospheric pressure and 25C and is
compressed to 3 atm in the first compressor. The compression is 65%
efficient based on a reversible adiabatic process. Prior to entering the second
stage of compression (where the air is compressed to 9 atm at the same
efficiency) the air flows through a water cooled heat exchanger where the
temperature is lowered. What is the optimum size of the heat exchanger?

Tair,2 Assume that pressure drop in pipe and exchanger are 0

cw U = 42 W/m2°C, Tcw,in = 30°C and Tcw,out = 40°C

Problem is specified by choosing Tair,2

Copyright - R. Turton and J.

Shaeiwitz, 2008 11
 Parametric Optimization – 1 variable
Parametric Optimization – Case Study No .2

Tair,2
cw

Tair,2 40
30
As Tair,2 decreases
Cexch
Ccw Formulate total cost as
EAOC and minimize
Cpower
Ccomp

Copyright - R. Turton and J.

Shaeiwitz, 2008 12
 Parametric Optimization – 1 variable
Parametric Optimization – Case Study No .3
Look at the distillation of light (volatile) materials – say propane from butane
in a depropanizer and consider the column pressure as the decision variable.
General trends as P 
Separation becomes harder – equilibrium curve moves closer to x =y line
Column has more stages and a smaller diameter
Column temperature increases (top and bottom)

 

A
Objective function, OF = EAOC   , i, n  FCI  utility costs (reboiler, condenser, pumps)
P  tower
reboiler
condenser ,etc

Copyright - R. Turton and J.

Shaeiwitz, 2008 13
 Parametric Optimization – 1 variable
Parametric Optimization – Case Study No .3
Depropanizer
At some pressure, we can
change from refrigeration to cw
in the condenser

As P increases, we must
change from lps to mps in
the reboiler
EAOC

refrig cw Column pressure

Copyright - R. Turton and J.
Topological change – eliminate refrig’n unit Shaeiwitz, 2008 14
 Parametric Optimization – Multivariable
Parametric Optimization – multivariable
When considering multivariable problems it is tempting to change one variable at a
time and proceed in a stepwise manner.
For example: Consider an OF and two variables T and P

First hold T constant at T1 and vary P to get OF T1

Popt P
Then using Popt, we vary T to get

However Topt, Popt is not the OF Popt

“global” optimum

Copyright - R. Turton and J. Topt T

Shaeiwitz, 2008 15
 Parametric Optimization – Multivariable
Parametric Optimization - multivariable
What we should do is:

T4
OF T1
T2
Topt

Popt P

Copyright - R. Turton and J.

Shaeiwitz, 2008 16
 Parametric Optimization – Multivariable
Parametric Optimization - multivariable
For more than 2 variables, the previous approach becomes very tedious so we try a response
surface technique (or lattice approach).
Consider the simulation and associated cost analysis that provides the OF as the result of a
single “case.” We choose values of decision variables in a reasonable range and pick the end
points of the range as our independent variables. Thus for three decision variables we will
have 23=8 “cases”
x3+ OF1 = f (x1-, x2-, x3-)
Find best OF and
then refine the
OF2 = f (x1 , x2 , x3 )
- - +
grid or use linear
x1- x3 - regression
techniques to fit
x1+ the data and
predict optimum
x2 -
x2 +
OF
May include a “base Copyright - R. Turton and J.
case” or center point Shaeiwitz, 2008 17
 Presentation of Optimization Results
In general, when plotting optimization results the curves
should be smooth if all the variables are continuous. For
example, if we plot conversion as a function of reactor
temperature then we should get something like this
not this

x x

T T
Copyright - R. Turton and J.
Shaeiwitz, 2008 18
 Presentation of Optimization Results
Beware of expanding the y-axis

EAOC, $1,000/yr
7,375
EAOC, $1,000/yr

8,000
7,000 7,370
6,000 7,365
5,000
4,000 7,360
3,000 7,355
2,000 7,350
1,000 7,345
0
0.0 2.0 4.0 6.0 8.0 0.0 2.0 4.0 6.0 8.0
Reactor Pressure, bar Reactor Pressure, bar
Be aware of the accuracy of
your estimates

Copyright - R. Turton and J.

Shaeiwitz, 2008 19
 Presentation of Optimization Results

Understand the main effects 3

C3H 6  O2  C3H 4O2  H 2O
2
OF 5
acrylic acid

c C3H 6  O2  C2 H 4O2  H 2O  CO2

d 2
b acetic acid
e
?
9
C3H 6  O2  3H 2O  3CO2
a f
2

0 Reactor Pressure, P 1 c

Yield d
b
e
a
f

0 Conversion in the reactor 1

Copyright - R. Turton and J.
Shaeiwitz, 2008 20
 Presentation of Optimization Results
Example - 1 for Class
Consider the reactor preheat exchanger for the DME process given in Appendix B.
T1 and T5 are fixed by process
R-201 Conversion and pressure in the adiabatic reactor are
fixed (80%)
How many variables are there in this optimization?
How would you solve this problem?

T2 T3 cw
E-202 T5
T1 T4
E-203
Copyright - R. Turton and J.
Shaeiwitz, 2008 21
 Presentation of Optimization Results
Example - 2 for Class
Consider the recovery of acetone from the reactor effluent
water T1 and composition are fixed by process
Reactor effluent contains acetone, IPA,
and hydrogen.
What are the key decision variables in
this optimization?
How would you solve this problem?
cw
Reactor
effluent

T1 T2
E-203
Copyright - R. Turton and J.
Shaeiwitz, 2008 22
 Summary

▪ The establishment of a base case is imperative when optimizing a

chemical process
▪ Topological optimization should be attempted before parametric
optimization
▪ Response surface techniques work well for processes with 3 or more
decision variables
▪ It is important to understand the reasons for trends in results

Copyright - R. Turton and J.

Shaeiwitz, 2008 23