Chapter 2

BATCH PROCESSING DEVELOPMENT

The batch mode of operation seems to have been consolidated in the manufacture of
fine and speciality chemicals such as API and pharmaceutical products, as shown in
Chapter 1. We use concepts that were introduced in Chapter 1 to describe the process
development workflow and the many methodological advances made in the synthesis of
batch processes. Then, we explain a manual procedure for sizing batch equipment that
demonstrates the philosophy behind the existing algorithms and the tools developed to
solve more complex cases. In these cases, the presence of branched recipes or highly
flexible structures, such as those found in multiproduct and multipurpose plants make
the problem intractable by manual procedures. Next, the preliminary design procedure
for sizing single product and multiproduct batch plants is presented. Then, we address
the retrofitting of these plants. A tutorial example is developed throughout the chapter
and introduces the reader to batch plant preliminary design. Finally, we present a real
case taken from the pharmaceutical industry sector and analyze various solution
strategies. All of this will be complemented in Chapter 5, in which we present the
detailed design of multipurpose plants under uncertainty.

December 2009
OUTLINE
2.1 INTRODUCTION ............................................................................................................................... 3
2.2 PROCESS DEVELOPMENT WORKFLOW .................................................................................. 4
2.3 SYNTHESIS AND PRELIMINARY DESIGN ................................................................................. 5
2.4 MANUAL SIZING PROCEDURE .................................................................................................... 7
2.5 THE SINGLE PRODUCT CASE .................................................................................................... 12
2.5.1 Sizing Solution Strategy ............................................................................................................... 16
2.5.2 The Synthesis Problem ................................................................................................................. 19
2.6 THE MULTIPRODUCT CASE ....................................................................................................... 20
2.6.1 The optimization algorithm .......................................................................................................... 22
2.6.1.1 Parallel equipment ............................................................................................................... 23
2.6.2 Retrofitting studies ....................................................................................................................... 25
2.6.3 Intermediate storage ..................................................................................................................... 26
2.6.3.1 Plant design without intermediate storage .......................................................................... 28
2.6.3.2. Intermediate Storage Location ............................................................................................ 28
2.6.3.3. Preliminary Design with Intermediate Storage .................................................................. 29
2.6.4. Optimum Design with Intermediate Storage ............................................................................... 31
2.7 CASE STUDIES................................................................................................................................. 31
2.7.1 Base case design ........................................................................................................................... 31
2.7.2 Retrofit design: plant capacity expansion .................................................................................... 33
2.7.3 Retrofit design: plant design modification ................................................................................... 36
2.7.4 Industrial case stydy: penicillin plant........................................................................................... 38
2.7.4.1 Base case design .................................................................................................................. 44
2.4.7.2 Incorporating intermediate storage ..................................................................................... 45
2.4.7.3. Reuse of equipment ............................................................................................................. 48
2.8 SUMMARY ........................................................................................................................................ 49
NOTATION............................................................................................................................................... 50
REFERENCES ......................................................................................................................................... 51
EXERCISES .............................................................................................................................................. 52
2.1 INTRODUCTION

Chemical plants are commonly designed for fixed nominal specifications, such as the
capacity of the plant and the type and quality of raw materials and products. In addition,
they are designed with a fixed set of predicted values for the parameters that specify the
performance of the system, such as transfer coefficients or efficiencies and the physical
properties of the materials used in the process. However, chemical plants often operate
under conditions that are quite different from those considered in the design. A plant
may have to process different feeds, produce several products and operate and meet
specifications at various levels of capacity. Alternatively, it may show significant
uncertainty in the parameter values. It is essential to take all of these factors into
account when we consider the design. Thus, a plant has to be designed with enough
flexibility to meet the specifications, even when it is subjected to various operating
conditions.
In practice, empirical over-design factors are widely used to size equipment, in the
hope that these factors will compensate for all the effects of uncertainty in the design.
However, this is clearly not a very rational approach to the problem, since there is no
quantitative justification for the use of such factors. For instance, it is not clear what
range of specifications an over-designed plant can tolerate. In addition, the economic
performance of an over-designed plant will probably not be optimum, especially if the
plant design has only been optimized for nominal conditions.
In the context of the theory of chemical process design, the need for a rational
method of designing flexible chemical plants stems from the fact that there are still
substantial differences between designs that are obtained with currently available
computer aids and designs that are actually implemented in practice. These differences
include the issue of introducing flexibility into a plant’s design. This is a very important
stage in the design procedure, since the main concern is to ensure that economically the
plant can to meet the specifications for a given range of operating conditions. Clearly,
the dynamic performance characteristics of the chemical processing system should be
considered in such a design procedure, to ensure smooth operation of the plant.
However, to provide a valid framework for such considerations, feasible steady-state
operation of the plant must be guaranteed under the various conditions that may be
encountered. Here, the term flexibility indicates precisely this characteristic, namely
that the existence of feasible steady-state operation is ensured for every allowable
situation that is specified.
Present trends indicate that true flexibility can be better achieved by designing
batch-processing networks that operate in a variety of expected and unexpected
production scenarios. This is particularly true in the pharmaceutical industry, which was
attracted for some time by the great economies of scale and high profits obtained by
continuous processing. Now, the batch mode of operation seems to have been
consolidated in the manufacture of fine and speciality chemicals such as API and
pharmaceutical products, as seen in Chapter 1.
Notwithstanding this trend in production economics, the batch process design has
remained poorly understood and still has fewer appropriate support tools and
methodologies than continuous processing. In this chapter, we use the concepts
introduced in Chapter 1 to describe the many methodological advances that have been
made in the synthesis of batch processes. We also present a manual procedure for sizing
batch equipment to illustrate the philosophy behind the algorithms and tools that have
been developed to solve more complex cases. In complex cases, the presence of
branched recipes or the highly flexible structures offered by multipurpose plants makes
the problem intractable by manual procedures. Then, we describe the preliminary design
procedure for sizing single product and multiproduct batch plants. Finally, we deal with
the retrofitting of these types of plants. A tutorial example in this chapter introduces the
reader to preliminary design of batch plants. Chapter 5 (Process design optimization)
complements this chapter with information on design issues that is related to the
integrated design of multipurpose plants with production scheduling and design under
uncertainty.

2.2 PROCESS DEVELOPMENT WORKFLOW

The design of batch processes for the manufacture of pharmaceutical products differs
essentially from the conceptual design of continuous processes. According to Sharrat
(1997), in a batch process, special consideration must be given to the aspects discussed
below.
The typical stages in batch process design for the manufacture of fine and speciality
chemicals are shown in Figure 2.1.

Route Process Site Detailed Construction

Selection Development Decisions Design & Production
tt

Figure 2.1 Main stages of batch process evelopment

Process chemistry is of essential importance in batch manufacturing, mainly in the

early stages of development when the route selection is made, in the use of solvents as
reaction media and in the high complexity of the substantial number of reactions
involved.
As indicated in Chapter 1, chemical route selection is particularly complex in the
case of pharmaceutical API processes and has a major impact on the final process
performance. The high number of alternative routes and starting materials to be
considered and the fact that early decisions have to be made with limited information
makes the route selection problem extremely complicated. Appropriate tests must be
carried out to identify potential health problems in later stages of process development.
Safety and environmental aspects should also be evaluated. The final decision on route
selection should integrate all these criteria.
Batch processes are also characterized by extensive use of solvents, e.g. solid
liquefaction and evaporation, reaction media and reactor temperature control. The use of
solvents influences batch process unit operations due to the safety and environmental
issues associated with waste, contaminated solvents and volatile organic compounds.
Selection of the appropriate solvents is again a decision that should integrate all of the
aspects related to their use, to improve process performance.
Having identified the route and the solvent that will give the best process
performance, an initial recipe is developed, usually by a chemist. The recipe includes a
detailed description of each process stage, its operating conditions and processing times,
the material transfer between stages, and its duration, with an indication of the source
and destination. In fact, the recipe contains all the necessary information for the
production of the final product at full scale. Thus, it constitutes the equivalent of a
flowsheet in a continuous process.
 One of the main problems at this stage of process development is the scale-up,
since the recipe is developed by a chemist at laboratory scale under operating conditions
that may cause inefficiency in the full-scale process, unless engineers get fully involved
at this early stage of development.
Further development of the batch process design should consider the specific
characteristics of batch processes examined in Chapter 1. Decisions should be made
regarding:

• Which units should be batch and which ones should be continuous.

• The assignment of units to tasks.
• The units’ modes of operation (in-phase, out-of-phase, etc.).
• The location and size of intermediate storage.

The next section describes a synthesis procedure that systematically considers the
alternatives that emerge from these decisions.
Process synthesis is complemented with scheduling and simulation tools. As a
good first approximation, scheduling considerations may be neglected in the
preliminary design. Production scheduling provides decision support, which helps to
improve designs so that plants operate smoothly, and to find and maximize process
integration opportunities. This is particularly true for the case of multipurpose plants in
which simultaneous processing of different products is a main feature of the design.
Consequently, scheduling considerations are involved. Dynamic simulators are
particularly useful to investigate possible control strategies and predict the accurate
behaviour of critical unit operations, as discussed in Chapter 5.

2.3 SYNTHESIS AND PRELIMINARY DESIGN

The design of batch plants can be considered at three different levels:

a) Overall process flowsheeting

b) Preliminary design of process equipment units
c) Detailed mechanical design of individual pieces of equipment.

The first level (a) is subject to single product recipe specifications and must take
into account potential interaction between product lines, items that are in parallel in-
phase or out-of-phase and the presence of intermediate storage, all of which must be
appropriately optimized to obtain the best possible design (the synthesis step). The last
level in the design procedure (c) will lead to general purpose equipment and standard
items, so no further optimization will be possible. Therefore, only the process synthesis
mechanisms and the preliminary design (the sizing step) will be considered in this
chapter.
Prior to the preliminary design of this type of process, the following information is
required:
• A list of products and the amount of each one to be manufactured.

• The available production time.

 • The individual recipes for each product.
• The size/duty factors for each task.
• The material flow balance for each task.
• The manufacturing process and the flow characterization.
• The available equipment to perform each task that should include
 the cost/size ratio
 the processing time for each task related to the batch
• A suitable performance function involving capital and/or operating cost
components

To determine:

i) The number of equipment stages and the tasks allocations

ii) The intermediate storage requirements
iii) The parallel equipment items in each stage
iv) The sizes and capacities of all equipment items

Thus, the objective of the predesign problem is to optimize the size of the
processing units by minimizing the selected performance function under specific plant
operating conditions.
The following assumptions are made at the predesign stage and are subsequently
modified on the basis of additional information from production planning:

• Only single product campaigns are considered. When storage costs are
substantial, the demand pattern will determine the proper ordering of production
campaigns.
• Each equipment unit is utilized only once per batch.
• Parallel equipment units are assigned to the same task and the out-of-phase mode
is also permitted.
• Only an overlapping mode of operation is considered.
• A continuous range of equipment sizes is assumed to be available.
• Multiple equipment items of a given type are identical.
• Instantaneous batch transfer mode (ZW transfer rule).

The aforementioned three variables (i, ii and iii) define the structure of the process
network and constitute the synthesis problem, while the last variable (iv) refers to the
sizing problem.
2.4 MANUAL SIZING PROCEDURE

A hand calculation procedure was described by Flatz (1980) to determine the

dimensions of the various items of equipment and to allocate available production time
to the various production processes (the sizing problem). In principle, the strategy can
be used for multiproduct plants to find the alternative production scenarios that have the
lowest investment costs (the synthesis problem). The procedure can be applied to:

• Find flowsheet alternatives.

• Allocate production time to two items of equipment.
• Optimize (e.g. to the minimum capital investment) the plant facilities for standard
equipment size.
• Size a plant when only some of the equipment sizes are known.
• Discuss certain procedures for capacity utilization and production planning.

The multiproduct plant-design problem can be stated as follows:

• Given:

 The quantities of the product to be produced during a given interval

• Determine:
 The dimensions of the various equipment items in the plant.
 The allocation of available production time to the various production
processes.

The solution strategy can be summarized in the following six steps:

Step 1. Sizing calculation: dimension equations.

Step 2. Equipment selection: which will be fully utilized?
Step 3. Equipment sizing.
Step 4. Safety aspects and standard sizes.
Step 5. Equipment use time calculation.
Step 6. Debottlenecking: size-limiting and time-limiting stages.

Example 2.1

A pharmaceutical company wishes to expand the production capacity of intermediate

products that are used in other sections of the same plant at another location. This
expansion is achieved by installing a new set of equipment units at the limiting stages of
the current plant structure.

To meet current demand, the manufacture of the intermediates A and B requires

immediate attention:
Product A

Currently, an external manufacturer supplies this product. The cost to the company is
2.90 $/kg, including transport and miscellaneous expenses.
A series of test runs have been carried out at laboratory scale to analyze the
feasibility and economy at industrial scale. The following recipe was obtained as a
result:

1) Dissolve 50 g of reactant R in 250 cm3 (197 g) of ethanol by stirring at room

temperature. The dissolution time is approximately 1 h.
2) Simultaneously, 250 cm3 of ethanol is introduced in a 1000 cm3 jacketed vessel
and heated to boiling point.
3) For 3 h, the solution of R drops into the flask, while the EtOH boils. During the
reaction, S is lost as a gas through a cooled trap. There is so much froth that the
flask is completely filled.

R( EtOH ) → A( EtOH ) + S( g ) ↑

4) At the end of the reaction, the solution of crude A is cooled to 40ºC. We obtain
25 g of crude A (0.907 kg A) in solution, which is equivalent to 22.7 g of pure
product A.

7,88 kg 7,88 kg
Et0H 2,00 kg R Et0H

0,25 h 0,25 h

1,00 kg S
0,16 kg Et0H

1,00 h

3,00 h

0,25 h
1,00 kg A (90.7%)
15,6 kg Et0H

Figure 2.2 Data corresponding to production of A

Additional information:

• Safety: the usual precautions for ethanol management and waste gas purification
are required.
• Environment: waste gas purification is required for the ethanol.
 • The actual volume of the reactor vessel of 1,000 cm3 nominal capacity (which
becomes fully occupied by foam formation) is 1,250 cm3. The addition of
chemicals to prevent foam formation is not considered, due to the product
sensitivity.
• The manufacturing, reactant, manpower, utility and miscellaneous costs
(excluding equipment amortization costs) amount to 1.80 $/kg.
• The penalty for not producing A is evaluated at 1.10 $/kg (the difference between
the purchase price and the production cost).
• As a first approximation, we consider that size factors, reaction times, etc. can be
directly extrapolated to full plant capacity. This hypothesis will have to be
revised in the detailed design stage.

Product B

Currently, this product is supplied by a subsidiary of the same company on a site that is
500 km away. Product B is mainly consumed in this other facility. However, the
production of B has now reached its upper limit.
Since the production of B is similar to that of A, it has been decided to produce B
locally using the same facilities as those used to produce A, as far as possible. The
reaction and economics of product B are well known:

X ( aq ) + Y( s ) → B( s )

The following recipe was obtained from experimental data collected at the site
where B is currently produced:

• Dissolve raw material: Initially, the raw material X is dissolved in water. If 100
kg of X are added to over 100 kg of water at room temperature under stirring
conditions, the dissolution time is about 1 h.
• Separately, 100 kg of Y are added to another 100 kg of water in a jacketed vessel
that is heated to 100°C.
• Reaction: the first solution is added to the jacketed vessel, which already contains
the second solution, at boiling point for 0.5 h (information about the heat of the
reaction is not yet available).
• After completion of the reactions, the reactor content is cooled to 30 °C (which
takes about 1.5h in the pilot plant). Thus, 200 kg of product B in water suspension
are obtained.

The purity of the product is about 90%, which is more than enough for most of the
applications envisaged. However, a B of higher purity is also planned for additional
applications of higher added value.

Additional information:

• Safety: there are no particular problems.

• Environment: there are no particular problems.
 • The volume of solids and thermal expansions has to be taken into account in the
design calculations, especially in the reaction step. These aspects can be
expressed in terms of the minimal density in the reaction system, which is
evaluated at 2,000 kg/m3 in the first step (dissolution) and 1,200 kg/m3 in the
second step (reaction).
• The production costs are estimated to be the same as those currently incurred in
the existing plant. However, the company will save transport costs, which amount
to 0.15 $/kg (raw material transport costs will remain the same).
• The production of B in the new plant will meet the demand of the present plant;
hence the availability of B will no longer be a limiting factor. Consequently, it is
essential to invest in manufacturing this product. Besides, the capacity to
overproduce B will give the company a predominant position in this competitive
market.
• It is assumed that this pilot plant data can be extrapolated to full-scale production.
0,50 kg X
0,5 kg Y

0,5 kg H2O
0,25 h 0,25 h

0,25 h

0,5 kg H2O
1,00 h
0,25 h

0,5 h 9,00 h
0,25 h
1.00 kg B
1.00 kg H2O

Figure 2.3 Data corresponding to production of B

Plant design specifications

Products A and B use the same kind of equipment units, thus a multiproduct plant seems
to be the most appropriate alternative. Equipment 1 will be used for the dissolution step
and the reaction step will take place in Equipment 2.
The demand is 32,000 kg/year of pure product A (or its commercial grade
equivalent) and 18,000 kg/year of product B of similar quality to that obtained in the
present facility.
The plant operates 24 hours/day, 5 days/week with 4 weeks/year of holidays, which
are taken at the same time as the rest of the plant and used for general maintenance.
Due to the inventory limitations and product characteristics, about 8 x 2 = 16
annual changes of the product will need to be made if the multiproduct plant alternative
is chosen. During each product change, a period of time will be lost through operations
such as cleaning and maintenance. This extra time can be evaluated as about half a week
for each change.
The plant has utilities for steam production (6 bars) and a cooling system (water at
14°C). It is assumed that any capacity expansion in the utilities due to the new
production patterns will not be acceptable, so no extra investment in utilities should be
needed.
For design purposes, it will be considered that the size factors and reaction times
obtained from the laboratory and/or industrial practice can be applied to the final
equipment units.

Cost of equipment units

The equipment unit used for dissolution has the following simplified cost equation:

Cost1 ($) = 15000 + 5000 V (m3 )

The cost of the unit used for reactions can be estimated as:

Cost2 ($) = 14000 + 6000 V (m3 )

where V is the nominal size in m3.

Solution

The data provided in the text and Figure 2.2 and Figure 2.3 are summarized in Table
2.1.
The time horizon is 52 weeks. Since 4 weeks are dedicated to maintenance and
another 8 weeks are lost due to product changeovers, there are 40 working weeks of 5
days per week and 24 h per day, which makes a total of 4,800 working hours.
The manual sizing procedure is shown in Table 2.2. In this case, the time-limiting
stage is the second (the reaction stage) for both products.
The results indicate that Unit 2 is the optimum size for Product 2 at Stage 2
(limiting cycle time), but is oversized for Product 1.
This manual strategy can also be used to attain the same batch size as the unit sizing
criteria. Calculations for this case are shown in Table 3.1 and clearly indicate a worse
solution, as both units are greatly oversized, and also underutilized in the case of
product B. These results are not surprising, given the large difference in the size factor
values.
Table 2.1 Data summary for the Example 2.1
Demand
Product A Product B
Demand kg 32000 180000
Penalty ($/kg) 1.10 ∞
Production horizon 4800 h
Size Factors Operation times
SAj (m3/kg) SBj (m3/kg) tAj (h) tBj (h)
Unit 1 0.011 0.00050 4.5 8.0
Unit 2 0.055 0.0017 2.0 12.0
Cost Límits
αj ($) βj ($/m3) γj Mínimum (m3) Máximum(m3)
Unit 1 15000 5000 1.0 0.2 10
Unit 2 14000 6000 1.0 0.2 10

This manual procedure has a number of limitations. Among others, the following
situations are not considered:

• The limiting cycle time stage could be different for each product.
 • The operating times could vary with equipment size.
• The presence of semi-continuous units and the corresponding transfer times.
• The increase in problem size (number of products and processing stages)

Table 2.2 Best plant sizing for the time limiting stage
Product A Product B
Time limiting stage (T) Stage 2
Size Factor (Si) 0.0551 m3/kg 0.001667 m3
Demand (Di) 32000 kg 180000 kg
Cycle Time (Ti) 8h 12 h
Time required TMi = TiDiSi 14100 m3 h 3600 m3 h
Time per m3 (TM=Σ TMi) 17760 m3 h
Available time (H) 4800 h (40 weeks)
Size (V=TM / H) 3.7 m3
Time distribution (Hi=TMi/V ) 3823 h 977 h
Size Factor (Si) 0.0110 m3/kg 0.000500 m3
Demand (Di) 32000 kg 180000 kg
Cycle time (Ti) 8h 12 h
Time required TMi = TiDiSi 2816 m3 h 1080 m3 h
Size (Vi=TMi / Hi) 0.74 m3 1.11 m3
Final decision 3.7 m3 for Unit 1 1.11 m3 for Unit 2

Table 2.3 Best plant sizing to achieve the same batch size
Product A Product B
Limiting Stage (T) Stage 2
Size Factor (Si1) 0.0110 m3/kg 0.000500 m3
Size Factor (Si2) 0.0551 m3/kg 0.001667 m3
Demand (Di) 32000 kg 180000 kg
Cycle Time (Ti) 8h 12 h
Batch Size Bi = TiDi/Hi 256000/ HA 2160000/ HB
Available time (H) 4800 h (40 weeks)
Time distribution (Hi=TiDi/Bi ) 508.6 h 4291.4 h
Size V1= Si1 Di /(Hi / Ti) 5.54 m3 0.25 m3
Size V2= Si2 Di /(Hi / Ti) 27.73 m3 0.84 m3
3
Final Decision 5.54 m for Unit 1 27.8 m3 for Unit 2

All the above indicates that the design procedure must be rationalized and
automated for optimum design. We will achieve this as follows. First, we introduce the
formulation and an automated solution procedure for the single product case. Then, the
multiproduct case is considered. Finally, we present the use of retrofit design for
capacity expansion situations.

2.5 THE SINGLE PRODUCT CASE

In the most general case, it is assumed that the product is manufactured using a process
that consists of M batch equipment types and K types of semi-continuous units. The K
types of semi-continuous equipment form L semi-continuous stages. Parallel batch units
operating in-phase or out-of-phase, which increase capacity or reduce the cycle time
respectively, will be allowed. Although the economy of scale dictates that parallel semi-
continuous units operating in-phase will always be more expensive than one large unit,
the formulation has been kept general for the case in which the capacity of available
equipment is exceeded. The size Vj of batch equipment of type j can be calculated once
the size factor Sj for the same equipment and the batch size of the product are known, as
follows.

V j = B·S j j = 1,..., M (2-1)

When precise dynamic models of the equipment are not available, processing times
Pj can be calculated by the general expression (see Chapter 1).
cj
Pj = a j + b j ·B j = 1,..., M (2-2)

Taking into account the relationships established in the previous chapter, it has
been shown (Yeh and Reklaitis, 1987) that the optimal sizing problem for the
overlapping mode of operation can be formulated as a nonlinear programming problem
(NLP) when a capital cost objective function is used:

Minimize f (Vj, Rk)

subject to

B·S j
Vj ≥ BS j j = 1,..., M (2-3)
mij

B ⋅ Dk
θl f ≥ k ∈ K jf l = 1,..., L (2-4)
Rk

B ⋅ Dk
θle ≥ k ∈ K ej l = 1,..., L (2-5)
Rk

θl f + Pj + θle
T≥ j = 1,..., M l = 1,..., L (2-6)
moj

T ≥ θl j = 1,..., L (2-7)

These constraints indicate that the sizing of batch units must be carried out in such
a way that the appropriate capacity for processing the product is met, taking into
account mij parallel units operating in-phase (2-3). In addition, filling and emptying
times are limited by the maximum semi-continuous time involved in semi-continuous
equipment k with an Rk processing rate and duty factor Dk (2-4) and (2-5) respectively.
The allocated time for processing batch size cannot be less than that required by any of
the batch operations involved and is calculated by the general expression (2-6) with m oj
parallel units operating out-of-phase. The cycle time cannot be less than that required by
any of the semi-continuous operations involved (2-7). Additionally, the total production
time should be less than or equal to the available production time H (Yeh and Reklaitis,
1985):

!Q"
% & ⋅T ≤ H (2-8)
'B(

where Q is the amount of product required over production time horizon H.

Finally, due to the available ranges of equipment sizes

V jmin ≤ V j ≤ V jmax (2-9)

Rkmin ≤ Rk ≤ Rkmax (2-10)

.

The function to be minimized is of the form:

γ γ
f (V j , Rk ) = ∑ moj mij (α j + β ·V j j ) + ∑ moj mij (α j + β ·R j j ) (2-11)
j k

In pure batch case, this is reduced to

γ
minimise∑ moj mij (α j + β ·V j j ) (2-12)
j

If moj and mij are fixed, at the optimum

QT
B= (2-13)
H

Then, the only variable is B, restricted to:

! mij ·V jmin " ! mij ·V jmax "

max $ %% ≤ B ≤ min $$ %% (2-14)
$ S
& j ' & Sj '

which is a single-variable optimization for the limiting batch size B:

min f ( B) (2-15)

subject to
cj
a j + b j ·B
T≥ j = 1,..., M (2-16)
moj

Equations (2-3) and (2-14)

 If processing times Pij are known and assumed to be independent of batch size B,
the calculation of cycle time becomes straightforward

T = max(a j ) j = 1,..., M (2-17)

and the batch size can be readily obtained

Q·T
B= (2-18)
H

Therefore, no optimization is performed, except the choice of moj and mij .

Example 2.2

A three-stage (reaction-separation-purification) batch plant produces 2,000 tonnes of

product C. The plant is assumed to operate 8,000 h a year. Transfer times are negligible
compared with processing times and will be disregarded. The plant data are summarized
in Table 2.4.

1 2
3
C

Figure 2.4 The simplified layout of the three batch stages Example 2.4 showing the
reaccion (1), separation (2) and purification stages.

Table 2.4 Plant data for Example 2.2

Stage, j 1 (Reaction) 2 (Separation) 3 (Purification)
Size Factor, Sj (m3/kg) 2.5 2.0 1.5
aj 5.0 8.0 12.0
bj 10-4 10-8 5·10-4
cj 2.0 1.0. 1.0
Q (kg) 2·106
H (h) 8000

The following case studies will be considered. First, processing times are assumed
to be independent of batch size and equipment in parallel is not allowed (Case 1).
Second, processing times are calculated by their expressions, which are dependent on
batch size, and equipment in parallel is permitted (Case 2).

Case 1 T ≠ f ( B), cj = 0, mj = 1
 T = max {5,8,12} = 12 h
(2 ⋅106 )12
B = QT / H = = 3000 kg
8000

Thus, the sizes of all three batch equipment items can be readily obtained

V = (2.5 ⋅ 3000, 2 ⋅ 3000, 1.5 ⋅ 3000) → (7500,6000, 4500)

Alternatively, if m j = 2 (two operating units working in parallel out-of-phase) at

the limiting stage (purification step, Unit 3), separation (Unit 2) becomes the
bottlenecking stage

T = max {5,8,12 / 2} = 8 h
(2 ⋅106 )·8
B= = 2000 kg
8000

The sizes of the three equipment units are:

V = (5000, 4000, 3000)

Although a substantial reduction in plant size is obtained, an investment must be

made in the second unit in parallel at the purification stage. A consideration of the cost
of equipment in the objective function plays a key role in determining the best plant
design.

Case 2 T = f ( B), c j ≠ 0, mj = 1,...,3

If mj = 1, then
8000
T= 6
·B = 4 ⋅10−3 B
2 ⋅10

Subject to the following batch size constraints:

5 + 10-8 B 2 serves to set UB on B

-3 -4
4 ⋅10 B ≥ 8 +10 B
12 + 5 ⋅10-4 B serves to set LB on B

This leads to the solution of an optimization problem in B and mj.

2.5.1 Sizing Solution Strategy

Previous examples have shown the increasing complexity of the solution to the sizing
problem. Alternative production routes must also be assessed in the overall plant design.
This constitutes the synthesis problem. During the course of production network
synthesis, the sizing problem must be solved repeatedly. Consequently, a very efficient
strategy has been proposed to alleviate the computational burden associated with the
synthesis step in preliminary plant design. This strategy uses an approximate method for
sizing the equipment items. The basic motivation for the algorithm is that, in an
economical design, the available production time H will be fully utilized and all batch
subtrains will be as close as possible to the limiting cycle time. Thus, the results will be
similar to the continuous process performance. The only requirement for the algorithm
is that the objective function should be separable, in terms of the contributions of the
capacities of the separate equipment items (Yeh and Rekalitis, 1987).
For a single product plant with a limiting batch size B, we can initially assume that
all of the semi-continuous equipment k is set at its maximum processing rate Rk = Rkmax .
We can use Equations (2-1) to (2-8) to calculate the total processing time. If H is
violated (Equation (2-8)), the given limiting batch size is infeasible. In all other cases,
the best sizing solution can be found by adjusting the Rk. Once a suitable performance
function has been established, an improved value of B can be obtained by using a single
variable search algorithm. This iterative procedure continues until the minimum value
of the performance function is reached. The algorithm proceeds in the following steps:

1. For a given B, compute V j and Pj by Equations (2-1) and (2-2).

2. Check the feasibility

a) set Rk = Rkmax ; compute θl for all semi-continuous stages l
b) compute t j and T
QT
c) if > H , infeasible (increase B)
B
BH
Otherwise, set T =
Q
3. Adjust Rk .
Rank semi-continuous subtrains by decreasing cost and begin with the first l in
the list
{ }
a) Set θ1 = min m jT − Pj − θ je , m j −1T − Pj −1 − θ jf−1 , T
in which the semi-continuous Stage 1 is between batch stage j and j-1.
Compute all associated Rk .
b) If any Rk < Rkmin , reset to Rkmin
c) Choose the next l in the list and go to (a).
If none, continue.

4. Compute the total cost.

5. Use a single variable search algorithm to determine an improved B value and
repeat Steps 2-5 to achieve the minimum cost.

Example 2.2

Consider a batch plant that produces a single product C in two main batch stages: 1) a
reaction and 2) a crystallization stage. Semi-continuous (S/C) equipment items A and B
are used to feed the reactant into the batch reactor and transfer the intermediate product
from the reaction to the crystallizer, respectively. Plant data are summarized in Table
2.5.

S/C Batch S/C Batch

2
A 1 B C

Figure 2.5 The simplified layout of the batch plant used in Example 2.2 showing the
semicontinuous (S/C) and batch stages.

Table 2.5 Plant data for Example 2.3

Stage, j A 1 B 2
Sj (m3/kg), Dj 2.5 2.5 1.0 2.0
Pj (h) 5.0 8.0
R max
j
(kg/h) 2500 4000
min
R j
(kg/h) 500 800
1/ 2 1/ 2
Costs ($) 600RA 1200V1 80RB 10V2
Q (kg) 2·106
H (h) 8000

1st Iteration

1. Choose B = 4000 kg
Batch equipment sizing

V1 = S j ⋅ B = 2.5 ⋅ (4000) = 104

V2 = S j ⋅ B = 2 ⋅ (4000) = 8 ⋅103
Processing times
2.5 ⋅ 4000
θ A = DA B / R = =4h
A
2500
1⋅ 4000
θB = =1 h
4000
t1 = θ A + P1 + θ B = 4 + 5 + 1 = 10
t2 = θ B + P2 = 1 + 8 = 9
Limiting cycle time
T = Max(t1 , t2 ,θ A ,θ B ) = 10 h
Total production time
Q·T 2·106 ·10
= = 5000 h
B 4000
2. Check feasibility
 QT
= 5000 h < H = 8000 h
B
Maximum allowable cycle time
8000·4000
T= = 16 h
2·106
3. Rk adjustment: Readjust θ for most expensive s/c unit, which is B

θ B = min {16 − 5 − 4, 16 − 8, 16} = 7 h

By comparing the value obtained for θ B = 1, it should be obvious that
semicontinuous unit B is running under minimum specification conditions,
namely
4000 ⋅1.0
RB = = 571 < RBMIN
7
Therefore, let us set
RB = RBmin = 800 kg/h
The adjusted time spent by θ B is now
4000 ⋅1.0
θB = = 5.0 h
800
Let us use the remaining time with A

θ A = min {16 − 5 − 5,16} = 6

4000 ⋅ 2.5 5 3
RA = = ⋅10 < RAMAX
6 3
4. Calculate cost:
CA + C1 + CB + C2 = $288, 495

5. Continue searching on B

For instance, by choosing B=3,000, the cost obtained using the above algorithm
is 256,700 $.

Any variable optimization method (e.g. the golden section) can be used to find the
minimum cost. The heuristic procedure presented here was evaluated by its authors
(Yeh and Reklaitis, 1987) against an optimal (but not discretized) solution obtained
using a GRG2 code. They concluded that their algorithm was faster (typically, 3 to 20
times faster), more robust and required much less computer storage.

2.5.2 The Synthesis Problem

As mentioned in Chapter 1, the key controlling element in batch process synthesis is the
limiting cycle time of the product (LCT). The LCT is in fact the bottleneck that limits
the use of equipment in the process. In an ideal network, all processing stages will be
fully utilized over the LCT (TL). However, differences in the processing times of
individual equipment units make it impossible to obtain the ideal network in real cases.
Nevertheless, TL and idle times can be reduced by the use of appropriate strategies,
including the following.

These are as follows:

• The use of parallel units in-phase and out-of-phase

• Task merging and splitting policies
• Task transfer policies
• The use of intermediate storage

Appropriate use of these strategies will ensure that the production network can
respond to variable demand and will enable new products or variants of the present
products to be introduced. The example used in the previous chapter compares the
advantages and disadvantages of the alternatives provided by the application of the
above strategies.
The limiting batch size also plays an important role in the retrofitting problem,
since the plant production rate can be increased either by increasing the limiting batch
sizes of the products or by reducing the limiting cycle times of the products. In order to
increase the limiting batch sizes, parallel batch units that operate in-phase can be added
at the size limiting stage.
Synthesis procedures will be introduced in the global algorithm for the general case
of a multiproduct plant, as described below.

2.6 THE MULTIPRODUCT CASE

The general multiproduct problem can be described in the same terms as the single
product case, although the computation time and complexity of the solution increases
significantly. Therefore, reasonable simplifications must be introduced. The
incorporation of appropriate heuristic rules helps to simplify the solution to this
problem. In the work of Espuña et al. (1989), the objective function only considers key
elements that affect optimum sizing at the preliminary stage and are associated with
plant investment costs. Such costs are the sum of individual process operational costs,
which are assumed to remain unaltered, regardless of the decisions taken during the
preliminary design procedure. The production costs that are incurred to meet a specific
market demand may eventually generate losses. Consequently, the objective function
contains an additional term that allows decisions to be made on whether an increase in
productivity is justified by a potential increase in profit or whether a reduction in
investment would be more appropriate, at the expense of leaving some demands
unsatisfied.
Therefore, the formulation of the objective function is:

M L N
γ
f ( Di ,V j , Rk ) = ∑ $(moj ⋅ mij %) ⋅ $(α j + β j ⋅V j j %) + ∑ mko ⋅ mki ⋅ $(α k + βk ⋅ Rkγ k %) + ∑αi ⋅ [Qi − Di ]
j =1 k =1 i =1
(2-19)
 Plant and process specifications and problem assumptions are as summarized in
Section 2.3. The minimization of the above Expression (2-19) is subject to the
reformulated restrictions:

mij ·V j
Bi ≤ , i = 1,..., N ; j = 1,..., M (2-20)
Sij

θik =
( Bi ·Dik ) / Rk (2-21)
mki

θiff ≥ θik , k ∈ Kijf ; i = 1,..., N ; j = 1,..., M (2-22)

θife ≥ θik , k ∈ Kije ; i = 1,..., N ; j = 1,..., M (2-23)

cij
!B "
Pij = aij + bij ·# ii $ (2-24)
%# m j &$

θijf + Pij + θije

tij = , i = 1,..., N ; j = 1,..., M (2-25)
moj

Ti = max {θijf , tij ,θije }, i = 1,..., N ; j = 1,..., M (2-26)

j

N
Qi
H ≥∑ ⋅ Ti ≥ 0 (2-27)
i =1 Bi
N
Qi
SP = H − ∑ ⋅ Ti (2-28)
i =1 Bi

V jmin ≤ V j ≤ V jmax , j = 1,..., M (2-29)

Rkmin ≤ Rk ≤ Rkmax , k = 1,..., K (2-30)

0 ≤ Qi ≤ Dimax , i = 1,..., N (2-31)

Again, these restrictions ensure that the sizing meets the intended level of
production (2-20); the filling and emptying times are limited by the maximum semi-
continuous processing time involved (2-22; 2-23); the allocated time for batch size of
product i cannot be less than that required by any of the semi-continuous operations
involved (2-25); the limiting cycle time for product i cannot be less than that required
for any of the batch/semi-continuous operations involved (2-26); and the total
production time cannot exceed the available time (2-27). The upper and lower bounds
on batch and semi-continuous units are set as before (2-29 and 2-30).
A simplified strategy has been proposed (Espuña et al., 1989) that summarizes all
the constraints enumerated above in a single decision parameter, which considers that
the makespan or completion time required to meet the specified production target
cannot be greater than the available time (2-28). Then, the only variables to be adjusted
are the equipment sizes. Logically, at optimum plant production levels, the batch size
for each product is:
i
#! m ·V #"
Bimax = min $ j j % , i = 1,..., N ; j = 1,..., M (2-32)
&# Sij '#
j

and the processing time for batch and semi-continuous units can be obtained

# Bi ·( Dik Rk ) $
θijfmin = max % & , i = 1,..., N (2-33)
k∈K j f
' mki (

# Bi ·( Dik Rk ) $
θijemin = max % & , i = 1,..., N (2-34)
k∈K je
' mki (
cij
! B max "
Pij = aij + bij ·# i i $ , i = 1,..., N ; j = 1,..., M (2-35)
#% m j $&

Once processing times are known, the cycle time is readily obtained

"& $ θijf + Pij + θije % e #&

Ti = max )θijf , ' (( , θij * , i = 1,..., N ; j = 1,..., M (2-36)
j
&- +
' moj , &.

It follows that the total batch processing time for each product will be the maximum
of all the times calculated above (in overlapping mode). Then, the time required to meet
the specified production levels can be determined and, consequently, we will also
discover the remaining time for the specific sizing that is used. The optimum sizing will
be obtained by minimizing the objective function and keeping the remaining time
positive.

2.6.1 The optimization algorithm

In order to determine the best equipment size for a specific set of parallel equipment
units, the objective function is to minimize the cost of the equipment units that are
required to meet the desired production target, and thus keep the surplus time positive.
Therefore, there is only one constraint, which is independent of the problem size. This
simplified formulation favours the solution of the optimization problem by heuristic
methods, even though the computation of some terms of the constraint can be more
laborious.

The overall design strategy is given in Figure 2.6. The optimization procedure is
based on the calculation of partial derivatives of the objective function and associated
constraints, with respect to the unit sizes.

∂SP ∂Vl
(2-36)
∂f Vl
 These values will be used to modify the size of unit l (batch or semi-continuous)
depending on the step size hl. The unit l to be modified is selected according to the
feasibility of the current sizing point. The unit that most improves the objective function
with no excessive loss of marginal time will be selected. When a non-feasible point is
reached, the unit is selected that gives the highest increase in the marginal time with the
lowest penalty cost (the objective function). Whenever completion time or boundary
restrictions are violated, the step length hl is decreased accordingly. The optimization
procedure ends when hl values become insignificant. Convergence is accelerated using
additional rules that take into account the size reduction for all units that are not actually
involved in processing time calculations and by keeping the step lengths within the
same order of magnitude for several computation cycles.

Figure 2.6 Optimization strategy

We have demonstrated the performance and robustness of the proposed algorithm

and its suitability at the preliminary design stage. We will now use this algorithm in the
following examples.

2.6.1.1 Parallel equipment

The addition of in-phase or out-of-phase parallel units to the appropriate production

stage will result in an increase in batch size or a decrease in cycle time respectively.
 Provided that the number of parallel units cannot be reduced below the initial set,
the following algorithm (Fig. 2.7) has been proposed to ascertain where and when a
parallel unit should be added or eliminated (Epuña & Puigjaner, 1989):

1. Carry out non-integer variable optimization from the current set of parallel
equipment units, as described previously.
2. Verify whether current optimum values are better than the previous global
optimum. If so, save the current values and parallel equipment unit set, and
continue to Step 3. Otherwise go to Step 4.
3. From the current optimum, select the best batch task candidate to add a new
parallel out-of-phase equipment unit. If this alternative was not selected before,
add it and go to Step 1. Otherwise continue to Step 4.

Figure 2.6 Optimization strategy: Parallel equipment selection

This selection is made by calculating, at the current optimum sizing point, the
“surplus time” when we add a new parallel out-of-phase equipment unit to all
batch operations. The batch operation that produces the maximum “surplus
time” is selected, as the objective of parallel equipment is to minimize cycle
time. Usually, this selection will be clear-cut, but if close alternate evaluations
are detected, we will select the batch operation that has the best ratio of
equipment cost/surplus time. In such cases, the surplus time cost of adding new
equipment in parallel is minimized.
 4. If any of the selected optimum sizes are near to the upper or lower bound (and
this alternative has not been considered before), a new parallel unit operating in-
phase should be added or eliminated, when possible. Then, go to Step 1.
Otherwise end the optimization process.

This procedure assumes that to attain a desired production capacity, it is cheaper to

install a single (large) equipment unit, although this could have a negative effect on
plant flexibility.
The optimization procedure (Step 1) for a specific set of parallel equipment units is
very fast, particularly in this case, in which the initial sizing is a set of coherent points
from previous optimizations. In addition, the number of modifications in this set will
usually be small. As a result, this optimization strategy may be repeated every time a
modification is introduced, at a very low cost in terms of computing time.
This optimization algorithm can be used for new plant designs as well as for retrofit
designs for existing plants, as mentioned in Section 2.6.3.

2.6.2 Retrofitting studies

The problem of optimal retrofit design of multiproduct batch plants considers the case
of an existing plant that has given sizes and types of equipment. Due to changing
market conditions and external economic pressures, it is assumed that new production
targets and selling prices are specified for a given set of products. The problem then
consists of finding design modifications that involve the purchase of new equipment for
the existing plant, in order to maximize profit.
The design modifications that are considered in the retrofit design of a multiproduct
plant involve the addition of new equipment in two ways: (1) to ease bottleneck stages
by operating in parallel out-of-phase, which decreases the cycle time of a product; (2) to
increase the size of the present batches by operating in parallel and in-phase with the
current equipment.
The retrofit design problem can be formulated as a mixed integer nonlinear
programming (MINLP) problem. The solution to this problem may require excessive
computing times when real cases are contemplated. The introduction of appropriate
heuristic rules should help to simplify the calculation procedure and simultaneously
ensure convergence within a reasonable time (Espuña & Puigjaner, 1989).
Most of the problem assumptions are the same as those used to solve the design
problem (see Section 2.6). However, the following should be stressed:

a) The variables set involves all the unit sizes that are not included in the initial
plant configuration and production levels for all products. From assumption (2-
35), it can be considered that all parallel equipment units are identical for batch
size calculation purposes.
b) When a non-identical parallel equipment unit is added, the resulting task
processing time is assumed to be the maximum associated with the unit set, if
the times differ. All sets of in-phase units operating out-of-phase should have
the same overall capacity. These sets are composed of identical “pseudo-units”
working in-phase. The “pseudo-units” may represent new units or existing
units, plus the extra ones needed to carry out production that is equivalent to
that attained by the new units.
 c) From Expression (2-32), at optimum plant production levels, the batch size for
each product Bi will correspond to its maximum feasible value Bimax , in order to
fully occupy the limiting equipment item. The same argument will also be valid
for semi-continuous units: from Expressions (2-33) and (2-34) the filling and
emptying times, θijf and θije , will be identified by θijfmin and θijemin respectively at
optimal production conditions.

Under these conditions, and taking into account all of the aforementioned
assumptions, it is possible to calculate all batch sizes Bi and cycle times Ti for a specific
set of variable values, and to determine the limiting units for all products. Once all these
values have been determined, the “surplus time” SP will be calculated from Expression
(2-28).
Therefore, when the plant is working at optimum production levels, all constraints
can be summarized by establishing that the time required to meet the specified
production cannot be greater than the available time H (2-27). Exceptional treatment
will be required to consider the bounds of Equations (2-29), (2-30) and (2-31). The
optimal solution will then be obtained by minimizing the objective function (2-19)
while keeping the “surplus time” (SP) positive. For a given number of items of parallel
equipment, standard NLP techniques can be used to solve this problem. However, it can
be handled more effectively in the heuristic way proposed by Espuña et al. (1987). The
number of parallel equipment units required (integer variables) is treated as described in
Section 2.6.1.1.

2.6.3 Intermediate storage

In Chapter 1, we indicated that intermediate storage can serve to decouple upstream and
downstream trains (Figure 2.7). This decoupling can occur in two different forms. If the
amount stored is of the order of an entire production campaign, then the trains can
operate as two independent processes. However, the storage capacity can be selected as
just large enough to decouple the cycle times, but not the batch sizes.
The problem of sizing, the location of intermediate storage and its influence on
overall equipment cost has been studied by Karimi and Reklaitis (1985). In
multiproduct plants two possibilities may occur:

• The same storage location is set for all products.

• A different storage location is allowed for each product.

If we assume that the storage cost is negligible compared to the equipment cost, as
given by Equation (2-19), the insertion of intermediate storage has the following general
consequences (Yeh and Reklaitis, 1985):

• Minimization of the sum of the individual subtrain costs is not equivalent to

minimization of the plant cost.
• The minimum cost with N storage locations is equal to the minimum cost with N-
1 storage locations.
• The minimum cost with storage locations that differ by product is equal to the
minimum cost at which all locations are the same
 STORAGE

STORAGE
STORAGE

REACTORS
REACTORS DRYERS
FILTERS

Figure 2.7. Use of storage buffers to smooth out the materials flow and
increase the overall productivity

The general formulation of the problem requires the introduction of location choice
variables of the type

X
! 1 if stage j of produc t i is assign ed to subt rain l
ijl =" (2-37)
# 0 otherwise.
This formulation can be used to resolve large MINLP problems. However, a
substantial reduction in computational effort can be obtained by introducing the
following design principles:

• The design for N-1 storage locations is feasible for N storage locations.
• The design with fixed storage is feasible for variable storage.

However, since subtrain designs do not decouple, a dynamic programming strategy

may be much too time-consuming, due to the number of possible storage locations. An
approximate method is proposed (Yeh and Reklaitis, 1985), which consists of
successive enumeration and comparison strategies that will be linear in the number of
storage locations. The algorithm must include the selection of parallel units (mj) as part
of the basic sizing procedure.

The general case of synthesis and sizing, including the merging and splitting of
tasks and the insertion of intermediate storage, requires further work. The special case
in which only pure batch units are considered and constant processing times are
assumed can be better handled by combining sizing and parallel unit selection for all
subtrains.
Let the subtrains be q = 1,..., r. Thus, the storage flow balance at steady state
requires that
q s
! Bi " ! Bi "
# $ =# $ (2-38)
% Ti & % Ti &
Then, given a limiting subtrain l,

q ! Ti q "
l
B = B ·# l $
i i
(2-39)
% Ti &
and for a given m j , Ti q Ti l has a fixed ratio, and the heuristics presented for the
sizing/parallel unit’s case can be adapted to the present case.
An enumeration scheme has been proposed to solve the complexities of the FIS
problem. The solution strategy consists of a comprehensive evaluation of the
alternatives, based on a cost-oriented objective function (Espuña et al., 1990). The
proposed strategy considers the following main stages:

a) Plant design without FIS

b) Intermediate storage location
c) Initial sizing with intermediate storage design
d) Final design with FIS

2.6.3.1 Plant design without intermediate storage

The preliminary design problem of multiproduct plants with no scheduling

considerations has been formulated by using a minimum capital cost criterion, which is
the sum of the individual process operation equipment costs (Espuña et al., 1989). All
problem constraints are summarized in a single decision parameter “surplus time” (SP),
which is defined as the difference between the present available time and the total
production time. Optimum sizing will be obtained by minimizing the objective function
while keeping the “surplus time” positive. The optimization procedure has been
described in detail previously (see Section 2.6, The multiproduct case).

2.6.3.2. Intermediate Storage Location

Intermediate storage site selection is determined by the initial sizing, according to the
maximum surplus time with minimum cost of equipment storage units and by following
basic strategies of batch plant design (Yeh and Reklaitis, 1987).
Potential candidates for intermediate storage selection sites are examined. Once the
batch size and cycle time have been calculated for each “subplant” as a result of
additional intermediate storage, the time horizon required to cover the production target
is determined and the additional cost due to storage is introduced. An optimum
intermediate storage location will maximize the return on investments in storage
equipment by decreasing the production time obtained per unit cost of the capital
outlay.
It is not possible to guarantee reliable unique solutions from the preceding criterion,
as it does not take into account the optimum sizing of equipment units when
intermediate storage is introduced. On the contrary, several valid alternatives must be
expected from situations offering similar efficiency. In such cases, the final decision at
this stage rests with the experienced plant engineer and is based on additional ad-hoc
criteria and one-off checks that are facilitated by the same algorithm.
 Usually, the most suitable location for intermediate storage is next to the cycle-
time-limiting batch equipment for one or several products. As a consequence, the
limiting cycle time may be increased by the storage filling and/or emptying time.
Therefore, the algorithm makes the necessary corrections to the overall process
operation time savings whenever possible. In any case, capital costs are increased not
only by the storage vessel cost itself, but also by the additional costs associated with
emptying and filling equipment.

2.6.3.3. Preliminary Design with Intermediate Storage

Once the most suitable location for intermediate storage has been determined in the
search for optimum system efficiency, the initial sizing point for all process equipment,
intermediate storage included, is calculated such that capital costs are kept to a
minimum and all process restrictions are met.
The process constraints are the same as those used in the design with no
intermediate storage. However, they are now applied to each subplant that results from
the addition of intermediate storage, with the additional production constraint.
U D
! Bi " ! Bi "
# $ = # $ , i = 1,..., N (2-40)
% Ti & % Ti &

where B is the batch size and T the cycle time for the upstream (U) and downstream (D)
subtrains processing N products. Constraint (2-40) ensures the same upstream and
downstream storage productivity (no storage depletion of accumulation occurs) for all
products.
Therefore, the previously used optimization strategy can be employed here. The
iterative procedure starts from an initial sizing point. Then, appropriate modifications to
the subplant data are calculated in such a way that all final resulting subplants are
equally productive for all products. The global algorithm flowchart is shown in Figure
2.8. The parameter chosen for subsequent modifications to the iterative procedure is the
processing time available in each subplant.
The calculation procedure employs the following steps:

1. Each subplant is independently sized over Hk = H available horizon (section

2.6).
2. Check for minimum productivity of each product in all subplants.
3. Determine the processing times that are required in each subplant to cover
the demand under minimum productivity conditions.
4. Correct the available production time for all of the subplants that result from
the preliminary design, according to the actual spare time and the processing
time needed to cover the demand.
5. Check that the processing time correction is less than some specified value.
Otherwise go to Step 1.
 General data:
Recipe and production characteristics
Economic Feasible storage location (s)
parameters and additional requirements

Sub-Plants setting

Site selection procedures

H1 = H

Subplants sizing using time horizon Hk

B/T selection
*
Processing time calculation (Hk )
*
Hk+1 = Hk + κ . (H - Hk )

Convergence?

Other
locations?

Feasible points

Location selection

Processing time transferring

Convergence?

OUTPUT

Figure 2.7 Global algorithm flowchart

The procedure quickly converges towards feasible solutions, as sizing calculations

are relatively simple and the initial feasible point is already close enough to the solution.
In any case, the results obtained at this stage are close to the optimum solution and lead
to:

 Further selection of an intermediate storage location: some equally efficient

sites, which were not discarded before, are now eliminated under the available
processing time criterion.
 Economic evaluation, according to the objective function, of the consequences
of the eventual addition of intermediate storage.
2.6.4. Optimum Design with Intermediate Storage

Once a feasible sizing point for processing units and intermediate storage vessels has
been reached in the preceding stage, an optimization procedure is initiated to minimize
the idle time of process equipment. This will maximize the process efficiency by
introducing appropriate productivity modifications for each product, which will not
necessarily be the same for all subplants.
The redistribution of overall plant productivity, taking into account all of the
products that are processed, requires a detailed analysis of each product’s individual
processing times. Therefore, productivity modifications are introduced using the
following algorithm:

Step 1: The objective function is calculated for each productivity increase in each
product. Then, calculate the term
ΔFobj
(2-41)
Δt
independently for each product by keeping constant the productivities of
the remaining products.
Step 2: Increase the productivity of one product, and decrease that of another.
These products are selected according to the previous step. Essentially,
processing time is transferred from the product that least affects the
objective function to the one that penalizes it most heavily, while the
overall processing time is kept constant.
Step 3: Carry out individual subplant equipment sizing optimizations for the
productivities that have already been found. Check for convergence and
finally go to Step 1.

2.7 CASE STUDIES

The following series of test case studies will be used to show the potential use of the
methods and algorithms presented for multiproduct batch plant design. The base case
design is built on the example proposed by Flatz (1990), which was used to illustrate the
manual sizing procedure in Section 2.6.

2.7.1 Base case design

The previous example (Example 2.1) will be used as the base case design for further
study to show the increasing complexity of decision-making when additional
information is incorporated into a more detailed and realistic design.
One of the limitations of using the manual sizing procedure is that the production
costs and market prices are not taken into account. Below, we will use the algorithm
presented in Section 2.6 to examine the influence of economic variables on production
decisions.
Results and discussion

The use of the previously presented optimization algorithm enables us to undertake an

economical appraisal of the convenience of producing A instead of buying it externally,
in order to achieve an optimal compromise. A trade-off is required, according to various
market scenarios. The alternatives are shown in Table 2.6 and Figure 2.8:

Table 2.6 Design results for the plant of Example 2.1 under different
market scenarios
Profit A > 0.68 $ / kg Profit A > 0.60 $ / kg Profit A > 0.56 $ / kg
3
V1 (m ) 1.11 0.98 0.22
3
V2 (m ) 3.69 3.29 0.75
TA (h) 8 8 8
TB (h) 12 12 12
BA (kg) 67 60 13.6
BB (kg) 2212 1970 450
QA (kg) 32000 27694 0
QB (kg) 180,000 180,000 180,000
Cost Equip.($) 56,654 53,691 34,625
Penalty. ($) 0 2,583 <17,920
F. Obj. ($) 56,654 56,274 <52,545

The design that satisfies the full demand for products A and B requires a tank
capacity of 1.11 m3 (V1) for the solution step and a reactor volume of 3.69 m3 (V2). The
overall capital cost of the plant is $56,654 (objective function).
A sensitivity analysis of the solution in terms of the production costs of product A,
which can be externally supplied, indicates that the design remains valid whenever the
profit (savings) of producing A is over 0.68 $/kg, which corresponds to a production
cost of 2.22 $/kg. However, when the profit reaches (or is below) 0.56 $/kg, the
production of A is no longer justified, since the production cost is 2.34 $/kg. In this
case, the best plant design only produces B, giving: V1=0.23 m3 and V2=0.75m3
(Column c of Table 2.6). The overall cost is now $52,545, which corresponds to an
investment of $34,625 in the new plant and an additional amount of $17,920 to buy A
from the external manufacturer.
The optimum design is found for the second scenario of Table 2.6. As can be seen,
the best design corresponds to a production scenario that involves full production of B
(which is mandatory) and 27.694 kg of A. Now, the size of the vessels has been reduced
to V1=0.98 and V2=3.29 and the total production cost is $56,274.
 [ Ton ] [ m3 ] [ M$ ]
3.00
32 Production of A
30 1.5
DEMAND PARTIALLY
COVERED

Size of unit “1”

2.00

20 1

1.00
15 0.5

Penalty cost
0
220 223 225 230 235

Production cost ($/kg)

Figure 2.8 Plant design under different market scenarios for the production of A

Additional considerations, which may be difficult to quantify in economic terms,

can influence decision making. For instance, the following factors could also be
considered:

• Tightening supply conditions independently of market fluctuations.

• Minimizing supply times.
• Improving market forecasting with proactive initiatives.
• Expected variations in price and cost.
• Forecasting future expansion and production needs.
• Forecasting future expansion and product diversification.
• Quality control.

2.7.2 Retrofit design: plant capacity expansion

A market survey indicates that a third product must be manufactured to stay
competitive. Therefore, next we will consider the production of a third product C.

Product C
The reaction is:

F(l) + G(s) + HCl(aq) → C(aq) + Q

Two production stages are required. Thus, the reuse of existing equipment is
considered, which would increase the equipment utilization. From laboratory
experimentation, the following recipe is available:
 1) Reaction: introduce 1000 cm3 (1.15 kg) of HCl (30% in weight or 345 g of pure
HCl) and 100 g of F (liquid) into a stirred tank reactor. Start stirring.
2) While the solution is stirred, slowly add 55 g of G for 1 h. The heat released by
the exothermic reaction increases the temperature of the solution by up to 30°C.
3) Crystallization: finally, a neutral solution is obtained, crystallization occurs and
500 g of product C is produced.

Additional information:

• Standard security measures and appropriate environmental protection are required

to handle HCl.
• The market demand for product C is 290,000 kg per year. The expected increase
in gross profit for producing C in the same plant has been evaluated as 0.8 $/kg.
• Other plant operating conditions are the same as before (processing periods,
penalty for non-production, maintenance, cleaning, etc.).

The following alternatives must be analyzed:

a) Two new units are added to produce C separately from A and B. In this case the
costs of the equipment are:

Cost3 ($) = 18000 + 7000·V (m3 )

Cost4 ($) = 15000 + 5000·V (m3 )

b) Crystallization occurs in Unit “1” of the previously designed section of the plant
(2.7.1. Base case design), using the same sizes obtained there for Units “1” and
“2” (identical units may be added). The cost of Equipment “3” is

Cost3 ($) = 18000 + 7000·V (m3 )

c) Crystallization occurs in Unit “1” as before, but new facilities must be designed
to produce all three products. The cost of the different units is as before.

Results and discussion

In cases b) and c), a total of 8 x 3 = 24 product changeovers will take place. However,
since B → C and A → C changes only require the cleaning of shared Unit “1”, the other
unit can be cleaned while the plant is in operation. The cleaning times will be the same
as in the base case design.
The results are shown in Table 2.7. The final design for retrofit case a) requires the
use of two additional vessels V3=0.21 m3 and V4=0.21 m3, along with the existing ones
(V1 and V2). The full demand for the three products is met at an additional cost of
$35,552, which means a total investment of $92,206 in the retrofitted plant. However,
two additional scenarios could be contemplated, as summarized in Table 2.7.
 Table 2.7 Design results for the different scenarios contemplated in the
retrofit example
(a) (b) Expansion (c)
C produced New
independently Optimum Qi = Di Plant design
3
V1 (m ) 1.11 1.11 1.11 1.36
3
V2 (m ) 3.69 3.69 3.69+1.16(ip) 4.55
3
V3 (m ) 0.21 1.11 1.11 1.12
3
V4 (m ) 0.21 - - -
TA (h) 8 8 8 8
TB (h) 12 12 12 12
TC (h) 1,6 1.6 1.6 1.6
BA (kg) 67 67 88 83
BB (kg) 2212 2213 2220 2729
BC (kg) 97 504 504 509
QA (kg) 32,000 24,300 32,000 32,000
QB (kg) 180,000 180,000 180,000 180,000
QC (kg) 290,000 290,000 290,000 290,000
Previous cost ($) 56,654 56,654 56,654 0
New equip cost ($) 35,552 25,770 46,722 88,970
Penalty. ($) 0 8,440 0 0
Obj. Function ($) 92,206 90,864 103,376 88,970

In retrofit case b), only one piece of additional equipment (a stirring tank with
V3=1.11 m3) is needed to produce C. Now, the total production cost has been reduced by
1.4% at the expense of only partially covering the demand for A. If full production of A
is required, a parallel reactor in-phase with V2 is necessary (V2ip=1.16 m3) at an
additional cost of 1.5% (Retrofit [b2]). The last column in Table 2.7 shows the results
obtained for the best grass-roots design, when production of the three products is
considered. This is indeed the best design (1.6% savings), as it should be.

Additional considerations:

The first alternative a) (product C produced in a separate line):

• Keeps separate processing lines, which may imply higher safety and quality
control (set points are not modified).
• The presence of duplicate equipment permits higher flexibility and reliability.
• Since no equipment is shared, productivity is higher and production planning
decisions are simplified.
The second alternative b) (sharing existing equipment) may result in:
 • Equipment sharing, which contributes to overall cost reduction.
• The use of existing equipment may cause production under-capacity, particularly
when size factors for the various products differ considerably.
The third alternative c) (new plant):
• Offers optimum equipment sharing, which reduces investment costs.
• Avoids production under-capacity.
• Has the lowest overall investment cost. This should be the best option in the long
term.

2.7.3 Retrofit design: plant design modification

The overall performance of some subsequent processes can be increased by obtaining

product B of higher purity. This can be achieved by dissolving contaminating impurities
in HCl as follows:

B(s) (90 %) + HCl (aq) → B(s) (100 %) + Impurities [HCl(aq)]

Laboratory test runs indicate that the following recipe is effective:

1) Reactor set-up: introduce 1000 cm3 (1.150 kg) of HCl (30% in weight or 345 of
pure HCl) into the reactor. Initiate stirring.
2) Reaction: once stirring has been initiated, add 220 g of B (with 10% impurities).
Stirring continues for 105 minutes. The dissolution heat is negligible. Finally,
200 g of higher purity B remain in the reactor and 20 g of impurities have been
dissolved.

Additional information:

• The same environmental protection precautions as before are required during the
handling and manipulation of HCl.
• The market demand is that of product B (now 100% pure) and the costs of the
various units are those indicated in the previous cases (HCl processing requires
type “3” units).
• Plant operation is the same as before (processing, cleaning time, etc.). A total of 8
x 3 = 24 product changeovers are envisaged. However, product change C → A
only requires the cleaning of unit “1”. The second unit can be cleaned while the
plant remains in operation. Therefore, it can be considered that the total cleaning
time is equivalent to 20 changeovers of the base case.

Results and discussion

This case study also offers several alternatives, which are indicated in Table 2.8.
 The first column (Retrofit [a]) shows the results when the existing facilities for
producing A and B (commercial grade) are kept. The penalty for under-producing A is
1.10 $/kg. Consequently, only 5,329 kg of A should be produced for the best retrofit
design, which has a total production cost of $131,674. Again, the full demand for A
could be met by adding another vessel in parallel out-of-phase with V2, which increases
the total cost by 1%.

Table 2.8. Retrofit of Case 2.7.2 to produce B purified.

Retr.(a) Retr.(b) New plant (a) New plant (b)
V1 (m 3 ) 1.11 1.11 1.17 1.83
V2 (m 3 ) 3.69 2(op) x 3.69 0.99 3.29
3
V3 (m ) 3.95 3.98 3.25 3.73
3
V4 (m )
TA (h) 8.00 4.50 8.00 4.50
TB (h) 12.00 6.00 12.00 6.00
TC (h) 1.60 1.60 1.60 1.60
BA (kg) 67.0 67.0 17.9 59.8
BB (kg) 719.0 724.2 591.5 68.1
BC (kg) 504.5 504.5 529.6 831.6
QA (kg) 5,329 32,000 0 32,000
QB (kg) 180,000 180,000 180,000 180,000
QC (kg) 290,000 290,000 290,000 290,000
Previous cost ($) 56,654 56,654 0.0 0.0
New equipment cost ($) 45,682 82,020 81,517 135,748
Penalty ($) 29,338 0.0 35,121 0.0
Obj. Function ($) 131,674 138,674 116,638 135,748

A totally different scenario would be to consider the best grass-roots design. New
Plant (a) results in a substantial saving of 16%, due to a significant reduction in
equipment sizes. The best design practically discards the production of A, since it is
more economical to buy it from the external manufacturer. Finally, the last column
(New Plant [b]) shows the grass-root design when all products must be produced to
meet full demand. In this last case, the results of retrofit and grass-roots design are quite
similar (2% savings in the grass-roots design). As a consequence, careful consideration
should be given to these situations, in which new technological designs represent a
competitive economical alternative to retrofit, and simultaneously offer a more robust
scenario (lower maintenance costs, etc.).

Figure 2.9 shows the final plant layout for manufacturing A, B, C and purified B.
Reuse of the reactor and crystallizer is considered in this case.
 Fig.2.9 Production layout for manufacturing products A, B, C and B
purified, where reuse of reactor and crystallizer is contemplated

The above case studies were selected to emphasize the fact that the design and
retrofit problems of multiproduct batch plants may not have a single clear-cut solution
from the perspective of design economics. Alternative “best” designs can be found,
depending on specific techno-economic scenarios. Present methods and available tools
offer powerful and “expert” support in decision-making, in which the most technically
appropriate and most economically profitable solution is selected in each case.

2.7.4 Industrial case study: penicillin plant

The following case study is based on a real case that is related to the pharmaceutical
industry. The aim is to design a plant for producing two different grades of penicillin-G,
according to their final destination: human or animal consumption. The forecasted
nominal production capacity of the plant is as follows:

• 11,000 kg per year for human consumption (therapeutic grade, product A).
• 61,000 kg per year for use in animal feed mixtures (veterinary grade, product B).

The production process will be the same for both products, except for the additional
purification stage required to produce product A for human consumption, which consists
in redissolving product B, followed by a final crystallization.
The inventory of the final product cannot exceed 6 months after it has been
produced. Therefore, the production horizon will be set to 4 months (approximately 100
working days in which 3,704 kg of therapeutic product and 22,700 kg of veterinary
product must be obtained.

Laboratory scale data

Laboratory experiments were carried out at full capacity in a 10-liter batch reactor using
the following procedure:

1) Before the culture is inoculated, the fermenter is filled with semi-synthetic

medium and sterilized with steam at 121ºC for 40 minutes.
2) Fermentation typically takes 5-7 days at 60ºC. An average of 6 days will be
considered. pH control is required to avoid formation of acid compounds.
Vigorous stirring is needed to convey the oxygen required during the
fermentation process.
After appropriate tests, it was decided to use a particular strain (XSD-1731) to
produce penicillin-G, whose output is 16,000 units of product per ml (1,500 units to 1
mg matching of penicillin-G).

Pilot plant results

From pilot plant experiments conducted in a 200-liter fermenter, we observed a small

increase in efficiency with respect to the results obtained in the laboratory. This was
attributed to greater uniformity in the concentration of oxygen in the fermenter.
However, in these cases, greater attention needs to be paid to the sterilization stage.
The minimum time for sterilization (at 121ºC) can be calculated with the following
correlation:

V2
test = t0 + 2.0·ln (2-42)
V1

where times are given in minutes.

To avoid product degradation and allow a minimum load in the fermenter itself, a
final disinfection stage has been added, which allows a timeout of up to a week in the
equipment, at room conditions, without causing significant degradation of the product.

Separation and purification

The process of separation and purification of penicillin-G involves the following steps:

1) Settling. The output of the fermenter is mixed with an equal volume of ethyl
acetate for 30 minutes. The mixture is left standing for 2 hours, so that the
aqueous phase (ethyl acetate and penicillin-G) is separated from the water
(which contains solid impurities). The aqueous phase is removed (to the water
treatment section). The organic phase contains 97% of the penicillin formed
during fermentation.
2) Washing. A second wash of the organic phase is performed with water (one
volume of water for two volumes of wash solution) for 30 minutes. Then, the
phases are separated again for only 1 h and the aqueous phase is again removed.
The organic phase contains 98% of the penicillin from the previous stage.
3) Crystallization. The mixture must be cooled to 10ºC to recover penicillin-G.
This temperature is then maintained for 1 h, which leads to the recovery of 98%
of the product in crystallized form.
 4) Centrifugation. The ethyl acetate is separated and sent to the solvent recovery
section.
5) Drying. The degree of purity of the resulting penicillin is suitable for veterinary
use. In this case, proceed to final drying. Ethyl acetate is also recovered from
drying.
6) Recrystallization. Additional recrystallization is required to obtain therapeutic
grade penicillin-G. The solids from the centrifuge are redissolved in a 75/25
mixture by volume of ethanol and water at room temperature by stirring for 1 h
approximately. The volume of solvent is the same as that of the ethyl acetate
used in Step 1.
7) Cooling. The solution must be cooled back to 10ºC. After 1 h, crystals are
obtained that contain 97% of the product that is still present.
8) Final centrifugation and drying. Finally, we proceed to centrifugation and
drying, as in Steps 4 and 5 above.

Except for the fermenter, the cleaning steps can be eliminated if there is no change
in product characteristics.
The times mentioned in all previous steps are independent of batch size, except for
the sterilization of the fermenter.
It is generally considered that the equipment for the extraction and crystallization
steps should not be above 80% of its rated capacity.

Tasks description, operating times and size factors

Fermenters
In the size range indicated below for these items of equipment (from 19.0 m3 to 114 m3),
the operation and cleaning times are considered constant (6 days for fermentation and 8
h for cleaning). With respect to sterilization time, if we apply the relationship (2.42) for
the extreme cases of this size range, the required time is between 55 and 59 minutes.
Therefore, the processing time is approximately 154 h for both product qualities (1 h
sterilization, 0.5 h loading, 6 days fermentation, 0.5 and 8 h discharge cleaning).
The size factors relate to the ratio of 1,500/16,000 l/g shown in laboratory tests,
corrected for the recovery rates of different stages of purification, and by the factor 0.8,
which reflects the maximum permitted filling level. The results for the therapeutic grade
product are:

1,500 1 1
· · = 0.1297 m3 /kg
16,000 0.97·0.98·0.98·0.97 0.8

The veterinarian grade product gives

1,500 1 1
· · = 0.1258 m3/kg
16,000 0.97·0.98·0.98 0.8

Storage tanks
The material should not be stored for more than 8 days. Cleaning time is over 4 h, and
should be done after each emptying. Since the tanks are used to store the fermentation
product, the same size factors will be considered.

Non-jacketed tanks
These are used to remove product from the exit of the fermenter. For both grades of
product, the processing time is estimated at 6.5 h (0.5 h loading, 0.5 h mixing, 2.0 h for
the first separation, 0.5 h drainage, 0.5 h for the second load, 0.6 h for the second
mixture, 1.0 h for the second separation, 0.5 h final drainage and 0.5 h discharge).
In the first separation, which uses a solvent volume that is equal to the material to
be treated, the size factor for the therapeutic grade product is 2.0·130 = 0.260 m3/kg.
For the washing step, this factor will be 1.5·130 = 0.195 m3/kg, which leads to a size
factor of 0.260 m3/kg. Similarly, for the veterinary product, the size factor is 2.0 126 =
0.252 m3/kg.

Jacketed tanks
These are used at different stages of crystallization, and in the redissolution stage in the
case of therapeutic grade product.
For the purposes of calculating the heat transfer, we considered an average
coefficient of heat transfer of 312 W/ (m2K) for heating, and 170 W/ (m2K) for cooling.
External heat losses are estimated at 10%.
If no cleaning is required due to product changeover, the first crystallization takes
about 2 hours (0.5 h of loading, 1 h of crystallization and 0.5 h discharge), regardless of
the product to be manufactured. The second crystallization should be preceded by a
redissolution stage that takes 1 h, so the duty cycle increases to 3 h. The cleaning time is
estimated at 4 h, but is only needed when there is a change of product.
We work with the same volume of material as in the fermenter (although in each
case, the solvent will vary depending on the stage). Consequently, the size factors are
the same as those for the fermenter: 0.130 m3/kg for the therapeutic grade product and
0.126 m3/kg for the veterinary grade product.

Centrifuges
Testing has been carried out on various models of centrifuges. The most satisfactory
results were obtained with a given model that had the following cycle:

1. When the centrifuge is in operation, the liquid feed pulls the solid formed in the
previous stage until a cake of about 0.50 m3 is attained. Once the cake has been
formed, the feed is interrupted and centrifuged to reduce the solvent content by
up to 15% by weight. The cake has a thickness of about 16 cm over the entire
surface of the basket (1.22 m in diameter and 0.762 m high).
2. Then, a rinse cycle is carried out with water and the solution is centrifuged
again to leave a cake with 15% solvent content by weight of water. Finally,
0.453 m3 of product is obtained, with a final density on a dry basis of 996 kg of
dry solids per m3.
3. The cake is then discharged into a suitable container (transportation cart).

A complete cycle takes about 45 minutes, whatever the product, which leaves the
equipment prepared to begin a new cycle of the same product. Every time the quality of
the product is changed, the equipment must be thoroughly cleaned, which takes about 4
hours.
If we take the maximum product volume in the cake as a basis for sizing, the size
factor for the final centrifugation is 1.0/996.0 = 0.001 m3/kg. For therapeutic grade
product, the first centrifugation has a size factor of 1.0/996.0/0.97 = 0.00104 m3/kg.

Transport carts
If all the equipment units work properly, there is no practical limitation on the time the
material can be held in the transport carts, so they can be used as intermediate storage.

Dryers
The drying time is between 9.0 and 15.0 hours per charge. This preliminary design
procedure will take an average of 12.0 h, to which half an hour must be added for each
of the loading and unloading operations. Each cleaning operation requires 4 h of further
work, but is only required when there is a change in product quality.
For each grade of product, the density of material at the inlet is indicated for the
end of the centrifugation cake (996 kg of dry solids per m3). If we consider that the
equipment cannot be filled to above 65% of its rated capacity, the size factor will be
1.0/996.0/0.65 = 0.154 m3/kg.

General information on design costs

Equipment costing

We know the cost of each machine up to a certain size. The relationship between cost
and size is considered an exponential correlation, according to Expression (2-43).
γ
Cost = α j + β j ⋅ V j j (2-43)

The costs below include implementation and installation costs, by means of the
addition of a factor of 8.0 to the cost of equipment. This factor, which was adopted to
take into account the characteristics of this case, the type of equipment and the control
needs, is much higher than that normally used in the chemical industry and does not
apply to transport carts, which do not require any installation or instrumentation.
For preliminary design purposes, it will be assumed that the equipment can be
found on the market in the desired size, within the range indicated for each case.

Fermenters
Fermenters are available in the market sizes of 19 to 114 m3. The cost of a 76 m3
fermenter is $800,000, including instrumentation and installation costs. An exponent of
0.5 and a constant term zero can be considered in the cost equation.
The energy cost for agitation can be estimated at three times that needed for a
jacketed deposit of the same volume. The main energy cost is that of pumping air
(oxygen). However, this will not be considered in a comparison of the alternatives, as it
is nearly constant.
Fermenters should not be used at over 80% of the volume.

Storage tanks
Available sizes are between 19 and 114 m3.
 The cost of a 76 m3 tank that meets the requirements is $64,500. The exponent is
0.46 and the constant term zero.
They should not be used at over 80% of the volume.

Jacketed vessels
These are used in the crystallization stages and can be found in sizes between 3 and 19
m3, with an area of exchange of 11.8 to 32 m3 and a stirring power of 10 to 40 HP (7.35
to 29.4 kW) respectively.
The cost of a tank of 3.8 m3 (1000 gallons) is $48,000. The exponent is 0.65 and the
constant term is zero.

Non-jacketed vessels
These are equivalent to jacketed vessels. However, they cannot be used for operations
that require heating or cooling, for example, product extraction. The dimensions are the
same as those of jacketed tanks.
The cost of a tank of 3.8 m3 (1,000 gallons) is $17,000. The exponent is 0.40 and
the constant term is zero.

Centrifuges
There is only one basket centrifuge that is suitable for automatic operation on the
market. It costs $130,000 and is capable of processing 0.453 m3 of product per cycle of
operation.

Transport carts
Wheeled carts will be used, which can be filled up to 80% of their rated capacity.
The cost is estimated at 10% of a storage tank of the same dimensions. In addition,
this cost will not be affected by the installation factor (8.0) mentioned above...
Given the low cost, nature of the operation and form of use (the size of the
centrifuge is not subject to optimization), carts on wheels of 1.8 m3 of rated capacity
will be used.

Dryers
A conical rotary dryer will be used. On the market, these can be found in sizes from 3
m3 (with a heating area of 10.5 m3) to 7.22 m3 (with an area of 18.8 m2). The cost of a 3
m3 dryer is $85,000. Consider an exponent of 0.2 and the constant term zero.

General utilities

Steam is generated at 40 bar at a cost of 7.7 $/ton (3.6 ·10-9 $/J). For low-temperature
heating, you can also use waste steam of 2.5 bar valued at 2.1 $/ton (1.0·10-9 $/J).
Cooling water at 25ºC is also available, which should not be eliminated without a
treatment over 50ºC. The cost is 0.20 $/m3 (total cost 1.0 ·10-9 $/J). A CaCl2 brine at -
15ºC can be heated up to -9 ºC, at a cost of 25.6 ·10-9 $/J.
The cost of electricity is 0.035 $/kWh (9.7· 10-9 $/J)

Environmental conditions

The process will take place in a building in which the temperature ranges between 15ºC
and 26ºC, depending on external conditions.
Data summary

All the above information is summarized in Tables 2.9, 2.10 and 2.11 and Figure 2.10.

Table 2.9. Penicillin plant: Forecasted nominal production capacity

Product Global demand
Penicillin therapeutic grade (H. Grade) 3784 kg
Penicillin veterinary grade (A. Grade) 22700 kg

Table 2.10. Penicillin plant: Size factors and processing times

Equipment unit Size factor ( m 3 / kg ) Processing time (h)
H. Grade A. Grade H. Grade A. Grade
Fermenter 0.1297 0.1258 154.00 154.00
Extraction 0.2594 0.2516 6.50 6.50
Crystallization 0.1297 0.1260 2.00 2.00
Centrifugation 1 0.0010 0.0010 0.75 0.75
Recrystallization 0.1297 3.00 0.00
Centrifugation 2 0.0010 - 0.75 0.00
Drying 0.0015 0.0015 13.00 13.00

2.7.4.1 Base case design

When we used the above data and the proposed design strategies, we obtained the
results shown in the first column (a) of Table 2.12 and Figure 2.11.
Given the differences between the operating times of the equipment that we
considered, especially between the fermentation stage and the remainder, a first solution
to the nonlinear problem indicated that a large number of minimum capacity fermenters
should be installed in parallel, which would match the maximum capacity of the units
assigned to the extraction stage. This would maximum the installed capacity of other
units. However, the algorithm for introducing equipment in parallel detected the
possibility of installing new equipment in-phase, thus reaching the solution shown. The
procedure is much faster when the maximum capacity restrictions are removed to obtain
a first estimate of the units that need to be installed, as shown in Section 2.6.1.

Table 2.11. Penicillin plant: Cost factors and size range of available
equipment
Equipment units Cost factors Limiting zize
αj βj γj V jmin V jmax
Fermenter 0.0 91,947.0 0.50 19.0 114.0
Non-jacketed reactor 0.0 9,982.0 0.40 3.8 19.0
Jacketed reactor 0.0 20,207.0 0.65 3.8 19.0
Centrifuge 130,000.0 0.0 0.00 0.453 0.453
Dryer 0.0 68,071.0 0.20 3.0 10.5
Storage tank (I.S.) 0.0 8,814.0 0.46 19.0 114.0
Transport cart (I.S.) 0.0 110.0 0.46 1.8 1.8
 XSD - 1731 ETHYL ALCOHOL
H2O
ETHYL ACETATE
FERMENT H2O ETHYL ACETATE
PENICILLIN-G
ETHYL ACETATE CRYST.
PENICILLIN-G

CRYSTAL
EXTRAC. E-15

E-9
CENTRIF.

P-44

LIQUID PHASE CENTRIF.

SOLIDS
DRYING

ETHYL ACETATE

Figure 2.10. Plan layout for the production of Penicillin-G in two

grades: therapeutic (product A) and veterinary (product B).

An obvious alternative, given the time data, the availability of equipment and the
previous design results, is to install proper storage equipment between the fermentation
and extraction stages, and thus divide the production line into two sub-trains.

2.4.7.2 Incorporating intermediate storage

The algorithm proposed in Section 2.6.3 indicates that when intermediate storage
equipment is introduced, its location is ideally between the fermentation and extraction
stages. This conclusion was obvious, in view of the cycle times and batch sizes
calculated above.
The results of applying the sizing strategies shown in Section 2.6.4 can be seen in
the second column of Table 2.12 (Column b) and are also shown in Figure 2.12. The
fermenters, which are the most expensive piece of equipment that is subject to
optimization, work very similarly for both products and the time distribution varies
little. However, the downtime savings in the extraction and purification steps allow a
reduction in the size of equipment and a remarkable decrease in the number of units
working in parallel.
The optimization in this case has been influenced by the large price difference
between the fermenter and the other units, so that the system respects the optimal
productivity relationship for the fermenters compared with the rest of the plant.
The sizing of the intermediate storage tank was calculated using the following
relationship, which provides a very conservative value:

Vs = max {Sis ·( Biu + Bid )} i = 1,..., M (2-44)

As the intermediate product for storage is the same regardless of the final product in
this case, it is not necessary to ensure that productivity is the same in both subtrains.
Therefore, if measures are taken to ensure that the intermediate product remains no
more than a week in the storage tank, both subtrains can be designed independently.
This results in the dimensions shown in Column c) of Table 2.12.
 Table 2.12. Penicillin plant: Design results without reuse of equipment
a) Base Case b) I.S.(1) c) I.S.(1) d) I.S.(2) e) I.S.(2)
Design B/T=const. B/T ≠ const. B/T=const. B/T ≠ const.
BA (kg) 413.9 827.9 827.9 827.9 827.9
BB (kg) 426.7 853.5 853.5 853.5 853.5
TA (h) 38.5 77.0 77.0 77.0 77.0
TB (h) 38.5 77.0 77.0 77.0 77.0
V1 (m 3 ) 4op x 53.7 2op x 107.4 2op x 107.4 2op x 107.4 2op x 107.4
I .S. Tank (m ) 3 _ 116.4 116.9 111.9 112.1
BA (kg) 413.9 69.9 55.10 34.9 29.3
BB (kg) 426.7 72.0 75.50 36.0 37.3
TA (h) 38.5 6.50 6.50 3.25 3.25
TB (h) 38.5 6.50 6.50 3.25 3.25
V2 (m 3 ) 6ip x 17.89 18.13 19.00 2op x 9.06 2op x 9.37
3
V3 (m ) 3ip x 17.89 9.06 9.50 4.53 4.69
3
V4 (m ) 0.45 0.45 0.45 0.45 0.45
V5 (m 3 ) 3ip x 17.89 9.06 7.15 4.53 3.80
3
V6 (m ) 0.45 0.45 0.45 0.45 0.45
Cart (m ) 3 _ _ _ 1.80 1.80
BA (kg) 413.9 69.9 55.10 1,800.0 1,800.0
BB (kg) 426.7 72.0 75.50 1,800.0 1,800.0
TA (h) 38.5 6.50 6.50 13.0 13.0
TB (h) 38.5 6.50 6.50 13.0 13.0
V7 (m 3 ) 3.0 2op x 3.00 2op x 3.00 3.0 3.0
H A (h) 352.0 352.0 N/A 352.0 N/A
H B (h) 2048.0 2048.0 N/A 2048.0 N/A
Obj. Function ($) 4,019,400 2,614,300 2,605,600 2,483,200 2,479,300

XSD - 1731 ETHYL ALCOHOL

H2O
ETHYL ACETATE
FERMENT
H2O ETHYL ACETATE
54 m3
4 op PENICILLIN-G
ETHYL ACETATE CRYST.
PENICILLIN-G 17.9 m3
3 ip
CRYSTAL
EXTRAC. 17.9 m3 E-15

17.9 m3 3 ip
6 ip
E-9
CENTRIF.
0.45 m3

P-44

LIQUID PHASE
CENTRIF.
SOLIDS
0.45 m3
DRYING
3 m3
COST = $3,494,300

ETHYL ACETATE
 Figure 2.11. Base case design for the production of Penicillin-G in two
grades: therapeutic (product A) and veterinary (product B).

Although it was not considered necessary to maintain productivity between

different subtrains, an analysis of the results showed that it would be advantageous to
increase the production rate of the veterinary grade product after the storage tank. This
could be achieved by increasing the capacity of the non-jacketed tank (extraction) and
that of the first jacketed tank (crystallization). This could lead to some time savings
during the development of the veterinary grade product, which would make it feasible
to reduce productivity while work is undertaken on the first product. As a result, the
required capacity of the second jacketed tank could be reduced, as this tank is only used
to obtain the therapeutic grade product. This strategy is economically favorable as long
as new equipment does not need to be incorporated in parallel to obtain further
productivity increases in the manufacture of the veterinary degree product.

However, if the storage tank is not installed or the aim is to maintain productivity
between the subtrains, any increase in the plant’s capacity to produce the veterinary
grade product involves an increase in the capacity of the fermenter. In this case, the
incremental associated cost is not compensated by the savings achieved by reducing the
capacity of the second jacketed tank.

XSD-1731 ETHYL ALCOHOL

FERMENT H2O
ETHYL ACETATE
108 m3 H2O
2 op ETHYL ACETATE
PENICILLIN-G
PENICILLIN-G CRYST.
ETHYL ACETATE 9.1 m3
3 ip
CRYSTAL
TANK EXTRAC. 9.1 m3 E-15

116 m3 18.1 m3 3 ip
6 ip

E-9
CENTRIF.
0.45 m3

LIQUID PHASE
SOLIDS CENTRIF.
0.45 m3
DRYING
3 m3
2 op
COST = $2,614,300

ETHYL ACETATE

Figure 2.12. Design with intermediate storage for the production of

Penicillin-G in two grades: therapeutic (product A) and
veterinary (product B).

In all cases, the proposed dryer size is the least available on the market. It may be
advisable to install two dryers in parallel to reduce cycle time in the second subtrain. In
this case, a second decoupling point could be introduced before the dryer, through an
intermediate storage solution. This situation is also clearly detected by the proposed
localization algorithm, which indicates that the benefits will be much lower than those
for the first location (after the fermenter). Technically, the transport carts can be used
for storage.
As a result of the analysis of this option and the similarities between the two
products, the optimal design was obtained with the productivity that allows the plant to
work with maximum occupancy of the fermenters. These results are summarized in
Table 2.12 (Column d).
As in the previous case c), when we consider a single location of intermediate
storage and remove the restriction that forces the same productivity to be maintained in
both subtrains (case e), enhanced economic performance favors a reduction in the
productivity of therapeutic grade in the plant. This also reduces the capacity needed in
the second crystallizer, and thus cuts costs. In this case, the economic advantage will be
maintained until the minimum allowable size of such equipment is reached. Table 2.12
summarizes the results obtained.

2.4.7.3. Reuse of equipment

As noted in the therapeutic grade product recipe, the equipment used in the first and
second centrifugation has the same characteristics, as in the case of the two
crystallizations. Therefore, a possible improvement to the production schedule could be
brought about by reusing these pieces of equipment when therapeutic product is
produced.
Cleanup operations must take place between the first and second use of each
equipment unit in the same batch to avoid contamination of the purified product. In this
particular case, the processing times are amended, as shown in Table 2.13. The size
factors will remain unchanged. In each case, they will be the greatest of all the tasks that
are undertaken in each equipment unit. The calculation of equivalent times was
performed in a semi-overlapping mode of operation (Chapter 1, Section 1.4.4.3).

Table 2.13 Processing times for reuse of equipment

Processing time (h) Equivalent Processing time (h)
Task TAj (h) TBj (h) Equipment H. Grade A. Grade
Fermentation 154.00 154.00 Fermenter 154.00 154.00
Extraction 6.50 6.50 Non-jacketed tank 6.50 6.50
Crystallization 6.00 2.00 Jacketed tank 14.00 2.00
Centrifugation 1 4.75 0.75 Centrifuge 12.75 0.75
Recrystallization 7.00 - - - -
Centrifugation 2 4.75 - - - -
Drying 13.0 13.00 Dryer 13.00 13.00

The results are summarized in Table 2.14 and show that, in all cases, this option is
more appropriate than the equivalent option in which there is no equipment reuse.
The proposed size of the dryer is, again, the lowest available. Hence, we again
consider decoupling the drying stage from the rest of the purification steps, using
transport carts as intermediate storage.
From the calculations, we can see that this solution is the same as in the case of
single intermediate storage (Column d vs. Column b in Table 2.14), except with respect
to the cost of new equipment cabinets. This is because the cycle time for the therapeutic
grade product is not affected by the introduction of storage equipment. With the new
situation, we intended to use the time gained by decoupling processes to lower
productivity and thus reduce the size of the units. However, this is not possible for one
of the products without increasing the size of units that are already at their limiting
capacity or without putting new equipment in parallel, which is expensive. Under these
conditions, there is a change in relative productivities. However, this alternative is not
economically acceptable as the change will spread to other areas, thereby causing a
decrease in the occupational level of the fermenters, which are the most expensive
pieces of equipment.

Table 2.14. Penicillin plant: Results in case of reuse of equipment

a) Reuse b) Reuse c) Reuse d) Reuse e) Reuse
Base Case I.S.(1) I.S.(1) I.S.(2) I.S.(2)
Design B/T=const. B/T ≠ const. B/T=const. B/T ≠ const.
BA (kg) 413.9 805.9 827.9 805.9 827.9
BB (kg) 426.7 857.5 853.5 857.5 853.5
TA (h) 38.5 77.0 77.0 77.0 77.0
TB (h) 38.5 77.0 77.0 77.0 77.0
V1 (m 3 ) 4op x 53.7 2op x 107.9 2op x 107.4 2op x 107.9 2op x 107.4
I .S. Tank (m ) 3 _ 126.1 125.7 126.1 114.1
BA (kg) 413.9 146.5 141.3 146.5 51.9
BB (kg) 426.7 144.8 145.7 144.8 53.5
TA (h) 38.5 14.0 14.0 14.0 14.0
TB (h) 38.5 13.0 13.0 6.5+6.5 3.25
V2 (m 3 ) 6ip x 17.89 2ip x 19.00 2ip x 18.33 2ip x 19.00 2op x 13.46
3
V3 (m ) 3ip x 17.89 19.0 18.33 19.00 6.73
3
V4 (m ) 0.45 0.45 0.45 0.45 0.45
Cart (m 3 ) _ _ _ 1.80 1.80
BA (kg) 413.9 146.5 141.3 146.5.0 1,800.0
BB (kg) 426.7 144.8 145.7 144.8 1,800.0
TA (h) 38.5 14.0 14.0 13.0+1.0 13.0
TB (h) 38.5 13.0 13.0 13.0 13.0
V7 (m 3 ) 3.0 3.0 3.00 3.0 3.0
H A (h) 352.0 361.6 N/A 361.6 N/A
H B (h) 2048.0 2038.4 N/A 2038.4 N/A
Obj. Function ($) 3,494,300 2,410,800 2,399,100 2,410,900 2,324,200

If we accept that we can work with varied productivity in each subtrain, the
presence of a second intermediate storage location allows for further savings (Column e
versus Column c of Table 2.14).

2.8 SUMMARY

In this chapter, the fundamentals of batch plant design have been presented in increasing
complexity: from single product and purely batch equipment considerations, to
multiproduct and various material transfer policies, including the use of intermediate
storage. Finally, a real case from the pharmaceutical sector has been presented and
diverse solution strategies analyzed, including the reuse of available equipment. In
Chapter 5, advanced concepts on the design engineering of more flexible structures will
be presented, which will expand on the fundamentals and simplified structures
discussed in this chapter. The next chapter will cover:

• The incorporation of production planning at the design stage

• The design of multipurpose batch plants
• Design and operation of energy efficient batch processes
• Environmental considerations in batch plant design
• Integrated design and production control

NOTATION

aij; bij; cij Parameters for calculating processing time of stage j of product i.
Bi Batch size of product i.
Ck Expresion to calculate cost of eqquipment k.
Di Demand forecast for product i.
Dik Duty factor for product i in semi-continuous unit k.
Hi Time horizon for the production of product i.
ip Equipment in-phase.
Iit Quantity of product i in inventory at end of period t
K ej Total number of semicontinuous equipment associated to emptying
equipment performing discontinuous task j.
K jf Total number of semicontinuous equipment associated to filling equipment
performing discontinuous task j.
N Total number of products i to be produced in a given time horizon Hi.
ni Number of batches to be produced for product i
mi j Number of parallel batch units operating in-phase at stage j.
mo j Number of parallel batch units operating out-of-phase.at stage j.
M Total number of batch equipment
op Equipment out-of-phase.
Pij Processing time for a batch of product i on the equipment at batch stage j.
PTi Total production time for product i.
Qi Required production for product i.
Rk Processing rate of semicontinuous equipment k.
Sij Size factor (characteristic size of unit j needed to produce unit mass of
product i).
SP Surplus time: difference between available time and total processing time
Ti Cycle time for product i.
TL Limiting cycle time
tij Processing time of batch stage j of product i.
Vj Capacity of batch equipment at stage j.
Greek symbols
α, β, γ Parameters for calculating equipment cost.
θik Processing time of semicontinuous stage k of product i.
e
θ :
ij Emptying time of discontinuour equipment at stage j of product i.
f
θ:
ij Filling time of discontinuour equipment at stage j of product i.
τi Processing time of product i in a subtrain.

Subscripts
i It makes reference to the product.
j It makes reference to the batch equipment or task/stage carried out on it.
k It makes reference to the semicontinuous equipment or task/stage carried
out on it.
s It makes reference to the Indice que hace referencia intermediate storage.
t It makes reference to time period.

Superscripts
d It makes reference to the intermediate storage downstream train.
e It makes reference to the emptyimg equipment of a batch unit.
f It makes reference to the filling equipment of a batch unit.
i It makes reference to the equipment operating in-phase.
o It makes reference to the equipment operating out-of-phase.
u It makes reference to the intermediate storage upstream train.

REFERENCES

Biegler, L.T., Grossmann, I.E., & Westerberg, A.W. (1997). Systematic Methods of
Chemical Process Design. Upper Saddle River, New Jersey: Prentice Hall PTR Inc.
Chapter 6, pp.180-203 & Chapter 22, pp. 719-747.
Espuña, A. and Puigjaner, L. (1989). On the Solution of the Retrofitting Problem for
Multiproduct Batch/Semi-continuous Chemical Plants. Computers and Chemical
Engineering, 13, 483-490.
Espuña, A., Lázaro, M., Martínez, J.M., and Puigjaner, L. (1989). An Efficient and
Simplified Solution to the Predesign Problem of Multiproduct Plants. Computers and
Chemical Engineering 13, 163-174
Espuña, A., Palou, I., Santos, G., and Puigjaner, L. (1990). Adding Intermediate Storage
to Non-continuous Processes. In “Computer Applications in Chemical Engineering”
(edited by H. Th. Bussemaker and P.D. Iedema) Elsevier, Amsterdam, pp. 145-152.

Faqir, N.M. and Karimi, I.A. (1989). Optimal Design of Batch Plants with Single
Production Routes. Ind. & Eng. Chem. Res. 28, 1191.

Flatz, W. (1980). Equipment Sizing for Multiproduct Plants. Chemical Engineering, 87,
71.
Graells, M., Espuña, A., Santos, G., and Puigjaner, L. (1991). Improved Strategy in
Optimal Design of Multiproduct Batch Plants. In “Computer Oriented Process
Engineering” (edited by L. Puigjaner and A. Espuña) 10, pp. 67-74.
Karimi, I.A. and Reklaitis, G.V. (1985a) Variability Analysis for Intermediate Storage
in Non-continuous Processes: Stochastic Case. I. Chem. Eng. Sym. Series 92, 79.
Karimi, I.A. and Reklaitis, G.V. (1985b). Deterministic Variability Analysis for
Intermediate Storage in Non-continuous Processes. AIChE J. 31, 1516.
Korovessi, E., & Linninger, A.A. (2006). Batch Processes. Boca Raton, Florida: CRC
Press, Taylor & Francis Group. Chapter 2, pp.7-39.
Mah, R.S.H. (1990). Chemical Process Structures and Information Flows. Boston:
Butterworths Pub. Chapter 6 & 7, pp. 241-347.
Puigjaner, L. (1999), Handling the Increasing Complexity of Detailed Batch Process
Simulation and Optimisation. Computers & Chemical Engineering, 23 S, pp. S929-
S943.
Rauch, J. (2003). Multiproduct Plants. Wiley-VCH, Weinheim: Verlag, GmbH & C.
KgaA.
Reklaitis, G.V., Sunol, A.K., Rippin, D.W.T., & Hortaçsu, O. (1996). Batch Processing
Systems Engineering. Berlin: Springer-Verlag.
Sharratt, P.N. (1997). Handbook of Batch Process Design. London: Blackie Academic
and Professional, Chapman & Hall. Chapter 2, pp 24-60.
Vaselenak, J.A., Grossmann, I., and Westerberg, A.W. (1987b). Optimal Retrofit
Design of Multiproduct Batch Plants. Ind. Eng. Chem. Res. 26, 712.
Yeh, N. and Reklaitis, G.V. (1987). Synthesis and Sizing of Batch/Semicontinuos
Process. Computers & Chemical Engineering, 11, 639-654.

EXERCISES