ArticlePDF Available

Abstract and Figures

The objective of this paper is to present a new formulation for the optimal scheduling of multipurpose batch plants where equipment redesign is considered simultaneously with the scheduling decisions. The equipment redesign is characterized by the implementation of modifications in the existent processing units so as to change their suitability to perform certain tasks, while regarding tasks’ characteristics inside a given scheduling horizon. This approach may be advantageous in cases where no schedule solutions are found with the existent equipments and where, with minor technology modifications on the processing units, feasible schedules can be obtained. Each of these changes has a cost and requires a certain time to be implemented. In order to model such problem a simple Mixed Integer Linear Programming formulation (MILP) is proposed having as basis the unified Resource-Task Network (RTN) representation presented by Pantelides (1994). An example motivated by a chemical-pharmaceutical industry is used to demonstrate the applicability of the proposed formulation.
Content may be subject to copyright.
SCHEDULING WITH EQUIPMENT REDESIGN IN
MULTIPURPOSE BATCH PLANTS
Samuel Moniz
1,3
, Ana Paula Barbosa-Póvoa
1
*
, Jorge Pinho de Sousa
2,3
1
Centro de Estudos de Gestão, Instituto Superior Técnico,
Universidade Técnica de Lisboa, 1049-001 Lisboa, Portugal
2
INESC Porto
3
Faculdade de Engenharia da Universidade do Porto
Abstract
The objective of this paper is to present a new formulation for the optimal scheduling of multipurpose
batch plants where equipment redesign is considered simultaneously with the scheduling decisions. The
equipment redesign is characterized by the implementation of modifications in the existent processing
units so as to change their suitability to perform certain tasks, while regarding tasks characteristics
inside a given scheduling horizon. This approach may be advantageous in cases where no schedule
solutions are found with the existent equipments and where, with minor technology modifications on the
processing units, feasible schedules can be obtained. Each of these changes has a cost and requires a
certain time to be implemented. In order to model such problem a simple Mixed Integer Linear
Programing formulation (MILP) is proposed having as basis the unified Resource-Task Network (RTN)
representation presented by Pantelides (1994). An example motivated by a chemical-pharmaceutical
industry is used to demonstrate the applicability of the proposed formulation.
Keywords
Multipurpose batch plants, simultaneous scheduling and design, equipment redesign
*
To whom all correspondence should be addressed, apovoa@ist.utl.pt
Introduction
The chemical-pharmaceutical industry has been facing
an increasing demand for the production of a high variety
of low volume products at a minimum cost. Such pressure
leads to the need of production systems that run efficiently
both in terms of cost and time. Consequently, production
flexibility is required so as to accommodate the customers’
orders within acceptable response times and costs usually
imposed by the market. To compete in such environment,
the chemical industry has been using multipurpose batch
plants that are characterized by having a set of resources
(processing units, raw materials, utilities, manpower, etc.)
that can be shared, so as to produce several products.
These plants are especially attractive in situations where
product demands and formulations change rapidly, since
they can be easily adapted to the production specificities of
each product. Moreover, changes in a plant such as the
addition of new processing units or connections and the
removal of old inefficient units are decisions that can also
be considered. In this context, planning and scheduling
become important functions of the production system
enabling a flexibility increase of the multipurpose batch
plants while minimizing costs. This problem has been
addressed in the literature as the design and retrofit of
multipurpose batch plants. For the most recent review on
these issues see Barbosa-Póvoa (2007). The design of
batch plants from scratch is referred as a grassroot problem
while the redesign of an existing plant is denoted as a
retrofit problem. Two additional concepts have been used
to categorize these research problems: basic design and
extended design. As stated by Barbosa-Póvoa (2007) the
former refers to the simple choice of equipments and
associated scheduling, while the latter goes further and
addresses scheduling and detailed design where not only
the choice of the equipment is considered but also
topology and operational aspects are explored. A number
of papers have been published on these topics and the
proposed models cover a large number of problem features
such as: the selection of the processing units and their
sizes; addition of storage vessels; storage policies; design
of equipment units connections; operating mode cyclic
and non-cyclic; campaign structure; and 2D and 3D layout
design.
Furthermore, when looking into the batch scheduling
problem as a standalone problem, the aim is to operate a
set of resources so as to produce a set of products within a
defined scheduling period. For a detailed review on this
topic the work of Mendez et al. (2006) should be analyzed.
Batch scheduling problems need to deal with a great
variety of aspects that are intrinsically linked to the
problem structure. Some of the most important of these
aspects are: multiproduct and multipurpose batch
topologies; equipment connectivity; inventory storage
policies; material transfer; batch size and batch processing
time; and changeovers. When modeling such problems one
of the most important issues is the time representation,
which can be discrete or continuous. Discrete formulations
have been shown to be a good approach for those
scheduling problems that can be represented with a
reasonable, not too large, number of time intervals (Castro
et al., 2003). Continuous formulations explicitly represent
the timing decisions as a set of continuous variables, as a
way to define the exact time at which the events occur.
Typically, this results in the reduction of the number of
variables of the model. Despite the added flexibility,
continuous formulations tend to increase the models
complexity by means of the use of big-M constraints.
As mentioned before most of the work performed on
the scheduling problem of multipurpose batch plants
mainly addresses the optimal utilization of a set of existent
resources so as to produce what the customers need. On
the other hand, the design and retrofit of multipurpose
batch plants looks into the need of designing a plant from
scratch or redesigning the existing plant, by adding new
units or connections. Nevertheless, an intermediate
problem, somewhere between the design and the
scheduling problem, is often faced by multipurpose
process companies when trying to produce a new set of
products, see Figure 1. This problem is related to the need
of performing changes in the existing processing units
equipment redesign so as to improve the existent
equipment suitability, thus providing more flexibility to the
plant. The timing of the equipment redesign decisions is
similar to the scheduling decisions since their scope is also
of short-term. Furthermore, the retrofit and grassroots
design take time to be implemented in the shop-floor and
may require large investments, hence these decisions must
be considered in the long-term planning. The equipment
redesign assumes more relevance in industries that perform
process development, since the production recipes evolve
with it and for that reason it may be necessary to modify
the processing units.
Figure 1. Impact of scheduling, planning and
design decisions over the time horizon.
As an example, we have the addition or removal of
cleaning-in-place (CIP) systems as well as the addition or
removal of temperature or sampling systems. Such
operations allow for changes in the equipment’s suitability
so as to perform new process recipe tasks. Doing this, new
design and scheduling alternatives are then generated at
lower cost and with smaller time consumption.
This problem is addressed in the current paper and has
emerged from a real problem that is been addressed by the
authors in a chemical-pharmaceutical industry. Unlike the
previous research on this topic, that has been addressing the
plant design as grassroot or retrofit problems at the global
plant level, we consider that performing specific changes in
the processing units can be an alternative to tackle
scheduling and design problems simultaneously. A Mixed
Integer Linear Programming (MILP) model is proposed
based on the Resource-Task Network (RTN) representation
presented by Pantelides (1994).
The remaining of the paper is structured as follows. We
first present the problem definition as well as the modeling
framework that is being used. Two ways of modeling the
equipment redesign problem are then characterized. One
uses the original RTN formulation and the other is an
extended RTN formulation. We present the computational
results of a scheduling problem motivated by the chemical-
pharmaceutical industry under study, where equipment
redesign is a regular approach when performing the
production schedule. We finish the paper with the
conclusions and some future work is also suggested.
Problem Definition
As referred above the generic scheduling problem
assumes that, when performing scheduling, there must be a
perfect match between the tasks requirements and the
existent processing units’ characteristics. Clearly, this is not
easy to do due to the large number of processing units
existing in the plant and due to the various recipes
requirements. Finding a schedule solution without relaxing
any of these inputs is often difficult to accomplish, mainly
when the plant operates close to the maximum capacity and
when new products are frequently being introduced. In these
cases to get feasible schedules usually requires re-
negotiating new order due dates with the customers.
Nevertheless, new alternatives for the schedules can also be
generated with some equipment modifications involving
little costs and time.
The use of multipurpose reactors is indeed
advantageous in these situations since such units are very
flexible and can often perform several tasks. Additionally,
their operating range can be increased by doing small
equipment modifications. The same reasoning can be
applied to all processing units whose suitability to execute
tasks can be changed quickly. The redesign problem takes
into account the setup-time to perform the equipment
modifications and, at the same time, the resources that are
needed to do the modification. This approach transforms the
processing units into more generic units capable of
executing more tasks. From the point of view of the
operations this adds flexibility, since more scheduling
alternatives can be explored. Such scheduling with
equipment redesign is modeled in the present work, and can
be described as follows:
Given:
the RTN representation of the process (tasks and
resources);
the number of processing units available, and
their maximum and minimum capacity;
the scheduling granularity and time horizon;
the production requirements during the time
horizon;
the auxiliary equipments that can change the
suitability of the processing units;
the cost and setup-time to add and remove
auxiliary equipments;
Determine:
a process schedule such that the processing units
suitability change during the time horizon;
an equipment modification plan to respond to the
above schedule, taking into account the setup
times for adding and removing the auxiliary
equipments and their limited availability;
Minimize:
the processing units modification costs plus the
operational costs, while respecting the delivery
due dates.
Problem Modeling
The problem considered here is modeled with a
discrete time formulation based on the Resource-Task
Network representation proposed by Pantelides (1994).
The scheduling of a set of products is performed in a set of
existing equipments allowing for modifications in some
resources. The set of modifications is identified
simultaneously with the definition of the production
schedule, within a pre-defined time horizon.
Resource Task Network discrete formulation
The Resource-Task Network representation proposed
by Pantelides (1994) involves two types of entities, tasks
and resources. A task is an abstract operation that
consumes and/or produces a specific set of resources
(material, equipment items, utilities, etc.). For the purposes
of the discrete time formulation presented in this paper, the
time discretisation is made fine enough so that all tasks can
be considered to start and end at a time interval boundary.
Each task has a fixed duration
k
and the execution of task
k starting at time t is characterised by its “extent” - a pair
of variables (N
kt
,
kt
). N
kt
is the number of instances (either
0 or 1) of task k starting at time t while,
kt
is the total
amount of material that is processed by all these instances.
Resources are produced and consumed at discrete times,
during the execution of the task. The amount of resource r
produced or consumed by a task k at different times over
its duration
k
can be obtained from the values of the
extent variables. Changes to the resource utilisation can
occur only at interval boundaries. The amount of unused
(“excess”) resource r, held over time interval t, is denoted
by R
rt
.
As presented by Pantelides (1994) the RTN discrete
scheduling problem can be described by the following
three types of constraints:
HtRr
NRR
rt
Kk
tkkrtkkrtrrt
r
k
,
0
,,1,
(1)
HtRrRR
rtrt
, 0
max
(2)
HtKkErNVNV
rktkrktktkr
,,
maxmin
(3)
Constraints (1) express resource balancing through the
variables R
rt
, that state the availability of resource r at time
t. The amount of resource r consumed and produced at
each time is expressed by the integer and continuous part
of constraints (µ
krϴ
N
k,t-ϴ
krϴ
ξ
k,t-ϴ
). N
k,t-ϴ
is a binary
variable that takes the value 1 if task k starts at time t, and
ξ
k,t-ϴ
indicates the amount of material being produced at
each time period, i.e., the batch size. The parameters µ
krϴ
and
ν
krϴ
represent the fixed and variable resource
consumption/production respectively. Constraints (2) limit
the availability the resources to the maximum value
max
rt
R
during the time horizon. And constraints (3) set the
batch sizes within the limits of the resource
capacity
min
kr
V
and
max
kr
V
, where E is the subset of R for the
processing units, and K
r
is the set of tasks that use resource
r.
Equipment redesign problem using the RTN
Applying the existing formulation to the equipment
redesign problem requires the explicit representation of all
possible modification alternatives. Hence, we need to
create new tasks to explicitly take into account all steps
required to modify the processing units, i.e. to model the
addition and removal of auxiliary equipments. This
approach will make the network of processing tasks very
complex and more difficult to tackle.
Figure 2 shows how the RTN formulation can deal
with the equipment redesign problem. To consider the
setup time for adding and removing the auxiliary
equipment CIP on Reactor1, we need to create two
additional tasks (Add_CIP and Remove_CIP), and one
extra resource (Reactor1_CIP). This allows us to model the
availability of Reactor1 after the modification, i.e., having
Reactor1 with a CIP system installed.
Figure 2. RTN of the equipment redesign
problem (reversible modification)
If the modification is irreversible there is no removing
task; if the modification is reversible it is necessary to
create two tasks: one to add the auxiliary equipment to the
processing unit, representing the equipment modification,
and another task to remove the previously installed
auxiliary equipment, providing the processing unit with its
initial suitability. The network of processing tasks requires
the explicit representation of all possible combinations of
auxiliary equipments (e.g. CIP, sampling devices and
temperature systems) and processing units (e.g. reactors,
filters, dryers). In the case of the reversible modifications,
two additional tasks and one extra resource will be added
to the model for each equipment modification needed. For
these reasons, the model complexity for representing the
problem using the RTN formulation rises. The same
obviously happens with the computational time needed to
obtain a solution.
Equipment redesign problem using an extended RTN
formulation
An alternative approach to tackle this problem is to
create two additional sets of binary variables to control
when the processing unit needs to be modified in order to
be suitable for the task execution, see constraints (4).
HtRr
MM
NRR
rt
Kk
s
u
utk
rukutkruk
Kk
tkkrtkkrtrrt
k
r
k
,
'
r
'
'
0
,'
',''
0
,,1,
(4)
To express the redesign of the processing units, we
will use the binary variables
kt
M
and
kt
M
that will be
equal to 1 if a modification (addition or removal
respectively) occurs by means of the task k at the time
interval t. The parameter λ
kru
denotes the resources r that
will be consumed (e.g., CIP and Reactor1) by an
equipment modification required by a task k during the
interval u, once the modification has started. The
parameter γ
kru
denotes the reverse operation. It consumes
the modified resource (e.g., Reactor1) and releases back
the resources (e.g., CIP and Reactor1). The setup-time
required for each modification is given by the parameter s
k
.
Constraints (1) are modified and a third term is added to
reflect this behavior. The
utkkru
M
,
expression enforces
the modifications to be done by each task k, while the
utk
kru
M
,
part denotes the removal of the auxiliary
equipment from the processing units.
The entire formulation also guarantees that the
auxiliary equipment cannot be removed during the task
execution and that the setup-times s
k
for modifying the
processing units are respected. K’
r
is a subset of K
r
that
denotes the tasks that require redesign through the
resources r. More specifically, for the example given in
Figure 2, we get the λ
Reaction,Reactor1,0
= λ
Reaction,CIP,0
=-1 and
λ
Reaction,Reactor1,1
=1 and γ
Reaction,Reactor1,0
= -1 and
γ
Reaction,Reactor1,1
= γ
Reaction,CIP,1
=1, see Figure 3.
Figure 3. Equipment redesign modeling with
the alternative formulation
An additional constraint type is also needed for the
correct assignment of the
kt
M
and
kt
M
binary variables.
Since the equipment modification needs to be done before
the task starts, constraints (5) guarantee that the auxiliary
equipment has been previously installed. A is the subset of
R which has auxiliary equipments needed to modify the
processing units, and K
kr
is the set of tasks that share the
auxiliary equipment r.
HtKkArMMN
r
t
Kk
k
ktk
kr

,,
0 '
',
',,
(5)
When the binary variables N
kt
are equal to 1, the right
hand side of the constraints needs also to be 1, therefore
having at that time instant a sum (involving the
kt
M
and
kt
M
variables) equal to 1. In practice, this means that the
auxiliary equipment needs to be previously consumed by
that task, or by other task that was executed in the past and
that required the same auxiliary equipment in the same
processing unit.
With this formulation, there is no need to explicitly
write the modification tasks. Instead two sets of additional
binary variables are added to the model to express the
addition and removal of auxiliary equipments to the
processing units. The resources are still treated uniformly
as they are in the original RTN formulation.
Finally, for both formulations the objective function
considered in this work is the minimization of the
processing units modification costs
k
C
and,
k
C
as well as
the operational costs
k
O
, see equation (6).
'
,,,,
,
min
Kk
t
trkktrkk
k t
tkk
MCMC
NO
(6)
Case Study
A real world problem from a chemical-pharmaceutical
industry is solved using both presented formulations. The
company performs the development and production of
complex and fine chemicals to the pharmaceutical industry
and biotechs. Its core business is the development and
manufacture of new active pharmaceutical ingredients
(APIs). In this business, the chemical industry is
continuously challenged to respond within short time
windows. On the one hand, the company needs to manage
small batches of under development products and, on the
other hand, needs to produce large batches of products in
commercialization. Thus, operations flexibility is required
to respond to this heterogeneous demand. This adds extra
complexity to operations management especially to the
planning and scheduling functions.
The product object of our analysis goes through a
sequence of tasks such as reaction, precipitation,
crystallization, filtration, suspension, drying, quality
control and packaging, which can be performed by the
following resources: four reactors, one vessel, one filter,
one dryer and a packaging room. The typical production
time is around ten days. For illustration purposes, we will
focus here on the multipurpose reactors since these are the
most difficult resources to schedule, thus imposing the
schedule of the remaining resources. Devices such as CIP
and temperature systems (TS) are considered auxiliary
equipments that can be used for the reactors redesign. The
reaction, precipitation, crystallization and suspension tasks
can either be executed in reactors that do not require
modifications but have small capacity, or can be executed
in reactors with higher capacity but need to be modified at
a certain cost. The product must be delivered at a date and
quantity agreed with the customer. The objective is to get
the optimal schedule for this product, minimizing the
global operation and modification costs, while respecting
the product delivery date.
Case Study Results
The scheduling problem was solved for a time horizon
of ten days. The time was discretized to one shift of eight
hours, which resulted in a scheduling horizon of 30 time
intervals (three shifts per day). We have considered an
operational cost for each task depending on the processing
unit that is used. Tasks that take place in low capacity
reactors (capacity of 4,000 liters) have an operational cost
of 70 mu (monetary units) and tasks that are performed in
high capacity reactors (capacity of 10,000 liters) have a
cost of 100 mu. In the course of the recipe production the
tasks’ characteristics may change requiring the processing
units redesign. For instance, precise temperature control is
needed on Mixing and Precipitation tasks at Reactor1 and
Reactor2, and a CIP system must be available in Reactor2
and Reactor3 when performing Reaction and Stirring tasks,
respectively. The costs to modify a reactor with a CIP and
TS are, respectively, 3 mu and 5mu. The setup-time to
modify the reactors with a CIP is 8 hours, while for a TS is
16 hours. The time required to remove those systems from
the reactors in order to restore their original suitability is
equal to 8 hours for both auxiliary equipments. One final
product delivery of 2 tons is scheduled for the entire
schedule horizon. The optimal schedule obtained for our
example is depicted in Figure 4. This optimal solution has
a value of 2074 mu. Although this test instance is relatively
simple, it allows us to understand the tradeoffs existing in
the equipment redesign problem, between equipment’s
suitability and the setup-time and costs to perform the
equipments modifications. As can be seen in Figure 4, to
respect the delivery date, equipment redesign tasks must
take place. To perform the Reaction task in Reactor2 it is
necessary to add a CIP, and to do the Precipitation task in
this same reactor it is necessary to add a TS. These tasks
can be seen at the time interval 0 and 5 of the schedule,
respectively. The same reasoning applies to the Mixing
task at Reactor1 and to the Stirring task at Reactor3. But
note that no auxiliary equipments were defined for the
Crystallization task at Reactor2 and for the Cooling task at
Reactor3, that nevertheless were modified previously. In
the end of this schedule Reactor2 had a TS installed, while
Reactor3 had a CIP mounted. The MILP model using
Pantelides formulation resulted in 1178 binary variables,
2202 continuous variables and 5085 constraints.
Optimality could be proved in 3.15 seconds. The extended
formulation has 775 binary variables, 1396 continuous
variables, 2853 constraints and reached the optimal
solution in 1.78 seconds.
REACTOR1
Mix
RTS
REACTOR2
ACIP
RCIP
REACTOR3
ACIP
Stir. Coo Coo
REACTOR4
VESSEL1 Hold
FI LTER1
DRYER1
Dry. Dry
QUALITY C.
PACK.ROOM
Pck Pck
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
time
Quality contr.
Quality contr.
Suspen.
Suspen.
Suspen.
Fil tration
Fil tration
Fil tration
Fil tration
ATS
ATS
Reaction
Precip.
Crystallization
Precip.
Crystallization
Legend:
ATS - a dd temperature system
RTS - remove temperature s ystem
ACIP - a dd CI P
RCIP - remove CIP
Reactor2_CIP
Reactor2_TS
Reactor3_CIP
Figure 4. Optimal production schedule with the equipment redesign plan
The model was implemented using ILOG/CPLEX
version 12.2 on an Intel Core i7 at 2.67GHz with 4 GB of
RAM. The extended formulation has less binary and
continuous variables and a smaller number of constraints.
When analyzing these results some disadvantages can
be pointed to the original RTN formulation when using it
in the redesign problem. It requires the representation of
all modification tasks, which results in a complex network
of processing tasks. One needs to create additional
resources to manage the modified equipments, such as for
instance: Reactor2_CIP and Reactor2_TS; these are two
additional resources that define Reactor2 modified with a
CIP and a TS, respectively. At the same time, since we are
assuming the redesign process increases the processing
units’ suitability such that more tasks can be performed, we
must represent all new production alternatives. For
instance, the Crystallization task does not require any
change on Reactor2, nevertheless if this reactor is modified
with a CIP or TS, becoming Reactor2_CIP, Reactor2_TS
or Reactor2_CIPTS, we need to create several additional
tasks to allow for the possibility of the task being executed
in one of these resources. This kind of tasks needs to be
created for all resources that can be modified, thus
increasing the model size. These drawbacks are overcome
in the proposed formulation by replacing the redesign tasks
by
kt
M
and
kt
M
binary variables. The resulting model is
smaller and it is easier to write since it does not require the
representation of additional tasks. The redesign tasks are
simply modeled by the
kt
M
and
kt
M
variables. For that
reason, the resulting MILP has less binary and continuous
variables. Nevertheless, the use of the
kt
M
and
kt
M
variables limits the equipment modification to one
auxiliary equipment per task. The possibility of doing more
than one modification per task would clearly be an
interesting extension of our model.
Conclusions
This paper has addressed a new type of problem that is
being faced by the chemical-pharmaceutical industry using
multipurpose batch plants, and performing simultaneous
design and scheduling within a short period of time. The
equipment redesign problem concerns the need to perform
changes in the processing units such that their suitability is
increased and therefore the units are capable to perform
additional tasks. The redesign tasks can be seen as an
additional way to increase flexibility of these plants. The
redesign problem was formulated using the RTN
formulation introduced by Pantelides and an extension to
this formulation was also proposed in this work. While the
RTN formulation requires the explicit representation of all
production alternatives, taking into account the different
states of the modified resources, the extension here
developed deals with the equipment redesign decisions
through two extra groups of binary variables. Preliminary
computational results show that the proposed formulation
has better performance. The formulation applicability was
tested in an industrial example and the achieved results are
promising but improvements should be further explored.
Namely, it would be interesting to extend the formulation
to deal with multiple modifications per task. Also more
comprehensive tests need to be performed to further
compare the two analyzed formulations.
Acknowledgement
The authors would like to thank Pedro Duarte from
Hovione FarmaCiencia SA for his continuous help and
gratefully acknowledge the financial support of Hovione
and Fundação para a Ciência e Tecnologia under the grant
No. SFRH/BD/33970/2009.
References
Barbosa-Póvoa, A. P. (2007). A critical review on the design and
retrofit of batch plants. Computers & Chemical
Engineering, 31, 833-855.
Castro, P. M., Barbosa-Póvoa, A. P., & Matos, H. A. (2003).
Optimal periodic scheduling of batch plants using
RTN-based discrete and continuous-time formulations:
A case study approach. Industrial & engineering
chemistry research, 42, 3346-3360.
Mendez, C. A., Cerda, J., Grossmann, I. E., Harjunkoski, I., &
Fahl, M. (2006). State-of-the-art review of
optimization methods for short-term scheduling of
batch processes. Computers & Chemical Engineering,
30, 913-946.
Pantelides, C. C. (1994). Unified frameworks for optimal process
planning and scheduling. In (pp. 253-274): Cache
Publications New York.
Article
The chemical-pharmaceutical sector is facing an unprecedented fast-changing environment, with new market and technological trends impacting the companies’ operational strategies. Managing the pharmaceutical supply chain (PSC) operations is, therefore, ever more complex and challenging. The goal of this work is to present a comprehensive overview of the current state of the industry and research developments; and then, to develop a new decision-making reference framework to assist in the creation of optimization-based decision support models. This will be achieved through a multi-perspective analysis that encompasses strategic and tactical planning decision-making, in the current and future business context of the chemical-pharmaceutical industry. The findings reveal a lack of research addressing the most prominent trends currently driving this sector, such as patient centricity or new technological developments, thus highlighting the disruptive nature of the expected changes in a highly conservative industry.
Thesis
Full-text available
The main objective of this thesis was to development scheduling models for multipurpose batch plants operating in the context of the pharmaceutical industry. The production scheduling problem is commonly recognized as being very difficult since it must deal with several potential conflicting objectives. The primary goal of production scheduling is to produce the right amounts of product at the right time, cost, and quality. For that purpose, model-based approaches can be applied so as to obtain optimal (or close to optimal) scheduling solutions. Several challenges that arise at this level are related to implementation, modeling issues, and computational efficiency when solving large-scale problems. Nevertheless, the application of such models in real world scheduling problems clearly creates improvement opportunities for logistics and manufacturing activities. In this thesis, an innovative methodology is introduced for efficiently representing and solving the integrated scheduling problem, based on a new general discrete-time model. The characteristics of the chemical-pharmaceutical industry led to the definition of an extended view of the scheduling problem that accounts for units redesign decisions. Decomposition methods and reformulation strategies are also introduced to address the computational complexity of the models. The effectiveness of the proposed methods is illustrated by solving several real world instances. Practical implementation issues of the scheduling methodology are also discussed so as to demonstrate its application potential.
Conference Paper
The retrofit design of a multipurpose batch pilot plant is studied so as to accommodate the addition of a new product. A model is utilised which represents in detail structural and operational characteristics as required for production scheduling, while also accounting for cleaning in place (CIP) integration. The State Task Network concepts for defining process recipes and procedures are augmented by the definition of the equipment states (unit dirty, connection dean, etc.). A superstructure of several proposed plant modifications is considered to debottleneck the plant and provide increased flexibility, including piping reconfigurations and the replacement of the existing flow plates with double seat valves manifolds. The retrofit problem is formulated as a Mixed Integer Linear Program (MILP) and solved using a branch and bound technique. The best compromise between extra capital cost (plant reconfiguration) and operating cost/revenues (operating schedule and production amount of the new product) is obtained by maximising the annualised profit. Results show that a small capital expenditure permits the production of the new product within the original time horizon.
Article
The retrofit design of a multipurpose batch pilot plant is studied so as to accommodate the addition of a new product. A model is utilised which represents in detail structural and operational characteristics as required for production scheduling, while also accounting for cleaning in place (CIP) integration. The State Task Network concepts for defining process recipes and procedures are augmented by the definition of the equipment states (unit dirty, connection clean, etc.). A superstructure of several proposed plant modifications is considered to debottleneck the plant and provide increased flexibility, including piping reconfigurations and the replacement of the existing flow plates with double seat valves manifolds. The retrofit problem is formulated as a Mixed Integer Linear Program (MILP) and solved using a branch and bound technique. The best compromise between extra capital cost (plant reconfiguration) and operating cost/revenues (operating schedule and production amount of the new product) is obtained by maximising the annualised profit. Results show that a small capital expenditure permits the production of the new product within the original time horizon.
Article
This paper considers the problem of deriving an optimal periodic schedule for an industrial batch plant. Both discrete and continuous-time formulations, based on the general resource task network process (RTN) representation, are employed. For a given cycle time, the proposed discrete-time formulation results in a mixed integer linear programming (MILP) program that can be solved to optimality in a reasonable time, even for a fine discretization of the time grid. The optimal cycle time is then determined by solving a sequence of fixed cycle time problems. On the other hand, the continuous-time formulation results in a mixed integer nonlinear (MINLP) problem that under the assumption of constant throughput becomes a MILP. This can be solved to optimality within reasonable computational effort only for a relatively small number of event points, making it practically impossible to find the global optimum. These results favor the use of the discrete-time formulation over its continuous-time counterpart.
Article
There has been significant progress in the area of short-term scheduling of batch processes, including the solution of industrial-sized problems, in the last 20 years. The main goal of this paper is to provide an up-to-date review of the state-of-the-art in this challenging area. Main features, strengths and limitations of existing modeling and optimization techniques as well as other available major solution methods are examined through this paper. We first present a general classification for scheduling problems of batch processes as well as for the corresponding optimization models. Subsequently, the modeling of representative optimization approaches for the different problem types are introduced in detail, focusing on both discrete and continuous time models. A comparison of effectiveness and efficiency of these models is given for two benchmarking examples from the literature. We also discuss two real-world applications of scheduling problems that cannot be readily accommodated using existing methods. For the sake of completeness, other alternative solution methods applied in the field of scheduling are also reviewed, followed by a discussion related to solving large-scale problems through rigorous optimization approaches. Finally, we list available academic and commercial software, and briefly address the issue of rescheduling capabilities of the various optimization approaches as well as important extensions that go beyond short-term batch scheduling.
Article
This paper presents a review on the design of batch plants where both the grassroots and the retrofit design problems are analysed. Two types of plants are considered, the multi-product and the multipurpose plants. The characterisation of the design problem is made and the key decisions and elements involved are identified. Limitations and gaps in the existing published approaches are discussed. Although, a considerable number of studies have appeared in the last 15 years there is still a need for further development on the models generalisation and applications. Further attention should be given to the multi-objective design where different objectives are present at the design level. Great investment is also required on the solution methods so as real case problems can be solved in reasonable computational time.The paper concludes with the identification of future research challenges in the design and retrofit of batch plants.