=Paper=
{{Paper
|id=Vol-2486/icaiw_aiesd_1
|storemode=property
|title=A Novel Approach to Address Process Plant Layout based on a Bacterial Genetic Optimization
Algorithm
|pdfUrl=https://ceur-ws.org/Vol-2486/icaiw_aiesd_1.pdf
|volume=Vol-2486
|authors=Brandon Y. Morales-Calvache,Santiago Vásquez-Méndez,Fabian C. Prada-Ariza,Camilo Mejı́a-Moncayo
}}
==A Novel Approach to Address Process Plant Layout based on a Bacterial Genetic Optimization
Algorithm==
https://ceur-ws.org/Vol-2486/icaiw_aiesd_1.pdf
A Novel Approach to Address Process Plant
Layout based on a Bacterial Genetic
Optimization Algorithm
Brandon Y. Morales-Calvache , Santiago Vásquez-Méndez , Fabian C.
Prada-Ariza , and Camilo Mejı́a-Moncayo
Universidad EAN,
Calle 79 # 11-45, Bogotá, Colombia
{bmorales8762,svasquez4323,fpradaar2767,cmejiam}@universidadean.edu.co
Abstract. The process plant layout is a determining factor for the effi-
cient and secure operation of it. In this sense, define the best position of
each process unit contributes to reducing transport costs and the explo-
sion and fire risk. This issue has been addressed in the past by different
methods like expert criteria, simulation, and optimization. In this con-
tribution, a hybrid bacterial genetic optimization algorithm for process
plant layout is introduced, which solves a mathematical model that min-
imize interconnection cost and land cost. In addition, the model takes
into consideration facilities dimensions, minimal distances among facil-
ities, and minimum distance up to the property boundary. The perfor-
mance evaluation was carried out with a case of study of literature, and
the results expose the advantages of this proposal in performance and
time.
Keywords: Process plant layout · Bacterial optimization algorithm ·
Genetic Algorithms · Metaheuristics · Chemical Engineering
1 Introduction
The optimal design of processes plants layouts is one of the main issues in chem-
ical engineering, which is in continuous development as is described by Xu and
Papageorgiou [7], which expose different approaches to address this problem.
Despite the advances in the computational matter, to achieve that all the design
factors converge in a single model is impossible, and the same process variables
are not used in all cases. Among the most used factors are the economic, oper-
ability and flexibility, availability for future extensions, security and reliability,
and finally the environmental.
Plant design is a crucial step for a new production plant; but in some cases
during the design process, some risks are dismissed, due to the need to produce
more efficiently. However, to minimize the risks and possible emergencies has
been more relevant because they represent a more significant amount of costs
for lawsuits or infractions to the norms.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0)
2019 ICAI Workshops, pp. 1–11, 2019.
2 B. Morales-Calvache et al.
Therefore, companies must evaluate the location and distribution of the plant
meticulously, because they must know in detail about what, how? With what?
Also, where? Provide their services, together with the details of the capacity to
achieve the best functioning of the facilities [5].
Poor distribution of spaces in the plant leads to problems such as hampering
areas necessary for movements, harming workers or staff in the vicinity, or limited
storage space. These because many facility designs lead to solutions that not
admit modifications in areas, forms, or orientation, for which in their design do
not take into account passages or interior walls.
Regarding production process, the excellent management of spaces is essen-
tial, because it contains high-risk phases, where if the equipment is close to places
where workers are located, it can cause fatal damage and even death, either by
the uncontrolled reaction, pressure leaks or by inhalation of toxic reagents during
production.
Having said the above, the main interest in the realization of this work is to
define the best position of a set of facilities in a process plant by mean of a hybrid
metaheuristic algorithm based on artificial bacteria and sexual reproduction of
Genetic Algorithms in the solution of study case from [4].
2 Theoretical framework
In the design of process plants, it is essential, the definition of the location of each
equipment into the plant. Due to it is the starting point to optimize each of the
sub-processes that make up the production plant. That is why these problems,
make up a disjunctive to which every organization faces, in which it is necessary
to define the allocation of the equipment or process and the required connections
among these. In this sense, a high number of different approaches have been
presented. The initial approximations were based on the heuristics, just as it
did [10] where they performed a solution without guaranteeing that it is the
most optimal. R. Jayakumar [2] proposed the location of units by sections and
defines that there are three methods of developing these problems, the exact
plans, heuristics, and meta-heuristics; depending on the mathematical model
formulation which fits the conditions that the study case needs [11].
In this way, heuristic meta-models began to appear, in which it is possible to
recover or take into consideration previous models. An approach applied to the
styrene industry is shown in [1], in which genetic algorithms are implemented and
gave an effective strategy to find excellent and practical solutions for a design
problem, although it did not guarantee to find the global optimum.
Thanks to the above meta-heuristic approaches, which are the first approach
in the hybridization of heuristics for the resolution of combinatorial problems [6],
allowing to arrive at the non-linear programming of mixed integers, as the model
proposed by Penteado and Ciric [9], where financial risks related to accidents
and spreading to neighboring units are taken into account for the first time, as
well as other terms such as pipes, land and the costs of protective devices.
A Novel Approach to Address Process Plant Layout 3
Once entered into the non-linear programming of mixed integers, we con-
tinue with the mathematical models which were presented by Papageorgiou and
Rotstein [7]; where the optimal location and orientation for each equipment is
determined, and simultaneously the performance criterion given in the design is
minimized, this model was generalized to take into account the design organiza-
tion of the process plant in the production sections.
Due to the multiple configurations of different plants, it is necessary to take
into consideration the study options carried out by Suzuki [10] which are the
assignment of equipment to different floors, that is, not to use a plant along
only, but multiple levels plant, which satisfies preferential number of equipment
and also takes into account vertical pumping and land costs.
That said, [8] presents a mathematical model where it considers the problem
through the design of multiple floors, which determines the number of levels,
the necessary land, the location of equipment and the optimal amount of equip-
ment per floor simultaneously, this in order to minimize total costs in the design
of the plant, this presents the formulation of the mathematical model, while
demonstrating its applicability to the demonstration in the non-linear model,
this document has been defined as fundamental basis for the generation of the
mathematical model presented by Jung [3]. In which explores the available meth-
ods for the design of chemical plants with better and safer designs, where he
incorporates the security process and with QRA approaches within the models
used to solve the problems of plant designs, where they managed to overcome
some of the difficulties associated with non-linearity, uncertainties, and precision
for modeling consequences.
3 Mathematical model
The mathematical model defined for this contribution takes into consideration
the study case formulated by Jung et al., [4]. The model minimizes the land
cost and the interconnection cost equation 1, and it is subjected to three con-
straints, which are described in equations 2 to 4. Equation 2 guarantees that
the distance d(i, j) between the facilities centers be more than the minimum
allowed dmin(i, j), using Euclidean distance. Equations 3 and 4 guarantees that
the facilities’ borders being separated higher than dminf (i), the minimal dis-
tance from the property boundary to the facility. Equations 5 to 7 defines how to
calculate the land cost. Equation 8 defines the interconnection cost and finally,
equation 9 establish how is calculated dmin(i, j). The model is shown as follow:
Minimize:
Z = LC + IC (1)
Subject to:
d(i, j) ≥ dmin(i, j) (2)
Lx(i)
x(i) ≥ + dminf (i) (3)
2
Ly(i)
y(i) ≥ + dminf (i) (4)
2
4 B. Morales-Calvache et al.
Where:
Land cost (LC):
Lx(i)
Xsize = max x(i) + + dminf (i) (5)
2
Ly(i)
Ysize = max y(i) + + dminf (i) (6)
2
LC = U L ∗ Xsize ∗ Ysize (7)
Interconection cost (IC):
m X
X m
IC = U IC(i, j) ∗ d(i, j) (8)
i=1 j=i
i, j : Facilities index; i = 1, . . . , m and j = 1, . . . , m
m : Number of facilities.
d(i, j): Euclidean distance between centers of facilities i and j.
dmin(i, j): Minimal distance between centers of facilities i and j.
dminf (i): Minimal distance from the Property Boundary to facility i.
s 2 2
Lx(i) Lx(j)
dmin(i, j) = +
2 2
s
Ly(i)
2
Ly(j)
2 (9)
+ +
2 2
+ mindisb(i, j)
mindisb(i, j): Minimal distance between the borders of facilities i and j.
Table 1: Data for study case [4]
Facility (i) type of facility Lx-Ly [m:m] dminf (i) [m]
control room
1 10-10 30
(nonpressurized)
administrative
2 20-15 8
building
3 warehouse 5-10 8
high pressure
4 10-10 30
storage sphere
atmospheric
5 flammable liquid 4-4 30
storage tank 1
A Novel Approach to Address Process Plant Layout 5
Table 1: Data for study case [4]
Facility (i) type of facility Lx-Ly [m:m] dminf (i) [m]
atmospheric
6 flammable liquid 4-4 30
storage tank 2
7 cooling tower 20-10 30
8 process unit 30-40 30
Table 2. Unit interconnection cost, U IC(i, j) [4]
1 2 3 4 5 6 7 8
1 0,1 0,1 10 10 10 10 10
2 0,1 0,1 0 0 0 0 0
3 0,1 0,1 0,1 0,1 0,1 0,1 0,1
4 10 0 0,1 0,1 0,1 100 0
5 10 0 0,1 0,1 0,1 100 0
6 10 0 0,1 0,1 0,1 100 0
7 10 0 0,1 100 100 100 100
8 10 0 0,1 0 0 0 100
Table 3. Minimal distance between the borders of facilities i and j, mindisb(i, j) [4]
1 2 3 4 5 6 7 8
1 5 5 30 60 60 30 30
2 5 5 60 60 60 30 60
3 5 5 60 60 60 30 60
4 30 60 60 10 10 30 15
5 60 60 60 10 4 30 5
6 60 60 60 10 4 30 5
7 30 30 30 30 30 30 30
8 30 60 60 15 5 5 30
Tables 1 to 3 provides the study case data which was used in this work.
4 Bacterial Genetic Optimization Algorithm for Process
Plant Layout BGOA-PPL
In this proposal, a bacterial genetic optimization algorithm is introduced to ad-
dress the process plant layout problem; bacterial algorithms model food-seeking
and reproductive behavior of common bacteria such as E. Coli as an optimiza-
tion process. Meanwhile, Genetic Algorithms simulate sexual reproduction as an
6 B. Morales-Calvache et al.
optimization process. In this sense, BGOA-PPL exploits exploration advantages
of bacterial algorithm and convergence of Genetic Algorithms to define process
plant layout.
The process starts randomly generating the initial bacteria population, next
objective function evaluation, chemotaxis process, sexual reproduction, and elim-
ination and dispersion. The loop continues repeating the sequence from objective
function evaluation up to fulfill the stop criteria, as is described in the algorithm
1.
Algorithm 1 BGOA-PPL
Generation of initial bacteria
for i = 1 to Max generation do
Objective function evaluation
Chemotaxis
Sexual Reproduction
Elimination and Dispersion
end for
Where:
i population index.
j chemotaxis index.
θ: is the bacteria population.
S: is the number of individuals of θ or the size of bacteria population.
J(θi ): is the value of objective function of θi .
pied : is a random number to realize the elimination and dispersion of θi .
Ped : is the probability of elimination and dispersion.
BGOAPPL seek optimum value through the bacteria chemotaxis and share
information through sexual reproduction, and the elimination-dispersion process
has been defined to avoiding falling into premature convergence by a locally
optimal, creating new bacteria dispersed or located in other positions different
to the originals which replaced.
4.1 Chemotaxis
This process simulates the movement of an E. Coli cell through swimming and
tumbling via flagella. Chemotaxis, explore search space in two stages. Firstly
tumbling, where randomly is defined the direction of a unitary vector which
represents a trace of a source of food. The second stage is swimming, where each
bacterium goes forward in the direction the unitary vector previously defined.
Suppose θi represents i − th bacterium at j − th chemotactic, k − th reproductive
and l − th elimination-dispersal step. C(i) is the size of the step taken in the
random direction specified by the tumble. Then in computational chemotaxis,
A Novel Approach to Address Process Plant Layout 7
the movement of the bacterium may be represented by:
∆(i)
θi (j + 1) = θi (j) + C(i) p (10)
(∆T (i)∆(i))
Where ∆(i) indicates a vector in the random direction whose elements lie in
[−1, 1].
4.2 Sexual Reproduction
Sexual reproduction includes parent selection and cross-process, firstly define an
elite or the best-fitted individuals of the population (the user establishes the
group size). Next, a selection process is carried out, including the elite as part
of parents and other individuals who are selected by the binary tournament. In
which, two individuals are randomly selected, and their objective function values
are compared, the tournament is won by the individual with a better amount of
the objective function.
Secondly, the group of parents is crossed to generate children or new indi-
viduals, which take features of their parents. The cross-process is carried out
in two ways; the first option is by exchanging parents segments with a simple
cross point to generate children. The second is making the new individual by a
weighted average of the parents.
4.3 Elimination and Dispersion
In the evolutionary process, elimination events can occur such that the bacteria
die by adverse conditions. In this context, bacteria to avoid their extinction,
randomly disperse into a new environment using some influence, like an organ-
ism, which transports bacteria to other locations or bodies where exist better
conditions to survive. This process has the effect of assisting in chemotaxis, since
dispersal may place bacteria near of suitable food sources and avoid in this way
local optimums. From the evolutionary point of view, elimination and dispersal
are used to guarantees the diversity of population and to strengthen the ability
of global optimization.
In BGOA-PPL, bacteria are eliminated with a probability of less than P ed
(probability of elimination and dispersion). To keep a constant bacteria popula-
tion, if a bacterium is removed, this is replaced by others in a random location
on the optimization domain.
4.4 Constraints handling
The constraint handling process was made by mean of two strategies. Firstly
penalization was applied to unfeasible solutions, in parallel, the second strategy
was implemented, in it if the equation two is not satisfied, the two facilities are
separated up to satisfying equation 2. The combination of the two strategies
provides feasible solutions and reduce the time to convergence.
8 B. Morales-Calvache et al.
5 Results
The performance evaluation was carried out running the proposed algorithm
in the solution of the study case [4] and solving the mathematical model in
GAMS using DICOPT solver, the best solutions achieved with BGOA-PPL and
GAMS are shown in Table 4, in which it is possible to appreciate how BGOA-
PPL achieve a better solution in terms of objective function. Although, the area
in greater in BGOA-PPL, a less interconnection cost produces a reduced in the
total cost in contrast to the GAMS solution. The coordinates of the best solutions
with the two methods are summarized in Table 5, meanwhile in Fig.1, a plot of
BGOA-PPL solution as a schematic layout.
Table 4. Results of GAMS and BGOA-PPL
GAMS BGOA-PPL
Area [m] 19147,5 20128
IC [U$] 34648,1 25685,05
LC [U$] 95737,8 100640
Total Cost Z [U$] 130386 126325,05
Fig. 1. Layout of the best solution of BGOAPPL.
A Novel Approach to Address Process Plant Layout 9
Table 5. Best solutions with GAMS and BGOA-PPL
GAMS BGOA-PPL
X Y X Y
35,48 40,05 37,58 48,05
33,45 15,50 19,96 15,50
10,50 13,00 54,32 13,00
35,00 95,06 90,85 120,52
41,62 113,82 79,20 104,34
49,68 108,49 88,60 100,59
83,47 35,00 85,06 56,71
81,67 101,16 45,00 109,82
In terms of the proposed algorithm performance, Fig.2 exposes in box-plots
graphs the distribution of the BGOA-PPL solutions in values of the objective
function and the time spent to achieve each one of them. As is possible, appreci-
ate Fig.2 a) shows an algorithm that could provide a set of feasible sub-optimal
solutions that will be used as a seed to achieve the optimal solution in subsequent
processes. Besides, the time spent (Fig.2 b)) in achieving this set of solutions
is low in the scale of the process plant design project which is a significant ad-
vantage to develop a process with a growing improving up to the final version.
Fig. 2. Boxplots of objective function (a) and time (b) spent in the solution of study
case.
10 B. Morales-Calvache et al.
Fig. 3. Example of BGOA-PPL run
Finally, Fig.3 exposes a decreasing graph in which the combination of the dif-
ferent operators implemented in BGOA-PPL works to find a better solution in
each iteration up to complete the maximum number of iterations or generations
in a reduced time, near to 18 seconds.
6 Discussion
The results presented above suggest that the proposed method is a viable option
for the solution of design problems such as the study case, in addition to taking
into account all variables such as minimum distances and regulatory standards
for its implementation. The proposed method allows seeing multiple options
concerning the reduced time spent and the number of iterations, in comparison
with others that can only show a viable option and take a little longer to stabilize
the response. Finally, the results are comparable with other works, in which other
meta-heuristics or non-linear methods are applied to solve such problems.
6.1 Conclusions
The proposed hybrid algorithm based on Bacterial and Genetic algorithms to
address the process plant layout, shows promising results to face this mixed-
integer nonlinearly constrained optimization problem and other kinds of issues
in chemical engineering.
A Novel Approach to Address Process Plant Layout 11
The lowest time spent running BGOA-PPL is an advantage, due to allows us
to generate a significant number of feasible solutions that provide different alter-
natives that expand the horizon of the designers’ team. Therefore, it is possible
to achieve more robust designs, reducing cost and increasing the performance of
the plant operation.
References
1. Castell, C., Lakshmanan, R., Skilling, J., Bañares-Alcántara, R.: Optimisation of
process plant layout using genetic algorithms. Computers & Chemical Engineer-
ing 22, S993–S996 (mar 1998). https://doi.org/10.1016/S0098-1354(98)00198-7,
https://www.sciencedirect.com/science/article/pii/S0098135498001987
2. Jayakumar, S., Reklaitis, G.: Chemical Plant Layout via Gaph Partioning - I.
Single Level. Computers and Chemical Engineering 18(5), 441–458 (1994)
3. Jung, S.: Facility siting and plant layout optimization for chemical pro-
cess safety. Korean Journal of Chemical Engineering 33(1), 1–7 (2016).
https://doi.org/10.1007/s11814-015-0242-4
4. Jung, S., Ng, D., Diaz-Ovalle, C., Vazquez-Roman, R., Mannan, M.S.: New ap-
proach to optimizing the facility siting and layout for fire and explosion sce-
narios. Industrial and Engineering Chemistry Research 50(7), 3928–3937 (2011).
https://doi.org/10.1021/ie101367g
5. de Lira-Flores, J., Vázquez-Román, R., López-Molina, A., Mannan, M.S.:
A MINLP approach for layout designs based on the domino hazard in-
dex. Journal of Loss Prevention in the Process Industries 30, 219–227 (jul
2014). https://doi.org/10.1016/J.JLP.2013.07.007, https://www.sciencedirect.
com/science/article/abs/pii/S0950423013001381
6. Osman, I.H., Kelly, J.P.: Meta-Heuristics: An Overview. In: Meta-Heuristics, pp.
1–21. Springer US, Boston, MA (1996), http://link.springer.com/10.1007/
978-1-4613-1361-8{_}1
7. Papageorgiou, L.G., Rotstein, G.E.: Continuous-domain mathematical models for
optimal process plant layout. Industrial and Engineering Chemistry Research
37(9), 3631–3639 (1998). https://doi.org/10.1021/ie980146v
8. Patsiatzis, D.I., Papageorgiou, L.G.: Optimal multi-floor process plant
layout. Computers and Chemical Engineering 26(4-5), 575–583 (2002).
https://doi.org/10.1016/S0098-1354(01)00781-5
9. Penteado, F.D., Ciric, A.R.: An MINLP Approach for Safe Process Plant Lay-
out. Industrial and Engineering Chemistry Research 35(4), 1354–1361 (1996).
https://doi.org/10.1021/ie9502547
10. Suzuki, A., Fuchino, T., Muraki, M., Hayakawa, T.: An evolutionary
method of arranging the plot plan for process plant layout (1991).
https://doi.org/10.1252/jcej.24.226
11. Vázquez-Román, R., Dı́az-Ovalle, C.O., Jung, S., Castillo-Borja, F.:
A reformulated nonlinear model to solve the facility layout prob-
lem. Chemical Engineering Communications 206(4), 476–487 (2019).
https://doi.org/10.1080/00986445.2018.1499095, https://doi.org/10.1080/
00986445.2018.1499095