CHEMICAL ENGINEERING TRANSACTIONS  
 

VOL. 52, 2016 

A publication of 

 

The Italian Association 
of Chemical Engineering 
Online at www.aidic.it/cet 

Guest Editors: Petar Sabev Varbanov, Peng-Yen Liew, Jun-Yow Yong, Jiří Jaromír Klemeš, Hon Loong Lam 
Copyright © 2016, AIDIC Servizi S.r.l., 

ISBN 978-88-95608-42-6; ISSN 2283-9216 
 

A Genetic Algorithm Based Plant Layout Design Methodology 

Considering Piping and Safety 

Yan Wu, Yufei Wang*, Xiao Feng, Shuo Feng 

State Key Laboratory of Heavy Oil Processing, China University of Petroleum(Beijing) 

18 Fuxue Road, Changping, Beijing 102249, China. 

wangyufei@cup.edu.cn  

Plant layout design is the key way to help enterprise improving production efficiency, operation safety and 

energy saving. At the level of facilities in a plant, the constraints of layout design have been gradually 

improved. However, the main constraints will change when the problem of design extends to the level of 

plants in a factory or an industrial park. Because the long distance between the plants cannot be neglected. 

Pipeline network and safety issues for the whole factory are deemed to be the more remarkable constraints. In 

this paper, the pipeline network is specified into two types: material pipeline and steam pipeline. The 

investments for both types are evaluated through temperature, pressure and the property of material 

individually. For the length of steam pipelines, the minimum length of steam pipelines is optimized by Kruskal 

algorithm. Additionally, the likelihood of safety accidents, such as explosion, is taken into account to make the 

plant layout design more practical. A genetic algorithm is used to solve the problem. The objective function is 

to minimize the total pipeline cost and the economic property damage which is caused by safety accidents 

simultaneously. This paper presents a mathematical model for the plant layout design and a case study 

reveals the effectiveness of proposed methodology. 

1. Introduction 

The designing of plant layout is a complex and multidisciplinary work. Koopmans and Beckman (1957) firstly 

presented this problem for increasing economic benefits. From then on, academics explored a lot of methods 

to solve plant layout. In the early years, the problem is solved by the experiences of designers and related 

national standards, such as Kern (1978). At present, the method of mathematical programming is widely used 

to optimize plant layout and it will be the trend of development. 

The plant layout problem is formulated as a quadratic assignment problem (QAP) by Koopmans and Beckman 

(1957). Then Montreuil (1991) proposed mixed integer programming (MIP) model includes connection of 

material flows. Patsiatzis and Papageorgiou (2002) optimized the process plant layout includes floor number, 

plant size and relative location by Mix integer linear programming (MILP) model. Besides, mix integer 

nonlinear programming (MINLP) is used to solve the layout design consider safety issue (de Lira-Flores et al., 

2014). 

There are many factors impacting the layout, such as the connection cost of material flows, relative location, 

shape of the plant, floors number, no-overlapping and safety problems. Caputo et al. (2015) optimized plant 

layout whose objective function consists of material transfer cost, land cost and safety cost using genetic 

algorithm (GA). The material transfer cost includes piping and pumping cost. Martinez-Gomez et al. (2014) 

proposed a general optimization model to determine the optimal plant layout with the simultaneous 

consideration of economic and safety issues. The safety objective is represented by the sum of plants caused 

by boiling liquid expanding vapor explosions (BLEVE) and vapor cloud explosion (VCE). And the 

overpressures for these explosions is calculated by the TNT equivalent model. Jung et al. (2011) determined 

safe locations of facilities by minimizing the overall cost including fire and explosion scenarios. Additionally, a 

flame acceleration simulator (FLACS) is used to provide substantial guidance for deciding the optimal layout. 

Safety constraint is also included in the work by XU et al. (2013), it consists of minimum distance and toxic 

                                

 
 

 

 
   

                                                  
DOI: 10.3303/CET1652005 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Please cite this article as: Wu Y., Wang Y., Feng X., Feng S., 2016, A genetic algorithm based plant layout design methodology considering 
piping and safety, Chemical Engineering Transactions, 52, 25-30  DOI:10.3303/CET1652005   

25



gas dispersion. The optimal targets are land cost and pipe length.  

For the present papers, the connection cost is just a rough estimation, and consists of material flow piping cost 

only, ignoring the cost of steam piping. Besides, the safety issues such as fire, explosion and toxic gas 

dispersion are mainly affected by safe distance. Few paper mainly considered the security influence of relative 

locations. In this paper, objective function includes the cost of connection and safety. The connection cost is 

enriched, which includes the cost of material flow piping and steam piping. And the cost is calculated by 

proposed model rather than estimation. Furthermore, VCE is considered in the model. It focuses on the 

influence of relative locations on safety cost. GA is used to solve the problem. 

2. Methodology 

2.1 Problem statement 
This method is used to optimize the plant layout in a factory or industrial park and reduce the investment cost. 

The number of plants, the connections between material flows, the steam usage of the plants and the costs of 

different types of piping are given. Some assumptions are made in order to simplify the problem. Each plant is 

considered as a point which has no size and shape. But it will be described as rectangle in the figures in order 

to express clearly. Besides, the distance between two adjacent plants are the same and are meet the 

minimum safety distance. And the distance between plant and pipe rack is neglected. It should be noticed that 

the piping can be laid horizontally or vertically only. 

2.2 Mathematical model 
The objective function is computed as Eq(1). 

min +
n m k

S MS M
C Ra aL L      

  

 (1) 

Where C is the total cost of piping of the factory, n is the number of steam levels, m is the number of material 

flow, while aαS and aβM are the unit prices of different steam piping and material piping including the costs of 

materials, LαS is the length of each level of steam piping which is calculated by Kruskal algorithm, and LβM is 

the length of each process material piping. Rγ is the safety cost of each plant which is composed of the loss 

caused by VCE. Specially, LβM is computed as Eq(2). 

M I O I O
L = x - x + y - y
    

 (2) 

Where ((xβI, yβI) are the coordinates of the plants which produce process materials and (xβO, yβO) are the 

coordinates of the process materials users. 

In order to obtain aαS and aβM, the diameters should be calculated and the materials of different piping should 

be determined according to the chemical properties of material flows and steam. Diameters of piping is 

calculated as Eq(3). 

4


W
D

 v
 (3) 

Where D is the diameter of pipe (m), W is the mass flow rate (kg/s), ρ is the density (kg/m3) and ν is velocity of 

the flow (m/s). In the case study, v is set to 1.5 m/s for liquid and 20 m/s for gas (for material flows). The 
annual on-stream time is set to 8,400 h. Then according to the diameters and materials, the unit prices can be 

found out in the Application Data Bases of Petrochemical Installation Engineering Budget Estimate Making. 

Besides, Rγ is computed as Eq(4). 

( )R E N F T   
    (4) 

Where Eγ indicates the property damage of incident, Nγ indicates the loss of casualties, F is the frequency of 

incident, which is 7.8 ×10-6, T is plant life time, which is 50 y. Eγ is determined based on the radius of property 

damage (RE). And Nγ is determined based on the radiuses of die radius (R0.5), serious injuries radius (Rd0.5) 

and minor injuries radius (Rd0.01). The sum of the plant value in different radius is RE and it is same as Nγ. The 

radiuses are computed by TNT equivalent model (2014). 

1.8
f c

TNT

TNT

W H
W

Q



 (5) 

 
TNT TNT

E Q W  (6) 

26



Where λ is the equivalent coefficient of vapor cloud of flammable gas, Hc is combustion heat of flammable 

gas, Wf is the fuel quality that leaks into the air, QTNT is the combustion heat of TNT, E is the explosion energy, 

WTNT is the TNT equivalent of combustible gas. R0.5, Rd0.5, Rd0.01 and RE are computed by Eqs(7-9). 

0.37

0.5
13.6

1000

 
  

 

TNT
W

R  (7) 

3 2 1

0

0.5 0.01

1/ 3

0

0.137 Z 0.119 0.269 0.019

(or )

  
   



 


 
  
  

S
P

Z Z
P

Rd Rd
Z

E

P

 (8) 

1/ 3

1/ 6
2

3175
1


  
   
   

TNT
E

TNT

KW
R

W

 

(9) 

Where P0 is the ambient pressure, 101,300 Pa, Z is dimensionless distance and △Ps is overpressure, 44,000 

Pa for Rd0.5 and 17,000 Pa for Rd0.01, K is a constant which is 9.6. 

2.3 Optimization algorithm 
GA is used through MATLAB in this paper to obtain the optimal scheme of plant layout. It mimics the process 

of biological evolution. GA starts from an initial population and repeats the operations of selection, crossover 

and mutation until the optimal result is obtained.  

For the calculation of steam piping, Kruskal algorithm is used to obtain the shortest length. It is a kind of 

minimal spanning tree algorithm which calculates the connected graph with weight. The weight in this paper is 

set to be 1. And the steam usage is described by binary variables. In order to optimize the shortest path of 

steam piping, all the plants which need steam and unpredictably other plants need to be combined. The length 

and arrangement of piping can be obtained through the calculation and verification of Kruskal algorithm.  

3. Case study 

The case study has three scenarios based on the different optimization target in order to illustrate the 

proposed method. They are the cost of material piping, the total piping cost and the sum of piping cost and 

safety cost. There are nine plants in the factory. The steam is distributed at three levels: high pressure steam, 

medium pressure steam and low pressure steam. 

3.1 Data acquisition 
The design data of material flows piping and steam piping are shown in Table 1 and Table 2. 

Table 1: The design data of material flows piping 

Number Name of material flow Transport direction Mass flow (104 t/y) Density (kg/m3) 

1 material 1 from 7 to 2 1,200 900 

2 material 2 from 7 to 2 140 800 

3 material 3 from 3 to 7 50 720 

4 material 4 from 2 to 3 280 820 

5 material 5 from 4 to 6 400 910 

6 material 6 from 2 to 3 260 760 

7 material 7 from 3 to 5 260 760 

8 material 8 from 3 to 5 60 800 

9 material 9 from 3 to 4 100 920 

10 material 10 from 2 to 4 250 1,000 

11 gas 1 from 7 to 3 10 580 

12 gas 1 from 7 to 4 15 580 

13 gas 1 from 7 to 6 15 580 

14 gas 2 from 5 to 6 10 1.4 

15 gas 2 from 2 to 5 8 1.4 

27



Table 2: The design data of steam piping 

Pressure (MPa) Temperature (℃) Velocity (m/s) Mass flow (104 t/y) Density (kg/m3) 

3.5 450 55 504 10.8842 

1.1 350 40 252 3.903 

0.4 210 30 168 1.678 

 

According the data above, the results for unite prices are obtained, as shown in Table 3. 

Table 3: The results about the calculation and selection of piping 

Number Diameter (m) Size Type Unit price (￥/10 m) 

1 0.612 Sch40 DN600 seamless carbon steel 1,657.46 

2 0.222 Sch40 DN250 seamless carbon steel 879.09 

3 0.140 Sch40 DN150 seamless carbon steel 770.45 

4 0.310 Sch40 DN300 seamless carbon steel 907.28 

5 0.351 Sch40 DN400 seamless carbon steel 1,207.07 

6 0.310 Sch40 DN350 seamless carbon steel 989.34 

7 0.310 Sch40 DN350 seamless carbon steel 989.34 

8 0.145 Sch40 DN150 alloy steel 2,096.89 

9 0.175 Sch40 DN200 alloy steel 2,373.40 

10 0.243 Sch40 DN250 stainless steel 5,884.47 

11 0.070 Sch40 DN80 seamless carbon steel 584.47 

12 0.085 Sch40 DN100 seamless carbon steel 603.35 

13 0.085 Sch40 DN100 seamless carbon steel 603.35 

14 0.388 Sch40 DN400 alloy steel 3,892.26 

15 0.347 Sch40 DN350 alloy steel 3,090.16 

3.5 MPa 0.596 Sch80 DN600 seamless carbon steel pipe 2,319.41 

1.1 MPa 0.825 DN 900 welded carbon steel pipe 1,252.23 

0.4 MPa 1.186 DN 1200 welded carbon steel pipe 1,413.07 

 

Additionally, the steam usage and value of plants are shown in Table 4. In the table, 1 represents that the plant 
needs steam. On the contrary, 0 represents that the plant does not need any steam. Besides, VCE may take 

place in plant 7. 

Table 4: The steam usage and value of plants 

Number 3.5 MPa 1.1 MPa 0.4 MPa Number of people Value (104￥) 

Plant 1 1 1 1 25 30,000 

Plant 2 1 1 0 10 25,000 

Plant 3 1 1 0 20 106,000 

Plant 4 0 1 1 15 12,000 

Plant 5 1 1 0 8 43,000 

Plant 6 1 0 1 8 6,000 

Plant 7 0 1 0 2 5,000 

Plant 8 0 0 1 10 2,000 

Plant 9 0 0 1 60 3,000 

 

3.2 Results and discussions 
The results show that different costs decrease gradually with the increasing of generation. When the number 

of generation is greater enough, the target values of three scenarios are no longer change which shows that 

the stable results of optimization are obtained. For three scenarios, the optimization targets decreased 

respectively than the initial values, which are about 30.44 × 104 ￥, 24.40 × 104 ￥ and 73.74 × 104 ￥. These 

illustrate that the optimization of material flow piping, total piping and total cost considering safety are 

effective. After optimization using GA, the optimal layouts of different scenarios can be obtained. 

The specific calculation results are shown in Table 5, Table 6 and Table 7. 

 

28



Table 5: The results of three scenarios. 

 Material flow piping 

cost (104￥) 
Steam piping cost 

(104￥) 
Total piping cost 

(104￥) 
Safety cost 

(104￥) 
Total cost 

(104￥) 

Scenario 1 136.56 99.60 236.16 90.88 327.04 
Scenario 2 139.84 84.68 224.52 206.81 431.33 
Scenario 3 139.84 84.68 224.52 90.74 315.26 

Table 6: The summary of optimal relative locations 

 Relative location of each plant 

Scenario 1 6 5 3, 7 2 4, 1 8 9 
Scenario 2 6 4 8, 5 2 1, 3 7 9 
Scenario 3 8 6 4, 9 5 2, 1 3 7 

Table 7: The radius of safety 

R0.5 (m) Rd0.5 (m) Rd0.01 (m) RE (m) 

123.56 276.30 496.44 700.87 

 

In Table 5, the material flow piping cost of scenario 1 is the least, which is 136.56 × 104 ￥. But the cost of 

steam piping is 99.60 × 104 ￥ resulting in the total piping cost 11.64 × 104 ￥ more than that of scenario 2. In 

scenario 2, the cost of material flow and steam piping are 139.84 × 104 ￥ and 84.68 × 104 ￥. It reveals that 

optimization of material flow piping and steam piping simultaneously is better than a single piping optimization. 

Besides, the plant layout of scenario 2 is more practical than that of scenario 1 after enrich the content of 

pipeline network. 

Safety cost is added in the objective function of scenario 3. So the safety cost of it is the most economical 

which is 90.74 × 104 ￥. And the cost of steam piping is the same as scenario 2. It is the least cost. Though 

the material flow piping cost is not the least, the total piping cost is optimal. For the total cost, the value of 

scenario 3 is 315.26 × 104 ￥, 11.78 × 104 ￥ less than scenario 1 and 116.07 × 104 ￥ less than scenario 2. 

This displays the effect of the optimization including enriched piping and safety issue. Comparing calculation 

results of three scenarios, the scheme of scenario 3 is the most practical and economical layout. The optimal 

plant layout and pipeline network is shown in Figure 3 which corresponds to the scenario 3 in Table 6 and 

Table 7. 

 

Figure 1: The optimal layout scheme and pipeline network 

29



The three levels of steam piping are the simplest, as shown in Figure 1. Because the pipelines are arranged 

without passing the redundant plants. Besides, the piping of material flows is concentrated together, 

illustrating the decreasing of the piping cost. The different areas in Figure 1 indicate the different radius of 

personnel and property losses. The scope losses is greater in the area with more shallow color. Areas with 

deep color are contained by the shallow ones. Plant 7 is located in the bottom right corner in scenario 3. This 

is one of the important factors lead to reduced cost of safety. 

4. Conclusions 

The proposed method in this paper enriches the piping cost on the basis of improved calculation. The 

enriched piping consists of material flow and steam flow results in the economic cost. Additionally, the safety 

issue of VCE is added in the objective function. This paper investigated the effect of safety cost in different 

relative locations through three scenarios in the case study. It is different from other researches whose target 

is determining the safety distance of plant layout. Because the distance between plants in a factory is longer 

than that between facilities in a plant. GA and Kruskal algorithm is applied to solve the problem of plant layout. 

Finally, the optimal scheme of plant layout and pipeline network are obtained. The scenario considering the 

enriched piping and safety issue is illustrated to be the most practical and economical. Simultaneously, the 

case study reveals the effect of the proposed methodology. 

Acknowledgements 

Financial support from the National Basic Research Program of China (973 Program: 2012CB720500), the 

National Natural Science Foundation of China under Grant No. 21306228 and Science Foundation of China 

University of Petroleum, Beijing (No.2462013YJRC010) are gratefully acknowledged. 

References 

Caputo A.C., Pelagagge P.M., Palumbo M., Salini P., 2015, Safety-based process plant layout using genetic 

algorithm, Journal of Loss Prevention in the Process Industries, 34, 139-150. 

de Lira-Flores J., Vázquez-Román R., López-Molina A., Mannan M.S., 2014, A MINLP approach for layout 

designs based on the domino hazard index. Journal of Loss Prevention in the Process Industries, 30, 219-

227. 

Jung S., Ng D., Diaz-Ovalle C., Vazquez-Roman R., Mannan M.S., 2011, New approach to optimizing the 

facility siting and layout for fire and explosion scenarios, Industrial & Engineering Chemistry Research, 

50(7), 3928-3937. 

Kern R., 1978, Plant Layout, 10. How to Arrange Plot Plan for Process Plants, Chemical Engineering, 85(11), 

191-197. 

Koopmans T.C., Beckmann M., 1957, Assignment problems and the location of economic activities, 

Econometrica: Journal of the Econometric Society, 25(1), 53-76. 

Martinez-Gomez J., Nápoles-Rivera F., Ponce-Ortega J.M., Serna-González M., El-Halwagi M.M., 2014, 

Siting Optimization of Facility and Unit Relocation with the Simultaneous Consideration of Economic and 

Safety Issues, Industrial & Engineering Chemistry Research, 53(10), 3950-3958. 

Montreuil B., 1991, A modelling framework for integrating layout design and flow network design, Material 

Handling, 90, 95-115. 

Patsiatzis D.I., Papageorgiou L.G., 2002, Optimal multi-floor process plant layout, Computers & Chemical 

Engineering, 26(4), 575-583. 

Yuan X., Zhenyu W., Qunxiong Z., 2013, An improved hybrid genetic algorithm for chemical plant layout 

optimization with novel non-overlapping and toxic gas dispersion constraints, Chinese Journal of Chemical 

Engineering, 21(4), 412-419. 

 

 

30