Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Site Selection of New Facility Using Gravity Model and Mixed

Integer Linear Programming In Delivery and Logistic Company


Nafisha Herma Hanifha † Ari Yanuar Ridwan Prafajar Suksessanno
School of Industrial and System School of Industrial and System Muttaqin
Engineering Engineering School of Industrial and System
Telkom University Telkom University Engineering
Bandung, Indonesia Bandung, Indonesia Telkom University
nafisha.h3rma@gmail.com ari.yanuar.ridwan@gmail.com Bandung, Indonesia
prafajar37@gmail.com

ABSTRACT Bandung Raya, including Bandung City, Bandung Regency, West


Bandung Regency, and Cimahi City. Common problem occurs in
Delivery and logistic company is one of the businesses that include the company is the number of unfulfilled demand, means that the
in supply chain management business. There is one of companies goods that are supposed to shipped to customers are delayed. If
located in Indonesia that run in that field of business. One of cities goods are delayed, there are loss cost in the logistic cost.
that has the highest demand in West Java is Bandung with 50% of Meanwhile, logistic cost is one of the important roles in a company
total demand at the province. Bandung itself already has 13 [2]. These are the percentage of unfulfilled demand and the loss
delivery center in Bandung Raya, including Bandung City, cost of each delivery center:
Bandung Regency, West Bandung Regency, and Cimahi City.
However, there are still unfulfilled demands that cause loss cost for
Table 1 Unfulfilled Demand and Loss Cost Percentage
the company because the distance between delivery center to
Delivery Center Unfulfilled Loss Cost in %
customer is too far. Therefore, the company is willing to add one
Demand in %
more delivery center in Bandung Raya but outside Bandung City.
Asia Afrika 15 12.1
Gravity Model is used to determine the potential location of facility
Cikutra 4.6 3.7
in Soreang, Dayeuhkolot, and Ujung Berung. After the potential
Cipedes 6.5 5.2
locations are determined, Mixed Integer Linear Programming is
used for choose one of the potential locations to be one location that Situsaeur 15.6 12.5
has least cost between the three. Sekejati 9.4 7.5
Cimahi 6 7.2
CCS CONCEPTS Ujung Berung 8.1 9.8
Soreang 13.4 16.1
• Industrial Engineering • Supply Chain Management • Facility
Problem Dayeuh Kolot 12.3 14.9
Cikeruh 3.9 4.7
KEYWORDS Padalarang 2.5 3
Lembang 1 1.2
Facility Location, Delivery and Logistic, Gravity Model, Mixed
Majalaya 1.7 2
Integer Linear Programming
This company is willing to build a new delivery center to decrease
1 Introduction the number of unfulfilled demand. There are several reasons for
company to investing new facilities other than to fulfill demand.
In this new era, supply chain management is needed to support Several of these reasons are to increase its production capacity, to
businesses to develop rapidly [1]. Types of businesses that is extend its product range, or to enter new market [3]. Based on the
commonly found nowadays are delivery and logistic service table above, it can be seen that some of delivery centers have high
company. In Indonesia, there are a lot of company that provide percentage of unfulfilled demand and loss cost. So, the new
delivery and logistic service. There is one of companies that run in delivery center must be in the near area of the problematic delivery
that field of business in Indonesia. center in order to help that delivery center to fulfill the demand.
However, delivery centers that located in Bandung City are
excluded because the company prefer to build a new facility outside
Actually, in West Java, city that has the highest demand of the the Bandung City. Delivery centers that have the higher percentage,
company is Bandung, with average 570 thousand goods to send per than average, in unfulfilled demand and loss cost are Asia Afrika,
month, it is approximately 50% of total demand in West Java. Situsaeur, Ujung Berung, Soreang, and Dayeuh Kolot. However,
Bandung’s distribution center itself has 13 delivery centers in
APCORISE’20, June, 2020, Depok, West Java, Indonesia N. Herma Hanifha et al

Asia Afrika and Situsaeur are excluded because these two are will become the candidates of new facility locations. After the
located in Bandung City. potential location are determined, method that can be used is mixed
Build new facility locations are high cost and cannot be reverse, integer linear programming (MILP). The combination between
also has long time impact [4]. On the other hand, determining the gravity method and MILP itself had been done before by Brittany
location of facility has a positive impact on the distribution system Cher Collins and Hao Wang in 2019. The difference is in the
parameters including efficiency, cost, and time [5]. Alfred Weber objective of the gravity method.
was the first one that started the location theory, he considered how
to locate a single warehouse in objective to minimize the total MILP can be modelled when there is a set of candidate facility
distance between warehouse in 1990 [6]. locations problem [10]. The objective of MILP model is to identify
the optimal combination of location that leads to the minimum total
To determine the potential location for the new facility in this case, costs [11]. The input of MILP model are customer demand and
gravity model is used. This model has been a topic of research location, facility information and cost, and transportation cost. This
instituted in the field of economics. Therefore, there is a lot of
is the formula of MILP [12]
research related to gravity models [7]. It was first applied in
Tinbergen in 1962 and continued in 1966 by Linnemann [8]. This
model is part of the supply chain management network Minimize:
development strategy used to determine the location of a facility
(e.g. warehouse, factory, etc.). The input of gravity model are[9]: 𝑍 = ∑ ∑ 𝑐𝑖 𝑥𝑖𝑗 + ∑ 𝑓𝑖 𝑦𝑗
𝑖∈𝐼 𝑗∈𝐽 𝑗∈𝐽
1. Volume shipped from delivery center Subject to:
2. Transportation cost ∑ 𝑥𝑖𝑗 ≥ 𝑑𝑖 , ∀𝑖 ∈ 𝐼
3. Coordinate of existing and new facilities 𝑖∈𝐼
𝑥𝑖𝑗 − 𝑀𝑦𝑗 ≤ 0, ∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽
Formula that used for determining the coordinate of new facilities 𝑥𝑖𝑗 ≥ 0
candidate is:
𝑦𝑗 = {0, 1}

𝛴𝑖 𝑉𝑖 𝑋𝑖 𝐶𝑖 𝛴𝑖 𝑉𝑖 𝑌𝑖 𝐶𝑖 ∑ 𝑦𝑖 = 𝑛
𝑥= ;𝑦=
𝛴𝑖 𝑉𝑖 𝐶𝑖 𝛴𝑖 𝑉𝑖 𝐶𝑖
Notation:
𝑥𝑖𝑗 = Unit volume shipped from facility j to customer node i
Notation:
𝑑𝑖 = Volume of total demand for each customer node i
𝑋𝑖 = X coordinate of facility i
𝑦𝑗 = Indicates whether facility j is used or not
𝑌𝑖 = Y coordinate of facility i
𝑓𝑗 = Fixed cost of facility
𝑉𝑖 = Volume shipped from facility i
𝐶𝑖 = Transportation rate from facility i M = An arbitrary large number to link the volume with the
𝑥 = X coordinate of new candidate facility facility
𝑦 = Y coordinate of new candidate facility n = Number of facilities need to be decided in the model

The process of calculating the distance between two locations in


2. Methodology
this model is calculated as the geometric distance between two These are steps to be taken in solving problem in this study. The
locations using the following formula [9]: systematic problem solving is divided into five stages, namely (1)
Preliminary Stage that include Observation, Identification, and
𝐷𝑛 = √(𝑥 − 𝑋𝑖 )2 + (𝑦 − 𝑌𝑖 )2 Objective; (2) Data Collection Stage; (3) Data Processing Stage;
(4) Analysis; and (5) Conclusion Stage. This diagram below
Where (𝑋𝑖 ; 𝑌𝑖 ) is the candidate coordinates for each facility in a shows the systematic of problem solving:
particular area and (𝑥;y) is the facility considered. The purpose of
this model is to get the location of the facility that minimizes the
total shipping costs that can be formulated as follows [9]:

𝑇𝐶 = ∑𝑘𝑛=1 𝑉𝑖 𝐷𝑛 𝐶𝑖

Notation:
𝑉𝑖 = Volume shipped from facility i
𝐷𝑛 = Distance between location of facility and location i, in the
nth iteration.

The result of the coordinates of the model will be located in one of


the districts in the area. All of villages that located in that district
Site Selection of New Facility Using Gravity Model and Mixed
APCORISE’20, June, 2020, Depok, West Java, Indonesia
Integer Linear Programming In Delivery and Logistic Company

Figure 3 Customer Nodes Soreang

Figure 4 Customer Nodes Ujung Berung

3.2 Villages Coordinate

The data of customer nodes is classified into villages based on the


scope of each delivery center which are Dayeuh Kolot, Soreang,
and Ujung Berung. This table below shows the sample of villages
coordinate data in Dayeuh Kolot.

Figure 1 Systematic of Problem Solving Table 2 Sample of Villages Coordinate Dayeuh Kolot
No Village X Y
3. Data Collection 1 Ancolmekar -7.092341 107.670363
2 Andir -6.995162 107.616481
3.1 Customer Nodes Classified By Villages 3 Arjasari -7.064739 107.640131
4 Baleendah -7.015652 107.63212
Figures below show the big picture of the scope of each delivery
5 Banjaran Kulon -7.053783 107.5805
centers. To make it simple, the customer nodes are classified into
villages. 6 Banjaran Wetan -7.087402 107.609264
7 Baros -7.060888 107.629329
8 Batukarut -7.045685 107.597384
9 Bojongkunci -7.018277 107.567217
10 Bojongmalaka -6.993273 107.606791
.. …. … …
53 Wargamekar -7.02061 107.672832

3.3 Demand

Other than customer nodes and villages coordinate, data of


average demand per villages also needed. This table below depict
the sample of average demand per month of each villages and the
Figure 2 Customer Nodes Dayeuh Kolot total average demand per month for delivery center that located in
Soreang.
APCORISE’20, June, 2020, Depok, West Java, Indonesia N. Herma Hanifha et al

Table 3 Sample of Average Demand of Soreang in Arjasari District with the least transportation cost is
No Village Demand approximately Rp2,642,651.
1 Bandasari 508
4.1.2 Soreang:
2 Banyusari 1015
3 Buninagara 503 Just like Dayeuh Kolot, the number of iteration that needed for
4 Cangkuang 513 determine the new potential location in Soreang is also 17.
5 Cangkuang Kulon 1523 Table 5 Soreang Iteration Result
6 Cibodas 497 Iteration X Y Transportation
7 Cigondewah Hilir 518 Cost / month
8 Cilame 523 (Rp)
9 Cilampeni 1000 0 -6.995972 107.5497 3,255,380.92
10 Ciluncat 528
.. …. … 1 -6.99329 107.5524 3,219,455.85
46 Tanjungsari 559 2 -6.991439 107.5535 3,206,823.54
Total 40612 3 -6.990261 107.5539 3,202,497.59
4 -6.989638 107.5541 3,201,352.65
4. Result and Discussion 5 -6.989342 107.5542 3,201,080.90
6 -6.9892 107.5542 3,201,006.90
4.1 Gravity Model
7 -6.989128 107.5543 3,200,982.40
Gravity model used to determine candidate locations for the new 8 -6.989088 107.5543 3,200,973.28
facility by doing some iterations. The iteration stopped once the 9 -6.989065 107.5544 3,200,969.73
transportation cost reach the minimum. These are the result of 10 -6.989051 107.5544 3,200,968.32
gravity model of each area: 11 -6.989042 107.5544 3,200,967.75
4.1.1 Dayeuh Kolot: 12 -6.989037 107.5544 3,200,967.53
13 -6.989033 107.5544 3,200,967.44
For Dayeuh Kolot, 17 iterations are needed to get the minimum 14 -6.989031 107.5544 3,200,967.40
total transportation cost. 15 -6.98903 107.5544 3,200,967.39
Table 4 Dayeuh Kolot Iteration Result 16 -6.989029 107.5544 3,200,967.38
Iteration X Y Transportation 17 -6.989029 107.5544 3,200,967.38
Cost / month
(Rp) The table above depicts the least transportation cost for the new
0 -7.040229 107.6077 2,646,939.43 potential location is Rp3,200,967.38. With that least
transportation cost, the new location is coordinated with x -
1 -7.039595 107.6061 2,643,172.51
6.989029 and y 107.5544. This coordinate is located in Katapang
2 -7.039366 107.6056 2,642,813.81 District.
3 -7.039173 107.6054 2,642,709.59
4 -7.039036 107.6053 2,642,672.89 4.1.3 Ujung Berung:
5 -7.038946 107.6053 2,642,659.38
Different from Dayeuh Kolot and Soreang, Ujung Berung only
6 -7.038888 107.6052 2,642,654.33
needs 15 iterations to determine the potential location for the new
7 -7.038851 107.6052 2,642,652.42 facility.
8 -7.038829 107.6052 2,642,651.70
Table 6 Ujung Berung Iteration Result
9 -7.038814 107.6052 2,642,651.43
Iteration X Y Transportation
10 -7.038806 107.6052 2,642,651.32
Cost / month
11 -7.0388 107.6052 2,642,651.28
(Rp)
12 -7.038797 107.6052 2,642,651.27
0 -6.89536 107.658 7,757,043.08
13 -7.038795 107.6052 2,642,651.26
14 -7.038794 107.6052 2,642,651.26 1 -6.912717 107.6773 6,368,608.70
15 -7.038793 107.6052 2,642,651.26 2 -6.914564 107.6818 6,280,004.22
16 -7.038792 107.6052 2,642,651.26 3 -6.914564 107.6845 6,250,260.13
17 -7.038792 107.6052 2,642,651.26 4 -6.914257 107.686 6,240,430.05
5 -6.913993 107.6867 6,237,733.80
From the table above, it could be seen that the potential location 6 -6.913831 107.6871 6,237,064.16
in Dayeuh Kolot area is located in coordinate x and y -7.038792 7 -6.913742 107.6873 6,236,904.42
and 107.6052 respectively. To be exact, that coordinate is located 8 -6.913696 107.6874 6,236,866.92
Site Selection of New Facility Using Gravity Model and Mixed
APCORISE’20, June, 2020, Depok, West Java, Indonesia
Integer Linear Programming In Delivery and Logistic Company

9 -6.913673 107.6874 6,236,858.16 are Arcamanik Endah, Bina Harapan, Cisaranten Endah,
10 -6.913661 107.6874 6,236,856.12 Cisaranten Kulon, and Sukamiskin.
11 -6.913655 107.6875 6,236,855.65 From vary number of villages, two villages was chosen from each
12 -6.913652 107.6875 6,236,855.53 district with the minimum cost. The least total cost is performed
by the combination of Arjasari and Rancakole villages in Arjasari
13 -6.913651 107.6875 6,236,855.51
district (Dayeuh Kolot) with Rp459,061,630. While the least total
14 -6.91365 107.6875 6,236,855.50 cost of Arcamanik (Ujung Berung) is slightly higher than Arjasari
15 -6.91365 107.6875 6,236,855.50 with only 16 million difference, approximately. The last is
Katapang District (Soreang) with the minimum cost
The table above shows that the potential location for the new Rp685,689,988.
facility is located coordinate -6.91365 in x axis and 107.6875 in y
axis. This coordinate is located in Arcamanik District with the 5. Conclusion
least transportation cost Rp6,236,855.50.
4.2 Mixed Integer Linear Programming In conclusion, to find the location for the new facility in this case,
After the potential locations for the new facility are determined, gravity model and mixed integer linear programming (MILP) are
the mixed integer linear programming is performed by comparing needed. Gravity model used for determine potential location in
the total cost of each area of potential locations. Dayeuh Kolot, Soreang, and Ujung Berung. There are three
Table 7 MILP Results districts in result of gravity model. These are Arjasari, Katapang,
District Village Open/Not Open and Arcamanik. All of the villages that located in those 3 districts
are become the candidate of the new facility location. After got
Arjasari Open
the candidate locations, MILP is done to choose one of the
Mekarjaya Not Open locations that has the minimum total cost. As a result, combination
Ancolmekar Not Open of Arjasari and Rancakole villages with a total cost of nearly 500
Mangunjaya Not Open million rupiah was chosen. This villages are located in Arjasari
Wargaluyu Not Open District in Dayeuh Kolot. Therefore, the new delivery center will
Arjasari be built in that area.
Lebakwangi Not Open
(Dayeuhkolot)
Baros Not Open
Batukarut Not Open REFERENCES
Patrolsari Not Open [1] S. Chopra and P. Meindl, Supply Chain Management
Pinggirsari Not Open Strategy, Planning, and Operation-Pearsons. 2015.
Rancakole Open [2] R. Razafuad, A. R. Yanuar, and B. Santosa, “Development
Total Cost Rp459,061,630 of E-Kanban Application Using Stock-needs Rule
Prioritizing Policy To Reduce 0-Pick For Pharmaceutical
Cilampeni Not Open
Warehousing,” 2018.
Banyusari Open
[3] A. J. Arumugham, “Solving Supply Chain Network
Katapang Open
Gravity Location Model Using LINGO Investigations on
Katapang Pangauban Not Open
Design of Supply Chain Networks for Manufacturing
(Soreang) Sangkanhurip Not Open Industries View project Python View project,” 2015.
Sekarwangi Not Open [4] L. v. Snyder, “Facility location under uncertainty: A
Sukamukti Not Open review,” IIE Transactions (Institute of Industrial
Gandasari Not Open Engineers), vol. 38, no. 7. pp. 547–564, Jul-2006.
Total Cost Rp685,689,988 [5] J. J. Mwemezi and Y. Huang, “Optimal Facility Location
Arcamanik on Spherical Surfaces: Algorithm and Application,” 2011.
Not Open
Endah [6] R. Z. Farahani, M. Abedian, and S. Sharahi, “Dynamic
Arcamanik Bina Harapan Open facility location problem,” in Contributions to
(Ujung Berung) Cisaranten Endah Not Open Management Science, Springer, 2009, pp. 347–372.
Cisaranten Kulon Not Open [7] S. Shahriar, L. Qian, S. Kea, and N. M. Abdullahi, “The
Sukamiskin Open Gravity Model of Trade: A Theoritical Perspective,” 2019.
Total Cost Rp475,272,617 [8] R. Guo, “Determinants of spatial (dis)integration,” in
China’s Spatial (Dis)integration, Elsevier, 2015, pp. 67–
From the table above, it could be seen that each district has vary 105.
number of villages. Arjasari district with 11 villages which are [9] D. O. Effendi and N. Siswanto, “Determination of
Arjasari, Mekarjaya, Ancolmekar, Mangunjaya, Wargaluyu, Provincial Level of Hazardous Waste Collection Location
Lebakwangi, Baros, Batukarut, Patrolsari, Pinggirsari, and In East Java Province Using Center of Gravity Method.”
Rancakole. Katapang with 8 villages that are Banyusari,
Cilampeni, Katapang, Pangauban, Sangkanhurip, Sekarwangi,
Sukamukti, and Gandasari. Lastly, 5 villages of Arcamanik, these
APCORISE’20, June, 2020, Depok, West Java, Indonesia N. Herma Hanifha et al

[10] A. Klose and A. Drexl, “Facility location models for


distribution system design,” in European Journal of
Operational Research, 2005, vol. 162, no. 1, pp. 4–29.
[11] B. C. Collins and H. Wang, “Facility Location
Optimization for Last-mile Delivery,” 2019.

You might also like