Download as pdf or txt
Download as pdf or txt
You are on page 1of 50

Distance based model for facility

location

Lecture Delivered by:

Prof. SP Sarmah
Introduction
• A sports goods company (AI) wants to spread there
business.

• Their philosophy is “proximity to the customer means


everything in this business.”

• Therefore Company wants to look at adding and


optimizing warehouse locations across the country to
lower costs and provide the best service of any
sporting goods chain in the U.S.
Customer Map of the Company
Visualising customers on a map is a powerful way to
start to analyse any logistics network
Customer Map
Company has provided demand data of 200 cities in the
US where their stores are located. In the Fig 1,
customers are sized by their relative demand. In Fig 2,
deeper the shading, the more demand within the state

Fig 1 Fig 2
26 Possible Locations for Warehouses
Problem definition for “P” facilities
• Given a set of customer locations and their demand, find the
best “P” no of facilities that minimizes the weighted distance
facility to the customer,
• Assuming that each facility can satisfy the full demand of the
customer and that all the demand is always satisfied.

• ‘J’ denotes the set of customers.

• ‘dj’ stands for demand.

• ‘I’ set of potential facilities. 26 in case of AI Athletics company.

• ‘dist i,j’ stands for distance from facility i to customer j.

• If Xi = 1 facility is selected as location and 0 if not selected.

• Yij = 1 the facility i would serve the customer j and 0 if would not serve.
Formulation

• Objective function

Minimize  dist
iI jJ
i, j d jYi , j

• Subject to
Y iI
i, j = 1; j  J

X
iI
i =P

Yi , j  X i ; i  I , j  J
Xi and Yij are binary numbers
Formulation
• Objective function minimizes the total weighted
distance from warehouses to customers.

• The first constraint will stipulate that every customer


must be fully served.

• The second constraint stipulates that allowable


number of facilities are exactly P.

• Third constraint ensures that an assignment of


warehouse should not be made to a customer
until a warehouse is opened in that location.
Analysis of the Model

• We would run scenarios with different values


of P and compare the answers for the problem.
Analysis
• Our goal is to determine the best P warehouses but
we are not assuming that we know the best value of P.
We would run multiple scenarios with different
values of P and compare their answers.

• We ran the model for best 1, 2, 3…….., 10


warehouses and the resultant networks (with star
burst collection pattern) are shown in subsequent
slides.
1- City Warehouse
2- City Warehouses
3- City Warehouses
4-City Warehouses
Al’s Athletics’ Best 5 and Best 6 Warehouse Solution
Maps
Al’s Athletics’ Best 7 and Best 8 Warehouse Solution
Maps
Al’s Athletics’ Best 9 and Best 10 Warehouse
Solution Maps
Distance Band Comparison Across Solutions
In the 1 warehouse solution row, 1% of demand is within 100 miles, 4% of demand is
within 100 and 400 miles, 30% of demand is within 400 and 800 miles etc.

Distance band (Miles)


No. of Warehouses 100 400 800 3200 Weighted average
distance (miles)
1 1% 4% 30% 65% 1072
2 12 23 39 26 558
3 18 38 35 9 383
4 31 38 24 7 311
5 36 38 23 3 266
6 41 36 22 1 236
7 45 37 17 1 210
8 47 38 15 0 193
9 49 38 13 0 178
10 51% 40% 9% 0% 165
Results

Number of warehouses utilized


Inference
• Average Distance goes down as the number of
warehouses increases.

• Distance is a quick measure of your ability to quickly


deliver the materials. Hence response time to
customers being important, in SCM this model could
serve that purpose.

• You get an insight into how much value you gain by


using an additional facility and the resultant financial
implications thereafter.
Inference
• Distance and transportation cost are interrelated and this
model is a good approximation for minimizing the
transportation cost.

• Although we are keen to find out a single optimum


solution of a problem, but real world is more complicated
than what we visualize.

• We need to run multiple scenarios to understand trade offs


and marginal value of adding facilities. This information
could be coupled with strategic business objectives to
make a final decision.

• This model is the building block for more complicated


models apart from providing lot of values.
9 City warehouse location Problem
Demand for a 9-city problem
Set of Cities Demand j Xi (use = 1)
Chicago 28,70,000 1
Atlanta 5,72,000 0
New York 84,50,000 1
St. Louis 3,50,000 0
Detroit 9,01,000 1
Cincinnati 3,33,000 0
Pittsburgh 3,06,000 0
Charlotte 7,23,000 1
Boston 6,10,000 0
Total Facilites 4
Total Allowed 4
Distance Matrix
Distance New Cincin Pittsburg
Chicago Atlanta St. Louis Detroit Charlotte Boston
Matrix York nati h
Chicago 0 720 790 297 283 296 461 769 996

Atlanta 720 0 884 555 722 461 685 245 1099

New York 790 884 0 976 614 667 371 645 219

St. Louis 297 555 976 0 531 359 602 715 1217

Detroit 283 722 614 531 0 263 286 629 721

Cincinnati 296 461 667 359 263 0 288 479 907

Pittsburgh 461 685 371 602 286 288 0 448 589

Charlotte 769 245 645 715 629 479 448 0 867

Boston 996 1099 219 1217 721 907 589 867 0


• The problem is solved in CPLEX software and mapped in ArcGIS
for various scenarios which are explained in following slides.

• Scenario 1: When P=4

Selected cities for warehouse


Average demand weighted distance= 37
When the warehouse number is allowed to 2
Average demand weighted distance= 105
When the number of warehouse is 3
Average demand weighted distance = 56
Decrease in average distance = (105-56)=49
When demand for each of the cities to equal 100,000.
Average demand weighted distance = 186
Decrease in average distance = (186-56) = 137
• When every city is equal to 100,000 except New York and Chicago,
which are 200,000 each.
Distance based warehouse location Problem with
Indian Scenario solved through CPLEX and
ARC GIS
Demand of cities

Set of Cities Demand-j


Bangalore 165
Chennai 135
Delhi 280
Mumbai 200
Lucknow 125
Kolkata 155
Distance Matrix (Kms)
City Bangalo Chenna Del Mumb Luckn Kolkat
Warehouse re i hi ai ow a

Ahmedaba 1301 1444 881 458 1049 1766


d
Chandigarh 2202 2220 317 1550 724 1662

Ranchi 1706 1529 131 1753 847 198


3
Nagpur 962 958 979 695 740 1064
Vijayawada 544 415 154 910 1226 1118
1
CPLEX code
Input data
• The problem is solved in CPLEX software and mapped in ArcGIS
for various scenarios which are explained in following slides.
• Scenario 1: When P=4

Selected cities for warehouse


When warehouse number allowed is 2

Selected cities for warehouse


When warehouse number allowed is 3

Selected cities for warehouse


ISE | IIT Kharagpur.
9/2/2022 ISE | IIT Kharagpur.

You might also like