Transhipment Problem Formulation Applicability of the previous formulation

!!"#" … !#$%& Flow quantity Symbol

S1 D1 2 Set of vertices GOOD NEWS
definitions
3 Set of arcs in the network • We don’t need to change our vertex/arc
minimize 2!!"#" + !!"#$ + !!$#" S2 W1 D2 !') Flow on arc (-, /)
+ 2!!$#$ + !!&#" + 2!!&#$ Objective
definitions.
+ !#"%" + 2!#"%$ + !#"%& function S3 +') Cost of transport on arc (-, /)
D3
+ 2!#$%" + !#$%$ + 2!#$%& 1' Cargo production on vertex i
BAD NEWS
• Previous formulation can only handle a
subject to 5 = !!"#" + !!"#$ Enforce supply * +') !') single type of flow.
7 = !!$#" + !!$#$
minimize
limits ',) ∈+
• We are not currently able to impose a
3 = !!&#" + !!&#"
capacity limitation on our warehouses.
7 = !#"%" + !#$%" * !') − * !)' = 1' ∀ - ∈ 2
subject to
3 = !#"%$ + !#$%$
Enforce ): ',) ∈+ ): ),' ∈+
-7
5 = !#"%& + !#$%& demand limits 5

D1 0 ≤ !') ∀-, / ∈ 3
S1 2 1
A E
!!"#" + !!$#" + !!%#" = !#"&" + !#"&$ + !#"&%
1 W1 2
Flow 7
1 1
-3 C D
!!"#$ + !!$#$ + !!%#$ = !#$&" + !#$&$ + !#$&% conservation D2
2
S2 2
1 F
W2 B
!!"#" , !!$#" , !!%#" , !#"&" , !#"&$ , !#"&% ,
-5
Non- 3 1 2

!!"#$ , !!$#$ , !!%#$ , !#$&" , !#$&$ , !#$&% ≥ 0 negativity S3 2 D3 G

STEP 1 – Introducing a set of commodities STEP 2 – Handling capacity limitations

2 Set of vertices • We now added an additional set !, 2 Set of vertices • An variable "%& has been added to
3 Set of arcs which contains the different types of 3 Set of arcs represent the capacity limit on each arc.
8 Set of commodities commodity flows in our network. 8 Set of commodities

!')
.
Flow of commodity c on arc (-, /) !')
.
Flow of commodity c on arc (-, /) • A new constraint imposes an upper limit
• The decision variables and production on the sum of flows within a link.
+') Cost of transport on arc (-, /) +') Cost of transport on arc (-, /)
parameters now contain an additional
1'. Production of commodity c on vertex - 1'. Production of commodity c on vertex -
index to represent commodities.
;') Capacity limit on arc (-, /)

* * +') !')
. • Separate flow and demand variables * * +') !')
.
minimize will be used for each commodity. minimize
.∈/ ',) ∈+ .∈/ ',) ∈+

subject to * !') − * !)'. = 1'. ∀ - ∈ 2, c ∈ P subject to * !') − * !)'. = 1'. ∀ - ∈ 2, c ∈ P

. .

): ',) ∈+ ): ),' ∈+
A E ): ',) ∈+ ): ),' ∈+
A E
C D C D
F * !')
.
≤ ;') ∀ -, j ∈ E F
B .∈/ B
0 ≤ !')
0
∀-, / ∈ 3, c ∈ P G 0 ≤ !')
0
∀-, / ∈ 3, c ∈ P G
 Defining the model Continuous Facility Location Problem - Minisum
180

DECISION What are we trying Objective: Minimise the sum of weighted distances between customer locations
160 VARIABLES to determine? and the coordinates of the facility (as determined by the problem).
180

1 Set of customers
!, ? Coordinate values of facility 160
140
What do we >' Weight of customer -
PARAMETERS know? !' , ?' Coordinate of customer - 140
120
* >' ! − !' $ + ? − ?' $
minimize 120
OBJECTIVE How do we decide '∈%
100
FUNCTION what is best?
• In this case, distances are calculated over the 100

course of the solution of the problem.

80
• The formulation above has a non-linear 80
Are there any objective and does not contain any constraints.
CONSTRAINTS limitations?
60 • Since it is non-linear, it cannot be solved by 60
100 120 140 160 180 200
100 120 140 160 180 200 PuLP.

Single-Facility Network Location Problem Multiple-Facility Network Location Problem

In its simplest form, the network facility location problem needs to pick a single depot, among a 1 Set of customers
set of candidates. A Set of candidate facility locations
1 Set of customers !') Is customer - served by facility /?
A Set of candidate facility locations ?) Is a facility open in location /?
!') Is customer - served by facility /?
@') Distance from customer - and facility /
?) Is a facility open in location /?
B Number of desired locations
@') Distance from customer - and facility /
* * !') @') Chosen facilities
minimize
* * !') @') '∈% )∈! ($ = 2)
minimize
'∈% )∈! . !'( = 1 ∀* ∈+
subject to
. !'( = 1 ∀* ∈+ Every customer ! will be served by a single facility " (∈!
subject to (∈!
. '' = 1 We request multiple
open facilities here
. '' = 1 We require one and only one facility " to be open (∈!

(∈! !'( ≤ '( ∀ * ∈ +, , ∈ -

!'( ≤ '( ∀ * ∈ +, , ∈ - Facility " is only allowed to serve a customer if it is open !'( , '( ∈ {0,1} ∀ * ∈ +, , ∈ -
!'( , '( ∈ {0,1} ∀ * ∈ +, , ∈ - #!" and $" are boolean variables
 MFNLP - Limitations Capacitated Facility Location Model – Mathematical Model
. Set of customers
/ Set of candidate facility locations 1 Set of customers * * !') C') + * D) ?)
• The revised formulation gives us an additional level of minimize (1)
"#$ Is customer ) served by facility *? A Set of candidate facility locations '∈% )∈! )∈!
control over the problem. +$ Is a facility open in location *?
• The addition of the $ parameter in the 2nd constraint !') Quantity supplied to - served by / subject to . !'( = 3' ∀* ∈+ (2)
%#$ Distance from customer ) and facility *
allows us to define the number of selected depots. ?) Is a facility open in location /? (∈!
- Number of desired locations
C') Unit shipping cost from / to - . !'( ≤ 5( '( ∀, ∈- (3)
! ! "#$ %#$
B1'2
minimize '∈&
• However, we are still assuming that all customers and #∈( $∈' Minimum number of open locations
depots are identical. ! "#$ = 1 ∀) ∈. B134 Maximum number of open locations 1*'+ ≤ . '( ≤ 1*,- (4)
subject to $∈'
D) Facility / operating cost (if open) (∈!
! +# = - !'( ≥ 0 ∀ * ∈ +, , ∈ - (5)
• In the following versions of the NLP we will be introducing $∈' E' Material demand by customer -
'( ∈ 0,1 ∀ * ∈ +, , ∈ - (6)
the following operational aspects to the model: "#$ ≤ +$ ∀ ) ∈ ., * ∈ /
8) Total supply capacity in facility /
"#$ , +$ ∈ {0,1} ∀ ) ∈ ., * ∈ /
Ø Material demands by customers • The objective function (1) seeks to minimize the total cost of facility operation and material distribution
Ø Holding capacities at each facility • !') is not boolean, but rather defines the quantity of material flowing from facility / to customer -
Ø Facility operation costs • The sum of materials transported to any customer - must be equal to their demand E' (2)
Ø Customer service levels • The sum of materials transported from any facility / shall not exceed their total capacity 8) (3)
• If a facility / is not open '( = 0 , then all flows !') for / shall be equal to 0. (also 3)
• Constraint (4) limits the upper and lower number of open facilities – will be implemented as 2 separate constraints

Further insight – Logical Constraints (1) Further insight – Logical Constraints (2)
• This is a logical constraint, linking a logical decision (shall we
! "FH ≤ $H %H ∀& ∈ ( open a facility (?) to an operational quantify (flow of goods) ! "FH ≤ $H %H ∀& ∈ (
• If the facility ( is closed, then the !& )& = 0
F∈G F∈G
• As a result, every single value +%& for ( will also be equal to 0
EXAMPLE • In this case we are disabling facility 2 ()6 = 0).
• We have two depots that serve a set of four customers. The largest (1) can serve all demand. • Since we disabled one facility, the overall cost has decreased.
• However, its transportation costs from (1) tend to be higher than its smaller counterpart. • However, we can see that the logical ∑%∈7 +%6 ≤ !6)6 is not satisfied.
• As before, )& indicates whether a facility is open.
i j x ij t ij yj i Qi ΣQ OK? i j x ij t ij yj i Qi ΣQ OK?
1 1 25 1.5 1 1 192 192 TRUE 1 1 25 1.5 1 1 192 192 TRUE
1 2 167 1 1 2 140 140 TRUE 1 2 167 1 0 2 140 140 TRUE
2 1 140 1 1 3 104 104 TRUE 2 1 140 1 1 3 104 104 TRUE
2 2 0 0.5 1 4 129 129 TRUE 2 2 0 0.5 0 4 129 129 TRUE
3 1 0 2 1 3 1 0 2 1
3 2 104 1 1 j Pj Pj x y ΣP OK? 3 2 104 1 0 j Pj Pj x y ΣP OK?
4 1 0 1 1 1 600 600 165 TRUE 4 1 0 1 1 1 600 600 165 TRUE
4 2 129 0.5 1 2 400 400 400 TRUE 4 2 129 0.5 0 2 400 0 400 FALSE
Σ cost 613 Σ cost 563
 Further insight – Logical Constraints (3) Facility numbers - Trade-off analysis
• Compromise between operational costs and our capacity to serve customers.
! "FH ≤ $H %H ∀& ∈ ( • Increase in facility numbers has both advantages and disadvantages, which
F∈G can be managed through more elaborate models.

• We perform another reallocation of flows, while seeking to satisfy the logical constraint. Disadvantages Advantages
• We can see now that all constraints are satisfied: ∑%∈7 +%6 = 0 and all +%6 = 0.
• However, the overall cost is much higher (due to increased transportation costs).
• It therefore is evident that it made sense have a second facility open. • Facility costs (rent) • Improved response times
i j x ij t ij yj i Qi ΣQ OK? • Staffing costs (labor) • Customer satisfaction
1 1 192 1.5 1 1 192 192 TRUE
1 2 0 1 0 2 140 140 TRUE • Inventory costs • Market penetration
2 1 140 1 1 3 104 104 TRUE
• Transportation costs • Sales growth
2 2 0 0.5 0 4 129 129 TRUE
3 1 104 2 1 • Likelihood of stockouts • Increased level of service
3 2 0 1 0 j Pj Pj x y ΣP OK?
4 1 129 1 1 1 600 600 565 TRUE
Σ cost 815
4 2 0 0.5 0 2 400 0 0 TRUE