a logistic management system integrating inventory...
TRANSCRIPT
A Logistic Management System Integrating Inventory Management
and Routing
Ana Luísa Custódio*
Dept. Mathematics
FCT – UNL
July 2002
Rui Carvalho Oliveira**
CESUR/Dept. Civil Engineering
IST – UTL
Outline
1. Framing
2. The Case Study (Vagelpam)
3. Daily Demand
4. The Basic Model (Phase I)
5. Results (Phase I)
6. Sporadic Demand (Phase II)
7. Safety Stocks
8. Feasibility Analysis of the Solution
9. Conclusions
1. Framing
Classical ApproachesRoutingOptimization of frequency and replenishment quantities
Integrated approach of inventory management and routing
– Inventory Routing Problem (IRP) –
Strategical problem not daily management problem
2.1. The Case Study (Vagelpam)
Distribution– made from a central depot located near Lisbon– delivery points located south of Coimbra
Nestlé’s frozen products151 products321 delivery points 3937 items
FleetCapacity (pallets)
Number of vehicles
5 16 18 1
10 4
2.2. The Case Study (Vagelpam)
Delivery
Vagelpam
Nestlé
Order’s Database
Delivery Point Order Note
Order Note
Order Note
Stock
EmergencyOrder
Replenishment
2.3. The Case Study (Vagelpam)
variable transportation costdelivery fixed costinventory holding costorder fixed cost
Distance Matrixuse of VisualRoute
Cost Structure
Times involved:
transportation time – average traveling speed of 70km/h
waiting time for delivery
delivery time – related to quantity
Journey Duration- maximum of 13 hours
2.4. The Case Study (Vagelpam)
Orders
Quantity
01020304050607080
10 30 50 70 90 400
800
4000
1000
016
000
More
Orde r quantity (boxe s ) by de live ry point
Freq
uenc
y
0%10%20%30%40%50%60%70%80%90%100%
02468
101214161820
1003005007001700270037004700820014200
More
Or de r quantity (boxe s ) by product
Fre
qu
en
cy
0%10%20%30%40%50%60%70%80%90%100%
0
500
1000
1500
2000
2500
5 15 25 35 45 55 65
Number of orders by item
Freq
uenc
y
0%
20%
40%
60%
80%
100%
0
5
10
15
20
25
30
35
5 15 25 35 45 55 65 75 85 95More
Num ber of de livery points by product
Freq
uenc
y
0%10%20%30%40%50%60%70%80%90%100%
0
50
100
150
200
250
5 15 25 35 45 55 65 75More
Number of products ordered by delivery point
Freq
uenc
y
0%10%20%30%40%50%60%70%80%90%100%
0500
100015002000250030003500
50 200
350
500
650
800
950
3000
060
000
Order quantity (boxes) by item
Freq
uenc
y
0%10%20%30%40%50%60%70%80%90%100%
Result
Excluding 70% of items (2761 items)
Representing 5% of total quantity ordered
2.5. The Case Study (Vagelpam)Pareto Analysis on Quantity
10%
20%
70%
0
50
100
150
200
250
300
350
5 15 25 35 45 55 65
Number of records/Item
Freq
uenc
y0%10%20%30%40%50%60%70%80%90%100%
25% of items with 5 or less records
Power Law for
Variance Estimation
Daily Demand
3.1. Daily Demand
Power Law for
Variance Estimation
R² adjusted = 0,947F(1,660)=11813, p<0.000
Estimate t(660) p-levelD -0,133 -4,535 0,000P 2,165 108,687 0,000
Residuals Analysis
P2 Cµσ =
( ) ( )( )ClogD with
,µlogPDσlog 2
=×+=
Scatterplot (NEW.STA 4v*662c)
y=-0,133+2,165*x+eps
LOGMED
LOG
VAR
-8
-4
0
4
8
12
16
-3 -1 1 3 5 7
Normal Probability Plot of Residuals
Residuals
Expe
cted
Nor
mal
Val
ue
-4
-3
-2
-1
0
1
2
3
4
-3 -2 -1 0 1 2 3
3.2. Daily Demand
Sum of 4 lognormals
Expected
Variable VAR1 ; distribution: Lognormal
Kolmogorov-Smirnov d = ,0849710, p = n.s.
Chi-Square: ,5165555, df = 2, p = ,7723823 (df adjusted)
Category (upper limits)
No
of o
bs
���������
���������
���������
���������
���������
���������
�������������������
����������
����������
����������
����������
�������������������
���������
���������
������������������
���������
���������
����������
����������
���������
���������
02468
101214161820222426
0 220 440 660 880 1100 1320 1540 1760 1980 2200
Lognormal Distribution
Expected
Variable VAR1 ; distribution: Normal
K-S d = ,0243330, p = n.s. Lilliefors p < ,15
Chi-Square: 22,43448, df = 20, p = ,3174884 (df adjusted)
Category (upper limits)
No
of o
bs
����������
�������
��������
��������
��������
��������
��������
��������
��������
������
������
������
��������
��������
��������
��������
��������
��������
��������
�������
������
������
������
������
��������
��������
��������
��������
�������
������
������
������
������
�������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
��������
�������
������
��������
��������
����������������� ����
0102030405060708090
100110120
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 100
102
104
106
108
110
112
114
116
118
120
122
124
126
128
130
4.1. The Basic Model (Phase I)
Multiples Power of Two Policies
T2 – optimal replenishment period
of a multiple power of two
∈ *
*2 T2,
2TT
( )( ) 1.06TfTf
*
2
≤
{ }... 3 , , 2 , 1 , 0j, 2TT jB ∈×=
4.2. The Basic Model (Phase I)
Viswanathan and Mathur’s heuristic (1997)items allocated to clusters
different clusters replenished by distinct vehicles
allocation based on minimal replenishment period− EOQ formulae− power of two policies
Vehicles capacities − cyclic; decreasing order
TSP’s resolution:− nearest neighbor− minimal cost insertion− 2-optimal procedure
4.3. The Basic Model (Phase I)begin
create a new empty cluster
for all the items not yet allocated toclusters do
compute fixed replenishment cost of theitem in an empty cluster
for all the clusters with sufficient vehiclecapacity and available journey duration
compute fixed joint replenishment cost ofthe item in that cluster
choose the minimum fixed cost
compute T2
choose the minimum T2
create new route in the cluster and allocatethe item to the route and to the cluster
Is T2 of the new route> T2 of the last route in the
cluster ?
No merge routes compute replenishment timefor the new route
verify vehicle capacityand journey duration
Are all the itemsallocated?
Yes
endYes
Is there anempty cluster?
No
create a newempty cluster
No
Yes
Is vehicle capacityor journey duration
exceeded?
No
reduce replenishmenttime of the route
Yes
5.1. Results (Phase I)
89 clusters (75 seconds)
02468
1012141618
1 3 5 7 9 11 13
More
Jo u r n e y Du r atio n (h o u r s )
Freq
uenc
y
0%10%20%30%40%50%60%70%80%90%100%
Average vehicle occupation rate = 89%
Journey duration (including traveling, unloading and waiting times)
Average = 8.21 h
Maximum = 13.1 h
010203040506070
0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 More
Vehicle Occupation Rates
Freq
uenc
y
0%
20%
40%
60%
80%
100%
5.2. Results (Phase I)
maximum of 2 routes (87% only one)average daily demand is a key factor for route constructiongeographical proximity between localities belonging to the same cluster
Vehicles
0102030405060
1 2 3 4 More
Number of Localities
Freq
uenc
y
0%20%40%60%80%100%
0
10
20
30
40
50
60
1 2 3 4 5 6 More
Number of Delivery Points
Freq
uenc
y
0%
20%
40%
60%
80%
100%
low number of visited localities− 90% visit up to two localities
low number of delivery points− 80% supply one or two delivery
points
5.3. Results (Phase I)
number of products replenished variable− 58% up to 10 products− 92% up to 30 products
Vehicles
05
10152025303540
5 10 15 20 25 30 35 40 45More
Number of Products
Freq
uenc
y
0%
20%
40%
60%
80%
100%
6.1. Sporadic Demand (Phase II)
represents only 5% of total quantity ordered
sufficient available capacity to allocate the overall quantities to be delivered
Use of several heuristics
starting with items with:– the highest daily demand
– the lowest daily demand
choosing from within the routes, that serve the corresponding locality, the one with:
– the highest available capacity
– the lowest requirement
– the highest ratio available capacity/lowest requirement
6.2. Sporadic Demand (Phase II)
Best results Highest daily demand/
Lowest requirement
Only 38% of the remainder quantity allocated
Not allowed:route expansion
order splitting among routes
Daily operations management problem
7. Safety Stocks
∆ - safety stock
RT - demand during replenishment cycle Normal or Lognormal
62% less then 10 boxes
( ) ( ) 0.10R∆QP0Stock FinalP T <<−+=<
∑=
=T
1iiT RR
0100
200300
400500
600700
10 30 50 70 90 110
130
150
170
190
MoreSafety Stocks (boxes)
Freq
uenc
y
0%
20%
40%
60%
80%
100%
Future Work:Dimensioning of the safetystocks intern to model
8. Feasibility Analysis of the Solution
Heuristic Fleet
1 vehicle with routes every 2k days ⇔ 2j-k pseudo-vehicles with routes every 2j days (j ≥ k)
Total= 89
Capacity (pallets)
Number of vehicles needed
10 528 136 125 12
Capacity (pallets)
Number of vehicles needed
10 48 16 15 1
Total= 7
9. Conclusions
Feasible solution in an acceptable computational time
Key-factors for route construction:
geographical proximity
daily demand
Vehicles with:
high occupation rates
small number of distinct routes
visiting small number of different localities and delivery points
replenishing a high number of distinct products