Int J Adv Manuf Technol (2010) 50:775–788

DOI 10.1007/s00170-010-2555-0

ORIGINAL ARTICLE

“Supermarket warehouses”: stocking policies optimization

in an assembly-to-order environment
Daria Battini & Maurizio Faccio & Alessandro Persona &
Fabio Sgarbossa

Received: 20 October 2009 / Accepted: 26 January 2010 / Published online: 6 March 2010
# Springer-Verlag London Limited 2010

Abstract In modern assembly-to-order systems, sub- Keywords Stocking policies . Assembly to order .
assemblies warehouse locations, dimensions, and supply Centralization/decentralization . Warehouse . Supermarket
are critical decisional variables. Persona et al. (Int J Prod
Econ 110(1–2):147–159, 2007) affirm the structural diver-
sity between assembly parts can largely influence the 1 Introduction
stocking policies. On the other hand, Zhang et al. (2007)
and Hsu et al. (Nav Res Logist 54(5):510–523, 2007) The function of components and sub-assemblies inventories
substantiate that the problem gets more complex when it within a production plant is to keep production and
comes to choosing the correct physical locations for assembly lines operating efficiently, and the choices of
common assembly parts. The study at hand concentrates on inventory level, supply methods, and warehouse locations
components warehouse centralization/decentralization greatly impact global production costs. In fact, as evidenced
choices in an assembly-to-order (ATO) environment, which by Persona et al. [26], most companies continuously try to
typically supplies the assembly systems with components’ reduce stock inventory levels to free-up invested capital that
kits through decentralized warehouses called “Supermarkets”. does not generate any return. Manzini et al. [21, 22]
Specifically, this paper applies an innovative step-by-step highlighted that the component management policy is
procedure to support materials management decision-making important also in order to create an assembly systems that
policies in order to define when, how, and where it is is flexible and efficient. In the international literature, in
convenient to install a supermarket warehouse, considering order to calculate the optimal degree of centralization/
the typical aspect of an ATO environment: number, type, decentralization, several models have been developed;
position of the assembly systems, demand rate and common- some focus on whole supply chain network and consider
ality degree of components used, internal transportation different variables such as components demand (regular,
equipment and transportation costs, load unit capacity, space lumpy, intermittent, predictable or not, etc.), type of
cost for stocking, space availability in the plant, inventory component (raw materials of spare parts, etc.), specific
costs, and safety stock dimension. costs (downtime costs, inventory costs, transportation
costs), geographical position, etc. In reality, the decision
D. Battini : A. Persona : F. Sgarbossa to keep the number of storage areas low enables to supply a
Department of Management and Engineering, large number of production stations from one single
University of Padova, warehouse with a consequent reduction of average compo-
Stradella San Nicola, 3,
nents stock, and increase of free space near the production/
36100 Vicenza, Italy
assembly lines. On the other hand, such system requires
M. Faccio (*) higher procurement time, higher quantity of workstations
Department of Innovation in Mechanics and Management, refilling, and higher transport frequency. Moreover, for an
University of Padova,
assembly-to-order system, the inventory centralization/
Via Venezia 1,
35131 Padova, Italy decentralization problem is for some aspects more complex
e-mail: maurizio.faccio@unipd.it than for a supply network. In fact, it is related not only to
776 Int J Adv Manuf Technol (2010) 50:775–788

Fig. 1 Central warehouse

Assembly line

Central Assembly line

Internal Work Station- Warehouse
Supplier
Assembly line

Assembly line

the load unit inventory costs, or to the load unit where to supply the assembly line), it is important to
transportation cost, and to the position of warehouses and optimize the kit’s container in function of the different
assembly systems, but also to unit space availability for components and the different assembly systems to supply.
warehouses and supermarkets and ability to share compo- The first and the second image of Fig. 4 shows the kit’s
nents between products, in respect to the structural diversity container developed for this supermarket warehouse; the
of the products assembled. In different assembly systems third image shows a part of one of the three semi-automatic
(assembly line or single station assembly system), products assembly line supplied by the supermarket.
can be made of the same components, which are stored and The study at hand concerns components warehouse
supplied in different ways (Figs. 1, 2, and 3). centralization/decentralization choices and, in the case of
Figure 1 represents a system with assembly lines supermarkets solution, what kind of components to store
supplied from a central warehouse, Fig. 2 represents and how to create the assembly kits. The theoretical model
assembly lines supplied from a decentralized warehouse proposed involves multiple assembly systems, with known
normally placed close to the lines to serve, while Fig. 3 locations and consumption rates and variability of one
represents assembly lines supplied from an assembly station given assembly part. Generally, companies try to adopt the
warehouse, which is a typical solution obtained when an same management policy for the same component category
external supplier or an internal workstation supplies directly (for example, nuts and bolts) more than for each single
the assembly station without intermediate warehouses. component (for example, screw). For this reason, this paper
Figure 2 describes the use of a decentralized warehouse, analyzes the centralization/decentralization problem in an
created close to the assembly systems that typically refills ATO environment, using the concept of components
the stations with kits, i.e., with different materials stored in category defined as the grouping of different components
an apposite container that supply different assembly with the same use, application, function, and cost. The
stations for a certain time period. Obviously, as endorsed possibility to centralize/decentralize a certain components
by [3, 4] in a mixed model assembly line system, the kind category depends on the global stocking and supplying
of kit can be variable in function of the given number of policy costs in function of the commonality degree between
products assembled, as well as the balancing and the products assembled, i.e., by the utilization of the same
sequencing of the line. In fact, it is normally called components of the category.
“Supermarkets”, like a food supermarket in which ware- The aim of this paper is to determine the optimal
house operators, as well as customers, go through the stocking policy for a certain components category, to study
warehouse taking all the parts needed for a certain period of the possibility to manage stocking with kits, and tries to
time. Figure 4 shows a supermarket warehouse, for the minimize global costs. In addition, the study introduces an
assembly of gravimetric dispensers, which serves three innovative parameter, called the “components category
different semi-automatic assembly lines. In case of super- diversity index”, that integrates the possibility to group
market solution in an assembly-to-order (ATO) environ- together supplies destined to different assembly lines from a
ment, once the design of the warehouse is completed unique location, defining the kit creation for the supermar-
(including position, number and kind of components stored, ket solution and creating an optimal degree of warehouse
how to create the kit in the supermarket, how and from centralization/decentralization, taking into account products

Fig. 2 Supermarket warehouse

Super
Market Assembly line

Assembly line
Internal Work Station-
Supplier
Assembly line

Super
Assembly line
Market
Int J Adv Manuf Technol (2010) 50:775–788 777

Fig. 3 Direct assembly station

Assembly line
warehouse
Assembly line
InternalWork Station- Assembly
Supplier station
warehouses Assembly line

Assembly line

morphological diversity, and costs related to the compo- organizational aspects of such optimization; others like
nents managements policy adopted. Materials management Drake and Lee [11] studied the SCN purchase strategy or
policies defining the best decentralization/decentralization pointed their attention to the handle facilities design issues
degree, the best warehouses location, and how to supply such as warehouses and plant installation like Yeh [17] and
each assembly line must be decided case by case since Liu and Lin [20].
the needs of each assembly line is different depending on Lim et al. [19] looked at the SCN management as a
the product assembled, the quantity of output items, and the capacity plan; Thangam and Uthayakumar [30] and Ting et
production rate. al. [31] were concerned about replenishment policies
By summarizing the results for all components category, optimization and Saygin [28] and Li and Kuo [18] on the
in case of the supermarket solution, the methodology forecasting problem.
defines how to create and refill the kits, or what component During the last decades, many contributors studied the
category can be managed with a complete centralization or centralization/decentralization problem, offering material
decentralization of the warehouse. management solution models. Eppen [12], for instance,
studied the consolidation of demand from several facilities
and proposed an expression for the expected holding and
2 Literature review penalty costs as function of the demand considering
correlation between locations. Stulman [29], in addition,
The choice of the right stocking policy and of the optimal enhanced the study, determining the minimum inventory
inventory centralization/decentralization degree is a com- size needed at each location to assure the desired customer
mon problem in material management, and it can be as service level, and introduced a hybrid assembly system in
simple as a single stocking point inside a production plant order to meet the market requirements and studying the
or as complex as a supply chain network (SCN). Indepen- right material flows synchronization. The relationship
dent of its complexity, whether we are in the presence of an between aggregate safety stock and number of stocking
extremely simple stocking procedure or a very advanced locations has been studied by Ronen [27], who assumed
SNC system, an assembly production system has some that the demand at each location is equally distributed. The
specific characteristics that have to be considered. Howev- well-known “Square Root Law” has been deeply investi-
er, most of the recent studies on stocking policies are about gated by other authors as well. Evers and Frederick [13]
the optimization of a supply chain network. Some authors analyzed the effect of centralization on safety stock from n
like De Boeck and Vandaele [9] focused mainly on the to m locations (with m<n) considering the lead time

Fig. 4 Example of supermarket warehouse in an ATO environment

778 Int J Adv Manuf Technol (2010) 50:775–788

variable and a constant level of customer service, showing of satisfied orders, average level of finished). De Souza et
that reduction of safety stock depends only on the number al. [10] analyzed the severity of packing problems in
of decentralized facilities. assembly lines feeding system, in a real automotive plant.
Focusing on the assembly systems, some researchers Using an integer to program a model, heuristic procedure,
tried to solve the problems connected with components they formulated a singular approach that allowed choosing
allocation in the ATO, like [14, 15] who studied the the right packing policies to minimize holding and handling
commonality of components in a multiple period, and costs.
where components are replenished according to a (Q; r) From the analysis of the scientific contributions, the lack
policy, or Akçay and Xu [1] who formulated a two-stage of a useful framework for industrial engineers, which is
stochastic integer program and determined the optimal crucial in choosing the correct degree of centralization or
base-stock policy and the optimal component allocation type of supply system, is clear. Such void inspired a
policy for the ATO system. completely new vision of the feeding procedure which
Mohebbi and Choobineh [24] as well as other authors sparked a new set of research. In the next section, a novel
demonstrated that using a common component for different decision-making procedure is illustrated in order to assist
products allows a considerable reduction of the safety industrial engineers in their decision-making process on the
stocks without hindering service quality. Bollapragada et al. optimal degree of inventory centralization, or the most
[5] studied the inventory management problem for two end- appropriate time to adopt a kit-type feeding system, where
products ATO system; they considered a single-period ATO parts are supplied by supermarkets, without losing sight of
system with a manufacturer and three suppliers. Chu et al. the morphological difference of materials, unit transporta-
[7] investigated a two-component assembly system facing tion, or inventory costs.
stationary stochastic demand with constant component
replenishment lead times.
Many other aspects of the assembly system have been 3 The decision-making procedure
considered for the optimization of the component stocking
policies in an ATO environment: Hsu et al. [16] analyzed The choice of the optimal “degree of decentralization” and
an optimization model to determine the optimal component location of the components category warehouse must
stocking decision for a contract manufacturer facing an minimize inventory and transport costs.
uncertain future demand, where product price depends on
CTOT ¼ CI þ CT ð1Þ
delivery times in one single shipment. Nonas [25] consid-
ered the problem of finding the optimal inventory level for Where
components in an assembly system with random demand,
CTOT global cost
where multiple final products share common components.
CI inventory cost
The author focused his attention on inventory, considering a
CT transportation cost.
two-stage stochastic recourse problem, where the first stage
is to set the inventory levels to maximize expected profit,
Generally speaking, it can be stated that inventory costs
while the second stage is to allocate components to
tend to increase when the number of warehouses increases
products to fulfill demand.
due to the rise of safety stock level, while the transport cost
Other important aspects to consider during the analysis
decreases because the stock points are close to the assembly
and design of an assembly system to allow the maximization
systems. Tables 1 and 2 summarize the necessary data for
of efficiency and flexibility are possible feeding problems.
the analysis of a specific components category and for the
The scientific contributions examine different methods
position data. Where:
to supply the right component to the correct workstation of
a given assembly line, at the right time, especially in an Pi component i, which belongs to the analyzed
ATO environment. Choi and Lee [6] used a prototype category, i=1,…, N
system in a simulation environment, which was designed Lj assembly system j, in which a certain
for an automotive assembly line, and created a dynamic product family, j=1,…, M is assembled.
component feeding system in order to estimate the part This kind of system is determined by the
consumption that takes place during the production prog- layout (straight line, SL with parallel
ress. Matta et al. [23] studied the problem of evaluating the station, SL with serial station, parallel
performance of assembly systems managed with kanbans, lines, two-side lines, U line, fixed position)
analyzing different control policies (simultaneous and as highlighted in Battini et al. [3].
independent releases), and studying the queue and the Ui,j average utilization of the component i in the
system performance (throughput of the system, percentage assembly system j for the considered period,
Int J Adv Manuf Technol (2010) 50:775–788 779

(pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day)
component’s demand for a normal

utilization deviation
Standard
distribution.

0.75

3.75
σi,j

1.5
standard deviation of component i

0
0

3
3
utilization in the assembly system j for the
considered period
P
Average

M
Ui ¼ Ui;j global utilization of the component i in the
J ¼1
L5

20
20
10
25
assembly system j for the considered period
0
0
5
σi standard deviation of component i for the
utilization deviation
Standard

considered period
4.5 Ki kanban conveyor capacity for the

4.5
0
0

0
3

0
component i
Ci cost of the component i
Average

The finite possible positions of warehouses and assem-

L4

30
10
20
30
0
0

0
bly systems known are defined by a set of (x, y)
coordinates. The possible location of the warehouse always
utilization deviation
Standard

contains the position of the assembly system that corre-

1.5

4.5

sponds to the most decentralized solution (Fig. 3), with 0

3

0
0
0
3

transportation costs.
The assumptions of the proposed model, derived by the
Average

normal industrial operating methodologies, are:

L3

10
20
30

20
0
0
0

& The transportation cost is proportional to the distance

between assembly systems and warehouses, to the
utilization deviation
Standard

specific transportation cost (€/m) function of the system

4.5

4.5

used for the transports, and to the frequency of supply

3
0

0
6
0

of the average component demand and of the kanban

container capacity, in case of an ATO environment.
Average

& The inventory cost is the function of components

L2

30
20

30

40

consumption and the chosen warehouse centralization

0

degree. The different inventory cost between different

container (€/piece) (year) utilization deviation
Standard

stocking policies is related only to the safety stock and

not to the operative stock used for the assembly, which
7.5
0
0
0
0

3
3
Table 1 Inputs data of a components category: “Engines-Transmissions”

is the same in all the inventory centralization/decentral-

ization degree.
Average

The safety stock is calculated using the formula created

L1

50
20
20
0
0
0
0

by Persona et al. [26]:

pffiffiffiffiffiffiffiffiffi
0.3
0.3
0.3
0.3
0.3
0.3
0.3

SSi;j ¼ ki;j  s i;j  LTi;j ð2Þ

t

15,000
13,000
10,000
8,000
5,000
4,000
3,000

where:
Components category: Engines-Transmissions

Kanban C

ki,j factor that indicates the number of standard

capacity
(pieces)

deviations to be kept as safety stock for

component i in the assembly system j
15
15
15
20
20
20
20

σi,j standard deviation for component i in the assembly

(pieces/day)
utilization

system j
average
Part Description Global

LTi,j supplying Lead Time for component i in the

100
40
40
65
60
90

65

assembly line j. The LT is considered the same

for each component for the short transportation
Engine A

G
B
C

E
F

time assumed.
Engine
Engine
Engine
Engine
Engine
Engine

The components are considered independent, so the total

P1
P2
P3
P4
P5
P6
P7

standard deviation for a certain component i with a

780 Int J Adv Manuf Technol (2010) 50:775–788

Table 2 Position data

centralized solution is calculated using the Das and Thyagi & The distance between two points (a certain warehouse
formula, [8]: location Wp and a certain assembly system location Lk)
vffiffiffiffiffiffiffiffiffiffiffiffiffiffi is calculated using the rectangular distances.
uX
u m 2      
si ¼ t s i;j ð3Þ d Wp ; Lk ¼ xWp  xLk  þ yWp  yLk  ð4Þ
j¼1
Where p is one of the possible finite warehouse locations
& The inventory index ti calculates the specific inventory inside the area occupied by the plant; the position of each
cost for a certain component i, taking into account the Lk assembly lines are fixed and known. A certain assembly
financial investment, the space utilization in order to line k for a certain component category is supplied from a
store the component, and its obsolescence. These costs unique warehouse.
are calculated as a percentage of the component cost Ci & A certain product or a certain product family is
for the considered period as emphasized in Azzi et al. assembled in a unique assembly system inside a
[2]. production plant, and for this reason, the commonality

Fig. 5 Flow chart of the

decision-making procedure
Int J Adv Manuf Technol (2010) 50:775–788 781

Table 3 CCk,z matrix for “Engines-Transmissions” Where Ui,j is the average utilization of the component i
L1 L2 L3 L4 L5 in the assembly system j for the considered period, Ci is the
cost of the component i, ti is the inventory index for the
L1 0.000 0.795 0.641 0.154 0.000 calculation of the inventory cost for the component i, and yi,
L2 0.856 0.000 0.360 0.640 0.640 k,z is equal to 1 if component i is used by assembly system
L3 0.922 0.468 0.000 0.610 0.532 k, but not by assembly system z, and is equal to 0 if
L4 0.633 0.667 0.300 0.000 0.000 component i is used by assembly system j and z. This index
L5 0.494 0.529 0.706 0.176 0.000 is the ratio between the costs of components belonging to
the same category used by assembly system k, which are
not in common with the assembly system z, and the total
degree between the assembled products can be assim- analyzed components category used by assembly system k.
ilated to the commonality degree of the assembly lines. For example, if two assembly lines Lk and Lz are identical,
& A components category is supplied for a certain the index is equal to 0, because all components are
assembly system from a unique warehouse (the closest), common. If the two assembly lines are completely different
even if this components category is present in different (i.e., the products assembled in these assembly lines do not
stocking points inside the production plant. have common components), the index is equal to 1. When
& The ordering costs or the shortage costs are independent considering two assembly systems Lk and Lz with a low
and not considered. The handling costs from an inside value cck,z, it might be more economically convenient to
plant point of view are not considered because they are collect the components in one central warehouse which
the same for all the component management policies. serves both the assembly systems, while for an index value
The warehouses construction costs inside a plant have of 1, the realization of one or two warehouses appears in-
been ignored because they are roughly constant with different from an inventory costs’ point of view. However,
respect to the centralization degree. the transport cost of the components from the warehouse to
& The transportation cost for a given distance is calculated the assembly systems can make the collection of invento-
in €/trip as ries in a central warehouse that serves both inconvenient or
nor feasible, because savings in stock expenses could result
in elevated transportation costs. The lower the value of cck,z,
Ct z j ¼ dðW ; LÞ  T ð5Þ
the greater are the benefits derived from the stock
centralization estimating the inventory reduction.
Where T is a specific transportation cost €/m that CCk,z is a M×M matrix, where M is the set of different
depends on the system used for the transports. assembly systems in the plant and cck,z are the values in the
With a manual elevator assuming an annual cost, an matrix in the k, z. position. Introducing a binary variable
annual operator cost, and an average distance covered in a gk,q that assumes a value of 1 if the assembly line k is
year, it can be assumed as 0.025 €/m. This value could be supplied by the supermarket q, 0 otherwise, it is possible to
defined considering a certain transportation equipment, its define Gz,q, a M×G matrix, where G is the number of
annual cost, and the average distance covered. Assuming supermarket warehouses G assumes values from 1 (com-
that Lk and Lz are the two assembly systems, the plete centralization, unique group of assembly lines
components category diversity index between assembly supplied by a unique warehouse) to M (number of assembly
system Lk and Lz (i.e., between the two components systems, complete decentralization, where the warehouse is
category k and z) can be defined as: located in the assembly system itself). The problem of the
optimal grouping the assembly systems, for a set number of
0P
n 1
installed warehouses G, is given by:
Ui;k  Ci  ti  yi;k;z
B i¼1 C
cck;z ¼ B
@ P n
C0  cck;z  1
A ð6Þ X
G X
M

Ui;k  Ci  ti Minimize CCk;z  Gz;q ð7Þ

i¼1 q¼1 k¼1

Table 4 Group definition for “Engines-Transmissions”

1 group (complete 2 groups 3 groups 4 groups 5 groups (complete

centralization) decentralization)

(L1, L2, L3, L4, L5) (L1, L4, L5); (L2, L3) (L1); (L2, L3); (L4, L5) (L1); (L2, L3); (L4); (L5) (L1); (L2); (L3); (L4); (L5)
782 Int J Adv Manuf Technol (2010) 50:775–788

Table 5 Transportation cost in case of two supermarkets couple of coordinates that minimize the transportation cost
Supermarket A B C D E F G according to the formula
warehouse
location N 
XX  

Ui;k
Minimize  T  d Wp ; L k ð8Þ
k2q i¼1
Ki
L1–L4–L5 7.5 8.5 9.75 6.25 6.75 13.5 9.5
L2–L3 6.5 3 5.5 5.5 3.5 7.5 4
Where Ki is the kanban conveyor capacity for the
component i and T is a specific transportation cost €/m.
P
M The unique variable for a certain group q of the analyzed
Subject to gk;q ¼ 1; (each assembly system belonging to assembly systems is the position p of the warehouses Wp,
k¼1
one group) between the finite possible positions, defined as input. The
position of the assembly line Lk is fixed. The distance is
gk;q 2 f0; 1g; k ¼ ð1; :::; M Þ; q ¼ ð1; :::; GÞ calculated using formula (4). Now, for each possible
number of installed warehouses G and for the Wp position
where CCk;z  Gz;q is a matrix product, M is the total number defined with formula (8) for each assembly systems k
of assembly lines, and G is the total number of decided supplied by the supermarket q, the total transportation costs
supermarkets warehouses. The assembly systems served by function of G are the sum of each transportation cost for
the same supermarkets warehouse are defined as a group. each supermarket warehouse, given by:
The first constraint ensures that every assembly system G XX
X N
Ui;k  
belongs to a single group; the second ensures that the CT ¼  T  d Wp ; Lk ð9Þ
variable is binary. The gk,q with this minimization defined, q¼1 k2q i¼1
Ki
for a given G number of warehouses, how to group the
assembly lines and how to supply them. The third main step is to calculate the inventory costs
The model provides three main steps. After having related to the solution with G warehouses.
calculated the cck,z values, the first step collects the Changing the number of G warehouses should affect the
assembly systems M in G groups, with an array of inventory cost for each group q calculated as:
possibilities ranging from G=1 to G=M, minimizing the XX
N

sum of components category diversity index, using (7); this CI ¼ SSi;k  ti  C i ; ð10Þ
k2q i¼1
step ensures that the assembly systems are grouped
optimally according to the diversity of the components The total inventory cost in function of the decided
used, with respect to a certain category related to the number of warehouses G to install is given by:
inventory costs for a given number of warehouses. G XX
X N
The second step consists in the definition of warehouses CI ¼ SSi;k  ti  C i ; ð11Þ
locations, assuring that such location minimizes the q¼1 k2q i¼1
transport cost for the solution found in the previous step.
The location of every single warehouse is determined by Finally, it is possible to calculate the global cost using
the solution of an allocation problem with the objective formula (1) CTOT ¼ CI þ CT . The best components man-
function that minimizes exclusively the transport costs. agement policy for a certain components category will be
This is a well-known problem. Defined as inputs: found just selecting the number of warehouses from 1
(complete centralization) to M (complete decentralization)
1. number of defined warehouses G
that minimizes the global costs. How to supply the
2. the assembly line k supplied by the supermarket g
assembly systems, where to define the warehouses’
(output of step 1).
location, what kind of components stocks, and where and
Calculate the transportation cost in function of the what the SS is for each single components belonging to the
warehouse location Wp, choosing the Wp position as a analyzed components category are automatically defined by

Table 6 Warehouse’s location for “Engines-Transmissions”

1 group (complete 2 groups 3 groups 4 groups 5 groups (complete

centralization) decentralization)

(L1, L2, L3, L4, L5) (L1, L4, L5); (L2, L3) (L1); (L2, L3); (L4, L5) (L1); (L2, L3); (L4); (L5) (L1); (L2); (L3); (L4); (L5)
(E) (D); (B) (L1); (B); (B) (L1); (B); (L4); (L5) (L1); (L2); (L3); (L4); (L5)
Int J Adv Manuf Technol (2010) 50:775–788 783

Table 7 Inventory cost in case

of two supermarkets Component SS (L1, L4, L5) SS (L2, L3) Total Ci ti Total
(pieces/period) (pieces/period) (pieces/period) (€/piece) (/year) (€/period)

P1 0.00 4.74 4.74 15,000 0.3 21,345.37

P2 0.00 4.24 4.24 13,000 0.3 16,546.30
P3 4.56 4.50 9.06 10,000 0.3 27,186.22
P4 3.00 4.50 7.50 8,000 0.3 18,000.00
P5 8.62 0.00 8.62 5,000 0.3 12,925.27
P6 5.61 6.00 11.61 4,000 0.3 13,934.98
P7 4.80 3.00 7.80 3,000 0.3 7,022.11
116,960.25

the procedure in steps 1, 2, and 3. Figure 5 shows a flow and the total components of the analyzed category used by
chart of the complete procedure. assembly line 1 (Table 3).
In this case, the value is close to 1, so the assembly lines
are different from this component category point of view.
4 Applicative case PG PM
Using formula (7) CCk;z  Gz;q and changing the
q¼1 k¼1
An industrial application that confirms the model useful- number of groups, we obtain five possible solutions:
ness is presented in the following section. The company
Step 2: Best warehouse’s location
assembles motorcycles in five different straight assembly PP N   
Ui;k
lines, L1, …, L5 in an ATO environment, located as Minimizing formula (8) Ki  T  d W p ; Lk , we
k2q i¼1
described in Table 2. For different components categories,
define the optimal warehouse’s location in function of the
the components management has to be optimized. The
different groups obtained in the previous step (Table 4). For
study has been developed for different components catego-
example, in the solution with two groups (L1, L4, L5 and
ries choosing between the different options exposed in
L2, L3), we found the transportation costs showed in
Figs. 1, 2, and 3. The components category illustrated
Table 5 in €/period. The lowest transportation cost is given
below represents “Engines-Transmissions” for our motor-
with the supermarket in position D for lines L1, L2, and L5
cycle assembly study and it refers back to Table 1 for input
and in B for lines L2 and L3. Table 6 illustrates the best
data. σi,j has been considered as 15% of Ui,j, and LTi,j has
supermarkets location in function of the output of the step 1
been considered equal to 1 day.
of the procedure.
Step 1: Matrix calculation and groups’ definition The total transportation cost calculated using formula
PG PP N
Ui;k  
The 0calculation 1 the CC k,z , using formula (6)
of (9) CT ¼ Ki  T  d Wp ; Lk for each possible
Pn
q¼1 k2q i¼1
Ui;k Ci ti yi;k;z number of supermarkets is reported in Table 8.
B C
cck;z ¼ @ i¼1P
n A0  cck;z  1 and the input data
Ui;k Ci ti Step 3: Inventory costs calculation and transportation
i¼1
provided, defines the components category matrix illustrated costs calculation
in Table 3. The inventory cost in function of the decided number
For example, components category diversity index cc1,2 of supermarkets is calculated using formula (10) CI ¼
in the matrix is the ratio between the costs of components PP N
SSi;k  ti  C i , which results in the inventory costs
belonging to the same components category, used by k2q i¼1
assembly line 1 which are not shared by assembly line 2, showed in Table 7, expressed in €/period.

Table 8 Warehouse’s location,

inventory, transport, and global Number of Groups Inventory costs Transport costs Global costs
costs for “Engines-Transmissions” groups (€/year) (€/year) (€/year)

1 (L1, L2, L3, L4, L5) 97,976.08 20,191.67 118,167.74

2 (L1, L4, L5); (L2, L3) 116,960.25 19,225.00 136,185.25
3 (L1); (L2, L3); (L4, L5) 125,958.95 14,866.67 140,825.62
4 (L1); (L2, L3); (L4); (L5) 132,166.67 7,283.33 139,450.01
5 (L1); (L2); (L3); (L4); (L5) 144,675.00 0.00 144,675.00
784 Int J Adv Manuf Technol (2010) 50:775–788

(pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day) (pieces/day)
The total inventory cost calculated using formula (11)

utilization deviation
PG PP

Standard
N
CI ¼ SSi;k  ti  C i ; for each possible number of

2.5

2.5
q¼1 k2q i¼1

0
5

5
0
supermarkets is reported in Table 8.
Figure 6 shows the total cost data in function of the

Average
number of warehouses as calculated with the tested
procedure.

L5

10

10
5
0

0
The final step minimizes the global cost. In the case study,
the outcome is to supply the “Engines-Transmissions”

utilization deviation
Standard
components category to all the five assembly lines from a

2.5

10

15
central warehouses located in E, with a global annual cost

0
5
equal to €118,167.74.
As a second example, we present another components

Average
category used in all five assembly lines L1, L2, L3, L4, and

L4

20

10
30
L5: the addressed component is “Lights”. This category is

5
0

0
presented with six different components’ types, each type

utilization deviation
Standard
used differently in the assembly lines as illustrated in
Table 9. In this case, the value attributed to σi,j is 50% of Ui,j

25
5
5
0
0

0
and LTi,j has been considered equal to 1 day.
The calculation of the CCk,z for the component “Lights”

Average
0P
n
1
Ui;k Ci ti yi;k;z
is obtained using formula (6) cck;z ¼ B C
L3
A0 

10
10

50
@ i¼1P

0
0

0
n
Ui;k Ci ti
cck;z  1 and is reported in Table 10. i¼1

utilization deviation
Using formula (7) and changing the number of groups, Standard
we obtain five possible solutions (Tables 11 and 12). For

25

10

15
5
0

0
each one of these solutions, it is possible to obtain an
optimal warehouse location using formula (8) and the
Average

relative inventory and transportation costs using formula (9)

and (11). Figure 7 shows the total cost data in function of
L2

50
10

20

30
0

0
the number of warehouses for the components category
“Lights”.
container (€/piece) (/year) utilization deviation
Standard

According to the present study, for the components

7.5

category “Lights”, the optimal flow is reached supplying

25
10

25
0

the assembly line with direct assembly lines stations for L3,
L4, and L5. Due to the high commonality level between L1
Average

and L2 and the high cost of components between “light A”

L1

50
20

15

50

and “light B” both used by the two assembly lines, the

0

model proposes a decentralized warehouse placed in E that

0.3
0.3
0.3
0.3
0.3
0.3

supply L1 and L2 with the components “light A”, “light B”,

Table 9 Inputs data of components category “Lights”

“light D”, and “light F”.

600
500
450
300
200
170
Kanban C

capacity
(pieces)

160.000,00
140.000,00
40
40
40
40
50
50
COSTS (EURO/YEAR)

120.000,00
Components category “Lights”

Inventory Cost
(pieces/day)

100.000,00 Transport Cost

utilization

80.000,00 Total Cost

average
Part Description Global

60.000,00
120

110
40
30
40
70

40.000,00
20.000,00
0,00
A

D
B
C

E
F

1 2 3 4 5
Light
Light
Light
Light
Light
Light

NUMBER OF WAREHOUSES

Fig. 6 Total cost data in function of the number of warehouses for

P1
P2
P3
P4
P5
P6

“Engines-Transmissions”
Int J Adv Manuf Technol (2010) 50:775–788 785

Table 10 CCk,z matrix for “Lights” & Refill from the supermarket warehouse in B the
L1 L2 L3 L4 L5 Total assembly lines L2 and L5 for the “Fuel Tank”
components category.
L1 0.000 0.000 0.245 0.274 0.189 0.708
The results we presented are limited to the four
L2 0.000 0.000 0.241 0.239 0.219 0.698
components categories described. Since the number of
L3 0.476 0.476 0.000 0.238 0.238 1.429
components categories used in an assembly system is
L4 0.576 0.576 0.738 0.000 0.267 2.157
usually more than four, each component should be studied
L5 0.591 0.591 0.545 0.136 0.000 1.864
and grouped together in categories to extend the optimiza-
tion to all the components necessary for the complete
assembly. Using the concept of category instead of the
The procedure so far described was applied to other single component is one way to significantly simplify the
components categories for the same assembly lines, as problem, from thousands of components to tens of
summarized in Table 13, where the optimal components components categories. The concept of components cate-
category management solution is highlighted in gray for gory diversity index uses the components category in order
easy recognition. For example, the proposed procedure to define how to optimally group the assembly systems
shows that the “Fuel Tank” category needs to be managed from a stocking point of view. Once the common data
with two supermarket warehouses located in E and B, inputs for all the categories considered are standardized
supermarket E to supply assembly lines L1, L3, and L4, (assembly line position, warehouse possible position, type
and supermarket B to supply lines L2 and L5, while for the of transportation equipment and specific transportation cost,
“Nuts and Bolts” category it is convenient to adopt a etc.), it becomes very easy to implement the procedure. The
complete warehouse decentralization. concept of components category diversity index, together
Summarizing the results for all the components category with the concept of components category, introduces a great
studied, the methodology defines what components should simplification in a very complex problem with the
refill the station with kits and how the kit should be created consequent optimization of stocking policies, taking into
within the supermarket solution. account all the critical variables of an ATO environment.
The results are: In order to show the benefits obtained by the use of the
presented methodology, Table 13 offers a benchmark
& Create a central warehouse in E.
between optimal solution and possible sub-optimization
& Create two supermarkets warehouse in E and B.
given by a complete centralization, with various options: a
& Refill directly the assembly stations for “Nuts and
central warehouse, a complete decentralization, or the use
Bolts” components category.
of three supermarkets warehouses that refill the five
& In E, create kits in order to refill assembly lines
assembly lines. The table shows a considerable difference
composed by
in costs between the different solutions.
& “Engines-Transmissions”, “Lights”, and “Fuel
Tank”, for assembly line L1
& “Engines-Transmissions” and “Lights” for assem-
5 Conclusions
bly line L2
& “Engines-Transmissions” and “Fuel Tank” for
The proposed procedure, with a step-by-step process based
assembly lines L3 and L4.
on different and successive linear programming optimiza-
& Refill directly the assembly stations for lines L3, L4, tions, aims to support the materials management in case of
and L5 for the “Lights” components category. an ATO environment, indicating for what components

Table 11 Warehouse’s location, inventory, transport, and global costs for “Lights”

Number of Groups WH location Inventory costs Transport costs Global costs

groups (€/year) (€/year) (€/year)

1 (L1, L2, L3, L4, L5) (E) 14,183.45 7,940.00 22,123.45

2 (L1, L2, L3); (L4, L5) (E); (B) 16,563.37 6,860.00 23,423.37
3 (L1, L2); (L3); (L4, L5) (E); (L3); (B) 17,362.91 5,360.00 22,722.91
4 (L1, L2); (L3); (L4); (L5) (E); (L3); (L4); (L5) 18,317.90 3,745.00 22,062.90
5 (L1); (L2); (L3); (L4); (L5) (L1); (L2); (L3); (L4); (L5) 22,530.00 0.00 22,530.00
786 Int J Adv Manuf Technol (2010) 50:775–788

Table 12 Inventory costs, transportation costs, and global costs for different components categories
Inventory cost, transportation cost and global cost Total cost in function of the number of supermarket warehouses
Fuel Tank

Nuts and bolts

category it is advisable to adopt a decentralized solution

Inventory Cost with kits supply to the assembly systems from a supermar-
Transport Cost
25.000,00 Total Cost
ket warehouse, or if it is worth to pick a complete
centralization or decentralization of warehouses instead,
COSTS (EURO/YEAR)

20.000,00
minimizing global costs. The scientific literature is actually
15.000,00 more focused on the centralization/decentralization problem
10.000,00
regarding a supply chain network and usually does not
consider typical aspects of the assembly process. This
5.000,00
methodology considers all the critical variables for compo-
0,00 nents management in an assembly: type of components,
1 2 3 4 5
commonality degree between components and between the
NUMBER OF WAREHOUSES
assembly systems involved, load unit transportation costs
Fig. 7 Total costs in function of the number of warehouses for and transportation tool used, space availability in the plant,
“Lights” safety stocks values, introducing an innovative model to
group different assembly systems from the components

Table 13 Solutions’ benchmark

between optimal solution and Components category Optimal Complete Complete 3 supermarkets
sub-optimization for the four solution centralization (in E) decentralization solution
components categories (in E, D, B)
considered
Engines-Transmission 118,167 118,167 144,675 136,185
Lights 22,062 22,123 22,530 23,423
Fuel Tank 133,023 134,594 135,592 133,023
Nuts and Bolts 1,956 5,791 5,791 5,847
Total (€/year) 275,208 280,675 308,588 298,478
Costs increment 0 5,467 33,380 23,270
Int J Adv Manuf Technol (2010) 50:775–788 787

point of view, focusing on a fixed components category and References

for that, defining warehouses number (centralized, decen-
tralized, or supermarkets), location, and components stock- 1. Akçay Y, Xu SH (2004) Joint inventory replenishment and
ing methodology. component allocation optimization in an assemble-to-order sys-
tem. Manag Sci 50(1):99–116
Summarizing the results for all components category, the
2. Azzi A, Battini D, Faccio M, Persona A, Sgarbossa F (2009)
methodology defines what components refill the station Quantificare l’Indice Annuo dei Costi di Giacenza: Modello di
with kits and how to create the kit within the supermarket Calcolo e Analisi Comparativa di Casi Studio, Prooceding
solution. The model has been applied to different industrial ANIMP 2007. 36°Convegno Nazionale Ingegneria E Impiantistica
Italiana Rome, 11–12 June 2009. http://www.animp.it/news.aspx?
contexts: the motorcycle assembly application is described
sezione=Convegni
in this document, but the methodology has been success- 3. Battini D, Faccio M, Ferrari E, Persona A, Sgarbossa F (2007)
fully applied also to the assembly of injection and blow Design configuration for a mixed-model assembly system. Int J
molding dispensers. The methodology demonstrates to be Adv Manuf Technol 34:188–200
4. Battini D, Faccio M, Persona A, Sgarbossa F (2008) Balancing–
versatile and easily applicable to different industrial sequencing procedure for a mixed model assembly system in case
environments, plus it reveals to be a useful tool in decision of finite buffer capacity. Int J Adv Manuf Technol 44(3–4):345–
making for materials management policies in assembly 359
lines. 5. Bollapragada R, Rao US, Zhang J (2004) Managing inventory and
supply performance in assembly systems with random supply
In conclusion, we can confidently state the following:
capacity and demand. Manag Sci 50(12):1729–1743
6. Choi W, Lee Y (2004) A dynamic part-feeding system for an
& Material management in case of assembly environment automotive assembly line. Comput Ind Eng 43(1):123–134
has particular characteristics different from a generic 7. Chu H, Wang J, Jin Y, Suo H (2006) Decentralized inventory
control in a two-component assembly system. Int J Prod Econ
supply chain
102:255–264
& In an ATO environment, innovative material manage- 8. Das C, Tyagi R (1999) Effect of correlated demands on safety
ment policies are becoming increasingly common: the stock centralization: patterns of correlation versus degree of
supermarket warehouse, defined as a decentralized centralization. Int J Bus Logist 20(1):205–213
9. De Boeck L, Vandaele N (2008) Coordination and synchroniza-
warehouse created close to the assembly systems, that
tion of material flows in supply chains: an analytical approach,
typically refill the stations with kits (i.e., different International. J Prod Econ 116(1):199–207
material stored in an apposite container that supplies 10. De Souza MC, De Carvalho CRV, Brizon WB (2008)
different assembly stations for a certain time period.) Packing items to feed assembly lines. Eur J Oper Res 184
(2):480–489
& The choice of what component can be managed with
11. Drake PR, Lee DM (2009) Component prioritisation for strategic
supermarkets policy depends on factors that are typical purchasing and the case study of a South Korean elevator
of the assembly environment: number, type and position manufacturer. Int J Adv Manuf Technol 43:883–895
of the assembly systems, type and demand of each 12. Eppen GD (1979) Effects on centralization on expected costs in a
components, commonality degree of components be- multi-location newsboy problem. Manag Sci 51(1):498–501
13. Evers PT, Frederick BJ (1993) The portfolio effect and multiple
tween different assembly systems, internal transporta- consolidation points: a critical assessment of the square root law. J
tion equipment used and transportation costs, load unit Bus Logist 14(2):109–125
capacity, space cost for stocking and space availability 14. Hillier MS (2000) Component commonality in multiple-
in the plant, inventory costs, and safety stock period, assemble-to-order systems. IIE Transactions 32
(8):755–766
dimension.
15. Hillier MS (2002) The costs and benefits of commonality in
& In order to adopt the supermarket policy, it is important assemble-to-order systems with a (Q, r)-policy for component
to understand where the supermarket needs to be replenishment. Eur J Oper Res 141(1):570–586
placed, in function of the different assembly systems 16. Hsu VN, Lee CY, So KC (2007) Managing components for
assemble-to-order products with lead-time-dependent pricing: the
present on the line, how the components are used in
full-shipment model. Nav Res Logist 54(5):510–523
the different assembly systems, the kind of the 17. Yeh WC (2005) A hybrid heuristic algorithm for the multistage
components, and all the aspects described before. For supply chain network problem. Int J Adv Manuf Technol 26:675–
that the components category diversity index has been 685
introduced. 18. Li SG, Kuo X (2008) The inventory management system for
automobile spare parts in a central warehouse. Expert Systems
& For its important results in the industrial environment, Appl 34(1):1144–1153
future research are needed, focusing in particular on 19. Lim SJ, Jeong SJ, Kim KS, Park MW (2006) A simulation
the kit’s container optimization, which is a typical approach for production–distribution planning with consideration
problem in the ATO–JIT environment due to its given to replenishment policies. Int J Adv Manuf Technol 27:593–
603
variable transportation costs, its handling costs, and
20. Liu SC, Lin CC (2005) A heuristic method for the combined
its investment costs in function of different container location routing and inventory problem. Int J Adv Manuf Technol
structures. 26:372–381
788 Int J Adv Manuf Technol (2010) 50:775–788

21. Manzini R, Gamberi M, Persona A, Regattieri A (2004) A 27. Ronen D (1990) Inventory centralization/decentralization—the
framework for designing a flexible cellular assembly system. Int J ‘square root law’ revisited again. J Bus Logist 11:129–138
Prod Res 17(42):3505–3528 28. Saygin C (2007) Adaptive inventory management using RFID
22. Manzini R, Persona A, Regattieri A (2006) Framework for data. Int J Adv Manuf Technol 32:1045–1051
designing and controlling a multicellular flexible manufacturing 29. Stulman A (1987) Benefits of centralized stocking for the multi-
system. IJSOM 1(2):1–21 centre newsboy problem with first come, first served allocation. J
23. Matta A, Dallery Y, Di Mascolo M (2005) Analysis of assembly Oper Res Soc 38(9):827–832
systems controlled with kanbans. Eur J Oper Res 166(1):310–336 30. Thangam A, Uthayakumar R (2009) A two-level distribution
24. Mohebbi E, Choobineh F (2005) The impact of component inventory system with stochastic. Int J Adv Manuf Technol
commonality in an assemble-to-order environment under supply 41:1208–1220
and demand uncertainty. Omega, IJMS 33(1):472–482 31. Ting PS, Hou KL, Chung KJ (2009) An accurate and reliable
25. Nonas LS (2009) Finding and identifying optimal inventory levels solution algorithm for the (Q, r) inventory system with a fixed
for systems with common components. Eur J Oper Res 193 shortage cost. Math Comput Model 49(1):128–135
(1):98–119 32. Zhang X, Ou J, Gilbert SM (2007) Coordination of stocking
26. Persona A, Battini D, Manzini R, Pareschi A (2007) Optimal decisions in an assemble to order environment. Retrieved July, 10,
safety stock levels of subassemblies and manufacturing compo- 2007, from Kellogg Website: http://www.kellogg.northwestern.
nents. Int J Prod Econ 110(1–2):147–159 edu/