Professional Documents
Culture Documents
Excess Inventories Redeployment Strategy For Spare
Excess Inventories Redeployment Strategy For Spare
Abstract – Many leading companies are now offering Our research on how inventory redeployment strategy
global customers better spare parts services for system can reduce excess inventories of spare parts in a service
maintenance through a more complex service logistics logistics network contributes to the literature on two fronts:
network, extending beyond the traditional on-site stocking (1) expand research focus of reverse logistics management
management, to boost profit margin. One challenge these to spare parts inventory management, and (2) address the
spare parts service providers face is how to achieve desired
service levels at a low cost through minimization of excess
need to deal with items with non-stationary demand in
inventories in the global spare parts supply chain. To address spare parts inventory management research.
this issue, we demonstrate an inventory redeployment The organization of this paper is as follows. In Section
strategy to transform a conventional spare parts supply chain II, we review the literatures on spare parts inventory model
(with forward stocking facilities only) into a closed-loop, and logistics network design. In Section III, we discuss the
multi-echelon service network with the capability of methodology of how excess inventories can be redeployed
redeploying inventories from overstocking to understocking through a network optimization model based on real data.
facilities, reducing purchase of high-value spare parts. To In Section IV, we present results with sensitivity analysis.
assess the quality of our novel solution approach, we used a Finally, Section V presents the conclusion and future
network flow optimization model to analyze the proposed
excess inventories redeployment strategy of an international
research.
company’s service parts operations, and found significant
inventory cost savings.
II. LITERATURE REVIEW
Keywords – Inventory management, service logistics
network design, network flow model, spare parts. Review of spare parts inventory management literature
shows the need for more research on how to manage stock-
out or excess inventories of spare parts with non-stationary
I. INTRODUCTION demand in a service logistics network with multiple
stocking locations. Differed from before-sale supply chain
Spare parts management is crucial to the maintenance inventory management models, spare parts inventory
of critical systems in different sectors like healthcare, management are characterized by distinctive factors
military and aerospace. Instead of managing on-site spare including high spare parts cost, low demand with high
parts inventory, more customers have outsourced that uncertainty, and the critical effect of missing parts. Past
function to spare parts service providers. To stay studies of spare part inventory models examined on-site
competitive and earn a profit, these service providers have age-based replacement, repairable and non-repairable
on one hand to offer highly responsive services to fulfil items, and regular and emergent orders [1]. As keeping
next-day or even same-day delivery requirements, and on high-value spare parts for all systems on-site is cost
the other to minimize operating costs. To address these prohibitive, more users have outsourced spare parts
needs, we propose a new spare parts logistics model and management and shifted the burden to service providers
demonstrate its application in practice. who can provide the required parts on an emergency or
In this research, we study an international company short lead time basis. This trend has driven research on how
which manages 87 depots (which are forward stocking service providers manage spare parts in multiple locations
facilities) to provide spare parts services for customers’ of a service logistics network with forward stocking
system maintenance in Asia-pacific region. Our facilities [2]-[5]. These studies proposed an inventory
investigation shows that the company’s current operation pooling or sharing strategy with lateral supply to optimize
has suffered from excess inventories caused by non- the holding cost, expedited shipment cost and stock-out
stationary demand, base-stock inventory policy and cost. With this strategy, when a local stocking point is out
forward logistics network structure. To solve the problem, of stock, demand would be satisfied by a nearby stocking
we propose an excess inventories redeployment strategy by facility with an emergency shipment. As stationary demand
restructuring the forward logistics network into a closed- of forward inventory flows is generally assumed in these
loop network, aiming at better utilizing high-value models, there is a research gap on the management of
inventories and reducing purchase of new spare parts. stock-out or excess inventories of spare parts with non-
stationary demand.
Review of reverse logistics management literature The quantity and locations of excess inventories
indicates a promising way to reduce excess spare parts redeployment are determined based on considerations of
inventories. In this literature, some studies [6]-[9] applied forecast usage at sourcing depot and destination depot,
mixed integer programming models to solve reverse target inventory level, on-hand and on-order inventory,
logistics network design problems and determine the cost of a spare part, and transportation cost. We examine
optimal numbers and locations of remanufacturing this decision problem faced by a spare parts service
facilities, in both forward and reverse flows. Some recent provider. With the real data provided by the company, we
research [10]-[12] examined management of product develop excess inventories redeployment strategy by
returns and determined optimal acquisition quantities of optimization techniques.
used products with considerations of acquisition costs, As the target inventory level is named “approved
scrapping costs and remanufacturing costs in a closed-loop minimum” in the inventory planning system, the target
supply chain network. Although these studies focused on inventory level and approved min are used interchangeable
product returns and recycling, they inform us that spare in the following discussion. We next discuss the data
parts stocking facilities can handle both forward and more collection, identification of excess inventories and the
importantly reverse flows of spare parts. optimization model for redeploying excess inventories.
A. DATA COLLECTION
III. METHODOLOGY
In this research, we collected several reports from the
inventory planning system of the collaborated company.
To apply excess inventories redeployment strategy for
The data sets include information of past usage, part cost,
spare parts inventory and logistics management, we
on-hand, on-order, target inventory level of each service
propose to restructure a forward logistics network into a
part in each depot. Apart from inventory-related data, we
closed-loop logistics network. Take a typical global
also established a transportation cost matrix which
network with 3-tier forward stocking locations as an
indicates the express shipment cost among 87 depots. The
example, the structure consists of a set of directed flows
transportation cost data was collected from the UPS
from a centralized distribution center, to country depots,
Express website. The corporate freight rate with discount
and finally to city depots without reverse flow. See the
was used, according to the company.
forward flows with solid lines in Fig. 1. This forward
logistics network is usually implemented in an inventory
planning system for spare parts management.
B. EXCESS INVENTORIES IDENTIFICATION
Excess inventory is defined based on a calculation of
City Depot User
on-hand inventories, on-order inventories, forecasted
Country Depot City Depot User
monthly usage with a coefficient, target inventory level.
Excess inventory quantity is determined by the minimal
Distribution Center City Depot User value of two formulas listed below:
New Repair
1996
Proceedings of the 2017 IEEE IEEM
equal to 1.505 by Equation (2). However, the depot has no Generic Reverse Network Flow Model (GRFLM)
excess inventory according to Equation (1), in which 1 + 4
– 5 = 0. If one part is redeployed to another depot, a 𝑚𝑎𝑥 ∑𝑛𝑖=1 ∑𝑛𝑗=1(𝑝𝑖 𝑥𝑖𝑗 − 𝑐𝑖𝑗 𝑥𝑖𝑗 ) (4)
replenishment order of one part will be automatically 𝑖≠𝑗
placed from the preceding depot to bring the inventory up
to the target inventory level. This causes unnecessary s.t.
additional transportation cost. ∑𝑛𝑗=1 𝑥𝑖𝑗 ≤ 𝑠𝑖 ∀𝑖 ∈ 𝑁 (5)
∑𝑛𝑖= 1 𝑥𝑖𝑗 ≤ 𝑢𝑗 ∀𝑗 ∈ 𝑁 (6)
𝑥𝑖𝑗 𝑎𝑟𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
C. REVERSE LOGISTCS NETWORK FLOW MODEL
In this section, we discuss the development of excess Constraint (5) ensures that all excess inventories are
inventory redeployment strategy by restructuring the shipped from the origin to destination depots. Constraint
forward logistics network into a closed-loop network. (6) guaranties that the redeployed inventories should not
Based on the existing forward logistics network structure, exceed the maximum level accepted by destination depots.
we establish a reverse logistics network connecting all During the investigation, it was found that the solution
depots, and formulate network flow for excess inventories erected barriers for execution in practice. The major issue
redeployment as a network flow optimization model. was too many shipment routes. See Fig. 2 for the results
To represent the reverse logistics network flow, we solved by the generic network flow model for one part.
consider a graph G = (N, A) with a finite set of nodes (N)
and a finite set of arcs (A). A depot is represented by a node. Origin Destination
Each depot would be the origin depot i=1,…,n with excess
GUR
inventories and also the destination depot j=1,…,n to 5 5
receive excess inventories. The shipment route between
two depots is represented by (𝑖, 𝑗) ∈ 𝐴. The objective of HYD
5
the reverse network flow model is used to maximize cost 5
i j
1997
Proceedings of the 2017 IEEE IEEM
maximum flow of excess inventories to a single destination with the second lowest shipping cost. MRFLM offers a
depot. The modified model is presented below: solution with much fewer shipment routes, as the number
of shipment paths are reduced from 18 in the GRFLM
Min-Path Reverse Network Flow Model (MRFLM) solution to 5 only (or about 72% reduction). For cost and
benefit comparison, GRFLM outperforms MRFLM in the
𝑚𝑎𝑥 ∑𝑛𝑖=1 ∑𝑛𝑗=1(𝑝𝑖 𝑏𝑖𝑗 𝑧𝑖𝑗 − 𝑐𝑖𝑗 𝑏𝑖𝑗 𝑧𝑖𝑗 ) (7) value of excess inventories redeployed ($21,000 vs.
𝑖≠𝑗 $11,400) but incurs higher transportation cost ($5,146 vs.
$2,465). In this case, GRFLM attained higher cost saving
s.t. ($15,854 vs. $8,935). This indicates MRFLM is superior in
∑𝑛𝑖= 1 𝑧𝑖𝑗 𝑏𝑖𝑗 ≤ 𝑢𝑗 ∀𝑗 ∈𝑁 (8) reducing shipment paths and transportation costs but may
∑𝑛𝑖= 1 𝑏𝑖𝑗 = 1 ∀𝑗 ∈𝑁 (9) have less potential in inventory reduction. We next
𝑧𝑖𝑗 = min(𝑠𝑖 , 𝑢𝑗 ) ∀𝑖 ∈ 𝑁 , ∀𝑗 ∈ 𝑁 (10) evaluate how monthly usage coefficient used in defining
𝑏𝑖𝑗 ∈ {0, 1} ∀𝑖 ∈ 𝑁 , ∀𝑗 ∈ 𝑁 (11) excess inventories affects excess inventory reduction and
transportation cost of the two models.
Equation (7) is the objective function to maximize the
cost savings of excess inventories minus transportation IV. RESULTS
cost by binary decision variable, bij Constraint (8) ensures
that the inventory quantity in destination depot should not To evaluate the monthly performance of two reverse
exceed its acceptable limit. Constraint (9) restricts the network flow models, we first identify the number of parts
shipping of excess inventories to one destination depot with excess inventory and its inventory amount
only. Constraint (10) is to determine the min-max value of respectively by Equation (1) and (2) in a month. Second,
excess inventory quantity moved from the origin depot to we analyze the number of shipment paths. Third, we
the destination depot. It ensures that the shipment quantity evaluate the cost savings of inventory reduction and
will not be less than excess inventory in origin depot or transportation costs. We also study the change of the
greater than the maximum acceptable excess inventories in coefficient of monthly usage from one to five to evaluate
destination depot. Constraint (11) refers to the binary the impacts of changing the parameters of 𝑠𝑖 and 𝑢𝑗 . Table
decision variable, bij = 1 if the excess inventories in depot 1 shows the results that the number of parts per depot with
i will be redeployed to depot j; otherwise, bij = 0. This excess inventory and total dollar value of excess inventory
model satisfies the business requirements and objectives. decrease when coefficient of monthly usage increases.
Fig. 3 shows the shipment paths of excess inventories after
TABLE 1. EXCESS INVENTORY QUANTITY AND VALUE
solving the above model for the same part (i.e. the one
Coefficient 1 2 3 4 5
examined in GRFLM). (# of months)
# of parts per 271 242 185 132 78
Origin Destination depot with
excess inv.
OSA
2 TKO Excess inventory 2,211 1,535.6 870.6 520.3 376.4
2 25 value ($K)
BKK 9 SSY
9 19
Based on these excess inventories, we first investigate
the number of shipment paths of the two models. Table 2
HCU 7 MEL shows the number of shipment paths which are solved by
10
7
two network flow models.
JKT 7 HKG
7 9 TABLE 2. NUMBER OF SHIPMENT PATHS
Coefficient 1 2 3 4 5
MNL 13 SEL (# of months)
13 16 GRFLM 21 54 150 130 101
MRFLM 18 40 86 90 77
i j
The number of shipment paths increases when the
Fig. 3. Reallocation of excess inventories in modified model coefficient of monthly usage increases, because a larger
coefficient allows more excess inventories to be received
Fig. 3 shows that each destination depot can only in the destination depot (uj). However, the number of
receive excess inventories from one origin depot due to shipment paths drops afterwards because the increase of
constraints (8)-(9) in MRFLM. Take MNL as an example, coefficient can also reduce excess inventories of the origin
although it is cheaper to ship MNL’s excess inventories to depot (si). Fig. 4 shows the cost and benefit comparison of
HKG than to SEL, HKG does not have enough capacity to GRFLM and MRFLM, and Table 3 shows the cost savings
absorb all excess inventories from MNL. Thus, all of value and percentage (against total excess inventory value).
MNL’s 13 units of excess inventories are shipped to SEL
1998
Proceedings of the 2017 IEEE IEEM
50 ACKNOWLEDGMENT
We thank anonymous reviewers for their constructive
0 comments. This research is supported by the grant of RGC
1 2 3 4 5
FDS (UGC/FDS14/E05/16) and supported in part by the
Month
Fig. 4. Comparison of Two Reverse Network Flow Models
grant of RGC IIDS (UGC/IIDS14/E01/16).
1999