08 16 A Serial Inventory System With Supplier Selection and Order Quantity Allocation Considering Transportation Costs

Applied Mathematical Modelling 40 (2016) 612634
Contents lists available at ScienceDirect
Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm
A serial inventory system with supplier selection and order

quantity allocation considering transportation costs
Subramanian Pazhani a,1, Jos A. Ventura a,, Abraham Mendoza b,2
a
b
Harold and Inge Marcus Department of Industrial & Manufacturing Engineering, The Pennsylvania State University, University Park, PA 16802, USA
Department of Industrial Engineering, Universidad Panamericana, Zapopan, Jalisco 45010, Mexico
a r t i c l e
i n f o
Article history:
Received 24 September 2013
Received in revised form 30 April 2015
Accepted 18 June 2015
Available online 3 July 2015
Keywords:
Supply chain management
Multi-echelon inventory system
Supplier selection
Order quantity allocation
Transportation costs
a b s t r a c t
Given the importance of incorporating transportation costs in inventory replenishment and
supplier selection decisions, this article addresses the issue of supplier selection and order
quantity allocation in a multi-stage serial supply chain system with multiple suppliers considering inventory replenishment, holding, and transportation costs simultaneously. We
propose a mixed integer nonlinear programming model to determine the optimal inventory policy for the stages in the supply chain and allocation of orders among the suppliers
at the initial stage. Transportation costs between consecutive stages are modeled using a
piecewise constant setup cost structure arising from a full truckload freight rate cost
model. Vehicles of different capacities are available to transport materials from the suppliers to the manufacturing facility and between the remaining stages of the supply chain. The
usefulness of the model is discussed with an example. Sensitivity analysis is carried out to
determine the effect of cost parameters on supplier order allocation. The analysis shows
that the selection of suppliers and the corresponding order quantities are affected by the
variations in supply chain costs parameters. In addition, the advantages of using an integrated approach versus a sequential approach for inventory replenishment and supplier
selection decisions are shown. Computational results show that the integrated approach
yields average savings of 4.88% in total cost and 15.31% in logistics costs over the sequential approach.
2015 Published by Elsevier Inc.
1. Introduction
In todays competitive environment, companies are forced to optimize business processes and improve the performance
of their entire supply chains. Successful performance of a supply chain depends on every single organization involved, and an
efcient and exible supply chain enables the rm to select the right suppliers at the right time for the right materials, not
only signicantly reducing purchasing cost, but also greatly improving corporate competitiveness [1]. Among the various
drivers in a supply chain, sourcing, inventory, and transportation are recognized as the major ones [2]. In the rst place,
partnering with the right suppliers has been increasingly recognized as a strategic and crucial component of supply chain
management [3]. Selecting the right suppliers can impact the overall purchasing cost (the cost of raw materials and component parts), which accounts for a large percentage of the nal product cost. According to Burton [4], the cost of components
Corresponding author. Tel.: +1 (814) 865 3841; fax: +1 (814) 863 4745.
1
2
E-mail addresses: sathish.subramanian84@gmail.com (S. Pazhani), jav1@psu.edu (J.A. Ventura), amendoza@up.edu.mx (A. Mendoza).
Tel.: +1 (814) 865 3841; fax: +1 (814) 863 4745.
Tel.: +52 (33) 136 82221; fax: +52 (33) 136 82296.
http://dx.doi.org/10.1016/j.apm.2015.06.008
0307-904X/ 2015 Published by Elsevier Inc.
S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634
613
and parts purchased from suppliers may total more than 50 percent of sales for large automotive manufacturers, and purchased material and services represent up to 80 percent of the nal total product cost for high technology organizations.
Heberling [5] points out that over 80 percent of the purchases in the manufacturing sector are for materials and supplies.
Hence, it is essential for organizations to keep supplier-related costs to a minimum. According to Weber and Current [6],
the objective of supplier selection is to identify suppliers with the highest potential for meeting demand needs. This implies
not only to choose the best suppliers to do business with but also to determine the exact quantities for each order placed
with a selected supplier.
In the second place, since manufacturing processes are not instantaneous and to keep up with unpredictability in businesses, inventory has traditionally been viewed as a necessity to meet customers demand. Higher levels of inventory result
in increased responsiveness of the supply chain, but decrease the cost efciency because of the cost associated with holding
inventory. Therefore, inventory management is one of the important aspects of supply chain management because the cost
of inventory can represent anywhere between 20% and 40% of the total value of the product [7].
In the third place, when computing the order quantity of a product, the transportation cost is also an important aspect to
be considered. However, many inventory management models in the literature assume that the transportation cost is
included in the purchasing cost of the product [8]. Considering transportation costs in determining order quantities becomes
imperative to improve the efciency of the supply chain. Blumenfeld et al. [9] discussed the need for analyzing the trade-offs
between inventory and transportation costs. Their model focuses on minimizing the total inventory and routing cost by
simultaneously determining optimal routes and shipment sizes of vehicles.
Research on supply chain optimization has mainly focused on two problems: (1) the manufacturer has to determine its
optimal production, distribution and inventory policies considering its capacity, setup costs, distribution costs and operating
costs, and deliver the nal products to customers; and (2) the manufacturer has to determine the suppliers/vendors from
which to purchase raw materials as well as the corresponding order quantities (supplier selection). Solving these two problems separately (in sequence) may yield to sub-optimal solutions for the entire supply chain. Therefore, this article considers
an integrated approach consisting of a multi-stage serial supply chain system that simultaneously addresses the problems of
supplier selection and inventory replenishment considering purchasing, setup, holding, and transportation costs. We propose a mixed integer nonlinear program (MINLP) for the integrated inventory planning and supplier selection problem.
The advantages of optimizing the supply chain considering inventory and transportation costs simultaneously are illustrated
using a numerical example. Sensitivity analysis is carried out to determine the effect of cost parameters on supplier order
allocation. The analysis shows that the supplier selection and the corresponding order quantities are affected with changes
in the supply chain cost parameters. Moreover, the proposed integrated approach is compared with the sequential approach,
where the inventory planning and supplier selection problems are solved sequentially. The results from the analysis show
that the integrated approach yields signicant savings in terms of logistics and overall supply chain costs.
The remainder of this article is organized as follows. Section 2 provides a review of the literature. In Section 3, a discussion
on full truckload (TL) freight rates is presented. The vehicle selection problem in the supply chain is formulated as a minimization version of the knapsack (KS) problem. In Section 4, we provide a detailed statement of the integrated inventory
planning and supplier selection problem for a serial supply chain, and propose a new MINLP model for the problem. In
Section 5, we present an example to illustrate the proposed mathematical model. Sensitivity analysis is performed in
Section 6 to show the effect of the echelon holding cost, setup cost at the stages, and transportation cost between consecutive stages. Section 7 compares the proposed integrated approach to the sequential approach, where the problem is solved
in two stages. We also provide conditions under which the integrated approach gives better results than the sequential
approach. Section 8 presents some conclusions on the work and future research directions.
2. Literature review
The supplier selection problem has been widely addressed by many researchers. Various decision models and solutions
have been proposed over time. Weber et al. [10] presented a detailed review of different decision techniques and criteria used
for supplier selection provided in 74 articles that appeared in the literature since 1966. Degraeve et al. [11] adopted the concept of Total Cost of Ownership (TCO) as a basis for comparing supplier selection models. The TCO approach basically considers all relevant costs involved in the purchasing process of a good or service from a particular supplier. De Boer et al. [12]
studied the supplier selection literature in a more comprehensive manner. They proposed a framework that includes four
main steps in the supplier selection process: problem denition, formulation of selection criteria, pre-qualication (preliminary screening), and nal selection. Ho et al. [13] surveyed 78 journal articles from 2000 to 2008 related to multi-criteria
decision making approaches for supplier evaluation and selection. Most recently, Chai et al. [14] provided a systematic literature review on articles published from 2008 to 2012 on the application of decision-making techniques for supplier selection.
Some of the approaches summarized in these reviews are based on mathematical models that integrate the selection of
suppliers and calculation of order quantities for the selected suppliers using linear programming (LP), mixed integer programming (MIP), and multi-objective programming. Some of the papers that used LP models are summarized below.
Anthony and Buffa [15] developed a strategic level single objective LP model to minimize the total purchasing cost considering limitations on purchasing budget, supplier capacity, and buyers demand. Ghodsypour and OBrien [16] used analytical
hierarchy process (AHP) and LP to determine the best order quantity allocation while considering qualitative and
614
quantitative criteria into the analysis. Talluri and Narasimhan [17] proposed an approach called maxmin that incorporated performance variability into the evaluation process. Two LP models were developed in order to maximize and minimize the performance of a supplier against the best target measures. Ng [18] developed a weighted LP model for the
multi-criteria supplier selection problem with the objective of maximizing the supplier score.
As for MIP models, Narasimhan and Stoynoff [19] proposed a single objective MIP model to a manufacturing company for
allocating procurement among a group of suppliers. Chaudhry et al. [20] developed LP and MIP models for supplier selection.
In their models, price, delivery, quality and quantity discounts were included. Kasilingam and Lee [21] proposed an MIP
model to select suppliers and to determine the order quantities based on quality, cost of purchasing and transportation, xed
cost of establishing suppliers, and the cost of poor quality. Jayaraman et al. [22] considered a supplier selection and order
quantity allocation problem with multiple products, buyers and suppliers, and formulated a mixed integer linear programming (MILP) model for solving the problem. Ghodsypour and OBrien [23] used total cost of ownership for the supplier selection problem. Recently, Hammami et al. [24] developed an MIP model for the supplier selection model that considered
inventory decisions, inventory capacity constraints, specic delivery frequency, and transportation capacity for a
multi-product, multi-period scenario.
Given the intrinsic multi-criteria nature of the supplier selection problem, some studies considered multiple objectives.
Chaudhry et al. [20] considered lead-time, service and quality performance as objectives, and suggested goal programming
to select suppliers and allocate specic order quantities to these suppliers. Weber and Ellram [25] developed a
multi-objective programming approach for supplier selection in a just-in-time (JIT) setting. Price, delivery and quality criteria are considered in their model. Weber et al. [26] developed a multi-objective model and data envelopment analysis model
for evaluating and selecting suppliers. Kumar et al. [27] considered the objectives of minimizing net cost, net rejections, and
net late deliveries, and formulated a fuzzy mixed integer goal programming for the supplier selection problem. Wadhwa and
Ravindran [28] presented and compared several multi-objective optimization methods (weighted objective, goal programming and compromise programming) for solving the supplier selection problem. Xia and Wu [1] used the analytic hierarchy
process (AHP) and a multi-objective MIP model for the supplier selection problem. Mendoza and Ventura [29] proposed a
two-phase method for integrating supplier selection and order quantity allocation. In phase 1, pre-qualication or preliminary screening is done in order to reduce the large list of potential suppliers to a manageable number. In phase 2, a mathematical model is used to allocate order quantities while minimizing the ordering, holding and purchasing costs per time
unit. Ravindran et al. [30] proposed a multi-objective model for solving the supplier selection problem in a global IT.
Their goal was to minimize the following: purchasing cost, lead-time, value-at-risk (VaR) -risk of disruption due to natural
events-, and miss-the-target (MtT) -risk of missing quality requirements. Some other recent works that have adopted
multi-objective modeling can be found in Amid et al. [31] and Feng et al. [32].
Most of the models discussed above have been designed for single-stage systems. In order to determine inventory replenishment policies for multi-stage systems, a more realistic approach is to consider multi-echelon models to capture the effect
of inventory decisions on the entire supply chain. Clark and Scarf [33] initiated the research on multi-echelon inventory
models and addressed the issue of optimal ordering and inventory policies in multi-echelon production and inventory systems. We also refer to Clark [34] for a survey on multi-echelon inventory theory. Federgruen and Zipkin [35] analyzed an
order-up-to level echelon policy by considering an innite time horizon. Several mathematical models have been proposed
to determine optimal lot sizes and safety stocks in multi-echelon systems, such as those presented by Sherbrooke [36],
Svoronos and Zipkin [37], Axsater [38], Nahmias and Smith [39], Gallego and Zipkin [40], and Chen [41]. Daniel and
Rajendran [42] considered a single-product serial supply chain operating with a base-stock policy and addressed the problem of optimizing the inventory levels using a genetic algorithm. Some researchers used simulation methodologies to
develop and evaluate inventory policies in serial supply chain systems: Lee and Billington [43], Glasserman and Tayur
[44], Axsater [45], and Sethupathi and Rajendran [46]. These studies have not considered freight rates or supplier selection
aspects in their models. Mendoza and Ventura [47] considered a serial supply chain system with multiple suppliers and proposed an MINLP model to determine an optimal inventory policy that coordinates the transfer of items between different
stages of a serial supply chain and, at the same time, determines the allocation of orders to selected suppliers. However, their
model does not include transportation costs. Yang et al. [48] implemented economies of scale and continuous-state approximation to a two-stage inventory system. A heuristic algorithm was proposed to nd a near optimal policy. Sana [49] proposed a model for a three-echelon supply chain with the objective of minimizing the total cost, from an integrated
perspective, among the supplier, the producer, and the buyer. They also considered imperfect quality of deliveries in the
analysis. A three-echelon supply chain with similar characteristics was modeled by Ben-Daya et al. [50]. Some other authors
that have recently presented integrated inventory system approaches are Rezaei and Davoodi [51] and Song et al. [52].
The literature on inventory management models incorporating transportation costs in decision-making has been often
focused on single-stage environments. Baumol and Vinod [53] proposed an inventory theoretic model integrating transportation and inventory costs. Unit shipping cost, lead time and variance of lead time are considered in their model. Das
[54] and Buffa and Reynolds [55] also considered transportation costs in their inventory models. Lee [56] extended the classical economic order quantity (EOQ) model with a setup cost that included a xed cost and a freight cost with quantity discounts. Swenseth and Godfrey [57] discussed the need for incorporating transportation cost in inventory replenishment
decisions and developed a freight rate function, which could be incorporated in inventory replenishment decision models.
According to the authors, 50% of the total annual cost of a product is attributed to transportation. Rieksts and Ventura
[58] investigated models with truckload (TL) and less than truckload (LTL) transportation costs and derived optimal policies
615
for both innite and nite planning horizons that allow a combination of the two transportation modes in lieu of using a
unique option. Toptal [59] formulated a general replenishment cost model that included stepwise freight costs and
all-units quantity discounts, and proved useful properties that were used to develop an efcient approach to nd optimal
order quantities. Zhang et al. [60] generalized the standard newsboy model to include a freight cost proportional to the number of containers used, and showed that the optimal order quantity was either the newsboy solution or a multiple of the
containers capacity. Zhang and Zhang [61] developed a mathematical model to implement the newsvendor inventory model
for a single rm with a xed selection cost and limitations on the minimum and maximum order size under stochastic
demand. Some other recent multi-stage systems have been analyzed. For instance, a single-manufacturer, multi-buyer
inventory system was taken into account by Hoque [62], while Chen and Sarker [63] analyzed the opposite case of single
buyer and multiple vendors using a single truck to collect the products from the vendors in a milk run. Their model tackled
minimization of product weights and traveling distances.
Mendoza and Ventura [8,64] studied the case of a single manufacturer and multiple suppliers, and proposed an MINLP
model for order quantity allocation and supplier selection, considering inventory and transportation costs simultaneously.
In their 2009 paper, LTL shipments were assumed and two existing continuous functions were used to estimate the actual
freight rates of the shipments. In their 2013 paper, actual transportation costs were considered and modeled. While both
papers consider the issue of order quantity allocation in supplier selection as well as inventory and transportation costs
simultaneously, they do not consider multiple echelons. The present article proposes an integrated approach that addresses
the issue of supplier selection and order quantity allocation in a multi-stage serial supply chain system with multiple suppliers, considering inventory and transportation costs simultaneously.
3. Transportation cost analysis for full truck load (TL)
TL, LTL, and small packages are three common modes of freight transportation [65]. For the TL case, the shipment cost is
independent of the size of the shipment for a given truck. For example, given a truck with a capacity of c units and cost of f
(dollars) for a given trip, any shipment size less than or equal to c units will be charged a at rate f. Partial loads are charged
at full load capacity. TL transportation cost is a piecewise constant and discontinuous function [65,66].
We will illustrate the cost function with an example. Table 1 shows freight rates from a source (manufacturer) to a destination (warehouse), for four different vehicle types. Under this freight rate structure, any shipment between 1 and 250
units is charged at a rate of $1393 for vehicle type 1, between 251 and 500 units is charged at a rate of $2200 for vehicle
type 2, and so on. Note that, if x vehicles of type i are used, then the total cost for the shipment is fi x.
In order to select the best vehicle combination for a particular shipment, suppose an order quantity Q of a certain type of
raw material or product needs to be transported between two consecutive stages of the supply chain. There are n types of
commercial vehicles (cargo vans and trucks) that can be used for transportation. Let ci be the capacity of a truck type i and f i
be the transportation cost of the truck for a round-trip between the two stages, for i = 1, . . ., n. It is assumed that Q, ci and f i
are positive integers, 1 c1 < c2 < < cn , and 1 f 1 < f 2 < < f n . Note that, if ci < ci1 and f i f i1 , then truck type i
can be eliminated from consideration, because truck type i + 1 has more capacity and the same or lower cost. It is also
assumed that Q > c1 ; if not (Q c1 ), a single truck of type 1 would solve the problem. Note also that f i =ci is the cost of
one unit of capacity of truck type i, as shown in the last column of Table 1. The objective of this problem is to determine
the commercial vehicles to be used in this trip that minimize the total transportation cost. This vehicle selection problem
can be formulated as a minimization version of the knapsack (KS) problem as follows:
KS minimize TC
n
X
f i xi ;
i1
subject to
n
X
c i xi Q ;
i1
xi 0;
integer; i 1; . . . ; n;
where xi is the number of type i commercial vehicles used in the trip.

This is a pseudo-polynomial problem that can be easily solved by dynamic programming (DP) [67]. The problem can be
modeled using a forward DP formulation. Let f(q) be the optimal value function representing the minimum transportation
n
o
cost when the order quantity is q units. Let r arg min cf ii : i 1; . . . ; n be the truck type with the smallest unit capacity
Table 1
TL freight rates (example).
Vehicle type (i)
Vehicle capacity, ci (in units)
Shipment cost, fi (in $)
Unit cost, fi/ci
1
2
3
4
250
500
1000
2000
1393
2200
2570
3521
5.57
4.40
2.57
1.76
(van)
(van)
(truck)
(truck)
616
cost. Note that UB f r ^

xr is an upper bound on the optimal solution (minimum transportation cost), where ^
xr dQ =cr e is the
number of type r commercial vehicles used. Note that this upper bound actually solves for three special cases: (a) n = 1, (b)
cr = 1, and (c) Q is a multiple of cr. The following DP formulation for the KS problem can be used to solve the general problem.
Step 1 (Initialization):
Set f q 0 , q c1 cn 1; :::; 0 ,
and f q f 1 , q 1; :::; c1 .
Step 2 (Recurrence Relation):
Compute f q min ff i f q ci : i 1; :::; ng; for q c1 1; :::; Q .
Step 3 (Answer):
Minimum Transportation Cost: TC f Q .
If we solve the example problem given in Table 1 using the DP formulation, we obtain the results provided in Table 2 for
different ranges of Q. These results are plotted in Fig. 1.
For cases where Q is very large, Theorem 1 provides conditions under which it possible to solve the problem without rst
computing f(q), for c1 + 1 < q < Q.
Theorem 1. In the vehicle selection problem, let r arg min
fi
ci
: i 1; . . . ; n
, cr P 2, and ci cn , i = 1, . . ., n 1. Then,
(a) Vehicle type r is guaranteed to be part of every optimal solution, for all qPq00 , when vehicle type r is in all optimal
solutions for q00 , q00 + 1, . . ., q00 + cn 1 (i.e., cn consecutive values of q).
(b) An optimal solution for any order quantity Q, Q P q00 , can be found by assigning s cr units to s vehicles type n, where
00
s bQ qcr 1c, and the remaining Q s cr units are assigned to the vehicles determined in the calculation of
f (Q s cr).
Proof. The proof of part (a) is similar to the one presented in Section 8.3 of Dreyfus and Law [67] for a maximization version
of the problem.
Concerning part (b), an optimal solution for an order quantity Q, Q P q00 , can be determined by rst identifying the
smallest integer s such that Q s cn < q00 , which is equivalent to
satises
Qq00
cr
Q q00
cr
< s . Since cr P 2, nding the smallest integer s that

j
k
00
< s can be expressed in the following closed-form expression: s Q qcr 1 . Thus, in any optimal solution, s
Table 2
Results from the DP formulation (example).
Q (units)
TC ($)
Solution
(x1 , x2 , x3 , x4 )
1250
251500
5011000
10012000
20012250
22512500
25013000
30014000
1393
2200
2570
3521
4914
5721
6091
7042
(1,
(0,
(0,
(0,
(1,
(0,
(0,
(0,
0,
1,
0,
0,
0,
1,
0,
0,
0,
0,
1,
0,
0,
0,
1,
0,
0)
0)
0)
1)
1)
1)
1)
2)
Fig. 1. TL transportation cost function (example).
617
vehicles of type r are assigned to the rst s cr units. The vehicles for the remaining units, Q s cr, are determined by
solving the reduced problem. h
Now, let us consider the example given in Table 1 with Q = 23,000 units. Note that n = r = 4 and cn = cr = 2000. As shown in
Table 2, commercial vehicle type 4 is part of the optimal solution for 1001 6 q 6 4000, which includes the cases
q = 1001, 1002, ..., 3000 (i.e., 2000 consecutive values of q). Therefore, based on Theorem 1, q = 1,001 units and
s b23;00010011
c 11 vehicles. Now, we solve the problem with the remaining quantity: 23,000 11 2000 = 1000 units.
2000
From Table 2, the optimal solution for an order quantity of 1,000 units is (0, 0, 1, 0). Thus, the optimal solution for
Q = 23,000 units is (0, 0, 1, 11) commercial vehicles. The minimum transportation cost is f (23,000) = 2570 + 11
3521 = $41,301.
In some applications, the number of commercial vehicles of each class may be limited. If this is the case, the DP formulation can be generalized (see [68,69]. In addition, TL cost structures can be adopted for small packages and LTL as well.
Package carriers use a xed charge for packages up to certain number of units (or lbs), like in the TL model. Similarly, the
LTL cost structure can be approximated by dening different ranges of the order quantity and by assigning a xed charge
for each range.
4. Integrated inventory model with supplier selection
The proposed inventory replenishment model with supplier selection in a serial supply chain structure, incorporating transportation cost is presented in this section. Let J = {1, 2,. . ., nJ} be the set of stages in a serial supply chain. We
dene the stages as follows: the manufacturer is in stage 1 followed by a set of intermediary stages 2 {2, 3. . ., nJ 1},
and the distribution center (DC) is in stage nJ. The intermediate stages can be additional manufacturing or warehousing facilities. Stage 0 is the supplier stage and stage nJ + 1 represents the retailers. Note that we optimize the decisions between stages 1 and nJ, with supplier selection at stage 0. Let R = {1, 2, . . ., nR} be the set of suppliers,
C = {1, 2, . . ., nC} the set of retailers, and K = {1, 2, . . .,nK} the set of vehicles with different carrying capacities. Fig. 2
shows the serial supply chain network with three main stages (nJ = 3), multiple suppliers at stage 0 and retailers
at stage 4.
The retailers face demand at a constant rate per time unit, which is supplied by the distribution center (DC) at the last
stage. Thus, the demand rate at the DC (stage nJ) is the sum of the demand rates for all the retailers. The manufacturing plant
procures the raw material required for production from the set of selected suppliers. Inventory at the manufacturing plant is
replenished periodically from the selected suppliers. Since the nature of the supplier selection problem is intrinsically
multi-criteria problem [28], the selection process considers different criteria into the analysis, namely capacity and quality.
The product ows from the manufacturing plant through a series of distribution stages, managed by a single decision
maker, i.e., centralized control system. There may be more than one facility within the manufacturing plant for processing.
In this case, each facility would be a separate stage of the supply chain. Inventory at any stage in the supply chain is replenished by its immediate upstream stage. The raw material from suppliers to the manufacturing stage and products from the
manufacturer to the other stages in the supply chain are carried out by a set of vehicles, with different carrying capacities. A
piecewise constant setup cost structure arising from a full truckload freight rate cost structure is considered to model transportation costs.
The proposed model is developed to: (i) select the appropriate suppliers, and determine the number of orders and replenishment quantities allocated to the selected suppliers per order cycle, (ii) determine the replenishment order quantities for
the remaining stages of the supply chain, and (iii) determine the type and number of vehicles needed to transport materials
between stages. The objective of the model is to coordinate the inventory from the point of supply to the point of consumption by minimizing the total cost per time unit associated with the entire system. The total cost includes purchasing, setup,
echelon inventory holding, and transportation costs between the stages.
Fig. 2. A serial supply chain structure with multiple suppliers and retailers.
618
The list of parameters, decision variables, and cost function components for the model are provided below:
Parameters
Demand for all retailers (units) per time unit
d
ci
Capacity of ith supplier (units) per time unit, for i 2 R
qi
Perfect rate of ith supplier, for i 2 R
qa
Minimum acceptable perfect rate
cvk
Capacity of vehicle type k (units), for k 2 K
Decision variables
Qj
Order quantity for stage j (units), for j 2 J
Ji
Number of orders to the ith supplier per order cycle, for i 2 R
NVik Number of vehicles of type k used to transport an order from supplier i to manufacturer, for i 2 R, k 2 K
SVjk Number of vehicles of type k used to transport an order from stage j 1 to j, for j 2 J {1}, k 2 K
T
Time between consecutive orders (time units)
Tc
(Repeating) order cycle time (time units)
Nj
Multiplicative factor for order quantity from stage j to j + 1, for j 2 J {nJ}
M
Number of orders allocated to all suppliers per order cycle
Cost function components
k1i
Setup cost at stage 1 for placing an order to ith supplier ($/order), for i 2 R
kj
Setup cost at stage j ($/order), for j 2 J {1}
pi
Unit price of ith supplier ($/unit), for i 2 R
fsik
Transportation cost per vehicle of type k from supplier i to the manufacturing plant ($/vehicle), for i 2 R, k 2 K
fcjk
Transportation cost per vehicle of type k to transport an order from stage j 1 to stage j ($/vehicle), for j 2 J {1},
k2K
hj
Conventional (unit) holding cost at stage j per time unit ($/unit/time unit), for j 2 J
ej
Echelon (unit) holding cost at stage j per time unit ($/unit/time unit), for j 2 J
Since products increase in value as they move forward in the supply chain system, unit holding costs also increase, i.e.,
h1 < h2 < . . . < hnJ . In the proposed model, echelon holding costs (ej) are used instead of conventional holding costs (hj).
The echelon holding cost (ej) is dened as the increase in unit holding cost between stages j-1 and j:
e1 h1 ; e2 h2 h1 ; . . . ; enJ hnJ hnJ 1 . Considering conventional holding costs, the inventory at a stage j depends on the
order quantities of its downstream stages. When echelon holding costs are used, the inventory at a stage j depends only
on the order quantity of that stage. Using echelon holding costs rather than conventional holding costs simplies the analysis. A detailed discussion of why echelon holding costs are used can be found in Zipkin [70], and Mendoza and Ventura [47].
It can easily be shown that the total inventory holding cost per time unit is the same whether using echelon or conventional
holding costs.
Note that, since the demand rate is constant, the time between consecutive orders to the selected suppliers becomes
P
P
T = Q1/d, and the length of the repeating order cycle can be dened as T c T i2R J i , where i2R J i is the total number
of orders placed to the selected suppliers. We will refer to Tc as the order cycle from now on in this article.
The following assumptions are considered in the development of the proposed model:
i. Demand rate per time unit for the retailers is constant and no shortages are allowed. This assumption is reasonable
given that the proposed model is intended to serve as a strategic decision-making model.
ii. The production rate of the system is assumed to be greater than or equal to the cumulative demand rate for all
retailers.
iii. The size of the order quantity from suppliers to the manufacturer is constant. For an N-stage serial system, an optimal
policy must be nested and the inventory replenished only when the inventory level is zero [71,72]. A nested policy is
one in which a stage and its downstream stages order at a same time i.e., Qj = Nj Q(j+1), j 2 J fnJ g, where Nj is a positive
integer; If a policy is not nested, shortages might occur. The zero-inventory ordering policy implies that stage j should
replenish its inventory only when the inventory level reaches zero and it is time to supply Qj+1 units to its downstream
stage, i.e., stage j + 1. A constant order quantity is necessary to generate a nested and a stationary inventory policy
(allowing coordination) in the serial supply chain system.
iv. Suppliers can have nite capacity. Some suppliers may have nite capacity rates, but the total capacity rate should be
less than or equal to the demand rate for all suppliers.
v. The production process of suppliers is imperfect and, thus, defective parts could be produced. The combination of the
selected suppliers should meet the minimum acceptable perfect rate imposed by the manufacturer.
vi. The manufacturer has an innite production rate.
619
vii. Sufcient number of vehicles of each type is available for transporting goods. We assume that transportation between
the stages is done using third party logistics carriers (3PLs). The rms usually have contracts with a primary carrier
and one or two back-up carriers to transport goods. If the primary carrier is out of capacity, back-up carriers are used.
In the current business scenario, with sufcient number of 3PLs, the assumption on sufcient number of vehicles is
valid.
viii. Transfer times of products between the suppliers and the manufacturing stage are considered to be very small (or
insignicant). Thus, in-transit inventory is not taken into account in the model. However, the proposed model is still
valid when transfer times are deterministic and the agreement with the suppliers is freight-on-board (FOB) destination, freight collect [73]. In this case, the model can still be used by increasing the unit price of each supplier as follows: Dpi = tx0i h1, where tx0i is the deterministic transfer time from the ith supplier to the manufacturing stage and
Dpi is the increase in the unit price of ith supplier ($/unit). This increase in price corresponds to the unit holding cost
during the transfer time applied to the manufacturing stage. When the agreement with the suppliers is
freight-on-board (FOB) destination, freight prepaid [73], the in-transit inventory costs are paid by the suppliers and
are already included in the unit price.
When transfer times between consecutive stages are deterministic, the model can also be used because in-transit
inventory costs per time unit are constant costs that do not alter the optimal solution. Note that the in-transit inventory cost per time unit between stages j and j + 1 is txj hj+1 d, which is the holding cost of d units for txj time units
applied to stage j + 1.
ix. Supplier lead-times are presumed to be insignicant. However, when lead-times are positive (deterministic), the
model can still be used by ordering from suppliers at the corresponding reorder points. Reorder points can be calculated as follows. Let li be the lead-time of ith supplier and zi the reorder point of ith supplier. Then,
zi =(d li) (wi Q1), where wi is the largest integer less than or equal to li/T1. The value wi represents the number
of orders that will be received before the order that is submitted at the current reorder point. If the lead-time is less
than the time between consecutive orders (T1), then wi = 0.
x. Retailers use their own means to transport materials from the DC to their locations. Thus, transportation costs
between the DC and retailers are not in our model. Note that, if the distributor were responsible for the transportation
costs, then a vehicle routing problem could be solved at the tactical level to generate optimal distribution plans.
4.1. Proposed mixed integer nonlinear programming model
Considering the purchasing, setup, holding, and transportation cost components, the problem can be formulated as an
MINLP, where the objective is to minimize the total cost per time unit:
M1 minimize Z
X kj 1 X
d X
dX
J k1i d
ej Q j
J i pi
Q 1 M i2R i
2
M
Q
j2Jf1g j
j2J
i2R
X X fcjk SV jk
d XX
fs NV ik J i d
;
Q 1 M k2K i2R ik
Qj
k2K j2Jf1g
subject to d J i ci M; 8 i 2 R;
X
J i qi qa M;
i2R
Q1
cv k NV ik ;
5
6
8 i 2 R;
k2K
Qj
cv k SV jk ;
8 j 2 J f1g;
k2K
Q j N j Q j1 ;
X
J i M;
8 j 2 J fnJ g;
9
10
i2R
Q j 0; 8 j 2 J;
J i 0 and integer; 8 i 2 R;
11
12
NV ik 0 and integer; 8 i 2 R; 8 k 2 K;
13
Nj 1 and integer; 8 j 2 J fnJ g;
14
SV jk 0 and integer 8 j 2 J f1g; 8 k 2 K:
15
Notice that the number of orders allocated to all suppliers in one order cycle (M) that minimizes Z may result in an excessively large order cycle time. In this case, a company may be interested in restricting M to a reasonable small value to reduce
the entire order cycle time. Short cycle times simplify the inventory planning process and facilitate the interaction with suppliers. To reduce the entire order cycle time, an upper bound on M can be added or M can be xed to a small integer value.
620
Besides controlling the length of the order cycle, setting M to a xed value simplies the mathematical model since the variables in the objective functions denominator are reduced to just order quantities, and the feasible region is further restricted,
as the value of M creates implicit upper bounds on the Ji variables.
The terms in the objective function (4) of Model (M1) are described as follows. The rst term represents the setup cost at
P
stage 1, which is obtained by dividing the total setup cost per order cycle,
i2R J i k1i , by the length of the cycle,
P
T c Q 1 =d i2R J i Q 1 =d M . The second term represents the setup cost for stages j 2 J f1g. Since, the order quantity
at these stages is different, the setup cost is obtained by multiplying the setup cost at stage j, kj, by the total number of orders
per time unit, d=Q j . The third term accounts for the total holding costs per time unit. At stage j, this is expressed as the average inventory at that stage, Q j =2, times the echelon holding cost at that stage, ej, and is then summed over all stages j 2 J.
P
P
The fourth term is the total purchasing cost of raw materials, where i2R J i pi = i2R J i , represents the average price of a purchased unit for all suppliers. The fth term corresponds to the transportation cost between suppliers and the manufacturer,
P P
which is obtained by dividing the total transportation cost per order cycle, k2K i2R fsik NV ik J i , by the length of the order
P
cycle, T c Q 1 =d i2R J i . The last term corresponds to the transportation cost between stages. The transportation cost at
P
these stages is obtained by dividing the total transportation cost at one specic stage, k2K fcjk SV jk , by the length of the cycle
at that stage, d=Q j . Therefore, the transportation cost at each stage is summed over the stages j 2 J f1g to obtain total transportation cost at the stages.
Constraint set (5) ensures that the rate at which raw materials are ordered from each supplier does not exceed its capacity. The term d J i =M , in the constraint, represents the proportion of demand per time unit that is allocated to the ith supplier,
which is constrained by its maximum offering capacity, ci. Constraint (6) ensures that the minimum acceptable perfect rate is
P
met by the selected set of suppliers. The term i2R J i qi =M represents the average perfect rate offered by the suppliers, which
is required to be greater than or equal to the minimum acceptable perfect rate, qa.
Constraint set (7) ensures that the appropriate number of vehicles is selected to transport materials between the supplier
P
stage and the manufacturer. The term k2K cv k NV ik represents the total capacity of vehicles allocated to transport raw materials, which should be greater than or equal to the order quantity at stage 1, Q1. However, for some supplier i with J i 0, this
constraint would give positive values for NV ik . But, the vehicles assigned to the ith supplier will never be used and the
cost components related to these vehicle allocations in the objective function will be zero. So, we could conclude that the
vehicles NV ik will not be used if J i 0.
Constraint set (8) ensures that the appropriate number of vehicles is selected to transport materials to stage j. The right
P
hand side term, k2K cv k SV jk , represents the capacity of vehicles allocated to transport parts to stage j, which has to be
greater than or equal to the order quantity at that stage j 2 J f1g.
Constraint set (9) represents the nested inventory constraint. This constraint ensures that the order quantity at stage j is
either equal to or a positive multiple of the order quantity at stage j + 1. Constraint set (10) ensures that the number of orders
sum up to M. Finally, the sets of inequalities and integral conditions (11)(15) describe the nature of the variables considered
in the model.
5. Illustrative example
In this section, an illustrative example of a serial supply chain system with ve stages (nJ 5), and four potential suppliers is presented and analyzed. The capacity, setup cost, unit cost, and perfect rate of each supplier used in the illustrative
example are shown in Table 3. The setup and unit echelon holding costs for each stage are shown in Table 4.
Stage 5 expects a demand rate of d = 30,000 units/month. Raw materials required at the manufacturing stage can be procured from the four potential suppliers. The manufacturer has established a minimum perfect rate of 0.95. This implies that
the mix of raw materials procured from selected suppliers at each time unit should have an average perfect rate of at least
0.95.
Four types of vehicles with different capacities are available. TL transportation costs for all four types of vehicles between
suppliers and the manufacturer are provided in Table 5.
Notice that supplier 3 offers the lowest transportation costs for the different types of vehicles. In addition, it has the lowest setup cost (see Table 3). However, supplier 3 also offers the highest unit procurement cost. This prevents a dominated
solution in the analysis of the problem. The transportation costs between stages (for each vehicle type) are shown in Table 6.
5.1. Optimal M value
The illustrative example was solved using LINGO 13.0 with the global optimizer on a PC with INTEL(R) Core (TM) 2 Duo
Processor at 2.8 GHz and 2.0 GB RAM. We rst run the model with M as a decision variable (M 1 and integer). Then, the
model was run again and analyzed with different values of M (as input).
By treating M as a decision variable, the problem was solved with a run time limit of 8 h. The optimal total cost per time
unit for the problem was $1,851,468/month, with M = 36. This corresponds to an order cycle length at stage 1of 9.60 months.
Then, the problem was run by specifying M as an input (for values of M = 425, and 36) with a run time limit of 1 h (see
Table 7). In most of the instances, the problem converged to the global optimum solution. For instances in which the solution
did not converge to the global optimum within the run time limit, the best objective and the lower bound values given by
621

Table 3
Data for suppliers (illustrative example).
Supplier i
Capacity (ci) (units/month)
Setup cost (k1i) ($)
Unit cost (pi) ($/unit)
Perfect rate (qi)
1
2
3
4
10,000
14,000
12,000
9000
3200
4600
2900
2300
32
47
54
49
0.96
0.93
0.97
0.95
Table 4
Cost data for stages (illustrative example).
Stage j
Setup cost (kj) ($)
Echelon holding cost

(ej) ($/unit/month)
1
2
3
4
5
1780
820
360
280
4
15
25
27
31
Table 5
TL transportation costs between suppliers and manufacturer (illustrative example).
Vehicle type k
1
2
3
4
(van)
(van)
(truck)
(truck)
Capacity (cvk) (units)
250
500
1000
2000
Transportation cost from supplier i to the manufacturer ($/vehicle)

Supplier 1 (fs1k)
Supplier 2 (fs2k)
Supplier 3 (fs3k)
Supplier 4 (fs4k)
2825
4412
6046
8278
3500
6007
9600
14,250
2189
3500
4,039
5533
3150
5407
8640
13,268
Table 6
TL transportation cost between stages (illustrative example).
Vehicle type k
1
2
3
4
(van)
(van)
(truck)
(truck)
Capacity (cvk) (units)
250
500
1000
2000
Transportation cost from stage j-1 to j ($/vehicle)

Stage 12 (fc2k)
Stage 23 (fc3k)
Stage 34 (fc4k)
Stage 45 (fc5k)
1393
2200
2570
3521
1393
2200
2570
3521
1393
2200
2570
3521
1393
2200
2570
3521
LINGO were used in the analysis (see Table 7). Notice that when M is 12, the order cycles length at stage 1 is 3.20 months and
the optimal cost is $1,851,468/month. When M is 24, the order cycles length is two times greater (6.40 months), yet the optimal cost remains the same ($1,851,468/month). The optimal costs for other values of M between 13 and 23 are greater than
$1,851,468/month. Based on Theorem 1 [47], there exists an optimal solution where Jis are a set of relatively prime numbers
(see solution for M = 12 in Table 7).
Note that when M = 6, the optimal solution is $1,851,503/month. This is higher than the optimal solution when M = 12 by
less than 0.1%. But, when M = 6, the order cycle time is reduced by half (1.60 months). Using a lower M simplies the analysis
and the inventory management process. In the present analysis, and for simplications purposes, we x M to 12 (the least M
optimal value). As observed in Table 7, this is an optimal solution to the problem (same as when M = 24 and 36). The solution
for M = 12 is highlighted in Table 7. In this solution, four type 4 vehicles (capacity of 2000 units) are used for transporting raw
materials from the suppliers to the manufacturer. Also, one vehicle type 4 is used for transportation between the rest of the
stages.
In the optimal solution with M = 12, Supplier 3, with the lowest setup and transportation costs has only been assigned
25% of the total orders. Supplier 2, with perfect rate of 0.93, below the minimum acceptable perfect rate, and second lowest
unit cost, has the highest percentage allocation of orders (41.67%). However, Supplier 2 offers the highest capacity. Supplier
1, with the lowest unit cost has been assigned 33.33% of the total number of orders, utilizing 100% of its capacity. From this
analysis, it is evident that the characteristic of the optimal solution for this problem depends upon the capacity, perfect rate,
unit cost, setup and transportation costs.
5.2. Solution without transportation cost + actual transportation cost
In order to analyze the effect of optimizing inventory and transportation costs simultaneously versus separately, in
this section we perform a comparison of these two approaches. The problem is rst solved without considering the
622
Table 7
Coordinated inventory policy for different values of M.
G represents that Global solution.

#
Bounds achieved after 1 h of run time.
transportation cost. Let this optimal cost of the model be Z s and their corresponding order quantities be Q j . Then,
based on the cost data given in Tables 5 and 6, we determine the transportation cost and the number of vehicles
using the DP formulation presented in Section 3. The actual transportation cost related to Q j is determined, using
Eq. (16):
XX
Q 1 M k2K i2R
fsik NV ik J i d
X X fcjk SV jk
;
Q j
k2K j2Jf1g
16
Where fsik is the transportation cost from supplier i to the manufacturer for using a vehicle of type k for transporting Q 1 , and

NV ik is the number of vehicles of type k used for transporting Q 1 from supplier i to the manufacturer. fcjk and SV jk are the
corresponding values for transporting materials from stage j 1 to stage j, for j 2 J f1g .
Table 8 shows the results for the illustrative example solved without considering transportation cost.
Based on the order quantities, the DP formulation is used to determine the appropriate type of vehicle and the number of
vehicles. The results are shown in Table 9.
The total transportation cost per time unit calculated using Eq. (16) is $537,878.05/month. Thus, the solution to problem
without transportation cost + actual transportation cost is equal to $1,994,928.05/month (i.e., 1,457,050.00 + 537,878.05).
Table 10 shows the comparison between the model with transportation cost and the model without transportation cost + actual transportation cost.
The analysis shows that the proposed model considering inventory and transportation cost simultaneously yields a savings of 7.75%. Therefore, one of the advantages of optimizing inventory and transportation costs simultaneously is that it
results in savings regarding the total cost per time unit.
Table 8
Solution to problem without transportation cost.
Z s ($/month)
1,457,050
Order allocation (number of orders)
Order quantity (units)
J1
J2
J3
J4
Q1
Q2
Q3
Q4
Q5
8447
2816
1408
704
704
623

Table 9
Vehicle allocation to each stage based on Q.
Stage
Supplier
Supplier
Supplier
Supplier
Stage 2
Stage 3
Stage 4
Stage 5
1
2
3
4
Q (units)
Solution (x1 , x2 , x3 , x4 ) (vehicles)
8447
8447
8447
8447
2815
1408
704
704
(0,
(0,
(0,
(0,
(0,
(0,
(0,
(0,
1,
1,
1,
1,
0,
0,
0,
0,
0,
0,
0,
0,
1,
0,
1,
1,
4)
4)
4)
4)
1)
1)
0)
0)
Table 10
Solution to the problem with and without transportation costs.
Scenario
Objective function
value ($/month)
% deviation = [(2) (1)]/(1) 100
Solution to problem considering inventory + transportation costs simultaneously (1)

Solution to problem without transportation cost + actual
transportation cost (2)
1,851,468.00
1,994,928.05
7.75%
6. Sensitivity analysis
In order to analyze the effect of changes in echelon holding, setup and transportation costs between stages over raw
material supply decisions, 36 different cost scenarios are studied. All these scenarios are run with M = 12. The changes in
cost parameters are expressed as percentage of their respective original values presented in Tables 36. The levels considered under each factor are shown in Table 11.
The cost scenarios are studied under two cases. In case 1, only the supplier stage costs are varied (9 scenarios). In case 2,
only the costs at the remaining stages, i.e., manufacturer, warehouse, and DC, are varied (27 scenarios) in order to study the
effects on supplier selection decisions.
6.1. Case 1: Varying costs at the supplier stage alone (stage 0)

Under this case, the setup costs at the suppliers and the transportation costs from the suppliers to the manufacturer are
varied. The naming convention used represents the cost scenarios as follows: S{set up cost level number}_T{transportation
cost level number}. For example, the cost scenario with 50% setup cost, and 100% transportation cost is represented as S1_T2.
The effect on the optimal costs, supplier selection, and order quantity allocation are discussed. Table 12 summarizes the
results.
The following are the inferences from the analysis (see Fig. 3):
1. The average order quantity allocated to the selected suppliers increases with an increase in setup costs at the supplier
stage (lower setup cost, smaller order quantity). The average order quantity with 50% set-up cost is 5333.33 units, with
100% set-up cost is 8000 units, and is 9333.33 with 150% set-up cost.
2. The average order quantity decreases with increase in transportation costs. The average order quantity for 50%, 100% and
150% transportation costs are 8000, 7333.33, and 7333.33, respectively.
The capacity utilization of the suppliers is given in Table 13.
Table 11
Cost parameters levels for the sensitivity analysis.
Cost parameter (label)
Levels* (as a % of the original values)
Echelon holding cost (E)

Setup cost (S)
Transportation cost (T)
50, 100, 150

50, 100, 150
50, 100, 150
The levels are coded with level numbers: 50% using 1, 100% using 2, and 150% using 3.
624
Table 12
Results for case 1.
Cost
scenario
J1
S1_T1
S1_T2
S1_T3
S2_T1
S2_T2
S2_T3
S3_T1
S3_T2
S3_T3
4
4
4
4
4
4
4
4
4
J2
4
5
5
4
5
5
4
5
5
J3
2
3
3
2
3
3
2
3
3
J4
2
0
0
2
0
0
2
0
0
Order cycle at stage 1

(months)
Order quantity at stage j (units)

Q1
Q2
Q3
Q4
Q5
2.40
2.40
1.60
3.20
3.20
3.20
4.00
3.20
4.00
6000
6000
4000
8000
8000
8000
10000
8000
10000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
Z
($/month)
Lower
bound
CPU time
(s)
1,763,347
1,842,832
1,919,066
1,771,680
1,851,468
1,927,066
1,778,916
1,858,421
1,933,848
1,473,694#
G
G
1,643,001#
G
G
G
G
G
3600
101
38
3600
499
62
3301
20
55
G represents the Global solution.

Fig. 3. Case 1: average order quantity at supplier stage.
Based on the average capacity utilization levels, the capacity of supplier 1 is fully utilized in all cost scenarios. Allocation
to supplier 1 is constrained primarily by its capacity. To illustrate this, let us consider a case with supplier 1 having capacity
of 30,000 units (equal to the demand/time unit) in the S2_T2 scenario. The number of orders to supplier 1 is 12, to supplier 2
is 3 and to supplier 4 is zero. All the demand is allocated to supplier 1, because of its economic advantage (it meets the perfect rate and offers the lowest unit cost).
Supplier 2 is the next best supplier with 83.34% average utilization followed by supplier 3 and then by supplier 4. As discussed earlier, with original cost structure, 41.67% of the raw material orders are allocated to supplier 2.
Out of the nine scenarios, there are two different supplier selection solutions in the analysis. Solution 1 results from the
cost scenarios shown in Table 14. Fig. 4 depicts the percentage allocation to suppliers under solution 1.
Similarly, solution 2 results from the cost scenarios shown in Table 15. Fig. 5 depicts the percentage allocation to suppliers under solution 2.
Table 13
Case 1: percent capacity utilization.
Cost scenario
Supplier 1 (%)
Supplier 2 (%)
Supplier 3 (%)
Supplier 4 (%)
S1_T1
S1_T2
S1_T3
S2_T1
S2_T2
S2_T3
S3_T1
S3_T2
S3_T3
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
100.00
71.43
89.29
89.29
71.43
89.29
89.29
71.43
89.29
89.29
41.67
62.50
62.50
41.67
62.50
62.50
41.67
62.50
62.50
55.56
0.00
0.00
55.56
0.00
0.00
55.56
0.00
0.00
Average
100.00
83.34
55.56
18.52
625
Table 14
Cost scenarios under solution 1.
Solution
Setup cost (S)
50
100
150
50
50
50
Fig. 4. Percentage allocation of orders to suppliers under solution 1.
Based on the case 1 analysis, we observe that the change in transportation cost has an effect on supplier selection.
Solution 1 contains all the scenarios when the transportation costs are set at 50% from the base cost. Under these cost settings, supplier 4 has some economic advantage and therefore appears in the allocation. However, when the transportations
costs are set at 100% and 150% levels, supplier 4 has 0% allocations. In both solutions, supplier 1 has 33.33 % allocations indicating that this supplier is the best.
6.2. Case 2: Varying costs at the remaining stages (manufacturer, warehouse, and distribution center)
Under this case, the echelon holding costs and setup costs at the stages, and transportation costs between stages (other
than the suppliers and manufacturer stage) are varied. The naming convention used to represent the cost scenarios are as
follows: E{echelon holding cost level number}_S{set up cost level number}_T{transportation cost level number}. For example, the cost scenario with 100% echelon holding cost, 150% setup cost, and 100% transportation cost is represented as
E2_S3_T2. A summary of the results for the 27 different cost scenarios is shown in Table 16.
The following are the inferences from the analysis (see Fig. 6):
1. The higher the set-up costs at the stages, the higher the average order quantity at the suppliers. The average order quantity with 50% set-up costs is 7777.78 units, and 8666.67 units with 100% and 150% setup costs.
2. The higher the holding costs, the lower the average order quantity.
3. The higher the transportation costs, the higher the average order quantity.
Table 15
Solution
Setup cost (S)
50
50
100
100
150
150
100
150
100
150
100
150
626
Fig. 5. Percentage allocation of orders to suppliers under solution 2.
Table 16
Case 2 results.
Cost
scenario
J1
E1_S1_T1
E1_S1_T2
E1_S1_T3
E1_S2_T1
E1_S2_T2
E1_S2_T3
E1_S3_T1
E1_S3_T2
E1_S3_T3
E2_S1_T1
E2_S1_T2
E2_S1_T3
E2_S2_T1
E2_S2_T2
E2_S2_T3
E2_S3_T1
E2_S3_T2
E2_S3_T3
E3_S1_T1
E3_S1_T2
E3_S1_T3
E3_S2_T1
E3_S2_T2
E3_S2_T3
E3_S3_T1
E3_S3_T2
E3_S3_T3
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
4
J2
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
4
4
4
4
4
4
4
4
4
J3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
J4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
2
2
2
2
2
2
2
2
Order cycle at stage 1

(months)
Order quantity at stage j (units)

Q1
Q2
Q3
Q4
Q5
3.20
4.00
4.00
4.80
4.80
4.80
4.80
4.80
4.80
3.20
3.20
3.20
3.20
3.20
3.20
3.20
3.20
3.20
2.40
2.40
2.40
2.40
2.40
2.40
2.40
2.40
2.40
8000
10,000
10,000
12,000
12,000
12,000
12,000
12,000
12,000
8000
8000
8000
8000
8000
8000
8000
8000
8000
6000
6,000
6,000
6,000
6,000
6,000
6,000
6,000
6,000
4000
2000
2000
4000
4000
4000
4000
4000
4000
2000
2000
2000
2000
2000
2000
4000
4000
4000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
1000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
1000
2000
2000
1000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
2000
1000
2000
2000
1000
2000
2000
2000
2000
2000
Z
($/month)
Lower
bound
CPU time
(s)
1,665,333
1,769,386
1,874,986
1,682,322
1,787,982
1,893,582
1,699,947
1,805,607
1,911,207
1,721,508
1,827,168
1,932,768
1,745,808
1,851,468
1,957,068
1,765,083
1,870,743
1,976,343
1,762,026
1,882,536
1,988,136
1,791,576
1,906,836
2,012,436
1,825,476
1,931,136
2,036,736
G
G
G
G
1,697,706#
1,770,295#
G
G
G
G
G
G
G
G
G
G
1,802,763#
G
G
G
G
G
G
G
G
G
G
63
182
188
10
3600
3600
78
8
2853
122
184
5
61
6
396
12
3600
10
195
52
130
609
194
50
30
41
122
G represents that Global solution.

There were two different supplier selection solutions in the analysis. The two solutions obtained from these scenarios are
shown in Tables 17 and 18. The percentage allocation to the suppliers under solutions 1 and 2 are same as those presented in
Figs. 4 and 5, respectively.
Solutions 1 and 2 for case 2 are similar to that of case 1. However, the cost scenarios under each solution differ. Nine scenarios are under solution 1 and the remaining 18 scenarios are under solution 2 (see Tables 17 and 18). Under solution 1, the
percentage capacity utilization of supplier 1 is 100%, supplier 2 is 89.29%, supplier 3 is 62.50%, and supplier 4 is 0%. Under
627
Fig. 6. Case 2: average order quantity at stages.
Table 17
Solution
Setup cost (S)
150
150
150
150
150
150
150
150
150
50
50
50
100
100
100
150
150
150
50
100
150
50
100
150
50
100
150
Solution
Setup cost (S)
50
50
50
50
50
50
50
50
50
100
100
100
100
100
100
100
50
50
50
100
100
100
150
150
150
50
50
50
100
100
100
150
50
100
150
50
100
150
50
100
150
50
100
150
50
100
150
50
Table 18
solution 2, the percent capacity utilization of supplier 1 is 100%, supplier 2 is 71.43%, supplier 3 is 41.67%, and supplier 4 is
55.56%.
Solution 1 contains all the cost scenarios when the echelon holding costs are set at 150% from the base cost. Under this
setting, the order quantity at stage 1 is 6000 units and supplier 4 has some economic advantages due to having lower setup
and transportation costs. Therefore, supplier 4 appears in the allocation. For order quantities greater than 6000 units,
628
Table 19
Cost data for stages (setup and echelon holding cost).
Stage j
Setup cost ($), kj
Holding cost ($/unit/moth), hj
Echelon holding cost ($/unit/month), ej
1
2
3
4
5
5340
2460
1080
840
20.00
21.50
24.00
26.70
29.80
20.00
1.50
2.50
2.70
3.10
Table 20
Transportation cost between stages (for each vehicle type).
Vehicle type, k (capacity in units), cvk
1
2
3
4
Transportation cost from stage j 1 to j (cost in $/vehicle)
(250)
(500)
(1,000)
(2,000)
Stage 12 (fc2k)
Stage 23 (fc3k)
Stage 34 (fc4k)
Stage 45 (fc5k)
4179
6600
7710
10,563
4179
6600
7710
10,563
4179
6600
7710
10,563
4179
6600
7710
10,563
Table 21
Scenarios for integrated vs. sequential analysis.
Scenario
Echelon holding cost at stage 1 (%)
Echelon holding cost at the stages 25 (%)
1
2
3
4
100
200
100
200
100
100
200
200
supplier 4 has 0% allocations, as it is not benecial. Based on the analysis of case 2, we observe that the supplier selection and
order quantity allocation is affected by the variations in stage costs.
7. Comparing integrated and sequential approaches
Raw material procurement and production/distribution decisions could be modeled using either a sequential or an integrated approach [74]. So far, we have followed an integrated approach to solve the numerical example problem. Now, let us
discuss the difference between the integrated and sequential approaches.
The sequential approach seeks to solve the supplier selection (raw material procurement) and the production/distribution
in two consecutive stages. Assuming innite supply capacity of raw materials available to the manufacturer, the production/distribution problem (manufacturer to DC) is optimally solved in the rst stage. Then, in the second stage, the optimal
order quantity at the manufacturer stage is considered as an input to the supplier selection problem, where the optimal
order quantity allocation to the supplier stage is determined, and supplier selection problem is solved. The optimal solution
for the problem is the sum of optimal solutions of the two sub-models. The sub-models (sub-model 1 and sub-model 2) for
the sequential analysis are given in Appendix A.
The integrated approach seeks to optimally solve both the supplier selection and production/distribution decision problems simultaneously, considering the entire supply chain (from suppliers to customers). The optimal solution from the
sequential approach is always an upper bound for the integrated approach (Z seq Z int ).
In order to illustrate the integrated and sequential approaches, let us consider 4 scenarios. The (base) data set for the analysis is given in Tables 19 and 20. The supplier data in terms of capacity, setup cost, unit cost, perfect rate, and transportation
costs between suppliers and the manufacturer are the same considered in the illustrative example in Section 5 (refer to
Tables 3 and 5).
The four scenarios studied in this section are shown in Table 21. The changes in cost parameters are expressed as percentage of their respective values from the base data set.
In addition to these 4 scenarios, the analysis is also carried out for the illustrative example. The results of the analysis are
given in Tables 22 and 23. The raw material cost is subtracted from the total cost to show the savings in logistics cost at the
stages. Raw material costs include the purchasing cost, set up cost, transportation cost from the suppliers to the manufacturer, and the holding cost at the manufacturers stage. Raw material cost in the integrated approach is always equal to or
less than the raw material cost in the sequential approach. This is because the integrated approach solves the problem optimally. In order to compare the savings in logistics costs between the sequential and the integrated approach, the raw material cost in the integrated approach is calculated and this is subtracted from the total costs of the corresponding approaches.
Scenario
1
2
3
4
Illustrative
example
*
Sequential analysis cost ($)

Sub-model
(S2) (Supplier
stage)
Sub-model (S1)
(Manufacturer to
DC stage)
Total cost
(Sequential)
1,630,654.00*
1,790,654.00*
1,592,972.00*
1,712,972.00*
1,493,608.00*
701,017.50
701,017.50
731,180.00
731,180.00
357,860.00
2,331,671.50
2,491,671.50
2,324,152.00
2,444,152.00
1,851,468.00
Integrated
analysis cost
($)
%
Savings
in total
cost
(1) Raw material

purchase cost in
integrated analysis ($)
(2) Raw material

purchase cost in
sequential analysis ($)
Minimum of
{1,2} ($)
2,253,756.00
2,296,923.00*
2,276,623.00
2,316,623.00
1,851,468.00*
3.46
8.48
2.09
5.50
0.00
1,541,976.00
1,570,642.00
1,530,642.00
1,570,642.00
1,493,608.00
1,630,654.00
1,790,654.00
1,592,972.00
1,712,972.00
1,493,608.00
1,541,976.00
1,570,642.00
1,530,642.00
1,570,642.00
1,493,608.00
Logistics cost ($)

Integrated
analysis
Sequential
analysis
7,11,780.00
7,26,281.00
7,45,981.00
7,45,981.00
3,57,860.00
7,89,695.50
9,21,029.50
7,93,510.00
8,73,510.00
3,57,860.00
Global optimum solutions achieved within one hour of run time; Note: All other values except * are values obtained after solving the model with one hour of run time limit.
% Savings
in
logistics
cost
10.95
26.81
6.37
17.10
0.00
Table 22
Results of the integrated vs. sequential analysis.
629
630
Scenario
1
2
3
4
Illustrative example
Order quantity at stage j in the integrated

approach (units)
Order allocation to the suppliers

in the integrated approach
Order quantity at stage j in the sequential

approach (units)
Order allocation to the suppliers in

the sequential approach
Q1(int)
Q2(int)
Q3(int)
Q4(int)
Q5(int)
J1(int)
J2(int)
J3(int)
J4(int)
Q1(seq)
Q2(seq)
Q3(seq)
Q4(seq)
Q5(seq)
J1(seq)
J2(seq)
J3(seq)
J4(seq)
6000
4000
4000
4000
8000
6000
4000
4000
4000
2000
6000
4000
4000
4000
2000
6000
4000
4000
4000
2000
6000
4000
2000
2000
2000
4
4
4
4
4
4
4
4
4
5
2
2
2
2
3
2
2
2
2
0
16,000
16,000
12,000
12,000
8000
16,000
16,000
12,000
12,000
2000
8000
8000
4000
4000
2000
4000
4000
4000
4000
2000
4000
4000
4000
4000
2000
4
4
4
4
4
5
5
5
5
5
3
3
3
3
3
0
0
0
0
0
Table 23
Order quantities for the integrated and sequential approaches.
631
The results on the four scenarios indicate that there is signicant savings in total cost using the integrated approach (see
Table 22). On an average, there is a 4.88% savings in the total cost and 15.31% savings in the logistics cost at the stages.3
However, the total cost in the illustrative example remains the same for both sequential and integrated analysis.
Notation
Q jseq
Optimal order quantity at stage j under the sequential approach, for j 2 J
Q jint
Optimal order quantity at stage j under the integrated approach, for j 2 J
J iint
J iseq
ZQ 2seq
Number of orders to the ith supplier per order cycle in the integrated approach, for i 2 R
Number of orders to the ith supplier per order cycle in the sequential approach, for i 2 R
Logistics cost at stages under the sequential approach (solution of Sub-model (S1))
ZQ 2int
Corresponding logistics cost at stages under the integrated approach
7.1. Conditions under which both integrated and sequential approaches give the same cost
Q 2seq Q 2int : Both sequential and integrated approaches give the same optimal cost. Solving the Sub-model (S1) gives a
feasible solution to the integrated model. When this order quantity to the manufacturer is the same as that of the integrated
approach, Sub-model (S2) yields optimal supplier selection and order quantity allocation to the supplier stage (as we are
giving a portion of the optimal solution to Sub-model (S2)). This can be observed from results of the illustrative example
in Table 23. In the illustrative example, both Q 2seq and Q 2int are 2000. Under this condition, the coordination constraint
does not restrict the solution of the Sub-model (S2), and the optimal cost in sequential approach is the same as the integrated
approach.
7.2. Conditions under which integrated approach is better than the sequential approach
Q 2seq Q 2int : Integrated approach is better than the sequential approach. This situation arises when the optimal order
quantity to the manufacturer stage in the integrated approach is not equal to the optimal order quantity to manufacturer
stage in the sequential approach. In such cases, Sub-model (S2) gives a sub-optimal solution. This increases the cost of
the entire supply chain under the sequential approach when compared to the integrated approach. This is seen from the
solutions for the scenarios 1 through 4 in Table 22. In all the four scenarios, Q 2seq Q 2int and the integrated approach yields
a better solution than the sequential approach (see Table 22). The suppliers selected under the two approaches are also different (see Table 23).
When Q 2seq Q 2int , the optimal cost given by Sub-model (S1) in the sequential approach is lower than the corresponding
logistics cost of the integrated approach. For example in scenario 1, logistics cost of the sequential approach is $701,017.50/month (i.e., 2,331,671.501,630,654.00) and the logistics cost of the integrated approach $711,780.00/month (i.e., 2,253,756.
001,541,976.00): if Q 2seq Q 2int ; then ZQ 2seq < ZQ 2int : This is because Sub-model (S1) solves the manufacturer to DC
stage problem optimally without considering the cost associated with the supplier stage. However, the optimal solution of
Sub-model (S2) will be worse than the corresponding supplier stage cost in the integrated approach. Thus, the total supply
chain costs are lower in the integrated approach and the savings in the total supply chain costs are higher than the savings in
the logistics cost at the stages.
8. Conclusions and future research
In todays increasingly cost-conscious and competitive market, companies strive to manufacture and distribute products
in an efcient manner. In this article, we have studied the impact of transportation costs on supplier selection, inventory
replenishment, and distribution decisions in a serial supply chain. We have proposed an MINLP model to determine the optimal inventory levels for the stages in the supply chain, allocation of orders among the suppliers at the initial stage, and shipments from the suppliers to the manufacturer and between the stages using truckloads. Since the supplier selection problem
is intrinsically multi-criteria in nature, our model minimizes the cost criterion in the objective while simultaneously considering quality as a goal constraint, specifying a minimum acceptable perfect rate from the selected suppliers, and capacity
constraints.
An illustrative example of a ve-stage serial supply chain with four suppliers has been solved to show the cost savings
obtained when transportation costs are included in the objective function. Moreover, sensitivity analysis was performed
by varying the holding, setup, and transportation costs. The results show that these cost parameters have a considerable
impact on the selection of suppliers and order quantity allocation to suppliers.
3
Logistics cost at the stages = Total cost of the supply chain (purchasing cost + set up cost at suppliers + transportation costs between suppliers and the
manufacturer + holding costs at the manufacturer stage).
632
We have also studied the benets of the integrated approach over the sequential approach for solving this problem. The
results indicate that the integrated approach may generate substantial cost savings over the sequential approach, improving
the supply chain efciency. Out of the four scenarios studied, the integrated approach yielded, on average, savings of 4.88%
on total cost and 15.31% in logistics costs over the sequential approach. We have also provided the conditions under which
the integrated approach is more benecial than the sequential approach.
Scope for future work includes analyzing a multi-product system with joint replenishment costs and quantity discounts
from suppliers in a general multi-stage network with a single or multiple nodes (facilities) at each stage, where the objective
is to determine a schedule of orders for each product over an innite horizon that minimizes the total cost per time unit. In
this case, an inventory policy could be applied, in which each product uses a stationary interval of time between orders and
the ratio of order intervals for a product at different nodes or for any two different products at the same node is an integer.
When the integers are restricted to be powers of two, this type of policy is called power-of-two [75].
The model could also be extended by considering multiple objectives, such as supply chain risk (disruptions during the
transfer of products due to uncontrollable events, uncertain supply yields, uncertain supply lead-times, etc.), and different
service levels from suppliers.
Appendix A
For the sequential approach the original model is split down into two sub-models. In Sub-model (S1), we consider the
stages from the manufacturer to the DC and solve the production/distribution problem. In this sub-model, we assume innite supply of raw materials from the suppliers.
Sub-model (S1): manufacturer to DC stages
S1 minimize Z d
subject to Q j
X kj 1 X
X X fcjk SV jk
ej Q j d
;
2
Q
Qj
j2Jf1g j
j2Jf1g
k2K j2Jf1g
cv k SV jk ;
8 j 2 J f1g;
A:1
A:2
k2K
8 j 2 J f1; nJ g;
A:3
Q j Nj Q j1 ; 8 j 2 J f1; nJ g;
Q j 0; 8 j 2 J f1g;
A:4
A:5
Q j Nj Q j1 ;
SV jk 0 and integer; 8 j 2 J f1g;
8 k 2 K:
A:6
The objective function (A.1) for Sub-model (S1) includes the setup cost at the stages, inventory holding cost at the stages,
and transportation cost between the stages (note that the stages in this sub-model refer to the entities from the manufacturer to the DC). Constraint set (A.2) corresponds to the vehicle capacity allocation constraint between the stages. Constraint
set (A.3) is the nested inventory constraint between the stages and constraint set (A.4) ensures that the multiplicative factor
in the nested inventory constraint is a positive integer. The set of inequalities (A.5) and (A.6) describes the nature of variables
considered in the sub-model.
Order quantity to the manufacturer, Q2, is given as an input to the second Sub-model (S2). In Sub-model (S2), we only
consider the supplier stage and solve the order quantity to the suppliers, supplier selection and order allocation problem.
Sub-model (S2): supplier stage
S2 minimize Z
X
d
1
dX
J k1i e1 N1 Q 2
Jp
N1 Q 2 M i2R i
2
M i2R i i
XX
d
fs NV ik J i ;
N1 Q 2 M k2K i2R ik
subject to d J i ci M; 8 i 2 R;
X
J i qi qa M;
i2R
N1 Q 2
cv k NV ik ;
A:7
A:8
A:9
8 i 2 R;
A:10
k2K
N1 1 and integer;
X
J i M;
A:11
A:12
i2R
J i 0 and integer; 8 i 2 R;
A:13
NV ik 0 and integer; 8 i 2 R; 8 k 2 K:
A:14
633
This sub-model is solved using Q2 as an input. Note that variable Q1 is eliminated and replaced by N 1 Q 2 as Q 1 N 1 Q 2 ,
based on the nested inventory constraint.
The objective function (A.7) of Sub-model (S2) comprises the setup cost at the stage 1 for placing orders with the suppliers, inventory holding cost for order quantity at the suppliers, purchasing cost of the raw materials, and transportation cost
between suppliers and the manufacturer. Constraint set (A.8) corresponds to the capacity constraints at the supplier stage.
Constraint (A.9) is the minimum acceptable perfect rate constraint for the suppliers. Constraint set (A.10) corresponds to the
transportation capacity constraints between suppliers and the manufacturer. Constraint (A.11) ensures that the multiplicative factor is a positive integer. Constraint (A.12) corresponds to the sum of orders at the suppliers summing up to M. The set
of inequalities (A.13) and (A.14) describes the nature of variables considered in the supplier stage sub-model.
The solution to the sequential approach is the sum of objective function values of Sub-models (S1) and (S2).
References
[1] W. Xia, Z. Wu, Supplier selection with multiple criteria in volume discount environments, Omega 35 (2007) 494504.
[2] S. Chopra, P. Meindl, Supply Chain Management: Strategy, Planning, and Operations, Prentice-Hall, Upper Saddle River, NJ, 2007.
[3] C. Araz, P.M. Ozrat, I. Ozkarahan, An integrated multi-criteria decision-making methodology for outsourcing management, Comput. Oper. Res. 34 (12)
(2007) 37383756.
[4] T.T. Burton, JIT/repetitive sourcing strategies: Tying the knot with your suppliers, Prod. Invent. Manage. J. 29 (4) (1988) 3841.
[5] M. Heberling, The dollar magnitude of purchases in American business and government, in: Proceedings of the Fourth Purchasing and Materials
Management Research Symposium (National Association of Purchasing Management), Tempe, AZ, 1990.
[6] C.A. Weber, J.R. Current, A multi-objective approach to vendor selection, Eur. J. Oper. Res. 68 (1993) 173184.
[7] R.H. Ballou, Business Logistics Management, Prentice-Hall, Englewood Cliffs, NJ, 1992.
[8] A. Mendoza, J.A. Ventura, Estimating freight rates in inventory replenishment and supplier selection decisions, Logist. Res. 1 (2009) 185196.
[9] E.D. Blumenfeld, D.L. Burns, J.D. Diltz, Analyzing trade-offs between transportation, inventory, and production costs on freight networks, Transp. Res.
Part B Method. 19 (1985) 361380.
[10] C.A. Weber, J.R. Current, W.C. Benton, Vendor selection criteria and methods, Eur. J. Oper. Res. 50 (1991) 218.
[11] Z. Degraeve, E. Labro, F. Roodhooft, An evaluation of supplier selection methods from a total cost of ownership perspective, Eur. J. Oper. Res. 125 (1)
(2000) 3458.
[12] L. De Boer, E. Labro, P. Morlacchi, A review of methods supporting supplier selection, Eur. J. Purchas. Supply Manage. 7 (2001) 7589.
[13] W. Ho, X. Xu, P.K. Dey, Multi-criteria decision making approaches for supplier evaluation and selection: a literature review, Eur. J. Oper. Res. 202 (2010)
1624.
[14] J. Chai, J.N.K. Liu, E.W.T. Ngai, Application of decision-making techniques in supplier selection: a systematic review of literature, Expert Syst. Appl. 40
(2013) 38723885.
[15] T.F. Anthony, F.P. Buffa, Strategic purchasing scheduling, J. Purch. Mater. Manage. 13 (1977) 2731.
[16] S.H. Ghodsypour, C. OBrien, A decision support system for supplier selection using an integrated analytic hierarchy process and linear programming,
Int. J. Prod. Econ. 5657 (1998) 199212.
[17] S. Talluri, R. Narasimhan, Vendor evaluation with performance variability: a maxmin approach, Eur. J. Oper. Res. 146 (3) (2003) 543552.
[18] W.L. Ng, An efcient and simple model for multiple criteria supplier selection problem, Eur. J. Oper. Res. 186 (3) (2008) 10591067.
[19] R. Narasimhan, K. Stoynoff, Optimizing aggregate procurement allocation decisions, J. Purchas. Mater. Manage. 22 (1986) 2330.
[20] S.S. Chaudhry, F.G. Forst, J.L. Zydiak, A multi-criteria approach to allocating order quantity among vendors, Prod. Invent. Manage. J. 32 (1991) 8285.
[21] R.G. Kasilingam, C.P. Lee, Selection of vendors: a mixed-integer programming approach, Comput. Ind. Eng. 31 (1) (1996) 347350.
[22] V. Jayaraman, R. Srivastava, W.C. Benton, Supplier selection and order quantity allocation: a comprehensive model, J. Supply Chain Manage. 35 (1999)
5058.
[23] S.H. Ghodsypour, C. OBrien, The total cost of logistics in supplier selection, under conditions of multiple sourcing, multiple criteria and capacity
constraint, Int. J. Prod. Econ. 73 (2001) 1527.
[24] R. Hammami, Y. Frein, A.B. Hadj-Alouane, An international supplier selection model with inventory and transportation management decisions, Flexible
Serv. Manuf. 24 (2012) 427.
[25] C.A. Weber, L.M. Ellram, Supplier selection using multi-objective programming: a decision support system approach, Int. J. Phys. Distrib. Logist.
Manage. 23 (1993) 314.
[26] C.A. Weber, J.R. Current, A. Desai, An optimization approach to determining the number of suppliers to employ, Supply Chain Manage. Int. J. 2 (2000)
9098.
[27] M. Kumar, P. Vrat, R. Shankar, A fuzzy goal programming approach for vendor selection problem in a supply chain, Comput. Ind. Eng. 46 (2004) 6985.
[28] V. Wadhwa, A.R. Ravindran, Vendor selection in outsourcing, Comput. Oper. Res. 34 (2007) 37253737.
[29] A. Mendoza, J.A. Ventura, An effective method to supplier selection and order quantity allocation, Int. J. Bus. Syst. Res. 2 (2008) 115.
[30] A.R. Ravindran, U. Bilsel, V. Wadhwa, T. Yang, Risk adjusted multi-criteria supplier selection models with applications, Int. J. Prod. Res. 48 (2010) 405
424.
[31] A. Amid, S.H. Ghodsypour, C. OBrien, A weighted maxmin model for fuzzy multi-objective supplier selection in a supply chain, Int. J. Prod. Econ. 131
(1) (2011) 139145.
[32] B. Feng, Z.-P. Fan, Y. Li, A decision method for supplier selection in multi- service outsourcing, Int. J. Prod. Econ. 132 (2011) 240250.
[33] A.J. Clark, H. Scarf, Optimal policies for multi-echelon inventory problem, Manage. Sci. 6 (1960) 475490.
[34] A.J. Clark, An informal survey of multi-echelon inventory theory, Nav. Res. Logist. Q. 19 (1972) 621650.
[35] A. Federgruen, P. Zipkin, Computational issues in an innite horizon, multi-echelon inventory model, Oper. Res. 32 (1984) 818836.
[36] C.C. Sherbrooke, Metric: a multi-echelon technique for recoverable item control, Oper. Res. 16 (1968) 122141.
[37] A. Svoronos, P. Zipkin, Estimating the performance of multi-level inventory systems, Oper. Res. 36 (1988) 5772.
[38] S. Axsater, Simple solution procedures for a class of two echelon inventory problems, Oper. Res. 38 (1990) 6469.
[39] S. Nahmias, S.A. Smith, Optimizing inventory levels in a two-echelon retailer system with partial lost sales, Manage. Sci. 40 (1994) 582596.
[40] G. Gallego, P. Zipkin, Stock positioning and performance estimation in serial production-transportation systems, Manuf. Serv. Oper. Manage. 1 (1999)
7788.
[41] F. Chen, Optimal policies for multi-echelon inventory problem with batch ordering, Oper. Res. 48 (2000) 376389.
[42] S.R.J. Daniel, C. Rajendran, A simulation-based genetic algorithm for inventory optimization in a serial supply chain, Int. Trans. Oper. Res. 12 (2005)
101127.
[43] H.L. Lee, C. Billington, Material management in decentralized supply chains, Oper. Res. 41 (1993) 835847.
[44] P. Glasserman, S. Tayur, A sensitivity analysis for base-stock levels in multi-echelon production-inventory systems, Manage. Sci. 41 (1995) 263281.
[45] S. Axsater, Approximate optimization of a two-level distribution inventory system, Int. J. Prod. Econ. 8182 (2003) 545553.
[46] P.V.R. Sethupathi, C. Rajendran, Optimal and heuristic base-stock levels and review periods in a serial supply chain, Int. J. Logist. Syst. Manage. 7 (2010)
133164.
634
[47] A. Mendoza, J.A. Ventura, A serial inventory system with supplier selection and order quantity allocation, Eur. J. Oper. Res. 207 (2010) 13041315.
[48] P.C. Yang, H.M. Wee, S. Pai, Y.F. Tseng, Solving a stochastic demand multi-product supplier selection model with service level and Budget constraints
using genetic algorithm, Expert Syst. Appl. 38 (12) (2011) 1477314777.
[49] S.S. Sana, A production-inventory model of imperfect quality products in a three-layer supply chain, Decis. Support Syst. 50 (2011) 539547.
[50] M. Ben-Daya, R. Asad, M.E. Seliaman, An integrated production inventory model with raw material replenishment considerations in a three layer
supply chain, Int. J. Prod. Econ. 143 (1) (2013) 5361.
[51] J. Rezaei, M. Davoodi, A joint pricing, lot-sizing, and supplier selection model, Int. J. Prod. Res. 50 (16) (2012) 45244542.
[52] D.-P. Song, J.-X. Dong, X. Jingjing, Integrated inventory management and supplier base reduction in a supply chain with multiple uncertainties, Eur. J.
Oper. Res. 232 (2014) 522536.
[53] W.J. Baumol, H. Vinod, An inventory-theoretic model of freight transport demand, Manage. Sci. 16 (7) (1970) 413421.
[54] C. Das, Choice of transport service: an inventory theoretic approach, Logist. Transp. Rev. 10 (1974) 181187.
[55] F.P. Buffa, J.I. Reynolds, The inventory-transport model with sensitivity analysis by indifference curves, Transp. J. 2 (1977) 8390.
[56] C.-Y. Lee, The economic order quantity for freight discount costs, IIE Trans. 18 (3) (1986) 318320.
[57] S.R. Swenseth, M.R. Godfrey, Incorporating transportation costs into inventory replenishment decisions, Int. J. Prod. Econ. 77 (2002) 113130.
[58] B.Q. Rieksts, J.A. Ventura, Optimal inventory policies with two modes of freight transportation, Eur. J. Oper. Res. 186 (2008) 576585.
[59] A. Toptal, Replenishment decisions under an all-units discount schedule and stepwise freight costs, Eur. J. Oper. Res. 198 (2) (2009) 504510.
[60] J.-L. Zhang, C.-Y. Lee, J. Chen, Inventory control problem with freight cost and stochastic demand, Oper. Res. Lett. 37 (6) (2009) 443446.
[61] J.L. Zhang, M.Y. Zhang, Supplier selection and purchase problem with xed cost and constrained order quantities under stochastic demand, Int. J. Prod.
Econ. 129 (1) (2011) 17.
[62] M.A. Hoque, Synchronization in the single-manufacturer multibuyer integrated inventory supply chain, Eur. J. Oper. Res. 188 (2008) 811825.
[63] Z.X. Chen, B.R. Sarker, Multi-vendor integrated procurement-production system under shared transportation and just-in-time delivery system, J. Oper.
Res. Soc. 61 (2010) 16541666.
[64] A. Mendoza, J.A. Ventura, Modeling actual transportation costs in supplier selection and order quantity allocation decisions, Oper. Res. Int. J. 13 (2013)
525.
[65] K.L. Croxton, B. Gendron, L.T. Magnanti, Models and methods for merge-in-transit operations, Transp. Sci. 37 (2003) 122.
[66] A. Cintron, A.R. Ravindran, J.A. Ventura, Multi-criteria mathematical model for designing the distribution network of a consumer goods company,
Comput. Ind. Eng. 58 (2010) 584593.
[67] S.E. Dreyfus, A.M. Law, The Art and Theory of Dynamic Programming, Academic Press, New York, NY, 1977.
[68] P.C. Gilmore, R.E. Gomory, The theory and computation of knapsack functions, Oper. Res. 14 (1966) 10451074.
[69] G.L. Nemhauser, Z. Ullmann, Discrete dynamic programming and capital allocation, Manage. Sci. 15 (1969) 494505.
[70] P. Zipkin, Foundations of Inventory Management, Irwin/McGraw-Hill, New York, NY, 2000.
[71] S.F. Love, A facilities in series inventory model with nested schedules, Manage. Sci. 18 (1972) 327338.
[72] L.B. Schwarz, L. Schrage, Optimal and system-myopic policies for multi-echelon production/inventory assembly systems, Manage. Sci. 21 (1975) 1285
1294.
[73] Hughes Network Systems, Routing guides: FOB terms, 2009. <http://www.hughes.com>.
[74] C.H. Fine, B. Golany, H. Naseraldin, Modeling trade-offs in concurrent product, process and supply chain design, J. Oper. Manage. 23 (2005) 389403.
[75] R. Roundy, 98%-effective lot-sizing rule for a multi-product, multi-stage production/inventory system, Math. Oper. Res. 11 (4) (1986) 699727.

08 16 A Serial Inventory System With Supplier Selection and Order Quantity Allocation Considering Transportation Costs

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

08 16 A Serial Inventory System With Supplier Selection and Order Quantity Allocation Considering Transportation Costs

Uploaded by

Copyright:

Available Formats

Applied Mathematical Modelling 40 (2016) 612634

Contents lists available at ScienceDirect

Applied Mathematical Modelling

A serial inventory system with supplier selection and order

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

where xi is the number of type i commercial vehicles used in the trip.

Vehicle capacity, ci (in units)

Shipment cost, fi (in $)

Unit cost, fi/ci

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

cost. Note that UB f r ^

< s . Since cr P 2, nding the smallest integer s that

Fig. 1. TL transportation cost function (example).

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Nj 1 and integer; 8 j 2 J  fnJ g;

SV jk 0 and integer 8 j 2 J  f1g; 8 k 2 K:

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Capacity (ci) (units/month)

Setup cost (k1i) ($)

Unit cost (pi) ($/unit)

Perfect rate (qi)

Setup cost (kj) ($)

Echelon holding cost

Capacity (cvk) (units)

Transportation cost from supplier i to the manufacturer ($/vehicle)

Capacity (cvk) (units)

Transportation cost from stage j-1 to j ($/vehicle)

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

G represents that Global solution.

Order allocation (number of orders)

Order quantity (units)

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Solution (x1 , x2 , x3 , x4 ) (vehicles)

% deviation = [(2)  (1)]/(1)  100

Solution to problem considering inventory + transportation costs simultaneously (1)

6.1. Case 1: Varying costs at the supplier stage alone (stage 0)

Cost parameter (label)

Levels* (as a % of the original values)

Echelon holding cost (E)

50, 100, 150

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Order cycle at stage 1

Order quantity at stage j (units)

G represents the Global solution.

Fig. 3. Case 1: average order quantity at supplier stage.

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Setup cost (S)

Transportation cost (T)

Fig. 4. Percentage allocation of orders to suppliers under solution 1.

Setup cost (S)

Transportation cost (T)

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Fig. 5. Percentage allocation of orders to suppliers under solution 2.

Order cycle at stage 1

Order quantity at stage j (units)

G represents that Global solution.

S. Pazhani et al. / Applied Mathematical Modelling 40 (2016) 612634

Fig. 6. Case 2: average order quantity at stages.

Nj 1 and integer; 8 j 2 J fnJ g;

SV jk 0 and integer 8 j 2 J f1g; 8 k 2 K:

% deviation = [(2) (1)]/(1) 100

Transportation cost from stage j 1 to j (cost in $/vehicle)

SV jk 0 and integer; 8 j 2 J f1g;