Anda di halaman 1dari 16

Applied Mathematical Modelling 39 (2015) 63596374

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

A comparative study of productioninventory model


for determining effective production quantity and safety
stock level
_
Glsen Aydn Keskin a,, Sevin Ilhan
Omurca b, Nursen Aydn c, Ekin Ekinci b
a

Department of Industrial Engineering, Kocaeli University, 41380 Kocaeli, Turkey


Department of Computer Engineering, Kocaeli University, 41380 Kocaeli, Turkey
c
Faculty of Engineering and Natural Sciences, Sabanc University, 34956 Istanbul, Turkey
b

a r t i c l e

i n f o

Article history:
Received 11 December 2012
Received in revised form 9 January 2015
Accepted 19 January 2015
Available online 11 February 2015
Keywords:
Safety stock level
Production capacity programming
Product capacity optimization
Mathematical modeling
Greedy algorithm
Genetic algorithm

a b s t r a c t
This study presents a comparative study to determine ideal stock levels of a multi-national
tire manufacturing company. The conventional inventory models can not be sufcient to
optimize the production, the inventory quantity and the backorder simultaneously. Therefore, it is not possible to obtain a production policy by considering these objectives for all
produced parts concurrently. In this paper, a production problem with three objectives is
solved with mathematical modeling, greedy algorithm and genetic algorithm considering
production constraints of a company. While existing inventory models based on conventional methods were applied for safety stock level determination, our proposed model uses
the mathematical programming based optimization methods based on mathematical programming. Furthermore, the production planning policy is obtained with the optimum
production amount and the stock is determined by considering the constraints dened
by the rm. Finally, in our numerical results, we compare each solution methodology with
respect to each objective criteria.
 2015 Elsevier Inc. All rights reserved.

1. Introduction
Production planning considers matching available capacity with the forecasted demand for varying customer orders from
3 to 18 months. Production planning aims to set the production level of all product categories to satisfy uctuating or uncertain demand while making decisions regarding the backorders and the inventory levels and dening suitable resources to be
used [1]. Under todays competitive conditions, customer satisfaction is one of the key points that should be addressed.
Hence, the objectives of production planning can be summarized as delivering orders at the time of the customers request,
organizing production and inventory levels parallel to sales level, optimum resource usage, and decreasing cost. Concordantly, backorder and ideal inventory level directly affects the optimum resource usage and decreasing cost/cost reduction.
The inventory theory assumes that there are only two possibilities for stocking out when a customer demand arrives.
Either the customer leaves the system cancelling the demand resulting in loss of prot, or the customer can wait until

Corresponding author at: Department of Industrial Engineering, Engineering Faculty, Kocaeli University, Umuttepe Campus, 41380 Kocaeli, Turkey.
Tel.: +90 262 3033326; fax: +90 262 3033003.
_
E-mail addresses: gulsenaydin@gmail.com, gaydin@kocaeli.edu.tr (G.A. Keskin), silhan@kocaeli.edu.tr (S. Ilhan
Omurca), nursenaydin@sabanciuniv.edu
(N. Aydn), ekin.ekinci@kocaeli.edu.tr (E. Ekinci).
http://dx.doi.org/10.1016/j.apm.2015.01.037
0307-904X/ 2015 Elsevier Inc. All rights reserved.

6360

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

the stock is available and then completes the transaction. This event is referred to as backorder [2]. When unsatised
demand occurs in real-life applications, we can generally observe that the customers refuse to wait for the backorder [3].
Therefore, the researchers have turned their attention to the models that do not allow backorders.
Organizations have to keep safety stock to avoid backorder in the cases of uncertain demand, deviation in lead time and
logistics, output gap and inaccurate knowledge [4],[5]. When the literature was examined regarding to these requirements, it
was observed that there have been several solutions for estimation of the safety stock level based on the inventory theory
from 1950s to the present [6]. Among the inventory problems, deterministic and stochastic models have been proposed. San
Jose et al. [7] studied an inventory model related to backorder. Rezaei and Davoodi [8] introduced imperfect items and storage capacity in the lot sizing with supplier selection problem and formulated the problem as a mixed integer programming
model to solve using genetic algorithm. Shah and Gor [9] aimed to minimize costs with the economic lot size model. Pentico
and Drake [10] proposed an approach to model economic order quantity with partial backorders. Chou et al. [11] proposed a
demand addicted unit cost and fuzzy economic order quantity model (EOQ) for an inventory problem with fuzzy constraints,
based on the maxmin operator. They reviewed their solution procedure by a method based on KuhnTucker approach to
point out their results. Maiti et al. [12] developed an inventory model based on stochastic lead time and price-addict demand
while considering advance payment. Chang and Lo [13] proposed an approach to overcome the shortcoming of traditional
methods for improving the continuous and discrete lead time with backorders. He et al. [14] developed productioninventory model for deteriorated items. Kang and Lee [15] presented an inventory model to minimize total cost that considers
warehouse capacity, revenue rate, and quantity discount for a certain product. Cardenas-Barron [16] proposed a practical
method for the economic production quantity and economic lot sizing models by using analytical geometry and algebra.
Cardenas-Barron et al. [17] presented an alternative heuristic algorithm to solve a vendor managed inventory system with
multi-product, multi-constraint based on EOQ with backorders considering linear and xed backorders costs. Omrani and
Keshavarz [18] developed an EOQ model to maximize the prot. Their model also determined the price, marketing cost
and lot sizing. The studies summarized above employ deterministic inventory model as the basis. Furthermore, numerous
studies based on stochastic inventory models exist in the inventory management literature. Moinzadeh and Aggarwal
[19] proposed an (s,S) production policy for such systems and developed expressions for the operating characteristics of
the system. Skouri and Papachristos [20] dened ve costs, which are deterioration, holding, backorder, opportunity and
renewal cost, to model continuous review inventory model. Ghalebsaz-Jeddi et al. [21] modeled a multi-item stochastic
inventory system with backorders when the estimation of marginal backorder cost is available, and payment is due upon
order arrival. Dutta et al. [22] presented a continuous review inventory system (Q,r) to nd reorder point and optimum order
quantity at fuzzy and random conditions. Al-Rifai and Rossetti [23] worked on the estimation of the reorder point at two
echelon inventory system with one warehouse and m identical retailers. Annadurai and Uthayakumar [24] worked on probabilistic inventory model with optimum backorder related to capital investment. Taleizadeh et al. [25] dened stochastic
replenishment arrivals at increased/decreased demand levels. Lee [26] studied an inventory problem for order and optimum
pricing decisions subject to service level for multi-priced two products. While the suggested analytical solutions in [2734]
can be easily applied, they failed at realistic supply chain problems that consider key features such as multi resource sharing
originating from multiple customers, capacity constraints and demands. Additionally, the safety stock leveling depending on
the probabilistic distribution of demand and production rate is insufcient in classical inventory models [35].
In this study, we consider a real life problem of a tire production company. The main objective is to develop a production
planning method which minimizes the amount of backorder and inventory by considering the production environment of
the company. This problem is modeled as a mixed integer linear programming (MILP) model and solved by an optimization
solver. Due to the computational time, we also propose efcient heuristic methods like genetic algorithm and greedy
algorithm and assessed the performance of each solution approach.
We make the following research contributions in this paper:
When inventory management literature is examined in depth, we observe that the production model which optimizes
production, inventory and backorder quantity together and achieves these objectives for all produced parts concurrently
has not been proposed [7,34,3639]. Rahmani et al. [40] studied a two-stage real world capacitated production system with
lead time and setup decisions. In their paper, they developed a mixed-integer programming (MIP) model to formulate the
robust production planning problem. Karimi-Nasab and Konstantaras [41] presented a new multi-objective production planning model for one product type. This model simultaneously optimizes three objectives by: (I) minimizing the total cost of
production plan, (II) minimizing the total variations in lot sizes, and (III) minimizing the distance of lot sizes to the customer
needs by mathematical model. Benjaafar et al. [42] considered the optimal control of a production inventory-system with
both backorders and lost sales of a single product. Mocquillon et al. [43] optimized production quantity and safety stock level
together. Sarkar [44] developed a productioninventory model in a two echelon supply chain management. Tian et al. [45]
proposed an iterative approach to jointly solve the problems of tactical safety stock placement and tactical production
planning. Zhou et al. [46] applied the joint replenishment strategy into an inventory system and builds a multi-product
multi-echelon inventory control model by using the genetic algorithm (GA) for solving the model. In brief, most of the
researchers studied only two concepts together as it is detailed above. Glasserman and Tayur [47] worked on the multi echelon production inventory systems considering base stock levels. They proposed a simulation based approach to estimate
inventory costs. In our study, we propose a mathematical programming model which optimizes production, inventory
and backorder quantities simultaneously for multi-product, multi period real-life problem.

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

6361

Contrary to existing studies, considered constraints affecting the production process is mandatory without any assumption and they are adopted from practical production environment.
The problem size we work in this paper is the largest among the studies we are aware of in the corresponding literature.
The original contribution of this paper is based on these points that mentioned above.
In this paper, besides a mathematical programming model, a greedy algorithm and a genetic algorithm are presented to
solve the production inventory problem in order to obtain alternative solutions. The computational results show that the
greedy algorithm performs better than the genetic algorithm and it is computationally more efcient compared to other
methods. The paper is organized as follows. In Section 2, the industrial problem is presented in detail. In Section 3, a mathematical model is presented in order to clearly state the constraints and objectives of the problem. In Section 4, the heuristics
with production constraints are presented. Finally, the experiments are evaluated in Section 5.

2. Industrial context
Brisa is the leading industrial conglomerate in Turkey, under the license agreement signed with American BF Goodrich
Company in 1974. The Company, a 100% Turkish investment, was named Lassa Tire Manufacturing and Trading Inc. at that
time. The production started in 1977 at the factory and the sales were extended to 60 provinces in Turkey through 186
organized dealers. Until 1988, the Company produced tires under the Lassa brand, expanding its product range from tires
for passenger cars, trucks and buses to farm and off-the-road vehicles.
In response to the developments in the world tire industry, a joint venture agreement was signed between the Bridgestone Corporation of Japan and the Sabanci Group. As a result of this agreement, the Company name was changed to BRISA
Bridgestone Sabanci Tire Manufacturing and Trading Inc. Today Brisa is the number one tire manufacturer in Turkey and the
sixth biggest tire producer in Europe [48].
The manufacturing process of the company consists of mainly six steps, which can be described as follows:
1. Rubber Mixing and Tread Stock Extruding Process: In a banbury mixer, various rubber compounds are mixed with
chemical agents to produce the type of material desired. This material is then sent to the next step in the manufacturing
process in the form of tread rubber or bead rubber. The tread rubber is shaped into a tire tread by extruder, and cut into
the size required by each type.
2. Rubber Coating and Cutting Process: Pre-heated and chemical treated nylon, rayon and polyester fabric cords and steel
cords are passed through the calendar which coasts both sides with carcass rubber. The rubber coated cord is cut and
jointed into one with a specied width and angle by a cutting machine and then sent to the next building process.
3. Bead Manufacturing Process: A predetermined number of high tensile steel bead wires are lined up to be coated with rubber while being extruded from the bead wire extruder. The rubber-covered bead wires are then formed by a bead former
into looped bead wires of the specic number of strands for a given size according to the nal use of the bead. After the
addition of a ipper and bead ller, the nished beads are sent to the building process.
4. Building Process: In the building process, a carcass (cord) and beads are afxed on a former to make what is called a green
case. Then, tread rubber is laminated onto the green case to make a green tire. The building process of radial tires is
basically the same as that of bias tires.
5. Vulcanizing (Curing) Process: Each green tire is then placed in a mould in which a given pattern has been engraved and
vulcanized for the required time at a specied pressure and temperature. The mould is heated by saturated steam to a
temperature ranging from 120 C to 170 C. Steam (at 150 C to 180 C) is also introduced into a rubber bag called a
bladder, located inside the green tire. Thus, the green tire is vulcanized from both inside and outside.
6. During the vulcanizing process, temperature, pressure, and vulcanizing time are automatically controlled by a fully
automatic vulcanizer. The vulcanized tires are then inated by post-cure inator and allowed to cool for a measured
time so that the desired tire dimension may be obtained after the rubber has set. The tires cooled by the post cure,
inator are sent to the nishing process for the removal of spews and other things and then onto the inspection
process.
7. Inspection Process: In the inspection process, the tires undergo various inspections such as external appearance inspection, uniformity inspection, and balance inspection, before being shipped.
The layout showing all parts of the factory is shown in Fig. 1. The material movements are shown with the help of arrows
and location of all parts.
The company produces 457 tire types but it is not possible to produce all of them simultaneously due to production
capacity and resource constraints. The company keeps inventory for all tire types at each month to prevent unexpected
backorder. However, this results in high holding cost. Therefore, a policy considering production, inventory and backorder
simultaneously will prevent high inventory and backorders costs.
A system providing a more efcient production and stock control model is aimed with the heuristic algorithms proposed
in the paper. For greedy and genetic algorithms, the safety stock levels of tires are required to be determined monthly by
calculating the production quantities. These approaches constitute acceptable and efcient solutions for the rms at different sectors because of their exibility and adaptivity.

6362

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

Fig. 1. Block plan and material ow schema of the company.

3. Mathematical programming model


The safety stock problem is a critical problem for a production process in different elds. The proposed stock control
system in this paper can be easily adapted to different companies with few modications. The scope of the study is shown
in Fig. 2.
In this section, we propose a mixed integer linear programming model (MILP) for safety stock problem in the tire manufacturing industry. First, we introduce the main points of the model and then, a detailed application is provided.
Due to the production of a wide variety of tires, the company encountered backorder and/or high inventory level. While
backorder will cause loss of prestige and hence, a possible prot loss, high inventory will require increasing the size of the

Mould capacity
Manufacturing capacity
Number of moulds

Producon
constraints

Keeping in stock
Backorder
Proposed
opmizaon
model

Number of machines
Product variaon

Monthly demand

Forecasngs

Fig. 2. The scope of the study.

Safety stock level

6363

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

storage area which may cause high holding cost. The company searches out an optimum solution for this potential situation.
In this problem, all constraints are determined with the production planning department of the company.
The company has to plan the production of 457 type tires denoted by n for 12-month period denoted by m. Due to the
machine and mould capacities and the company policy, at most 200 types of product can be produced in each month.
Production of each tire requires a different type of tire mould and manufacturing machine. Let w denote the tire mould
code and Ti denote the daily mould capacity. We also let Lm and Ni to denote the manufacturing machine type and associated
capacity, respectively. These variables are summarized in Table 1. Moreover, the available number of mould codes is limited.
We use Mw to present the number of available moulds which is presented in Table 2. Although the company has 527 different
moulds, it is stated that some of them are not used even if they are used in production of many tires.
For each tire, holding cost CLi and backorder cost CSi are dened and presented in Table 3.
The available number of tire manufacturing machines MAs is dened as in Table 4.
If we overproduce tires to prevent backorder, higher inventory may occur. On the other hand, producing tires regarding to
minimize holding costs may cause backorder. Our goal is to obtain a production policy that provides the optimal trade-off
between these costs.
Before proceeding to model formulation, we state our assumptions as follows. Lead time is ignored. Deteriorated products
are ignored. Transaction cost is not related to the quantity and product variation. Production demand is known through the
planning period.
Next, we present in detail the proposed mathematical formulation. Due to the company policy, we ignore the production
cost in this problem.
Parameters:
i = 1,. . .,n: number of tires
j = 1,. . ., m: month
w = 1,. . ., s: number of moulds
CLi, "i = 1,...,n: holding cost for one unit tire code i
CSi, "i = 1,...,n: backorder cost for one unit tire code i
Dj, "j = 1,...,m: working day in month j
Sij, "i = 1,...,n, "j = 1,...,m: demand of tire code i at month j
Ti, "i = 1,...,n: daily mould capacity of tire code i
Ni, "i = 1,...,n: daily manufacturing machine capacity of tire code i
Mw: number of mould code w
A: number of machine type A
B: number of machine type B
M: big number
Maxstock: maximum inventory level
Maxcode: number of tire type code at the same month

aik

if tire i uses machine type k; k 2 fA; Bg


0

otherwise

Variables:
Pij, "i = 1,...,n, "j = 1,...,m: the number of production for tire code i at month j
Eij, "i = 1,...,n, "j = 1,...,m: amount of inventory of tire code i at the beginning of month j
Bij, "i = 1,...,n, "j = 1,...,m: backorder of tire code i at month j


yij

if tire i is produced at month j;


0
otherwise

Table 1
Production constraints model for ideal inventory level.
Tire code

Mould code (w)

Daily mould capacity (Ti) (tire number/mould)

Tire manufacturing machine (Lm)

Tire manufacturing capacity (Ni)

X1
X10
X100
X101
..
.

K256
K291
K59
K501
..
.

94
90
101
101
..
.

B
B
B
A
..
.

693
693
728
728
..
.

6364

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374


Table 2
Resource capacity tables.
Mould code (w)

Mould number (Mw)

K1
K2
K3
K4
..
.

2
1
1
1
..
.

Table 3
Costs.
Tire code

Holding cost of one unit tire

Backorder cost of one unit tire

X1
X10
X100
X101
..
.

CL1
CL2
CL3
CL4
..
.

CS1
CS2
CS3
CS4
..
.

Table 4
Tire production machine constraints.
Tire manufacturing machine

Machine number (MAs)

A
B

11
39

Mathematical model:
P Pn
minimize m
j1
i1 CLi Eij CSi Bij
subject to:

Eij Eij1 Pij1  Sij1 Bij1


n
X
Pij =Dj =T i  6 M w

8i; j;

8j; w;

i1
n
X
aiA Pij =Dj =Ni  6 A 8j;
i1
n
X
aiB Pij =Dj =Ni  6 B 8j;
i1
n
X
Eij 6 Maxstock 8j;
i1
n
X
yij 6 Maxcode 8j;
i1

Pij 6 Myij

8i; j;

Eij ; Pij ; Bij 2 Z ; yij 2 f0; 1g:


Available mould numbers were dened for 527 different mould codes that were essential for tire production. Constraint
(1) gives the inventory balance equations of each product. Constraint (2) ensures that the daily total mould usage can
2  n  m m  s 4m not exceed the number of available mould. Tires are produced by two machines denoted as A and
B codes respectively. Eleven A machines and 39 B machines are available in the company. Constraint (3) and (4) guarantee

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

6365

that the daily total manufacturing machine usage respect to the manufacturing machine code cannot exceed the number of
available machines. These constraints derive from the limited mould and machine capacity of the tire production process.
Constraint (5) ensures that the total amount of inventory cannot exceed the capacity of storage area. Constraint (6) presents
that the total type of tires produced at each month cannot exceed the limit dened by the company. Finally, constraint (7)
guarantees that tire i can be produced if and only if the tire type is selected for the production at month j. This model involves
4  n  m variables and constraints. The objective function aims to minimize the total stock quantity and backorder.
Due to the integrality constraints, this problem is difcult to solve. Therefore, in the subsequent discussion, we focus on
the linear relaxation of this problem. By relaxing the integrality constraint on variables Eij ; P ij andBij , we obtain the MILP problem. We do not relax the constraint on yij to guarantee the total type of tires do not exceed the predened company limit. The
optimal solution of the resulting MILP can be non-integer. To obtain a feasible integer solution, we can simply round down
the non-integer values of inventory and production (Eij andPij ) and compute the integer value of backorder with respect to the
constraint (1).
^ij ; P
^ ij andB
^ ij be the optimal solution of its relaxation.
Let E ; P andB be the optimal values of the mathematical model and E
ij

ij

ij

ij ; P
 ij andB
 ij ) is feasible for the capacity constraints and balance constraint in the mathClearly the rounded down solution (E
ematical model. Let the objective function values of the mathematical model, MILP and rounded down solution are denoted
by z ; ^zandz, respectively. Then, we have;

z

m X
n
X
CLi Eij CSi Bij ;
j1 i1

^z

m X
n
X
^ij CSi B
^ ij ;
CLi E
j1 i1

z

m X
n
X
ij CSi B
 ij
CLi E
j1 i1

Since MILP is a partial relaxation of the mathematical model, we have ^z 6 z . Moreover, the rounded down solution is feasible for the mathematical model but not necessarily optimal, we have z 6 z. Consequently, we obtain ^z 6 z 6 z. As we shall
see in our computational study, the gap z  ^z turns out to be quite tight for the real life case study.
4. Numerical experiments
At this section, three approaches used in this paper and the obtained results are examined in depth.
4.1. Optimization model
The company has to produce 457 type tires in a 12 month period. The forecasted demand and the beginning inventories of
these tires are presented as in Table 5. Because of the restricted machine and mould capacities and the organization policy,
only 200 of the tires can be produced in a month. Due to the dened production constraints, the company encountered backorder and/or high inventory levels. However, the capacity of current warehouse is insufcient for high level of inventory.
Moreover, the company rejects backorder to prevent loss of prestige and hence, a possible prot loss. Therefore, they prefer
to hold maximum possible stock and this requires usage of large scale warehouse. If our proposed policy enables the company to minimize its backorder, the production planning department of the company will then propose an automation
investment in the upcoming production period. Otherwise, it will have to use the investment funds to acquire a new warehouse. In this section, safety stock problem at tire manufacturing industry is modeled as a MILP and solved by IBM ILOG
CPLEX to nd optimal production, backorder and inventory levels for each tire in 12 months period.
Resulting monthly production quantities, possible backorders and safety stock levels are presented in Tables 68,
respectively.
One of the challenging constraints in the model is the capacity constraint for the number of tires per month that have to
be produced. As a result of the application, 200 tires are determined with respect to the optimum production conditions.
The mathematical model has been solved by ILOG Cplex 10.0. The experiment has been run on an Intel Core 2 Quad
2.40 GHz with 4.00 GB RAM and it takes about 1205.94 s to solve the problem optimally for 457 products at 12-month period. For the larger problem sizes (e.g. production plan over a year), CPLEX cannot efciently be applied. Due to the long computation time required by CPLEX, in the next section, we propose a time partitioning heuristic to solve the problem
efciently with lower computation time while maintaining the quality of solution close to optimality.
4.2. Heuristic models
Greedy and genetic algorithms are proposed to obtain alternative solutions.

6366

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

Table 5
Forecasting demands and beginning inventories of tires.
Tire code

Beginning inventory

Jan

Feb

March

Apr

May

June

July

Aug

Sept

Oct

Nov

Dec

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99

1
790
1346
149
3384
115
3220
1887
..
.
748
60
283
1599
379
..
.
724
2699
143
2524
960

1
1580
2691
298
6768
229
6440
3774
..
.
1495
120
566
3197
757
..
.
1447
5397
285
5047
1920

0
152
4630
338
1745
429
3283
643
..
.
885
87
1603
4249
449
..
.
3506
696
2691
527
1485

0
54
4169
454
4736
455
11722
2122
..
.
2757
471
1788
4153
656
..
.
2431
5048
1723
3601
20

0
0
3617
289
4263
338
7088
1968
..
.
1650
314
1451
4624
408
..
.
3479
1434
2073
3534
20

0
0
3821
449
3842
353
10663
2666
..
.
1860
466
998
3896
648
..
.
2820
1406
1949
2571
2021

0
1501
2686
496
3414
239
11564
2948
..
.
1239
511
1139
4080
719
..
.
2507
1568
1180
3185
0

0
762
2717
398
6232
215
9417
2285
..
.
1076
418
543
3036
583
..
.
1723
1120
1206
2548
1448

0
2420
1567
242
276
59
5678
3593
..
.
792
270
505
2301
350
..
.
1442
914
635
383
0

0
11653
3048
606
1515
761
14015
2766
..
.
1098
615
552
3570
896
..
.
2355
2830
1110
2269
0

0
9102
2575
491
1119
215
11125
2259
..
.
947
506
685
3033
722
..
.
1654
3
1030
3178
0

0
6768
2496
446
3144
331
10339
2064
..
.
851
462
550
2764
652
..
.
1916
2201
784
2456
0

0
4230
2088
493
744
162
11407
2933
..
.
840
510
905
2149
732
..
.
1379
1821
766
2160
0

Table 6
Production quantities.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99

0
790
1345
0
3384
0
3220
1887
..
.
747
0
283
1598
378
..
.
723
2698
0
2523
960

0
0
4630
0
1745
0
3283
643
..
.
885
0
1603
4249
0
..
.
3506
696
2691
0
1485

0
0
4169
0
4736
0
11722
2122
..
.
2757
0
1788
4153
656
..
.
2431
5048
1723
3601
0

0
0
3617
0
4263
0
7088
1968
..
.
1650
0
1451
4624
0
..
.
3479
1434
2073
3534
0

0
0
3821
0
3842
0
10663
2666
..
.
1860
0
998
3896
648
..
.
2820
1406
1949
2571
2021

0
1501
2686
0
3414
0
11564
2948
..
.
1239
0
1139
4080
719
..
.
2507
1568
1180
3185
0

0
762
2717
398
6232
215
9417
2285
..
.
1076
418
0
3036
583
..
.
1723
1120
1206
2548
1448

0
2420
1567
0
0
0
5678
3593
..
.
792
0
0
2301
0
..
.
1442
914
635
0
0

0
11653
3048
0
1515
0
14015
2766
..
.
1098
0
0
3570
896
..
.
2355
2830
1110
2269
0

0
9102
2575
0
1119
0
11125
2259
..
.
947
0
0
3033
0
..
.
1654
0
1030
3178
0

0
6768
2496
0
3144
0
10339
2064
..
.
851
0
0
2764
0
..
.
1916
2201
784
2456
0

0
4230
2088
0
744
0
11407
2933
..
.
840
0
905
2149
732
..
.
1379
1821
766
2160
0

4.2.1. Greedy algorithm


Greedy algorithm is generally used at scheduling problems for production planning. In this paper, we adapt the algorithm
to an inventory and a backorder problem in order to obtain the desired production level considering the dened inventory
and backorder constraints. As mentioned in Section 3, certain types of the tires have to be produced in a year period and due
to the machine and mould capacity limits, all of them cannot be produced in a month. This case may result in an excessive
amount of inventory or backlogged demand. This is a typical trade-off problem. Due to the mould and machine capacity
restrictions, maximum 200 tires have to be produced in a month. On the other hand, a warehouse capacity is limited to
800,000 tires.
Different than the proposed mathematical model, greedy algorithm decides the production plan for each month. Since it
does not consider the following months, this approach presents a myopic solution. Greedy algorithm starts with the rst
month. While designing an algorithm, rst 200 tires with the highest backorder cost are considered for production due to
the nature of greedy algorithm. By controlling the remaining mould and machine capacities, we choose the 200 tires for production. We determine their production amount with respect to the demand of each tire for that month. Then, we compute
the amount of inventory and backorder for all tires. Next months production is planned by considering the rst 200 tires

6367

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374


Table 7
Backorders.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

0
0
0
149
0
114
0
0
..
.
0
60
0
0
0
..
.
0
0
142
0
0
13395

0
152
0
338
0
429
0
0
..
.
0
87
0
0
449
..
.
0
0
0
527
0
32612

0
54
0
454
0
455
0
0
..
.
0
471
0
0
0
..
.
0
0
0
0
20
37267

0
0
0
289
0
338
0
0
..
.
0
314
0
0
408
..
.
0
0
0
0
20
57703

0
0
0
449
0
353
0
0
..
.
0
466
0
0
0
..
.
0
0
0
0
0
32160

0
0
0
496
0
239
0
0
..
.
0
511
0
0
0
..
.
0
0
0
0
0
49409

0
0
0
0
0
0
0
0
..
.
0
0
543
0
0
..
.
0
0
0
0
0
125682

0
0
0
242
276
59
0
0
..
.
0
270
505
0
350
..
.
0
0
0
383
0
50088

0
0
0
606
0
761
0
0
..
.
0
615
552
0
0
..
.
0
0
0
0
0
139806

0
0
0
491
0
215
0
0
..
.
0
506
685
0
722
..
.
0
3
0
0
0
72210

0
0
0
446
0
331
0
0
..
.
0
462
550
0
652
..
.
0
0
0
0
0
101950

0
0
0
493
0
162
0
0
..
.
0
510
0
0
0
..
.
0
0
0
0
0
51740

Table 8
Safety stock levels.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
3641

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
962

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
2170

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
1877

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
2659

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
52574

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
22836

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
4332

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
2279

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
2798

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

with the highest net forecasted demand after updating the initial inventory. This process continues until we reach the nal
month. As a result, we obtain inventory, production and backorder matrices by considering the minimum stock and backorder. The pseudo code of the algorithm is given in Appendix A.
The outputs of greedy algorithm such as production quantities, possible backorders and safety stock levels are presented
in Tables 911, respectively.
4.2.2. Genetic algorithm
In production planning, genetic algorithm (GA) is generally used in scheduling problems like greedy algorithm. Different
than the literature, we implemented the algorithm to an inventory and a backorder problem to obtain simultaneously the
desired production, backorder and inventory quantities for each product by considering the capacity constraints. The
algorithm has been executed with the following parameters under the same constraints:
Iteration: 200
Population size: 70
Mutation probability: 0.05

6368

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

Table 9
Production quantities.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99

0
790
1345
0
3384
0
3220
1887
..
.
747
0
283
1598
378
..
.
723
2698
0
2523
960

0
0
4630
0
1745
0
3283
643
..
.
885
0
1603
4249
0
..
.
3506
696
2833
0
1485

0
0
4169
941
4736
998
11722
2122
..
.
2757
0
1788
4153
1105
..
.
2431
5048
1723
4128
0

0
0
3617
0
4263
0
7088
1968
..
.
1650
0
1451
4624
0
..
.
3479
1434
2073
3534
0

0
0
3821
0
3842
0
10663
2666
..
.
1860
1398
998
3896
1056
..
.
2820
1406
1949
2571
2061

0
1707
2686
1234
3414
0
11564
2948
..
.
1239
0
1139
4080
0
..
.
2507
1568
1180
3185
0

0
0
2717
0
6231
292
9417
152
..
.
0
0
0
3036
1302
..
.
1723
292
292
2548
1448

0
3182
1567
0
0
0
5678
5725
..
.
1868
1199
0
2301
0
..
.
1442
1741
1548
0
0

0
11653
3048
0
1790
1672
14015
2766
..
.
0
0
1600
3570
0
..
.
1416
2830
0
2652
0

0
9102
2575
1737
0
0
11125
2259
..
.
2045
0
0
3033
1969
..
.
2592
0
2140
3178
0

0
6768
2496
0
4262
0
10339
1016
..
.
0
1583
0
2764
0
..
.
1916
2203
0
2456
0

0
4230
2088
0
0
0
11407
3981
..
.
1691
0
2140
2149
1384
..
.
0
1820
1550
2160
0

Table 10
Backorders.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

0
0
0
149
0
114
0
0
..
.
0
60
0
0
0
..
.
0
0
142
0
0
13395

0
152
0
487
0
543
0
0
..
.
0
147
0
0
449
..
.
0
0
0
527
0
44093

0
206
0
0
0
0
0
0
..
.
0
618
0
0
0
..
.
0
0
0
0
20
66426

0
206
0
289
0
338
0
0
..
.
0
932
0
0
408
..
.
0
0
0
0
40
97579

0
206
0
738
0
691
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
99749

0
0
0
0
0
930
0
0
..
.
0
511
0
0
719
..
.
0
0
0
0
0
122831

0
762
0
398
0
852
0
2132
..
.
1076
929
543
0
0
..
.
0
827
913
0
0
210525

0
0
0
640
275
911
0
0
..
.
0
0
1048
0
350
..
.
0
0
0
383
0
152170

0
0
0
1246
0
0
0
0
..
.
1098
615
0
0
1246
..
.
938
0
1110
0
0
238280

0
0
0
0
1119
215
0
0
..
.
0
1121
685
0
0
..
.
0
2
0
0
0
237055

0
0
0
446
0
546
0
1047
..
.
851
0
1235
0
652
..
.
0
0
784
0
0
249364

0
0
0
939
744
708
0
0
..
.
0
510
0
0
0
..
.
1379
0
0
0
0
201878

Cross over operator: one point crossover


Mutation operator: inversion
The solution steps of the algorithm are established as follows:
Step 1. Generating an initial population: Initial population is generated randomly. It includes 70 feasible chromosomes with
457 bits. Each bit denotes a tire type. As we mentioned before, at each month at most 200 tires can be produced.
Therefore, while 200 bits represent the produced tires, the remaining 257 bits present the ones that are not selected
for the production. During the selection process of 200 bits in a chromosome, the production amount of each tire is
determined by considering the remaining mould and machine capacities.
Step 2. Evaluating the tness: Fitness value of each chromosome is computed. In this study, the tness value is formulated
P
P457
as FV 457
i1 Inv entoryi Cost
i1 Backorder i Cost. Then, total tness value of the population is computed.
Step 3. Selection: Chromosomes are selected from the current population to be the parents of generated offspring for the
new generation according to their tness values. In this paper, roulette wheel selection method is used as the selection technique. Since the objective is to minimize the total amount of inventory and backorder costs, the quantile of
each chromosome is computed by taking the inverse of the ratio of its tness value and the total tness.

6369

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374


Table 11
Safety stock levels.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
35

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
7

0
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
0

Table 12
Production quantities.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99

1
0
2691
298
6768
229
6440
3774
..
.
0
0
0
3197
0
..
.
1447
0
0
0
1920

0
942
3284
0
0
314
63
0
..
.
0
147
1886
2650
0
..
.
2782
0
0
3050
0

0
54
0
643
0
455
11722
0
..
.
4389
471
0
4153
1483
..
.
2431
8442
0
3601
0

0
0
7786
0
7360
338
7088
2846
..
.
1650
314
3239
4624
0
..
.
3479
0
6629
3534
565

0
0
3821
738
3842
0
10663
2666
..
.
0
0
998
3896
1056
..
.
2820
2840
1949
2571
2021

0
1501
0
0
3414
0
11564
0
..
.
0
0
1139
0
719
..
.
2507
0
1180
3185
0

0
762
5403
0
0
0
0
0
..
.
0
1395
543
0
0
..
.
0
0
1206
2548
1448

0
2420
0
0
6507
866
15095
0
..
.
4967
270
0
9417
0
..
.
3165
3601
635
0
0

0
11653
0
0
0
761
14015
0
..
.
1098
615
0
0
0
..
.
0
0
0
2652
0

0
9102
7190
0
2633
215
11125
0
..
.
0
506
0
6603
0
..
.
0
2832
2140
3178
0

0
0
2496
2679
3143
0
10339
15915.59
..
.
1798
462
0
2764
0
..
.
5925
0
0
0
0

0
0
0
0
0
493
0
0
..
.
0
510
3197
0
0
..
.
1379
4021
0
0
0

Step 4. Cross over: Genes are swapped between two chromosomes to produce a pair of offsprings. We use one-point
crossover method. Generated offsprings are checked whether they are feasible or not according to the problem constraints. If they are not feasible, the production amount of randomly selected offspring is set to 0.
Step 5. Mutation: Mutation is usually dened as ipping of randomly selected genes of an offspring. In our paper, mutation
rate is 0.05.
Step 6. Accepting: In this step, we rst check whether each generated offspring is feasible with respect to capacity constraints or not. If a tire type in an offspring could not be produced due to the insufcient capacity, it is marked
and its production amount is set to zero. This process continues until we check all 200 bits in an offspring. We randomly select a tire from the set of unproduced tires to complete the total produced tire type amount to 200. During
the selection of new tires, we also check their feasibility. At the end of this process, the tness value of each offspring
is computed as in step 2. Then, their tness values are compared with the maximum tness value in the population.
If it is lower than the maximum value, that offspring is replaced with the corresponding chromosome.
Step 7. Stopping criteria: Assign the chromosome with the lowest tness value in the population as the production program
if the number of iterations is completed. Otherwise, turn back to step 2.

6370

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

Table 13
Backorders.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

0
790
0
0
0
0
0
0
..
.
747
60
283
0
378
..
.
0
2698
142
2523
0
77004

0
0
0
189
0
0
0
0
..
.
1632
0
0
0
827
..
.
0
3394
2833
0
525
99133

0
0
4169
0
3097
0
0
878
..
.
0
0
1788
0
0
..
.
0
0
4556
0
545
186163

0
0
0
289
0
0
0
0
..
.
0
0
0
0
408
..
.
0
1434
0
0
0
197204

0
0
0
0
0
353
0
0
..
.
1860
466
0
0
0
..
.
0
0
0
0
0
184898

0
0
2686
496
0
592
0
2948
..
.
3099
977
0
4080
0
..
.
0
1568
0
0
0
247442

0
0
0
894
6231
807
9417
5233
..
.
4175
0
0
7116
583
..
.
1723
2687
0
0
0
326983

0
0
1567
1136
0
0
0
8826
..
.
0
0
505
0
933
..
.
0
0
0
383
0
281614

0
0
4615
1742
1514
0
0
11600
..
.
0
0
1057
3570
1830
..
.
2355
2830
1110
0
0
456688

0
0
0
2233
0
0
0
13851
..
.
947
0
1742
0
2552
..
.
4009
0
0
0
0
459032

0
6768
0
0
0
331
0
0
..
.
0
0
2292
0
3205
..
.
0
2200
784
2456
0
424742

0
10998
2088
493
744
0
11407
2933
..
.
840
0
0
2149
3937
..
.
0
0
1550
4616
0
384148

Table 14
Safety stock levels.
Tire code

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

X1
X10
X100
X101
X102
X103
X104
X105
..
.
X212
X213
X214
X215
X216
..
.
X95
X96
X97
X98
X99
Total

1
0
1346
149
3384
115
3220
1887
..
.
0
0
0
1599
0
..
.
724
0
0
0
960
238485

1
0
0
0
1639
0
0
1244
..
.
0
0
0
0
0
..
.
0
0
0
0
0
27220

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
10806

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
10434

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
9529

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
3951

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
2265

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
1399

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
21267

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
16715

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
15686

1
0
0
0
0
0
0
0
..
.
0
0
0
0
0
..
.
0
0
0
0
0
14871

The obtained outputs such as production quantities, possible backorders and safety stock levels are presented in Tables
1214, respectively.
The optimization model and heuristic algorithms are compared with respect to backorder levels, safety stock levels and
computation times. In our numerical experiments, we denote the mathematical model with MathMod. As it is seen in Table
15, monthly inventories do not exceed the maxstock constraint of the model each of the three methods. The pivot table of the
results of each model are compared in Table 15.
As it is seen in Tables 15 and 16, in terms of the backorder and inventory criteria, greedy algorithm is considerably favorable than GA. The performances of the mathematical model and the greedy algorithm are 77% and 47% higher than GA in
backorder scale, respectively. When we check the results in terms of inventory cost, it is observed that greedy outperforms
all methods. It performs 99% higher than GA.
Next, we examine the error incurred by solving the relaxed mathematical model instead of solving the integer model.
Table 17 presents the error gap for each cost parameter. Comparing the total cost of mathematical model in Table 15 against
the cost parameters given in Table 17, we notice that the resulting error is signicantly small.

6371

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374


Table 15
The pivot table of the results of mathematical model, greedy and genetic algorithm.

Backorders MathMod
Greedy
GA
Inventory
MathMod
Greedy
GA

M1

M2

M3

M4

M5

M6

M7

M8

M9

M10

M11

M12

Total

13395
13395
77004
3641
0
238485

32612
44093
99133
962
0
27220

37267
66426
186163
2170
0
10806

57703
97579
197204
1877
0
10434

32160
99749
184898
2659
0
9529

49409
122831
247442
52574
0
3951

125682
210525
326983
0
35
2265

50088
152170
281614
22836
0
1399

139806
238280
456688
4332
0
21267

72210
237055
459032
2279
0
16715

101950
249364
424742
2798
7
15686

51740
201878
384148
0
0
14871

764022
1733348
3325055
96128
42
372628

Table 16
Percentage gap with respect to genetic algorithm.
Percentage Gap (%)
Backorders
Inventory

MathMod
Greedy
MathMod
Greedy

77.02
47.87
74.20
99.99

Table 17
Error gap in solving the relaxation of the mathematical model.
Error gap in costs
z  ^z

Backorder cost
26.93

Inventory cost
6.65

Total cost
20.28

We conclude this section by presenting the results related to the computation time of each method. Greedy has the fastest computation time between the solution methods. Mathematical model takes about 1000.5 s to solve the problem for 457
industrial instances provided by the company at a 12-month period. While greedy heuristic takes about 1.05 s, GA solves the
same problem at 781.47 s.
5. Conclusion
While excess in production stocks may create additional burden like extra cost and new warehouse for the company,
scarcity in inventory causes out of stock because of uctuating demands and machinery breakdown and hence, customer
dissatisfaction may occur. Many organizations in Turkey have not enhanced any effective stock policy to avoid this dilemma
and for these reasons, this study is benecial. In this paper, we have developed an optimum production policy for a leader
tire manufacturing companys production and stock control model. Then, the effectiveness of this solution approach is presented by comparing its results with greedy algorithm and GA solution separately. While the greedy algorithm provides better results regarding to the computation time and the inventory level, the solution of the proposed mathematical model
provides the lowest backorder level compared to other methods. Since the greedy algorithm determines the production level
of each product according to the total demand and backorder level, it nearly keeps zero inventory at the end of each month.
However, this production policy results in high amount of backorder. Although, backorder level of greedy algorithm is high, it
is always quite lower than the GA. The performance of GA is poor compared to the other methods. The efciency of the GA
depends on the number of iterations. However, as the number of iterations increases, its computation time increases
exponentially.
The contribution of this paper is based on three points which are summarized below.
Computing optimum production quantity based on the limitations and capacity of the company and product range: The
best solution cannot be found intuitively due to the large number of constraints and the constant capacity and hence, the
development of a successful mathematical model has become mandatory. To the best of our knowledge, multi-product
ranges and all production constraints have never been included in any model proposed in the literature.
Developing an effective safety stock control model: Contrary to the existing studies, constraints of the production process
are real and mandatory without any assumption. Accordingly, in multi-product companies, optimum production quantity
per product is determined by considering the production capacity. Correlatively, safety stock levels per product that minimizes backorder are calculated separately by regarding the demand.
In addition, it should be considered that the problem size, we work in this paper, is the largest we are aware of in the
corresponding literature.
Consequently, it is obvious that the companies using the stock control models efciently, can increase their prot rates
and minimize their payments through restricting their expenditures and controlling their costs. The greedy algorithm takes
into account multiple resource share, capacity constraints and total demand as the key features to solve real supply chain
management problems efciently in different elds.

6372

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

Acknowledgement
This study was funded by the University of Kocaeli Research Fund.
Appendix A
The pseudo code of Greedy Algorithm:
Input
MaxStock = 800000
MaxCode = 457
tire = 457
month = 12
Demandtiremonth: demand of tire code i at month j
Dmonth: working day in month j
Ttire: daily mould capacity of tire code i
Ntire: daily manufacturing machine capacity of tire code i
Output
Ptiremonth
Inventorytiremonth
Backordertiremonth
Begin
Input InvBeg: the beginning inventory
totalP = 0
j=1
for each tire i 2 [1,tire]
minP(i,j) = Demand(i,j)-InvBeg(i)
end
SminP = Sort(minP)
for each tire i 2 [1,tire]
if totalP<200
if SminP(i,j) > 0
if Mw(i)>0
Q1 = min(SminP(i,j), D(j) T(i) Mw(i))
else
Mw(i) = 0
Q1 = 0
end
if MAs(i)>0
Q2 = min(SminP(i,j), D(j) N(i) MAs(i))
else
MAs(i) = 0
Q2 = 0
end
if min(Q1, Q2) > 0
P(i,j) = min(Q1, Q2)
Mw(i) = Mw(i)-(P(i,j) /D(j)/T(i))
MAs(i) = MAs(i)-(P(i,j) /D(j)/N(i))
totalP = totalP + 1
else
P(i,j) = 0
end
if P(i,j)-Demand(i,j) + InvBeg(i,j))> = 0
backorder(i,j) = 0
else
backorder(i,j) = P(i,j)-Demand(i,j) + InvBeg(i,j)
end
inventory(i,j) = P(i,j)-Demand(i,j) + InvBeg(i,j)

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

6373

end
end
end
for each month j 2 [2,month]
totalP = 0
Input Mw: number of mould code w
Input MAs: number of machine code
for each tire i 2 [1,tire]
minP(i,j) = Demand(i,j) + backorder(i,j-1) - inventory(i,j-1)
end
SminP = Sort(minP)
for each tire i 2 [1,tire]
if totalP<200
if (P(i,j-1)-Demand(i,j-1) + inventory(i,j-1)-Demand(i,j)>0)
P(i,j) = 0
else
if SminP(i,j) > 0
if Mw(i)>0
Q1 = min(SminP(i,j), D(j) T(i) Mw(i))
else
Mw(i) = 0
Q1 = 0
end
if MAs(i)>0
Q2 = min(SminP(i,j), D(j) N(i) MAs(i))
else
MAs(i) = 0
Q2 = 0
end
if min(Q1, Q2) > 0
P(i,j) = min(Q1, Q2)
Mw(i) = Mw(i)-(P(i,j)/D(j)/T(i))
MAs(i) = MAs(i)-(P(i,j)/D(j)/N(i))
totalP = totalP + 1
else
P(i,j) = 0
end
end
if P(i,j)-Demand(i,j)-backorder(i,j-1) + inventory(i,j-1) > = 0
backorder(i,j) = 0
else
backorder(i,j) = P(i,j)-Demand(i,j) + inventory(i,j-1)-backorder(i,j-1)
end
inventory(i,j) = P(i,j)-Demand(i,j) + inventory(i,j-1)-backorder(i,j-1)
end
end
if totalP> = 200
P(i,j) = 0
end
End

References
[1] R.C. Wang, T.F. Liang, Application of fuzzy multi objective linear programming to aggregate production planning, Comput. Ind. Eng. 46 (2004) 1741.
[2] G. Janakiraman, S. Seshadri, J.G. Shanthikumar, A comparison of the optimal costs of two Canonical Inventory Systems, Oper. Res. 55 (2007) 866875.
[3] Y.J. Lin, An integrated vendorbuyer inventory model with backorder price discount and effective investment to reduce ordering cost, Comput. Ind.
Eng. 56 (2009) 15971606.
[4] T. Wild, Best Practice in Inventory Management, 2nd ed., Elsevier Science LTD., Butterworth-Heinemann, 2002.
_
_
Publishing, Extended 3rd Press, Istanbul,
2008.
[5] M. Tanyas, M. Baskak, retim Planlama ve Kontrol, Irfan

6374

G.A. Keskin et al. / Applied Mathematical Modelling 39 (2015) 63596374

[6] C.T. Chang, A linearization approach for inventory models with variable lead time, Int. J. Prod. Econ. 96 (2005) 263272.
[7] L.A. San Jose, J. Sicilia, J. Garcia-Laguna, Analysis of an inventory system with exponential partial backordering, Int. J. Prod. Econ. 100 (2006) 7686.
[8] R. Jafar, D. Mansoor, A deterministic, multi-item inventory model with supplier selection and imperfect quality, Appl. Math. Model. 32 (2008)
21062116.
[9] N.H. Shah, A.S. Gor, An integrated economic lot-size model for vendor buyer inventory system when input is random, Math. Comput. Model. 49 (2009)
13261330.
[10] D.W. Pentico, M.J. Drake, The deterministic EOQ with partial backordering: a new approach, Eur. J. Oper. Res. 194 (2009) 102113.
[11] S.Y. Chou, C.J. Peterson, K.C. Hung, A note on fuzzy inventory model with storage space and budget constraints, Appl. Math. Model. 33 (2009)
40694077.
[12] A.K. Maiti, M.K. Maiti, M. Maiti, Inventory model with stochastic lead-time and price dependent demand incorporating advance payment, Appl. Math.
Model. 33 (2009) 24332443.
[13] C.T. Chang, T.Y. Lo, On the inventory model with continuous and discrete lead time, backorders and lost sales, Appl. Math. Model. 33 (2009)
21962206.
[14] Y. He, S.Y. Wang, K.K. Lai, An optimal productioninventory model for deteriorating items with multiple-market demand, Eur. J. Oper. Res. 203 (2010)
593600.
[15] H.Y. Kang, A.H.I. Lee, Inventory replenishment model using fuzzy multiple objective programming: a case study of a high-tech company in Taiwan,
Appl. Soft. Comput. 10 (2010) 11081118.
[16] L.E. Cardenas-Barron, The derivation of EOQ/EPQ inventory models with two backorders costs using analytic geometry and algebra, Appl. Math. Model.
35 (2011) 23942407.
[17] L.E. Crdenas-Barrn, G. Trevio-Garza, H.M. Wee, A simple and better algorithm to solve the vendor managed inventory control system of
multi-product multi-constraint economic order quantity model, Expert Syst. Appl. 39 (2012) 38883895.
[18] H. Omrani, M. Keshavarz, An interval programming approach for developing economic order quantity model with imprecise exponents and
coefcients, Appl. Math. Model. 38 (2014) 39173928.
[19] K. Moinzadeh, P. Aggarwal, Analysis of a production/inventory system subject to random disruptions, Manage. Sci. 43 (1997) 15771588.
[20] K. Skouri, S. Papachristos, A continuous review inventory model, with deteriorating items, time-varying demand, linear replenishment cost, partially
time-varying backlogging, Appl. Math. Model. 26 (2002) 603617.
[21] B. Ghalebsaz-Jeddi, B.C. Shultes, R. Haji, A multi-product continuous review inventory system with stochastic demand, backorders, and a budget
constraint, Eur. J. Oper. Res. 158 (2004) 456469.
[22] P. Dutta, D. Chakraborty, A.R. Roy, Continuous review inventory model in mixed fuzzy and stochastic environment, Appl. Math. Comput. 188 (2007)
970980.
[23] M.H. Al-Rifai, M.D. Rossetti, An efcient heuristic optimization algorithm for a two-echelon (R, Q) inventory system, Int. J. Prod. Econ. 109 (2007)
195213.
[24] K. Annadurai, R. Uthayakumar, Reducing lost-sales rate in (T, R, L) inventory model with controllable lead time, Appl. Math. Model. 34 (2010)
34653477.
[25] A.A. Taleizadeh, S.T.A. Niaki, A.A. Naja, Multi product single-machine production system with stochastic scrapped production rate, partial
backordering and service level constraint, J. Comput. Appl. Math. 233 (2010) 18341849.
[26] J. Lee, Inventory control by different service levels, Appl. Math. Model. 35 (2011) 497505.
[27] H.M. Lee, J.S. Yao, Economic production quantity for fuzzy demand quantity and fuzzy production quantity, Eur. J. Oper. Res. 109 (1998) 203211.
[28] J.T. Teng, C.T. Chang, Economic production quantity models for deteriorating items with price- and stock-dependent demand, Comput. Oper. Res. 32
(2005) 297308.
[29] G.P. Sphicas, EOQ and EPQ with linear and xed backorder costs: Two cases identied and models analyzed without calculus, Int. J. Prod. Econ. 100
(2006) 5964.
[30] S. Minner, A note on how to compute economic order quantities without derivatives by cost comparisons, Int. J. Prod. Econ. 105 (2007) 293296.
[31] A. Eroglu, G. zdemir, An economic order quantity model with defective items and shortages, Int. J. Prod. Econ. 106 (2007) 544549.
[32] H.M. Wee, W.T. Wang, C.J. Chung, A modied method to compute economic order quantities without derivatives by cost-difference comparisons, Eur. J.
Oper. Res. 194 (2009) 336338.
[33] K.M. Bjrk, An analytical solution to a fuzzy economic order quantity problem, Int. J. Approx. Reason. 50 (2009) 485493.
[34] N. Kazemi, E. Ehsani, M.Y. Jaber, An inventory model with backorders with fuzzy parameters and decision variables, Int. J. Approx. Reason. 51 (2010)
964972.
[35] J.Y. Jung, G. Blau, J.F. Pekny, G.V. Reklaitis, D. Eversdyk, A simulation based optimization approach to supply chain management under demand
uncertainty, Comput. Chem. Eng. 28 (2004) 20872106.
[36] J. Li, S. Wang, T.C.E. Cheng, Analysis of postponement strategy by EPQ-based models with planned backorders, Omega 36 (2008) 777788.
[37] W.C. Lee, J.W. Wu, J.W. Hsu, Computational algorithm for inventory model with a service level constraint, lead time demand with the mixture of
distributions and controllable negative exponential backorder rate, Appl. Math. Comput. 175 (2006) 11251138.
[38] W.T. Hu, S.L. Kim, A. Banerjee, An inventory model with partial backordering and unit backorder cost linearly increasing with the waiting time, Eur. J.
Oper. Res. 197 (2009) 581587.
[39] H.C. Chang, C.H. Ho, Exact closed-form solutions for optimal inventory model for items with imperfect quality and shortage backordering, Omega 38
(2010) 233237.
[40] D. Rahmani, R. Ramezanian, P. Fattahi, M. Heydari, A robust optimization model for multi-product two-stage capacitated production planning under
uncertainty, Appl. Math. Model. 37 (2013) 89578971.
[41] M. Karimi-Nasab, I. Konstantaras, A random search heuristic for a multi-objective production planning, Comput. Ind. Eng. 62 (2012) 479490.
[42] S. Benjaafar, M. ElHafsi, T. Huang, Optimal control of a productioninventory system with both backorders and lost sales, Nav. Res. Log. 57 (2010)
252265.
[43] C. Mocquillon, C. Lente, V. TKindt, An efcient heuristic for medium-term planning in shampoo production, Int. J. Prod. Econ. 129 (2011) 178185.
[44] B. Sarkar, A productioninventory model with probabilistic deterioration in two-echelon supply chain management, Appl. Math. Model. 37 (2013)
31383151.
[45] F. Tian, S.P. Willems, K.G. Kempf, An iterative approach to item-level tactical production and inventory planning, Int. J. Prod. Econ. 133 (2011) 439450.
[46] W.Q. Zhou, L. Chen, H.M. Ge, A multi-product multi-echelon inventory control model with joint replenishment strategy, Appl. Math. Model. 37 (2013)
20392050.
[47] P. Glasserman, S. Tayur, Sensitivity analysis for base-stock levels in Multiechelon productioninventory systems, Manage. Sci. 41 (1995) 263281.
[48] Brisa/The Company. [online] http://www.brisa.com.tr/English/Brisa/Corporate/The-Company.aspx (Accessed 17 September 2012).

Anda mungkin juga menyukai