Journal of King Saud University – Engineering Sciences 30 (2018) 243–249

Contents lists available at ScienceDirect

Journal of King Saud University – Engineering Sciences

journal homepage: www.sciencedirect.com

Original article

Investigation of the effect of buffer storage capacity and repair rate on

production line efficiency
Wadea Ameen a,b,⇑, Mohammed AlKahtani a, Muneer Khan Mohammed b, Osama Abdulhameed a,
Abdulaziz Mohammed El-Tamimi a
a
Industrial Engineering Department, King Saud University, Saudi Arabia
b
Princess Fatima Alnijiris’s Research Chair for Advanced Manufacturing Technology (FARCAMT Chair), Advanced Manufacturing Institute, King Saud University, Saudi Arabia

a r t i c l e i n f o a b s t r a c t

Article history: The buffer allocation problem concerning the size and location of storage between the stages of a flow
Received 3 January 2018 line is a critical research area in the design of production lines. In this study, a production system consist-
Accepted 11 March 2018 ing of two unreliable stages and a buffer between them is considered; the first stage incorporates two
Available online 13 March 2018
identical machines in parallel and the second stage consists of a single machine. A simulation model is
designed and used to evaluate the effect of the buffer capacity and the repair rates of machines on pro-
Keywords: duction line efficiency. The results revealed the capacity of buffer that yields the maximum production
Buffer allocation
line efficiency with various repair rate values.
Flow line
Simulation model
Ó 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an
Production line efficiency open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction
A production line is defined as a series of workstations that are
linked together by the handling systems that transfer parts from
one station to the other. Stochastic flow lines are typically subjected
to disruptions due to the variations in the processing times and fail-
ure of the workstations. These disruptions cause the production line
to be idle and result in decreased throughput. In order to reduce or
eliminate the effect of these disruptions, buffers are placed in
between the machines. The buffer storage enhances the throughput
of the production line by reducing the blocking and starving time of
the workstations, thereby improving the flow line efficiency.
Increasing the size of the buffer space between machines results
in increasing the throughput of a stochastic production line but to
a certain extent. Production lines have a limit of space that can be
allocated for buffers and they add an inventory costs as well. More-
over material handling is required to maintain the operational pro-
cess; thus, the buffer allocation problem translates into an
optimization problem of stochastic system involving many vari-
⇑ Corresponding author at: Industrial Engineering Department, King Saud
University, Saudi Arabia.
E-mail address: wadeaameen@gmail.com (W. Ameen).
Peer review under responsibility of King Saud University.
Production and hosting by Elsevier
ables (Tezcan and Gosavi, 2001). In production line the buffer zones
divide the line into stages—two stages if one storage buffer is used
and three stages if two storage buffers are used at two locations
along the production line. In general, n
1 storage buffers exist for
an n-stage line, as illustrated in Fig. 1 (Roser et al., 2004).
The study of the effects of machine failures and storage capacities
on the efficiency of production flow lines has attracted significant
interest. Exact solutions have been found multistage lines consist
unreliable machines. Wijngaard (1979) investigated the effect of
interstage buffer on the output of two-stage flow production line
when the production rate, breakdown rate, and repair rate are
unequal (Wijngaard, 1979). El-tamimi and Savsar studied the avail-
ability of an automated production line involving two unreliable
machines connected by a buffer and serviced by one repairman.
The stochastic model was modeled by using the Markovian tech-
nique. The repair and failure rates were unequal, while the produc-
tion rates were equal. The result presented the effect of the buffer
size on the availability and on the total cost (El-tamimi and Savsar,
1987). Mohant et al. identified the optimal buffer storage to maxi-
mize profit of stages facilities with finite buffer storage between
them. The service time of both stages are exponentially distributed.
The results demonstrated that the mathematical model enables the
accurate calculation of the optimum size of the buffer stock between
two consecutive operations exhibiting exponentially distributed
operation times (Mohanty et al., 1989). Gupta et al. presented a
straightforward mathematical formulation of a two-station problem
with an interstage buffer and a central buffer for bulk storage, and its
effect on the system cost of a two-stage production line was consid-

https://doi.org/10.1016/j.jksues.2018.03.001
1018-3639/Ó 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
244 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249
Fig. 1. Buffer storages in flow lines.
ered for the case where the two production milts have unequal pro-
duction rates. A mathematical model was created to identify the
optimal buffer capacity level and number of material handler trips
while minimizing the total system cost. Two straightforward strate-
gies were compared. The created model presented a straightforward
methodology to determine the cost-effectiveness of two potential
alternate manufacturing strategies (Gupta and Houshyar, 1990).
Savsar considered an automated production system with two unre-
liable stages and a buffer, in which the first stage consists of two
identical machines in parallel exhibiting similar failure, repair, and
production rates, and the second stage consists of one machine.
Moreover, he considered the buffer as exhibiting finite capacity
and the production rate of the first and second stages to be equal
with one repairman for each stage. An analytical stochastic model
was developed and solved to determine the production line effi-
ciency. The presented model yielded exact solutions to the extended
buffer storage (Savsar, 1992). Gopalan et al. considered a two-stage
flow-line production system (two machines) has an inter-stage, end
buffer and end inspections. The inter-stage and end inspections are
instantaneous, and the end buffer is of finite capacity. The failure
of machines I and II is not considered. A stochastic model is created
to model the system and analytical expressions for the idle time, per-
centage utilization, blocked duration with arbitrary distributions of
processing times at both stages of production, and exponential sup-
ply and demand pattern have been obtained by using the state-space
method and the regeneration point technique (Gopalan and
Anantharaman, 1992). Lau considered two stages flow production
system with an interstation finite buffer. His purpose to identify
the optimum combination of station processing-times means and
variances that yield the maximum system utilization and to deter-
mine the consequences of deviation from this optimum arrange-
ment. Numerous simulations were proposed for the production
line with various buffer capacities levels and combinations of means
and variances of station processing-times. The simulation results
demonstrated that the utilization found maximum when the means
and variances are balanced (Lau, 1994). Liu et al. modelled and ana-
lyzed the two-stage manufacturing system throughput with multi-
ple independent unreliable machines at each stage and a finite-
sized buffer between them. The matrix-analytic (MA) method is
used to calculate the system state probability and throughput.
Numerical cases showed the new model and the effectiveness of
the MA-based method (Liu et al., 2012). Diamantidis et al. used via
exact Markovian analysis to solve the buffer storage buffer in pro-
duction flow line. The considered production flow line consist two
stages and buffer of finite capacity and each stage consists of
multiple unreliable parallel machines. The failure times, repair times
and processing times of the parallel machines in each stage are expo-
nentially distributed with non-identical mean rates (Alexandros and
Chrissoleon, 2008). The results found that the proposed algorithm is
valid and practical. Nabil Nahas used an analytical decomposition-
type approximation to determine the buffer allocation and optimal
preventive maintenance policy that for a serial production line con-
sisting of n unreliable machines with n
1 buffers (Nahas, 2014).
Ouazene et al. proposed a new analytical approach to evaluate the
throughput of a buffered two-workstation production system with
unreliable parallel machines. The proposed method is based on
two steps. The first step consists of transforming each group of par-
allel machines into an equivalent one using different replacing tech-
niques. The second step consists of evaluating the throughput of the
obtained two-machine-one-buffer model using an equivalent
machine method (Ouazene et al., 2015). Zhou and Yu solved the buf-
fer allocation problem in serial flow lines with unreliable machines
by an improved ant colony optimization algorithm with an objective
of maximizing throughput rate (Zhou and Yu, 2016). Zhang proposed
a mathematical models for solving the joint workstation and buffer
allocation problem in production flow lines constrained to improve
the system throughput. The problem is formulated by an exact
model (mixed integer linear programming) and approximate mod-
els (linear programming). The both methods efficiency and accuracy
of the results (Zhang and Matta, 2016). Weiss et al. developed an effi-
cient rule-based local search algorithm that utilizes new individual
lower bounds to identify the optimal buffer sizes of production line
(Weiss et al., 2017). Dolgui et al. considered a multicriterial opti-
mization problem for the capacities of buffers in a flow line (Dolgui
et al., 2017). Mohamed-Chahir Fitouhi et al. evaluated the perfor-
mance of the flowline consisting of two-machine line with a finite
buffer and condition-based maintenance. They developed a method-
ology to analyze the trade-off between the contributions of the buf-
fer and the preventive maintenance to the system performance
(Fitouhi et al., 2017). The simulation method is a powerful tool for
designing and analyzing, improving manufacturing systems. It is
used to evaluate the effect of buffer space on the throughput of the
flow line production system (Sławomir et al., 2016). In current study,
a simulation model is designed to evaluate the effect of buffer size
and repair rate on flow line efficiency.
2. Problem description
In this study, a two-stage flow line with inter-stage finite buffer
(B) storage is considered and investigated. The first stage consists
of two machines in parallel (M1, M2), and the second stage consists
of one machine (M1). The output of the two machines in the first
stage flows into a buffer and then to the machine in the second
stage, as illustrated in Fig. 2.
2.1. Assumptions of the system
The system under consideration is assumed to have the follow-
ing operational characteristics:
 In the first stage, the failure rate ðk2 Þ, repair rate (l2 Þ, and pro-
duction rate ðq2 Þ of machine 1 (M1) and machine 2 (M2) are
respectively identical.
 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249 245

Fig. 2. Production line structure.

 In the second stage, machine 3 (M3) exhibits production rate,
repair rate and a failure rate q2; l2 ; k2 , and, respectively.
 q2 ¼ 2ðq2 Þ.
 Finite capacity of the buffer is z units.
 There are two repairmen, one for the first stage (M1, M2) and
the other for the second stage (M3).
 Machine repairs and failures are randomly distributed.
 The failure rate of machine in stage two (M3) reduces to
k2 ¼ k2 =2. when its operation rate is decrease from 2q2 to q2 .
 The failures are operation dependent.
 The system keep working till a machine fails. If any machine in
stage 1 fails, the system keep working till one of the following
events occur: the buffer reduced till reach zero value by
machine M3; the machine in stage two (M3) fails; machine
M2 (or Ml) fails. If the buffer level reduces to zero before the
repair of the failed machine is completed, the production rate
of the second stage (M3) reduces, and it operates at the rate
q2 instead of its normal rate q2 . If both the machines Ml and
M2 fail, the second stage continues operation until the buffer
is empty, at that time, M3 is forced shutdown due to the
unavailability of incoming components. The failure of M3
necessitates the shutdown (i.e., blockage) of Ml and/or M2
when the buffer attains its maximum level z. In all situations,
a forced shutdown machine will not fail.
this case study is to determine the effects of buffer capacity as well
as the machines’ repair rates on the line efficiency as shown in
Fig. 4. The simulation model is executed for several values of buffer
capacity and repair rate parameters of each machine.
3.1. Production line efficiency
The system efficiency є can be considered as the proportion of
time in which the last stage is in operation. However, in this sys-
tem, the last stage has two feasible production rates, normal rate
q2 and reduced rate q1. The proportion of time in the last stage dur-
ing which components are produced at the maximum rate q2 is
expressed as
X
2
e1 ¼ Pilx þ P210 þ P21z; ð1Þ
i¼0
The proportion of time that the last stage machine are produced
at the reduced rate q1 is expressed as Є2 = P110, where Pijx are the
state probabilities as defined before.
The production line efficiency is calculating as the actual pro-
duction rate in the last workstation, Be.
Be ¼ q2 e1 þ qle2 ð2Þ

The following notations are used to describe the state of pro-

duction system with above characteristics and operation policy.
State of the system at time t with buffer size x, t > 0, 0  x  z,
and i, j
Machines operating at the first and second stage respectively
(i = 0, 1, and 2 = 0,1).
X = Buffer size, 0  x  z.
Z = Maximum buffer capacity.
Fijx (t,x) = Probability distribution function of state Sijx(t, x)
with i = 0, 1,2; j = 0, 1;o < x < z, (x variable)
Pijx(t) = Probability of system state Sijx with x constant at
either x = 0, or x = z;
i == {0,1, 2;} = 0,1 (see Fig. 3).

3. Methodology

The simulation model has been developed by using AweSim 3.0

software, in which all the assumptions are considered as illustrated
in the network of the model. As mentioned above, the main goal of Fig. 3. Probability transition flow diagram.
246 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249

Fig. 4. Awesim simulation model.

Table 1
Flow line efficiency with buffer capacity case (a).

Case (a) Buffer size

No. run 1 5 10 20 30
1 0.778 0.942 0.949 0.949 0.949
2 0.723 0.832 0.861 0.861 0.861
3 0.745 0.839 0.839 0.839 0.839
4 0.788 0.927 0.934 0.934 0.934
5 0.818 0.927 0.927 0.927 0.927
6 0.745 0.81 0.81 0.81 0.81
7 0.839 0.927 0.927 0.927 0.927
8 0.876 0.92 0.92 0.92 0.92
9 0.723 0.854 0.854 0.854 0.854
10 0.657 0.803 0.803 0.803 0.803
(Efficiency) 0.7692 0.8781 0.8824 0.8824 0.8824

Thus, the actual line efficiency

e ¼ Be=q2 ð3Þ

4. Results analysis

To determine the effect of the buffer capacity and repair rates

on the line efficiency a case study from the previous work was used
(El-tamimi and Savsar, 1987). The production rate for stage 1
machines is assumed to be; q1 = 5 units/hr and for stage 2;
q2 = 10 units/hr. Failure rates and repair rates are assumed to be
exponentially distributed with a mean of 2.5 units. Moreover, four
cases with various repair rates and different levels of buffer
capacity were analyzed. Fig. 5. Flow line efficiency with buffer capacity case (a).
 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249 247

Table 2
Flow line efficiency with buffer capacity case (b).

Case (b) Buffer size

No. run 1 5 10 20 30
1 0.839 0.949 0.949 0.949 0.949
2 0.715 0.781 0.788 0.781 0.781
3 0.693 0.825 0.825 0.825 0.825
4 0.788 0.92 0.927 0.92 0.92
5 0.774 0.869 0.883 0.869 0.869
6 0.701 0.759 0.759 0.759 0.759
7 0.832 0.898 0.912 0.898 0.898
8 0.613 0.723 0.774 0.788 0.788
9 0.584 0.708 0.745 0.774 0.774
10 0.657 0.752 0.766 0.752 0.752
(Efficiency) 0.7196 0.8184 0.8328 0.8315 0.8315

Fig. 6. Flow line efficiency with buffer capacity case (b).

Fig. 7. Flow line efficiency with buffer capacity case (c).

4.1. Case (a): Equal repair rate l1 ¼ l2 ¼ 4
Table 1 and Fig. 5 illustrate the line efficiency of the production
line with various buffer sizes when the repair rates of both stages
are equal. The result indicates that the buffer capacity exerts signif-
icant influence on the line efficiency till z = 5, by which the effi-
ciency has increased by 0.113; marginal influence in the range
z = 5–10, over which the increase in efficiency = 0.004; and zero
influence on efficiency with further increase in buffer size.
4.2. Case (b): l1 ¼ 4l2 ¼ 2
Table 2 and Fig. 6 illustrate the variation of the production line
efficiency with respect to the buffer size when the repair rate of
stage 1 is higher than the repair rate of stage 2.
The result indicates that the buffer capacity has a significant
effect on the increase in the line efficiency till z = 5, by which
the efficiency has increased by 0.1064; marginal effect in the
range z = 5–10, over which the increase in efficiency = 0.0065;
and zero effect on the efficiency with further increase in the
buffer size.
4.3. Case (c): l1 ¼ 2; l2 ¼ 4
Table 3 and Fig. 7 illustrate the variation of production line effi-
ciency with respect to buffer size when the repair rate of stage 1 is
less than the repair rate of stage 2.
The result indicates that the buffer capacity has a significant
effect on the increase in the line efficiency till z = 5, by which the
efficiency has increased by 0.1064; marginal effect in the range

Table 3
Flow line efficiency with buffer capacity case (c).

Case (c) Buffer size

No. run 1 5 10 20 30
1 0.672 0.81 0.81 0.81 0.81
2 0.672 0.73 0.73 0.73 0.73
3 0.584 0.657 0.679 0.679 0.679
4 0.606 0.708 0.715 0.715 0.715
5 0.635 0.745 0.745 0.745 0.745
6 0.613 0.766 0.766 0.766 0.766
7 0.715 0.766 0.766 0.766 0.766
8 0.615 0.774 0.788 0.788 0.788
9 0.555 0.672 0.672 0.672 0.672
10 0.62 0.723 0.745 0.745 0.745
(Efficiency) 0.6287 0.7351 0.7416 0.7416 0.7416
248 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249

Table 4
Flow line efficiency with buffer capacity case (d).

Case (d) Buffer size

No. run 1 5 10 20 30
1 0.664 0.81 0.839 0.839 0.839
2 0.501 0.629 0.646 0.656 0.656
3 0.55 0.652 0.688 0.721 0.733
4 0.547 0.628 0.664 0.664 0.664
5 0.562 0.614 0.639 0.629 0.639
6 0.54 0.657 0.701 0.715 0.715
7 0.658 0.706 0.726 0.726 0.726
8 0.701 0.726 0.743 0.75 0.75
9 0.721 0.823 0.823 0.823 0.823
10 0.65 0.781 0.796 0.796 0.796
(Efficiency) 0.609 0.703 0.727 0.732 0.734

ciency has increased by 0.093; in the range z = 5–10, the increase

in the efficiency = 0.024; for increase in z beyond 10, the buffer
capacity exerts marginal effect.
The result indicates that the buffer capacity exerts significant
effect on the increase in line efficiency till z = 5, by which the effi-
ciency has increased by 0.093; in the range z = 5–10, the increase
in the efficiency = 0.024; for increase in z beyond 10, the buffer
capacity exerts marginal effect. Fig. 9 summarizes the effect of buf-
fer capacity and repair rate on the production line efficiency.
From Fig. 9, it can be concluded that the buffer capacity exerted
significant effect on the increase in the line efficiency for up to
approximately z = 5 units, in almost all cases; beyond that, the
effect is marginal in certain cases. Moreover, the repair rate exerts
significant influence (an increase of repair rate results in an
increase of efficiency) when the failure rates are fixed. The main
reason for this seems to be that as the repair rate is increased,
the equipment downtime is reduced, and thus, the line efficiency
is increased.
Fig. 8. Flow line efficiency with buffer capacity case (d).

5. Conclusions

In this study, a simulation model was designed to study the

Fig. 9. Effect of buffer capacity and repair rate on Flow line efficiency.
z = 5–10, over which the increase in efficiency = 0.0065; and zero
effect on the efficiency with further increase in the buffer size.
4.4. Case (d): l1 ¼ l2 ¼ 2
Table 4 and Fig. 8 illustrate the variation of production line effi-
ciency with respect to the buffer size when the repair rate of stage
1 is equal to the repair rate of stage 2 albeit less than that in case
(a). The result indicates that the buffer capacity exerts significant
effect on the increase in line efficiency till z = 5, by which the effi-
effect of buffer size and repair rate of machines on production line
efficiency. The results demonstrated that line efficiency increases
as the buffer capacity increases till certain level; beyond this level,
the buffer capacity exerts marginal or no effect on line efficiency
for example in case (a) increasing of buffer size from1 to10 resulted
in increased of the efficiency from 76.92% to 88.24%, where
increased of the buffer size from 10 to 20 and 30 resulted on the
same efficiency (88.24%). Moreover, as the repair rate increases,
the line efficiency increases for example in increasing of repair rate
from2 to 4 resulted in increased of the efficiency from 73.2% to
88.24%, and it can be concluded that the increase in the repair rate
of stage one results in a higher increase of line efficiency than that
from a similar increase in stage two.
The simulation is simple fast, accurate and modifiable approach
for optimization the buffer size compared to the other methods,
that will make the proposed approach most prefer and useful for
industry.
From the above results, it can concluded that it is essential to
consider the buffer capacity and repair rates to design production
lines with enhanced efficiency.
Future studies can be conducted to extend the developed model
to cover the following:
– Production lines with more than two stages.
– Parallel non-identical machines in the first stage.
– Optimum number of repair men needed for multi stages pro-
duction line.
 W. Ameen et al. / Journal of King Saud University – Engineering Sciences 30 (2018) 243–249
Conflict of interests
The authors declare that there are no conflicts of interest
regarding the publication of this paper.
Acknowledgments
The authors are grateful to the Raytheon Chair for Systems Engi-
neering for funding.
References
Alexandros, D.C., Chrissoleon, P.T., 2008. Exact analysis of a two-workstation one-
buffer flow line with parallel unreliable machines. Eur. J. Oper. Res. 197, 572–
580.
Dolgui, A.B., Eremeev, A.V., Sigaev, V.S., 2017. Analysis of a multicriterial buffer
capacity optimization problem for a production line”. Automat. Remote Control
78 (7).
El-tamimi, A., Savsar, M., 1987. The availability of a two-stage production line with
intermediate buffer. IASTED Reliability and Quality Control Conference, 24–26
June, Paris, France, pp. 117–120.
Fitouhi, Mohamed-Chahir, Nourelfath, Mustapha, Gershwin, Stanley B., 2017.
Performance evaluation of a two-machine line with a finite buffer and
condition-based maintenance. Reliab. Eng. Syst. Saf.
Gopalan, M.N., Anantharaman, N., 1992. Stochastic modelling of a two-stage
transfer-line production system with end buffer and random demand.
Microelectron. Reliab. 32 (1–2), 11–15.
Gupta, T., Houshyar, A., 1990. Consideration of instage and central buffers in two-
stage production system. Comput. Ind. Engng 19 (1–4), 427–431.
249
Kłos, Sławomir, Patalas-Maliszewska, Justyna, Trebuna, Peter, 2016. improving
manufacturing processes using simulation methods. Appl. Comput. Sci. 12, 7–
17.
Lau, H.-S., 1994. Allocating work to a two-stage production system with interstage
buffer. Int. J. Prod. Econ. 36, 281–289.
Liu, J., Yang, S., Wu, C., Hu, S.J., 2012. Multi-state throughput analysis of a two-stage
manufacturing system with parallel unreliable machines and a finite buffer.
Eur. J. Oper. Res. 219 (2), 296–304.
Mohanty, R.P., Dey, D., Kulkarni, R.V., 1989. A computer algorithm for determining
optimal buffer size in a serial flow production system. Comput. Ind. Eng. 16 (1),
75–86.
Nahas, Nabil, 2014. Buffer allocation and preventive maintenance optimization in
unreliable production lines. J. Intell. Manuf. https://doi.org/10.1007/s10845-
007-0030-z.
Ouazene, Y., Yalaoui, F., Chehade, H., Yalaoui, A., 2015. Analysis of an unreliable
series-parallel production system with limited storage capacity. Math. Methods
Syst. Sci. Eng.
Roser, Christoph, Nakano, Masaru, Tanaka, Minoru, 2004. Economic Optimization in
Manufacturing Systems through Optimal Buffer Allocation. Austria, Vienna.
Savsar, M., 1992. Modeling of a two-stage merge production with buffer storage.
JKAU: Eng. Sci. 4, 67–81.
Tezcan, T., Gosavi, A., 2001. Optimal buffer allocation in production lines using an
automata search. Proc. Inst. Indust. Engg. Conf.. Institute of Industrial Engineers,
Norwell, GA, Citeseer.
Weiss, S., Matta, A., Stolletz, R., 2017. Optimization of Buffer Allocations in Flow
Lines with Limited Supply. IISE Trans. 1–12. https://doi.org/10.1080/
24725854.2017.1328751.
Wijngaard, J., 1979. The effect of interstage buffer storage on the output of two
unreliable production units in series, with different production rates. AIIE Trans.
11, 42–47.
Zhang, M., Matta, A., 2016. Discrete event optimization: workstation and buffer
allocation problem in manufacturing flow lines. In: Proceeding 2016 Winter
Simul Conf, pp 2879–2890.
Zhou, B., Yu, J., 2016. Buffer allocation method of serial production lines based on
improved ant colony optimization algorithm. High Technol. Lett., 113–119