06003957
TRANSCRIPT
-
7/31/2019 06003957
1/5
*Tu Chuanqing (1968~), Corresponding author, Associate professor, doctor.
Excel-Based Simulation on Problems of Two-Stage Assembly Line
Queuing
Qiao xianxia Tu Chuanqing*
College of Computer and Information EngineeringJiangxi Agricultural University
Nanchang Jiangxi China
E_main: Qiao [email protected] [email protected]
Abstract: It is difficult to design mathematic
models for the problems of assembly line
queuing in the workshop, and simulation is
usually made in the computer. But it requires
that the researchers have high level of computer
proficiency to build up a simulation model with
a computer language. Taking Excel as a tool, we
have established a simulation model for the
problems of two-stage assembly line queuing.
With this model, we can not only easily examine
the production efficiency of this assembly line,
but also examine whether enlarging the space
between the two workstations could impose
some influences upon the productivity of
assembly lines and the waiting time.
Keywords: assembly line; queuing problems;
simulation model
0 Preface
It is often impossible to construct a
mathematic model because the problems of
assembly line queuing occur continuously and in
parallel. However, it is often no problem to
simulate these queuing problems. Presently,
there are plenty of literatures related to
simulation on problems of two-stage assembly
line queuing. But it requires that the researchers
have high level of computer proficiency to buildup a simulation model with the computer
language, which, to some extent, hinders the
application and popularity of such a model. In
this paper, taking Excel as a tool, we introduce a
method of setting up a simulation model on
problems of two-stage assembly line queuing.
1 Problems of Two-Stage Assembly Line
Queuing
Figure 1 represents two workstations on an
assembly line in which A is operating
Workstation 1 and B Workstation 2 respectively.
The products finished by A at Workstation 1 will
be passed to Workstations 2 at which B
continues to process. The volume of products is
an important factor for consideration in the
design and analysis of such an assembly line,
because the product storage of each workstation
will have effects on the workers efficiency. On
the assembly line in Figure 1, suppose that the
two workstations are connected and that there is
no place for the semi-finished products, two
possible cases will take place as follows: If Aworks more slowly than B, B is forced to wait;
but if A is faster than B on the contrary, then A
has to wait.
In the simulation of this problem, suppose
A is the first worker on the assembly line, then
he can work on the semi-finished products at any
time. Therefore, we focus our analysis on the
mutual influences between A and B.
Figure 1Two Workstations on an Assembly
Line
2011 International Conference on Business Computing and Global Informatization
978-0-7695-4464-9/11 $26.00 2011 IEEE
DOI 10.1109/BCGIn.2011.130
492
-
7/31/2019 06003957
2/5
1.1 Research Targets
About the assembly line, we hope that this
research will help solve the following problems:
i) how much is the average time for each worker
to complete the work? ii) what is the
productivity of this assembly line? iii) how longwill A wait for B? iv) how long will B wait for
A? v) how will it affect the productivity, the
waiting time and other factors that the space
between the two workstations are enlarged to
deposit semi-products and that workers
independence is increased subsequently?
1.2 Data Collection
For a systematic simulation, we need the
data about As and Bs processing time. To
collect the data, a feasible method is to divide
the total processing time into short periods, ineach of which each worker is observed
individually. In this example, we divide the
processing time, 10 seconds a period. A is
observed at work 100 times, but B only 50 times.
The times for the observation of the two at work
are different. We know that the more times for
the observation and the smaller division of the
time intervals, the more accurate the research
result. We make a table about the processing
time as shown in Table 1 below.
Table 1 The Data of Observation on the Two
Workers
In Table 2, we show the random intervals
distributed by the ratio of the data collected from
the actual observation. For example, four out of
one hundred times A has completed his
operations within 10 seconds. If one hundred
numbers are allocated, we should allocate four of
them to those operations completed within 10
seconds. These four numbers can be random,
such as 42, 18, 12 and 93, which though will
make our search very complicated. For a
convenient search, we allocate them with such
consecutive numbers as 00, 01, 02, 03. For the
fifty observed data of B, we can use two
methods to allocate the random numbers. In the
first method, we use fifty numbers like 00 to 49
for distribution and omit in the simulation all
those numbers larger than 49. However, this is a
great waste that we will discard half of the
numbers in the random sequence. In the secondmethod, we duplicate the frequency number. For
example, we distribute 00 through 07 to eight
out of one hundred observed data whose
processing time is within 10 seconds instead of
distributing 00 through 03 to four out of fifty
times whose processing time is within 10
seconds. In this way, the times for observation
are doubled while the ratio remains the same.
Table 2 The Random Intervals of A and B
2 Simulation Model on Two-Stage Assembly
Problems
2.1 Manual Simulation
Table 3 shows the results of manual
simulation of A and B processing 10 products.
The random numbers come from the Table of
Random Numbers, from the first column of
double-digit, the numbers are selected
downwards.
493
-
7/31/2019 06003957
3/5
If we count the time from zero and measure
it by the second, the first random number 56 is
corresponding to 50 seconds in which A
completes the first processing task. Then this
taskpiece is transmitted to B who starts at 50
seconds. The next random number is 83, and
according to Table 2, it takes B 70 seconds to
complete the taskpiece. Simultaneously, A is
processing a second taskpiece, starting from at
50 seconds and ending at 100 seconds (the
corresponding random number is 55). However,
A cannot start processing a third taskpiece
immediately afterwards as B completes the first
taskpiece at the 120th second. Consequently, A
has to wait 20 seconds before processing a
fourth one. With the same method of counting
above, we can calculate the rest data in the table:with a random number and the corresponding
processing time, we can figure out the waiting
time and the completing time. Then we can find
that the waiting time for both A and B is very
long because there is no storage place between
the two workstations.
Table 3: A and BSimulation of
Two-Stage Assembly Line
Now, we can answer some questions and make
some comments on the system. For example,
The average processing time of each
workpiece: 610/10=61 second;
The average processing time of worker A:
470/10=47 second;
The average processing time of worker B:
440/10=44 second;
The utilization of worker A: 470/540=87%;
The utilization of worker B:
440/560=78.6% (the initial 50 seconds for
waiting is removed)
Although we have explained how to build a
simple manual simulation on this problem, the
sample of 10 drawings is too small to warrant
the result reliability. A more reliable result could
be attained only through thousands of repeated
calculations on the computer.
2.2 Simulated by Excel
Figure 2 shows the partial results of the
simulation by Excel on the two-stage assembly
lines, which is in the same format as that of the
manual simulation in Table 2. This time, we
have used Excel to simulate 1, 000 times,
namely we suppose that A and B have processed
1,000 workpieces in total. Its specific operating
procedure is as follows:
Step one, generating random numbers by
the function RANDBETWEEN ()A fundamental step for any method of
simulation is to generate random variables
related to the distribution function. In this
example, the distribution function is about the
time for A and B to complete each taskpiece.
The function RANDBETWEEN () can generate
a random number for any pair of specified values.
What we need now is to generate the random
numbers between 0 through 99, which can be
attained by the function RANDBETWEEN (). In
Table 4, the 2nd
and 7th
columns are randomseries generated by =RANDBETWEEN(0,99),
representing the random numbers corresponding
to the time for A and B to complete each
taskpiece.
Step two, establishing the relation between
the random numbers and the processing time by
the function VLOOKUP
According to the distribution rules of
random numbers specified in Table 2 above, we
can establish the relation between the processing
time and the random number by the function
VLOOKUP. The method is to input 0 ~ 99
sequentially in the cell A3~A102, and then
according to the distribution rule of random
numbers to input in the cells B3~B102 and
C3~C102 separately the corresponding time for
494
-
7/31/2019 06003957
4/5
A and B to complete their processing work.
Figure 2 A and Bthe Simulation of
Two-Stage Assembly Lines by Excel
Table 4 Excel Cell Formula in Figure 2
The simulation technology being an
analysis tool, its dynamic characteristics
determine its advantages in quantitative analyses.
But an analytical method is different, which
indicates the average result of the long-term
operation of a system. As is shown in Figures 3
and 4, we can find a visible start-up. Figure 3
exhibits the average time for A and B to
complete 100 processing taskpieces. These data
are accumulative, i.e., the first datum is
generated randomly corresponding to thecompleting time for the first taskpiece. The
average completing time for two taskpieces is
the average value of the time to complete the
first taskpiece plus that to complete the second
one. By analogy, the average completing time
for three taskpieces is the average value of the
time to complete the first one plus those for the
second and the third ones respectively, and so
forth.
Note that the curve can be in different
shapes rather than exactly what is shown in thefigures below because the initial part of the
curve is determined by the flow of the generated
random numbers. What we are sure of is that
within a short period of time after the start-up of
the system, the average processing time is
fluctuant and then tends to be stable slowly.
Figure 3 Average Processing Time for Each
Workpiece
Figure 4 shows the average stay time in the
system, which includes the processing time and
the waiting time for each taskpiece. For
examplein Figure 2 the stay time for the first
taskpiece is 90 seconds in the system, which is
the sum of the cells G3, I4 and L3. And the stay
time for the 2nd
taskpiece is 120 secondsthesum of the cells G4, I4 and L4. And by analogy,
we can attain the stay time for the 1, 000 th
taskpiece in the system. At the start-up stage, we
can see in the curve trend of gradual increase,
because the system is started up from the idle
state when there is no interval in the process of
transferring the taskpieces from A to B.
However, as the 2nd taskpiece enters the system,
there is possibly a wait between two procedures
because of the inconsistent speed of A and B and
of no storage place between the two
workstations, which forces the taskpieces
entering afterwards to delay. As the time goes by,
the taskpieces transmission will tend towards
stability unless the working capacity in the 2nd
procedure is weaker than that in the 1st one.
Figure 4 Average Stay Time of Products in the
System
According to the simulation data in Figure
2, we can also produce some indices to
495
-
7/31/2019 06003957
5/5
investigate the system efficiency as is shown in
Table 5 below:
Comparatively, the result of the manual
simulation of 10 taskpieces is not too bad. The
average working time for A and B is 46.13 and
46.31 seconds respectively, which is very close
to the expected average value from the long-term
operations. The expected working time for A is
104+206+3010+/100=45.9 seconds,
and that for B is10 4+20 5+30 6+
/50=46.4 seconds.3 Research on the Storage Space Between Two
Workstations
A research on the storage space between of
two workstations is also very significant. We can
solve this problem by comparing the data of
productivity and utility in different storage
places. In the previous sections, we have built up
a simulation on the situation where there is no
storage space between two workstations. In a
second simulation, we create a storage space and
record some possible changes about the related
data. And then we set up other simulations on
the second, third, and so on, storage spaces.
With these data generated, decision makers can
calculate the cost increased to build more storage
spaces and the benefits brought about by the
improved productivity, and then make a
comparison of the cost and the benefits. More
storage spaces between two workstations
probably means a larger workshop, more
materials and taskpieces in the system, more
material handling equipment, transmission
equipment, and more use of heat and electricity,
as well as more maintenance of the workshop,
etc.
References
[1] Richard B. Chase, etc. Operations
Management for Competitive Advantage
[M].Beijing: Machinery Industry Press,2003.
[2]Hu Yunquan. Operations Research Tutorial
2nd edition [M]. Beijing: Tsinghua
University Press, 2003.
[3]Zhou Dehui Zuo Qi Li Changwen,
Application on Excel in Modern
Management [M]. BeijingPublishing House
of Electronics Industry1997.
[4]Chen Xuesong, Fang Rengcun, Cao Ju,
Research on Statistics Calculation of
Simulation Queuing System [J]. ComputerSimulation, 2003, (7).
496