Академический Документы
Профессиональный Документы
Культура Документы
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/4268293
CITATIONS
READS
11
301
5 AUTHORS, INCLUDING:
Qing Chang
Jun Ni
University of Michigan
SEE PROFILE
SEE PROFILE
Guoxian Xiao
Stephan Biller
General Electric
SEE PROFILE
SEE PROFILE
MoB2.1
I. INTRODUCTION
76
MoB2.1
where
is the blockage time for machine j;
is the
starvation time for machine j; j-1 is the nearest upstream
machine and j+1 is the nearest downstream machine.
We define the turning point to be a machine where the
trend of blockage and starvation changes from blockage
being higher than starvation to starvation being higher than
blockage. Furthermore, the sum of a turning point
machines blockage and starvation is smaller than its
neighboring machines. In this way, the turning point
machine has the highest percentage of the sum of operating
time and downtime compared to the other machines in the
segment.
This definition of turning point can be further illustrated
in Fig. 2, which plots the same actual data as in Fig. 1. In
Fig 2, within the segment from machine M8 to machine
M17 (a ten-machine segment), blockage is initially higher
than starvation (M8). Then, starvation becomes higher than
blockage (from M9 to M17). Furthermore, because the
relations can be obtained as:
MoB2.1
is mainly caused by
. Furthermore,
, and
is mainly
and
, and
is mainly caused by
caused by
and
. Therefore, four additional equations according
to these conditions can be obtained as
where
and
are
, then machine
is
failed, if there is a part on machine
blocked during the failure; else if there is no part on
machine
and machine
are
sensitivity
78
values
and
can
be
MoB2.1
machine-five-buffer model (6M5B), several-machine-nobuffer models, and models of real production lines from two
automotive plants. One case of a 6M5B production line is
shown to not only illustrate the simulation verification but
also highlight the main advantage of data driven method
over the traditional simulation-based bottleneck detection
method. The 6M5B production line is drawn schematically
in Fig. 4.
As a result,
Therefore,
and
as
TABLE I
SIMULATION PARAMETERS IN A 6M5B LINE
Cycle time
MTTR (min)
MTBF (min)
M1
10 sec
31
M2
10 sec
25
M3
10 sec
28
M4
10 sec
26
M5
M6
10 sec
10 sec
3
2
21
29
Therefore,
) is the
D. Simulation Verification
Besides the analytical verification on a three-machine-nobuffer serial line, simulation-based verification is utilized to
verify the turning point is a bottleneck in complex
production lines
More than five hundred cases have been studied, and over
95% of cases show good agreement that the data driven
bottleneck detection method works well and finds the actual
bottleneck. The simulation models studied include twomachine-one-buffer model (2M1B), three-machine-twobuffer model (3M2B), four-machine-three-buffer model
(4M3B), five-machine-four-buffer model (5M4B), six79
MoB2.1
E. Verification Discussion
In fact, the most important task for bottleneck detection is
to identify all the independent local bottlenecks, which are
often located in independent segments and divided by the
effective buffer locations. Buffer locations in the
production line are used to receive and store the finished
parts from the upstream machines and provide parts to the
downstream machines to be processed. Buffer content level
is an important factor which strongly affects system
performance. If a buffer content is seldom zero or full and
often close to a certain level (e.g., half of the total capacity),
then the machines ahead of this buffer are independent of
the machines after this buffer because the buffer can provide
or store enough parts and the production is not affected no
matter how good or bad the performance is upstream or
downstream. This kind of buffer is defined as an effective
buffer, and segments divided by an effective buffer can be
studied independently.
On the other hand, if the content of a buffer is often zero
or full, this kind of buffer is defined as non-effective
buffer and machines ahead of this buffer are dependently
related to machines after this buffer because the blockage
and starvation times of upstream machines affect the
blockage and starvation times of downstream machines (and
vice versa). However, the effective buffer and non-effective
buffer can only be defined qualitatively not quantitatively
because there is no unified standard accepted to determine
the boundary between effective and non-effective buffers. It
is obvious that a buffer in which the content is never zero or
full can be considered an ideally effective buffer. However,
it is less straightforward to define whether a buffer is
effective or non-effective when the buffer is full or empty 10
percent of the time. Fortunately, buffer dynamic movement
(including both effective and non-effective buffers) can be
considered as a blockage/starvation change, and the global
trend of blockage/starvation automatically decouples long
production lines into many independent segments. The
turning point is defined in both a local and a global sense; in
this way, it can identify independent local bottlenecks in
each independent segment of the production line.
Under the ideal condition, for the segment between two
effective buffers, if only one turning point exists, the
segment possibly can be aggregated into a three-machineno-buffer segment.
2,363
2,363
M2
2,403
2,371
M3
2,408
2,376
M4
2,431
2,384
M5
2,461
2,388
M6
2,378
2,378
2,480
2,480
M2
2,622
2,487
M3
2,500
2,486
M4
2,474
2,473
M6
2,461
2,461
to
, blockage is
definition, for each machine from
higher than its own starvation. However, for each machine
to
, blockage is smaller than its own
from
starvation. Therefore, the machines in the small segment
are blocked while the machines after
are
before
starved. Because of these characteristics, the m-machine
segment (including buffers) can be simplified and
aggregated into a three-machine-no-buffer segment
including
80
and
MoB2.1
524-532.
[9] D. Jacobs, and S. M. Meerkov, A system-theoretic property of serial
production lines: Improvability, Int. J. System Science, 1995, Vol. 26,
No. 4, pp. 755-785.
[10] J. Li, and S. M. Meerkov, Bottlenecks with respect to due-time
performance in pull serial production lines, Mathematical Problem in
Engineering, 2000, Vol. 5, pp. 479-498.
[11] J. A. Buzacott, and J. G. Shanthikumar, Stochastic Models of
Manufacturing Systems, Englewood Cliffs, NJ: Prentice Hall, 1993.
[12] S. Y. Chiang, C. T. Kuo, and S. M. Meerkov (2000), DT-bottlenecks
in serial production lines: theory and application, IEEE Transactions
on Robotics and Automation, 2000, Vol. 16, No. 5, pp. 567-580.
[13] S. Y. Chiang, C. T. Kuo, and S. M. Meerkov, c-bottlenecks in serial
production lines: identification and application, Mathematical
Problem in Engineering, 2001, Vol. 7, pp. 543-578.
[14] C. Roser, M. Nakano, and M. Tanaka, Comparison of bottleneck
detection methods for AGV systems, Proceedings of the 2003 Winter
Simulation Conference, 2003, pp. 1192-1198.
[15] Q. Chang, and J. Ni, Supervisory factor control based on real-time
production feedback, Ph.D dissertation, Program in Manufacturing in
the University of Michigan, 2005.
[16] J. M. Alden, L. D. Burns, T. Costy, R. D. Hutton, C. A. Jackson, D. S.
Kim, K. A. Kohls, J. H. Owen, M. A. Turnquist, and D. J. Vander
Veen, General Motors Increases Its Production Throughput,
Interfaces, 2006, Vol. 36, No. 1, pp. 6-25.
[17] S. R. Lawrence, and A. H., Buss, Economic Analysis of Production
Bottlenecks, Mathematical Problems in Engineering, 1995, Vol. 1,
No. 4.
[18] C. T. Kuo, J. T. Lim and S. M. Meerkov, Bottlenecks in serial
production lines: a system-theoretic approach, Mathematical Problem
in Engineering, 1996, Vol. 2, pp. 233-276.
81