Documentos de Académico
Documentos de Profesional
Documentos de Cultura
ISSN 1900-6241
Bogotá, Colombia.
info@revistavirtualpro.com
www.revistavirtualpro.com
2009
Pedro Luís González-Rodríguez
OPTIMISING JIT-CUSTOMIZED SYSTEMS
Universidad de Sevilla
Sevilla - España
e-mail:
pedroluis@esi.us.es
Abstract
In this paper the currently state of the art on JIT systems is introduced, while some
important issues concerning new trends in the JIT applicability are shown. In order to
explain every concept, the paper is mainly written to practitioners bringing close some
research aspects. Additionally, a case study based on a real automobile component factory
in Spain is shown.
Keywords: Optimisation, Just in Time, JIT, customized systems, token based, Kanban,
production control, simulation
Resumen
En este artículo se introduce el estado del arte de los sistemas JIT (Just in Time o Justo a
Tiempo), mientras que se muestran algunos aspectos importantes relativos a las nuevas
tendencias en la aplicabilidad de los sistemas JIT. Para explicar cada concepto, el artículo
es principalmente dirigido a profesionales, acercándoles algunos aspectos de investigación.
Adicionalmente, se muestra un caso de estudio basado en una fábrica real de componentes
de automoción en España.
1. Introduction
Before 70s, industrial companies were still influenced by the T.W. Taylor and H. Ford
production management concepts. After that, a new industrial revolution was switching
from mass production -and Just in Case (JIC) - to Just in Time (JIT). At the operational
level, JIT is implemented by means of the well-known Kanban production control system
(Monden, 1983). Kanban system was developed by Toyota and put into practice in a wide
variety of manufacturing environments. Nowadays is still successfully implemented in
many manufactures.
The Kanban system uses cards on every work station. For a line in one system
formed by an input buffer and a machine, the cards rule can be expressed as follows. A
kanban card is attached to a job (raw material) whenever it enters the input buffer, and the
card is withdrawn whenever this job is delivered to the next station. Whenever there are
available cards to be attached to new jobs, these can enter the station. Consequently,
kanban cards are capping the maximum amount of work in process (WIP) on every station.
Thus, two important observations can be formulated: first, the maximum WIP in one station
is given by the initial number of cards in the control panel in that station, and the latter is
that the sum of all the cards in the line gives us the maximum WIP in the line (see Figure
1).
Control Panel
from Material Flow
previous
station Cards flow
Kanban cards
In other production control systems - which also employ cards to operate- Kanban’s
influence can be observed, see e.g. Conwip, Generic Kanban, Generalized Kanban,
Extended Kanban or some hybrid systems, like the so called hybrid Kanban/Conwip (see
e.g. Karaesmen and Dallery (2000), for analytical results regarding comparison among
different Kanban systems, or Framinan et al. (2003) for a review of contributions
comparing Conwip with other production control systems).
Early in 90s, the Conwip system (Spearman et al., 1990) extended the original idea of
controlling the maximum amount of work for each station, applied to a completely line. In
this case, one production line was controlled by one only “control loop” of cards.
Therefore, this system is also known as Long Pull.
In this system, at the beginning of the manufacturing line, a kanban card is
attached to a job. The maximum work in process is equal to the total number of cards in the
system. When a job has finished its processing, the attached card is released and sent to the
beginning of the line, where it can be attached to another job (see Figure 2).
In order to obtain the “best” system’s performance, only one parameter (the number of
cards of the completely system) it is necessary to be adjusted in this case. This
characteristic makes the Conwip system easier to implement and, perhaps, it is more
accepted by practitioners than another more complex systems.
Which one of both systems reaches the best yield? There is a great controversy
regarding to this question, but in view of the fact that no one system outperforms the other
for all possible manufacturing conditions.
As can be observed, the Kanban system only gets local information (by using the
cards flow) about the needs of material on each station, while the Conwip system only
reaches global information about the state of the line from customer (end of the line) to the
beginning of the line. Hence, Conwip transfers the customer demand faster than Kanban,
while has no information about the particular state of their internal stations.
Both, Kanban and Conwip systems are two opposite cases of the JIT philosophy
operational systems. In the middle there exist different implementations which make use of
local and global information by using different card loops between stations. Hence, it is
possible to find some variations of the original Kanban systems, such as hybrid
Kanban/Conwip systems or even some card-based implementations of the well-known
Drum-Buffer-Rope introduced by Goldratt during 80s decade.
I’m the Process Control Engineer. What should I do? There is not a quickly
answer to this question, but as was previously stated, there is no one system which
outperform the others. In other words, in order to implement the most profitable system, the
pull system you select for your line should be the best under your environment conditions.
That means that the system performance depends on the following relation:
One of the emergent ideas was to consider every possible card-loop between every pair of
stations. In that case, each possible existing pull mechanism is contained into this new
control system. This idea was proposed by Gaury, Renault Engineer in 2000, who termed
this type of systems like Customized Systems (see Gaury 2000, Gaury et al. 2000, or Gaury
et al. 2001). There most important advantage of this approach is that the best pull system
for your environment is a subset of this general control mechanism. Initially it has every
possible information flows about the state of your system with different degrees of global
and/or local information. The main task is to optimize the production control system, which
is the main drawback of these systems. In this case the number of parameters to be
optimized has a quadratic growth with the number of stations and the correct number of
cards on each loop must be established. However, in the optimization process those loops
which do not give any important information about the state of the system under the current
environment will be detected, and later eliminated. In order to do that, we can address the
problem as a combinatorial optimization problem, trying to determine the loops and the
number of kanbans yielding the best performance while maintaining as lower as possible
the number of loops (see González-R and Framinan, 2009). Different standard algorithms
and optimizations techniques can be employed, such as e.g. Simulated Annealing, Genetic
Algorithm, Tabu Search, Variable Neighbourhood Search, Learning Automata Search
Technique (LAST), Ant Colony, the Cross Entropy (CE) method, etc. After the
optimization process, the number of parameters in the system is much reduced that in its
initial state. Furthermore, Gaury observed that only less than the half of the initial loops are
necessary to be implemented. As a result, the control mechanism may be simplified and
consequently results easier to put into real practiced industry.
3. Results
a heat treatment and a rebored process, for finally be assembled with other parts directly
from suppliers. Our studied process is the IML 3ª as shown in the figure. The materials
flow is done among machines, transportation process and stock on intermediate buffers.
IML 1ª
IML 2ª
Heat
IML 3ª Treatments
Rebored Assembly Gear Box
IML AP
IML CA
The studied line consists on ten stations in tandem. Programmed stops for maintenance
labors, important breakdowns and small breakdowns called ‘micro-stops’ are considered in
our study. Once the system is customized, the best yield was reached with the system
shown in figure 4 (circles means machines, while triangles represent intermediate buffers):
It is important to highlight that the current production control system in the factory is until
now, a Kanban system. The utilization of the line is 90% and the WIP close to 350 parts,
which involve a cycle time of 389 minutes per part. Using the JIT-Customized production
control system it is possible to reduce the work in process to 92.33 parts, the cycle time to
102.4 minutes, while the utilization rate would be close to 90.17%. This system results to
be easier to control than the current Kanban system, because of the number of used
parameters was nine, while in the JIT-Customized system is only of two.
It is important to note that the initial Kanban system was, perhaps, not optimized
and better results could be reached with the same control system. However, this system
results more difficult to be optimized in practice that the proposed one. Additionally, it is
necessary to stress that the real control mechanisms could be implemented by only using
the visual cards signals, while additional investment on information systems to support it
seems to be unnecessary.
4. Conclusions
Supply Chain context, like Total cycle time, set-up and lot-sizing reduction, or products
distribution may be applicable to the Supply Chain.
As a summary we can conclude that JIT philosophy offers a tool of interest to
engineers in order to improve their process on a local level in the manufacturing processes
by the JIT-Customized production control systems. Additionally, new trends in the JIT
philosophy shows us how are introducing in a global level by the JIT-Supply Chains.
5. Acknowledgements
This work stems from the participation of the authors in a research project funded by the
Spanish Ministry of Science, grant DPI2007-61345, title Advanced Systems for Integrated
Order Management, and SCOPE, grant P08-TEP-3630, funded by the Andalusian
Government.
6. References
1. Framinan, J. M., González, P. L., & Ruiz-Usano, R., (2003) The Conwip Production
Control System: Review and Research Issues, Production Planning and Control,
14(3), 255-265.
2. Gaury, E. G. A., (2000) Designing Pull Production Control Systems: Customization and
3. Gaury, E. G. A., Pierreval, H., & Kleijnen, J. P. C., (2000) An Evolutionary Approach to
Select a Pull System among Kanban, Conwip and Hybrid. Journal of Intelligent
4. Gaury, E. G. A., Pierreval, H., & Kleijnen, J. P. C., (2001) A Methodology to Customize
Customized Token Based Systems, Production Planning and Control, 20(3), 276-
287.
6. Karaesmen, F., & Dallery, Y. (2000) A Performance Comparison of Pull Type Control
Management Press.
8. Spearman, M. L., Woodruff, D. L., & Hoop, W. J. (1990) Conwip: A Pull Alternative to