THE PULL CONTROL SYSTEMS. A SYSTEM DYNAMICS PERSPECTIVE.
Authors
Adolfo CRESPO MARQUEZ
Rafael RUIZ USANO
José MANUEL FRAMINAN TORRES
Ricardo ZUBIRIA DE CASTRO
Dpto.Organizacién Industrial y Gestién de Empresas
Escuela Superior de Ingenieros
Universidad de Sevilla
Avda. Reina Mercedes s/n. 41012-Sevilla (SPAIN)
Fax: +34-5-4556883
E-Mail:crespo@cica.es
ABSTRACT
The goal of this article is to build a System Dynamics model of a production line under
long-pull control systems. These models are presently being completely tested under different
scenarios. The long-pull control systems appear to offer a higher performance than KANBAN
based systems in those environments where perfects material flow cannot be assured.
1. Introduction.
Selecting a suitable pull control system becomes an interesting concern for
those processes in which the complexity can greatly influence the behavior and
overall system performance (Ruiz-Usano, Crespo & Framifidn, 95).
The kind of “pull effect” is a consequence of the manufacturing organization
and describes the system behavior. For each particular manufacturing and business
environment, the choice of a proper behavior , i.e. how system reacts facing
exogenous uncertainty, permits the achievement of flexibility, lightness, lean
manufacturing, and therefore, increases the system performance.
This paper studies the behavior of a production system controlled by
different pull schemes, KANBAN/pull and CONWIP/long- pull systems.
2. The Pull Systems nowadays.
At present, pull systems are widely used for local production control and
they are generally integrated within the global plant production system. It is
common, for instance, to find these pull structures forming part of a higher MRP
company logic structure, with a lower indenture level. This type of configuration is
frequently named hybrid schemes, which generally require, push(orders) and
pull(cards) signals in order to release the jobs (YAMAHA Synchro MRP) (Hall,
1986).
KANBAN and CONWIP are cards driven pull production systems. The use
of cards (production / withdrawal) in the production/assembly line generates an
effect of pulling the materials flow from the last stages backwards until it reaches
the initial ones.
1o4
CONWIP (Spearman, Hopp & Woodruff, 1990) is an approach very similar
to KANBAN (Monden, 1983), the main differences between both systems are:
¢ CONWIP utilizes only one type of card to control all the processes (long-pull
system). KANBAN uses a different card per process.
« The number of CONWIP cards depends on system capacity features, not on the
production plan for a given planning period, as KANBAN does.
« CONWIP releases a job-only when the following conditions simultaneously
occur: there are free production cards and a positive orders backlog exist.
KANBAN only requires the existence of production orders (cards) and inventory
in the previous stage.
3. The System Dynamics Models.
3.1. KANBAN Model Casual Diagram.
production stage i
aN
(Production Rate)i pletion Rate)i
#
(aventory)i
} + (Work in Procdss yr
(Kanban Cycle)i
(Production Orders)i
+
~ (umber of ‘Banbans)i
(Safety Period)i ES Se
een J (Lead Time)i
(Units per Container)i Production Plan
&
ae Orders Backlog
This diagram shows interaction among variables of a production stage (the
global model includes three production stages), where three feedback loops can be
appreciated. The influence of the production plan in the stage production rate,
through the the variable “Number of Kanbans”, is also described.
Nki = (PP/Uci)(Lti + Cki)(1 +SSi)
Where:
Nki: Number of Kanbans in stage i
PP: Production Plan
Uci: Unit per Container in stage i
Lti: Lead time of the stage i
Cki: Kanban Cycle of stage i
Ssi: Safety Period for inventory of finished parts in the stage i
The variable “Number of Kanbans” is modified according to the planning
period considered in the plant. When a new planning period starts, the variable
WG
“Production Plan” is updated considering the available information about the
market demand.
3.2. CONWIP Model Casual Diagram.
roduction stage i
beeesese (Inventory)i-1 a
NN i
+ eee =" E Rate)i
+ Work in Process i -
- (Inventory)i .....
matey (safety period)i-1
production stage 1
Lead Times
+
Number of CONWIP cards Total Work in Process
_
Number of CONWIP free cards
: \
(Production Rate)|
ee
Orders Backlog
This diagram describes how the number of CONWIP cards only influences
the first stage production rate, and this rate will push the rest of the subsequent
production stages.
In case of CONWIP systems, it is demonstrated that the suitable number of
cards to use does not depend on the production plan, but on lead times. For
instance, it has been suggested (Ruiz-Usano, Framifidn & Crespo, 1995):
Nunber of CONWIP Cards = (2 Lti)/Lton
i
Where,
Lti: Lead time of stage i
Lton: Lead time of the bottleneck stage
4, Experiments.
The experiment presented in this paper contains several undesirables
situations which may appear in a manufacturing plant, like great demand
variations, bottlenecks or eventual demand higher than plant capacity. The
topology of the line is simple, consists of three production stages in line (flow
shop), which have an inventory of finished parts between each stage and the
subsequent one. Orders backlog and materials flow are exactly equal modeled for
both systems KANBAN and CONWIP. A planning period of one week is
considered for the KANBAN system.
\\\
ORDER BACKLOG
2,000
1,500 aN
1,000 |-—7 1
500 ra a
50.00 100.00
THROUGHPUT
100 777
50 fe
0
50.00 100.00 ) 50.00 100.00
CONWIP
LEGEND:
= KANBAN
5. Conclusions.
Simulation results demonstrate that CONWIP reacts earlier to great demand
variations, achieves lowers limits of total inventory in the system, and sustains
similar values for the throughput. Therefore, the average lead time for the whole
process, and the kind of selected scenarios, reaches better values with the utilization
of CONWIP systems instead of KANBAN.
6. References.
° Hall R.W., “Synchro MRP: Cmbining KANBAN and MRP: The YAMAHA PYMAC
System”. 1.E.M.P. 1986.
e Monden Y., “Toyota production System: Practical Approach to management”.
Industrial Engineering and Management Press. Norcross. 1983.
e Ruiz Usano R., Crespo Marquez A., Framifidn Torres, J.M.., “Advanced
Manufacturing System Dynamics”. ISDC’95 Proceedings (Parallel Program. pp. 888-
896). Tokyo. 1995.
¢ Ruiz Usano R., Framifidn Torres, JM. & Crespo Marquez A., “Determinacion del
Mimero Optimo de Tarjetas en un sistema de Inventario en Proceso Constante”. SEIO.
pp. 635-636. ISBN: 84-87215-46-7. Sevilla. 1995.
e Spearman M.L., Woodruff d.L. & Hopp W.J., “CONWIP: A Pull Alternative to
KANBAN”. International Journal of Production Research. Vol. 18. No.5. pp.879-894.
1990.
Wa
