Академический Документы
Профессиональный Документы
Культура Документы
Seminar Report
By
Lijo John
CONTENTS
1. INTRODUCTION
2. PARADIGM SHIFT
3. WLC PARADIGM
4. WLC CONCEPTS
6. CONCEPTUAL BACKGROUNDS
8. EXAMPLE
9. CONCLUSION
10. REFERENCE
ABSTRACT
Since late 1940’s Kanban, a card based control system was in use, but kanban is
still being used for the repetitive manufacturing. Thus in case of a make to order
job shop manufacturing this card based system fails. Thus a need for a new card
based system was realized for the job shops. The Cobacabana ( control of
balance by card based navigation) system is being proposed here by the
author. This system is based on the concept of work load control (WLC). Due to
recent developments in the understanding of the concepts of the work load
control it has become possible to convert the work load into a robust card
system. Here the card system is being divided into two. First is the release cards
which form a loop catering to release functions of the job to shop floor. Second
is the acceptance cards that forms the basis of accepting a order by the sales
department. Finally the implementation issues are also discussed in general
1. INTRODUCTION
Industrial practice shows that hardly any job shop is able to use the planning and
control modules provided in its ERP-package. Many solutions provided in ERP-
packages focus on Gantt chart or Leitstand scheduling, which is generally
doomed to fail in job shops because of the high data maintenance
requirements and because of their high sensitivity to uncertainty, resulting in
unstable schedules. Other ERP- packages only provide material-oriented
planning solutions such as MRP, while capacity planning and control is critical in
most job shops. To prevent from turning back to legacy systems, these
companies often opt for a planning and control system which can be
implemented with limited software support. This creates an obvious need for
card-based systems in job shops.
The popularity of card-based control systems has been rising since the
introduction of Kanban as a material control system for repetitive manufacturing
environments. During the last decennium new card-based systems such as
POLCA (Suri, 1998) have been developed, which can be implemented in
capacity-oriented control situations. The Generic POLCA system (Fernandes and
do Carmo-Silva, 2006) can be seen as an important step to make POLCA
principles suitable for the specific situations of job shops. It links POLCA card loops
with capacity allocations at the order release decision. Still, the basic idea of
POLCA to use card loops for each possible combination of successive work
centers will reduce its practical applicability in shops with high routing mix
variability. In a computerized system, POLCA principles have been adapted for
use in a job shop. But, as also concluded by Stevenson et al. (2005), none of the
currently available card-based systems will meet the dynamic requirements of
job shop manufacturing.
2. WLC PARADIGM
station.
WLC concepts do not release jobs to the shop floor if they are expected to
cause queue lengths to exceed certain workload norms. It results in a pool of
jobs waiting for release. As illustrated by Fig. 1 we refer to waiting time in the pool
as the pool time and to the interval between release and completion of a job as
the shop floor flow time. The shop floor flow time of a job can be subdivided into
station flow times. The pool is a new object of control. Unrestricted acceptance
of jobs at the entry could cause excessive pool times.
Fig. 1. Lead time components.
A hierarchical control concept emerges [Kingsman et al, 1989], with three levels
which respectively relate to job entry, job release and priority dispatching (Fig. 2).
At each level, we distinguish two means of control, input control and output
control. Input control regulates the allowance of jobs to the next stage,
respectively accepting jobs for entry into the pool, releasing jobs to the shop
floor, and dispatching jobs for processing (thus allowing a job to enter the queue
of its next operation). On the output side, capacity management contributes to
the control of workload through regulation of the outward flow, by means of
respectively medium-term, short-term and daily capacity adjustments.
Fig. 2. The hierarchical WLC concept.
The job entry level is very important, if one can influence the incoming orders. In
that case, order acceptance and due date assignment/acceptance can
support the release decision, providing it with a 'releasable' set of jobs, thus
keeping pool times small. In fact, the job pool between entry and release acts as
the visualized imbalance between job supply and production capacities. The
role of priority dispatching in WLC is a very modest one, because the choice
among jobs is limited due to short queues. Generally, WLC concepts favor
dispatching priorities such as first come- first-served (FCFS) which stabilize
operation flow times or due date oriented priorities which correct progress
differences among jobs. These kinds of priorities facilitate a good timing of job
release. However, the major strength of WLC concepts is withholding jobs from
the shop floor, reducing average queue lengths. Besides a reduction of work-in-
process, withholding jobs from shop floor has numerous additional advantages
as it enables management to delay final production decisions [Irastorza et al,
1974]. It reduces waste due to cancelled orders, facilitates later ordering of raw
materials, takes away the need of expediting of rush orders, etc. Fluctuations in
the incoming order stream should be absorbed by the pool. Altogether, it should
create a stable stationary situation on the shop floor. Only restricting queue
lengths is generally not sufficient. If average queue lengths decrease but
variances do not, the idle time at work stations will increase. This situation is not
allowable for the common job shop, where many work stations can be
temporary bottlenecks. The loads of potential bottlenecks should be kept close
to a norm level instead of below a norm level. The release function which aims at
short queue lengths and a reduced variability of queue lengths is called load-
balancing.
In summary, WLC concepts try to create a situation on the shop floor of short and
stable queues. A pool of unreleased jobs, buffers the shop floor against external
dynamics, the incoming non-stationary job stream. The queuing of jobs on the
shop floor is turned into a stationary process. Release performs a key-role in
reaching this stationary situation. It is the most elaborated function within WLC
concepts.
3. WLC CONCEPTS
In the preceding section we have seen that release should control the queue
lengths in front of each work station. The queues must be short and stable, the
load-balancing function. On the other hand, each job should be released timely
with respect to its planned due date and expected flow time, the timing
function.
Leaving out capacity decisions at the release level, two components of the
release decision are distinguished: a sequencing decision and a selection
decision. The sequencing decision can be described as the setting of priorities for
jobs to be released, 'selection' decides whether a job will be released or not at
some specific moment. Most WLC concepts focus the sequencing decision on
timely release and create due date based sequences. Taking into account this
sequence, release selects a set of orders that keep the workload of work stations
at certain norms. These workload norms are the main instrument of workload
control.
The workload considered in the concept of Bechte is the queue length at a work
station (in units of processing time). The workload is controlled by the load limit.
The load limit LLs of a work station s consists of two components: the planned
output during the release period and the planned queue length at the end of
the release period. The actual output Os, during the release period and the
actual queue length QEs at the end of the release period satisfy the balance
equation:
QEs + Os = QBs + Is
with
Is : the input to the queue from jobs arriving during the release period.
The release decision at the beginning of the release period must bring QEs + Os at
the norm level LLs. The above balance equation is used. QBs is known at the
moment of release, the queue input Is is influenced by the jobs on the floor
upstream of s and by the release of new jobs.
The workload definition of Bertrand covers the processing time of all jobs on the
shop floor which still have to be processed at the work station concerned. The
corresponding workload norm consist of two components: the planned work
station output during the release period and the planned quantity of work
upstream or in the queue at the end of the release period. An extended
balance equation can be used to determine the actual workload of a work
station s at the end of the release period:
with UEs as the processing time (on s) of jobs upstream at the end of the release
period, UBs as the processing time of jobs upstream at the beginning of the
release period, Rs as the processing time of jobs released at the beginning of the
release period. At the moment of release the right-hand side of this equation is
completely known. The processing times of all jobs which are newly released are
the input to the workload. Thus, the release of new jobs directly influences the
workload. The release decision can be made without a sophisticated estimation
procedure.
with DEs as the processing time (on s) of jobs downstream at the end of the
period, DBs as the processing time of jobs downstream at the beginning of the
period, Cs as the processing time of jobs which leave the shop during the release
period. All other variables as defined before. Again all right-hand-side
components are known at the moment of release. The WLC concept does not
clarify whether the shop output Cs from jobs fully completed during the release
period is included in the workload norm. Notice that the workload definition
further simplifies keeping up with the actual workload as it avoids the need for
data regarding the completion of single operations. The completion of the job
can be reported when it leaves the shop floor.
Most classical variants of the WLC concept take the release decision periodically
according to the following procedures. Orders in the pool are considered for
release in the sequence of their planned release dates. The order being
considered is added to the release selection as long as its release will not cause
any workload norm to be exceeded. Otherwise the order will have to wait in the
pool until the next release opportunity. An order with a later planned release
date maybe selected when it does fit in the norms. After this procedure is
completed, selected orders are sent to the capacity groups performing the first
operation and remain on the shop floor until all operations have been finished.
The five most distinguishing elements of the WLC approach to shop floor control
are the control point at release, the use of aggregate measures, resource
buffering, shop floor buffering, and central load buffering.
5.1 Control point at release
The main control point of the WLC concept is the release decision. This decision
precedes the first shop floor operation of the orders. At this point fitting the orders
into workload norms should create predictable operation lead times.
Downstream on the shop floor, simple priority rules at capacity group s are
sufficient (Bechte, 1994). Examples of priority rules are first-come-first served,
which guarantees the smallest variation of operation lead times, or due-date-
oriented rules to correct for individual progress disturbances among orders. No
sophisticated methods are used for controlling the downstream operations of the
orders. Although some of the orders arriving at a capacity group may come
directly from the pool, a significant amount may come indirectly via other
capacity group s which perform the upstream operations of the order ( Fig. 3).
The decision to allow an order for release depends on the shop floor situation,
which is reflected in workloads. Workloads are calculated as an aggregate of
individual processing times. Most workload definitions also count up the
processing times of orders waiting in front of a capacity group (direct load) and
those of orders upstream (indirect load), as shown in Fig. 4. The general
assumption is that variations within an aggregate measure of summed
processing times will be relatively small. Therefore, decisions will be rather
insensitive to individual processing time deviations.
Control within the WLC concept is not based on filling the capacities of resources
in a time-phased plan as for finite loading or deterministic scheduling
approaches. Instead it is based on maintaining a buffer for the resources in a
capacity group, by keeping workloads at norm levels. Although different types of
workload norms can be used,
the orders allowed on the shop floor after release will normally contain more
work than the capacity groups can handle before the next release moment (see
Fig. 5), resulting in queues of orders in front of the capacity group s. WLC is
essentially designed for situations where queues are inevitable, coping with
variations in order arrival and processing times
Even though resources are buffered by queues, these queues are kept small. As
far as possible the waiting time is placed before the first operation in the form of
pool waiting time. Thus, the main buffer is placed before the shop floor (Fig. 6).
The pool should absorb all kinds of fluctuations in the arriving order flow in order
to keep the resource buffers small and stable. Pool waiting times of orders may
vary according to their urgency, which is reflected in the slack to planned
release dates, and whether they fit well into the shop floor situation, which is
reflected in the workloads.
The main decisions of WLC are made centrally. The release decision compares
the urgency of orders and balances loads among capacity group s. This requires
a global view of the shop. As mentioned before, local decisions at individual
capacity group s can be based on simple priority dispatching rules not requiring
global information.
The central balancing of loads by fitting the orders from the pool into workload
norms (Fig. 7) will keep the resource buffers stable, despite variations in the
arriving order flow.
6. CONCEPTUAL BACKGROUNDS
The direct load LDst of workstation s at time t being the sum of processing time so
fall orders waiting or being processed can be specified by equation .In the
calculations below orders will be addressed as jobs.
with pjs is the processing time of job j at station s; J the set of all existing jobs; tQjs
the time of arrival of job j at station s; tCjs the time of completion of job j at station
s. The indicator function I(t) is defined as I(t) = 1 at the specified interval, I(t) = 0
otherwise.
Notice that the release of a job will not directly affect the direct load of a
workstation, unless the station performs the first operation in the routing of the
job. Therefore a group of classical release methods focuses on controlling
aggregate loads. Aggregate loads additionally incorporate all work which is still
upstream of the workstation being considered. The aggregate load LAst of station
s at time t can be specified by
with tjR time of release of job j (all other variables as defined before).
The control of aggregate loads does not necessarily leads to control of direct
loads in case of a fluctuating routing mix (Land ,2004). Moreover, the average
direct load LDst resulting from the set of jobs in the aggregate load of a
workstation will be at the level calculated as
with Ujs: the set of stations in the routing of j ‘up to and including’ station s.
In the simple case of equal planned values for all workstations the estimated
ratio will simply be equal to 1/njs with njs being the routing position of station s. In
that case each job being released will in crease the workload with pjs/njs for
each station s in its routing. After the respective operation at a station is
completed, the workload will decrease accordingly.
The above equations have been developed in WLC research to facilitate the
specification of a constant norm level independent of routing mix changes. The
norm level forestation can be specified as its planned direct load level and
according to Little’s result (specified in terms of processing time units, see Land,
2004) the planned direct load should be proportional to the planned station
throughput time. But even when the planned throughput times to be realized are
the same, each station may still require a different workload level, since
capacities may differ among stations. Therefore, further standardization can be
realized by depreciating the direct load contributions by the maximum output
O*Ds of station s during the planned station throughput time (T*Ds ). The maximum
output is a fraction of the station’s capacity as100%utilizationcannotberealized. It
is measured in processing time. After transforming this into a percentage, the
workload calculation results in the following:
Thus, when each job contributes an amount Cjs as specified to the calculated
workload, this can further simplify norm setting. The norm for the workload LDst
can simply be set at 100% for each station and planned station throughput times
T*Ds are the main parameters to be specified.
The sequence of considering jobs for release is classically based on their planned
latest release dates t*Rj which can be derived from the due date δj and the
planned station throughput times T*D .
By creating constant norm levels, the above definitions and calculations can be
used for translating workload requirements into numbers of cards. This forms the
starting point for development of the card-based system for job shop control.
The release cards authorize the planner to release new orders. To release an
order the planner has to attach the right amount of cards for each work station
in the order routing to the order guidance form. The cards relating to a certain
work station return to the planner after completion of the operations performed
by the considered station. The system balances the work load for all critical
workstations by allowing a fixed number of cards summing to 100% per station on
the shop floor.
The task of the planner is supported by a display for collecting and distributing
the cards. The simple display, as shown in Fig. 9, gives a quick overview of the
situation on the shop floor. In Fig. 9 each card represents 5 percent. Empty card
positions indicate the percentage of the planned station throughput time
already filled by released orders. The available cards on the display show the
possibilities for new releases. An even distribution of released cards a cross
stations indicates a workload which is well-balanced across stations. Similarly,
momentary bottle necks can be easily identified, which is mandatory for focused
decision making in a job shop. E.g. the lathes required for turning seem to be the
restrictive resource for new releases in the situation depicted in Fig. 9.
Fig 9:Display of available release cards, indicating the actual shop floor situation
The release system can be extended for support of the order acceptance and
due date promising function. A minimal delivery time djmin for an order can be
determined by adding an estimated waiting time T*Pj in the order pool before
release to the sum of planned throughput times T*Ds for stations in the routing of
the order
Fig 10 Cobacabana acceptance card loops, between order acceptance and release
Notice that the denominator O*Ds /T*Ds specifies the maximum output (in
processing time units) per day. Thus each card represents a fraction of a working
day.
The Cobacabana system, in a job shop in real time would work like as
summarized below. The steps are given in the order it has to be followed in the
job shop
The constant number of cards moving between the release function and the
workstation will keep the station throughput time within their planned levels,
allowing for a good timing of order releases and for predictable throughput
times when promising delivery dates. The procedure of selecting orders for
release will ensure that also the relative urgency of orders is considered .The use
of acceptance cards—withdrawn from the acceptance display—enables a
good estimate of the waiting time before release, which is the main variable
component to be estimated for quoting a realistic delivery date.
8. EXAMPLE
Consider a job arriving at the job shop which requires the following operation in
the sequence-
TIME
OPERATION (mins)
SAWING 20
TURNING 30
POLISHING 40
DRILLING 15
FINISHING 25
djmin = 20 + 30 + 40 + 15 + 25
= 130 min
= 20 min
3. Acceptance cards
= 5 uts/hr
= 4 cards
4. Contribution to workload
C js = 20%
Calculation for other operations are being summarized in the table below
TIME Ods/Tds
TURNING 30 180 6 5 4%
DRILLING 15 45 3 5 11%
FINISHING 25 125 5 5 5%
9. CONCLUSION
A new card based system has been designed by using the workload control as
the controlling factor. Due to the recent advances in understanding the
concept of workload control, it is possible to convert the workload into cards.
These cards can be effectively managed, controlled and navigated to obtain
the balance in the system. Cobacabana is essentially a link between the sales
department and the operations department of the industry. It has enabled the
sales department to quote realistic delivery dates with the help of the
acceptance cards and operations planner to release new orders with the help
of the release cards.
• Irastorza, J.C and Deane, R.H, 1974, Loading and balancing methodology for
job shop control, AIIE Transactions, 6, 302-307.
• Henrich,P.,Land,M.J.,Gaalman,G.J.C.,vanderZee,D.J.,2004b.Reducing
feedback requirements of workload control. International Journal of
Production Research , 42,5235–5252.