DRUM-BUFFER-ROPE SHOP FLOOR CONTROL
Eu SCHRACENHEM
The AY, Golda insite, 42 Orange Stet New Haver, CE OSS
Boaz RONEN
Faculty Of Mesegemet, Tel Asia
‘The drum-butfer-rope (DBR) methodology now
‘being implemented ina growing numberof manufac-
turing organizations enables better scheduling and
decision making on the shop foot * It's based on the
theory of constraints (TOC) [1,2], global managerial
‘methodology that helps the manager to éoncentrate
‘on the mest critical issues, TOC focuses onthe system's
constraints, their exploitation acording to the goal of
the organization, and the implications of exploiting
‘these constraints on the rest of the syste.
‘The concepts of optimized production technology,
DBR, and TOC were developed in 2,3, and 4]. Gold
ratt [2] delineates the connection of the OPT principles
to the steps of TOC, and a system overview of DBR,
‘TOC, and OPT is proposed in (6).
“The focus of OFT was the identiation of botle-
necks in the manufacturing process, withthe objective
fof basing the scheduling effets on these bottlenecks.
‘The essence of OPT philosophy was axpeessed by nine
rules [6], which have been replaced by Gve more gen=
eral and accurate steps constituting TOC, The term
“potteneck’ isreplaced by a broader term: constraint,
efined as anything that mits system from achieving
1igher performance relative tits goal.
“The five steps:
‘Step 1. Identify the system constraint(s)
Step 2, Decide how to exploit the consrait(s)
Step 3. Subordinate everything ese to the above de-
Step 4. Elevate the system constraints).
Step 5 If inthe previous steps, a constraint has been.
broken, go back lo Step 1, but do not let “ine
tia” become the system constraint
Drum-Buffer-Rope—A Scheduling Technique
‘A “drum isthe exploitation of the constant of,
the system: the constraint dictates the overall pace of
the system. The constraint may bea resource, market
* EOFTORS NOTE Ihave pont vere two sacetl i
‘lenentors in pn era, one of oc gear company
esl, Tel Anta 69978 nel
demand, sarce aw material, of management policy
Inmany cates, a drumhastoinclude a detailed Sehed-
ule of the constraint in order to ensure the exploitation
ofthe constraint
A “butler” is protection time, Buffers are used to
protect something from adjacent disruptions, The
protection is expressed in time units, since the parts
planned to reach the protected area some tine be-
fore they are scheduled to be processed, Disruptions
‘might sem from a variety of reasons breakdowns,
absenteeism, Guctuations in setup Himes, unreliable
vendors, sap, of just unavadabilty of a certain re
source because ti being used in other jobs. Bulfers
are planned only in crtcal areas that need tobe pro:
tected; it is obvious that the drum (Le, the schedule
fof a capacity constraint resource) should be protected
from disruptions on adjacent operations,
‘A “rope” is a mechanlem to force all the pets of
the system to work up to the pace dictated by the
drum and no mote. This is done, in DBR implemen-
tations, by creating a detailed schedule for releasing
raw materi into the shop Boor.
‘Scheduling in DBR
‘There are tree basic steps:
‘Step (a) Schedule your constraint(s) (in order to do
this you wil have to determine the master
production schedule atthe same tine). The
MPS is subjected to the capabilities of the
system constraints only. Next exploit the
Constraints according tothe organizational
goal. (A bottleneck that works all the time
{8 not necestanly exploited. lt should work
‘on products that are the most profitable and
only on those which are going to be sold
scon)
Step (9) Determine the buffer sizes,
Step (c) Derive the materials release schedule ac-
cording to steps (2) and (b).
‘These three base steps are repeated every time the
planning proceasis executed, Two mere general aeons
fare assumed tobe taken whenever necessary:
" [PRODUCTION AND INVENTORY MANACEMENT JOURNAL—Thind Quer, 190,+ Identify the system constraints, :
‘Determine a general subordination policy’ which
concerns the way the nonconstraints should be ace
tivated, This stepmay ormay not include a detailed
cor patial schedule forthe nonconstraint work cen=
ters.
Exploitation
Step 2 is exaclly the definition of a drum—if the
constraint Hes in the marke, then a detailed shipping
schedule isto be derived in order to satafy the whole
market. In this ase scheduling the shipping i the es-
sence ofthe constraints exploitation, and the detailed
shipping schedule isthe drum,
Step (a) of DER implies constructing the MPS and
the detailed shipping schedule according tothe con-
straint capabilites and maximizing the contribution
the bottom line of the constraints avalable time (1,
4]. Scheduling the constraint [Step a] is a farther im
pleation of the constraints exploitation. Inthe case
‘ofan internal constraint a detailed schedule is needed
‘because improvisations are risky to the exploitation,
and also because the subordination areas have to be
aware in advance, of the drum needs,
Subordination
‘The purpose of Step 3 isto keep the exploitation
intact. In a stochastic envionment, there is the risk
that the drum willbe exposed to disruptions that occur
elsewhere. The iamediate action that can be taken is
to plan butfers in front of the drum and the shipping
dock, The buffer sizes reflect the amount ofthe Hc:
tuations and the capscity level at nonconstraint re
sources, The capacity level is crucial because it deter~
‘ines the time needed fora nonconstraint resource to
catch up after a disruption,
Ropes, another mechanism derived from Step 3,
censure that the nonconstant wil be subordinate 0
‘he drum by forcing the system to contain only material
that is scheduled by a drum in the next buffer time
fame. A nonconstraint resource that is engaged in
processing parts that are not scheduled by a drum be-
haves in contradiction to the subordination rule, be-
cause it does not take advantage ofits exces capacity
to protect the drum.
‘The Environment Simulated
‘The simulation handles a plant with stable market
demand. The horizon is two weeks.
‘The Plant Resources
‘The plant uses six different machines. The physical
Jayout is shown in Figure 1 and the routing layout is
siven in Figure 2,
DRUM-AUFFER-ROPE SHOP FLOOR CONTROL
Product is made from raw materials A,B, and D.
Machine N processes mA (raw material A) 13 rain
con average (TPP), and delivers the part tothe W ma-
chine for 11 min, ete. Reeource Ris an assembly ma-
chine, each assembly operation beinga simple I: as
sembly. The upper row specifies the total weekly de-
sand for product (60 per week) and] (40 per week).
‘There are 20 pieces of WIP past E2 operation (ready
to-be processed by M at E3). Another 10 le past the
G2 operation, and another 10 past the G3 operation.
‘A table ofthe work load on the resources, along,
with more simulation details, can be obtained upon
written request ta ether author.
APPLYING THE DBR TECHNIQUE
Mis cleanly a capacty constraint ofthe system. The
‘market demand for the most proftable product is 3
‘constraint 2s well For a discussion of profitability see
uae
Step (a): Schedule /Exploit the Constraint
Exploitation decisions and the MPS must be deter-
mined. The current weekly demand reflects the max-
{mam capacity of M. A possible MPS for two weeks
‘would be 120 1 products and 80 J products, Shipping
wil be done in four equally spread shipments. The
total work load (not Inuding downtime) required
from M during the two weeks, including 12 setup, is
44760 min out of the 4800 min avaiable, The market
demand sto be exploited by supplying everything on
ue date,
lInscheduling the constraint itis ear that E3 should
Derthe fst operation, since it already has the pars for
wee VT) tee =] [R]
wom NF] mae ML
moe YZ] meme Wag]
Frcune
Resource layout Apart from the average
‘setup times, thor downtime mesoclated
‘wth each resource. The dowatine is 0%
‘on average excepten the Mmachine, which
has only 1% downtime, Downline in peo-
portional to the time he resource be busy,
In producton or setup.
wle TE
Product 1 Product J
Weekly Demand eae 40
[eee]
5 Y to |ixaniio|
6 R
4 R
LT,
3 M10 M12] M 25 w 8
mo [enon] fomgule Eman
2 wo w 4 {woos v2
io f=
4 R98 1s]
Raw Matertals rmA rnb rm rmF mG
A Bc Do «£ FOG
FIGURE 2: Layout ofthe routings for two products | ands ow a Bottom up Each box Is wn operation, done by a
rasource denoted by ts capital etr. Numbers Ganote average ane per part (TPP), Each operation fe
recognizable by its column/row, La the lafenest W operation, being in column A and row 2 fv A2.
PRODUCTION AND INVENTORY MANAGEMENT JOURAL—Mhind Qute,1990the fst batch, Developing the rest i straightorward.
‘The M schedule is given in Table 1. The lat line (for
3) is the demand for the third week, Its included
because the rope needs to beset a the second week
for week 3 demand.
‘Step (b) of DBR: Determine the Buffer Sizes
Only thre types of buffers are needed:
+ Constraint butfer: protecting M's operations Size
chosen: 9 he. The inal size ofthe constraint buffer
twas choaen intuitively onthe basis of several initial
"Smulated runs made to gve some sor of "gu fel-
ing.” The inital decision has to take isto consider-
aaton thatthe bufer size should be equal 08 fairly
Jong, yet sil realistic lead time tothe constraint. In
a stable envizonment, operating under DBR meth-
‘dology, where WIP is being limited by the rope
schedule, the butfer size may be three ties the av=
cxage lead time to the constraint. The “thre times
{based on experince and depends on realistic ead
time distibuton
«Shipping buffer protecting the dive dates. Size
‘chosen: 6h, Ths butfer protects the due dates from
disruptions on the way from the constaint bulfer
to the shipping dock.
* Assembly butfer Size chosen: 9 he. Ths type of
Dbatferis needed in eevain envizonments, Here itis
created for the F4 assembly operation only. Assem-
Diy operation i has two feding legs: one caries a
part that was processed by the constraint and the
ther caries a "nonconstant part.” The leg with
the nonconstant part neds buffering (protection),
otherwise the carly constraint parts will have to
wait for the nonconstalnt o reach the assembly.
ABLE 1: The M (Oram Schecule
Propcod
Operation unity ‘sett
s = 3
& 2 a
o 2 “
‘DRUM-BUFTER.ROFE SHOP FLOOR CONTROL
‘TABLE 2_ Rope Purchasing Schedule
Tew Maer nny *
. = °
x Ey 5
8 2 %
6 B °
e 2 =
5 Fy =
: 2 2
‘Step (c): The Rope Schedule
‘The calculation Is apparently simple: The M's
schedule minus the bulfer size, The rope for the G
rawematerial parts are tiggered by the assembly
buffer. The derived schedule for that buffer isthe same
as the schedule forthe E3 constraint operation. The
rope purchase schedule is given in Table 2."
‘As C3has to stat at about hour 8, the release should
bbedone at hour ~1. This means that forthe frst C3
‘operation, the actual buffers only 8h. This concludes
the scheduling stage according to DBR.
Subordination Policy f
the Nonconstraiat
Resources
‘The W resource is quite loaded, in spite of the fact
that its nota constraint. For this simulation run, a
schedule for W was derived bated on the constraint
and the extra operation on G3. A contol technique
tiled "buffer management” is usually used to monitor
the execution of the schedule and for spoting prob-
Jems which led tothe decision to teat W diferently
from the other nonconstaint (ste 7).
Tn order to keep full exbiity in subordinating the
‘other resources, the following working policy was in
corporatd into the simslaton‘you don’t have anything o do, do nothing!
‘= lf work appears somewhere and you are not busy,
setup to the new job.
‘If you have mote than one job to choose from,
chooce the one with the biggest work load, This rule,
‘which is oppeste tothe shortest processing Sine
decision rule, was chosen because it appears natural
toa foreman. The fact thatthe resource has excess
capacity enables usto let the foreman take decisions
inva simple manner. If and when this leads to a
‘wrong dedsion which might affect the system, baler
-management technique is supposed to reveal it on
time to make corrections.
‘= Once you start an operation, switch to another job
only if you are idle atthe moment.
Scheduling the capacity constaint did not include
the lead time from the constraint to the shipping dock.
“The only concem isthe last shipment. The shipping
buffer was chosen tobe 6hr and thus it was estimated
‘hat the average ead time, from the capacity constraint
to the shipping dock, will be less than that. The lst
M operation takes about 5 hr, 30 we can expect that
about half of the last shipment wil arrive when no
overtime is added (the average led time is estinated
tobe? 02.5 hr).
‘The Simulation Runt
Sinteen discrete runs were performed under the
above rules. On average 11.5 units, out of 30, were
‘missingin the last shipment Allinall the expectations
were met it was expected that the last shipment of,
product I would not be fished, Overtime of Ito 4
hr (in the R and Y work centers and sometimes also
1M), was needed in order to finish everything. The
results confirm the expectations af having, on average,
more than half of the shipment ready, even without
overtime.
ANALYSIS
‘The basie approach of DERis of exploiting the sys-
tem constraints and subordinatng the rst ofthe sy3-
tem to the exploitation of the constrains. This i the
‘essence of Steps 2 and 3 of TOC.
How Good Is the DBR Solution?
‘Any optimal solution regarding system performance
has to conform to Steps 2 and 3, This is logialy de-
rived from the definition ofa constraint (anything that
limits the system's performance). Any optimal sched
tule has to exploit the constraint and to protect the
explotation—subordinating tot. A DBR schedule cn
be improved by
+ Improving the exploitation.
' Improving the subordination.
‘Can the Exploitation be Improved?
“There are trade-offs which depend on the market
behavior and how the goal deiniion teats short-term,
profits elative to long-term pots (assuming profs
the system's objective function). One tadeatf con
cera the numberof shipments per week. If we ship
only one into weeks the plant wl be 2c pro-
duce more. On the other hand, a ead tine of two
‘weeks might cause a loss of market in the furre. This
trade-off decision & a key factor in the exploitation
stage Step (8).
(Can Subordination be Improved?
‘There is 4 need to find the balance in the buifer
sizes to ace that they are the smallest that protect the
‘constraint schedule (the drum). The way the noncor
straints are actualy activated might induence the nec
‘esary minimum size of the buffers, Buffer mana
‘ment Isa contol technique, which provides a better
way to handle the nonconstraints (7).
ACKNOWLEDGMENT
‘The authors wish to thank Eli Goldtat fr his en=
Lightening comments,
[REFERENCES
1, Fou RE, "Tee Const They” NAA Coforc 1988,
Monnaie (1988).
2. Gaunt EM, “Conpeteid Shop Rane Schad”
‘ero oa of Precin Rae VoL 3,8, 3 (1988),
3 Clee Mand Cox}, The Gat Nath ive Pre tone
tudo, (1985),
4, Cou EM. and Fox RE, The ac, Nach Rv Pres
‘Cees, NY 98),
5. Hone Band Poy 5, "MMS: Manag Managenent
lnfroton Syste” working pape Fey of Management
‘Teany Univerty (1989),
(6 Ronen Band Sat, MLK, “Syed Manuacurg #8
{Wopr,feem Pncce t Thea” Comptes and Ideal
Engrg (989),
7, Shrngenein EM. and Ronen, 8, “Blfer Management A
‘uaa Toso Proc Cn wen esl
of Management Tate Unie (1989).
[Note:The two working papers [8] and {7] 25 well as the
‘Smulton deals, may be obtained yp request ro ther
thor (adarese on tile page).
PRODUCTION AND INVENTORY MANAGEMENT JOURNAL—Thind Quer, 1950,