An algorithm for finding optimal or near optimal

solutions to the production scheduling problem

hy George IH. Brooks

Anwoiate Professor of Industrial Engincering, Purdue Unisersits

and Charles R. White®

Operations Kesearch Analyst, Armour Company

ABSTRACT @ A procedure has been developed for finding optimal solutions to the scheduling prob-
lem for a wide variety of possitle criteria and constraints, Ay most practical problems are muck
larger than could be soloed economically by this. procedure, the idea of using the lower bound as a
decision rule Jor developing a single near optimal solution was developed. The experiments on
these decision rules showed that they might profitably be used in place of other rules for certain
applications,

@ lv industrial situations the need to determine the se- all commodities will be combined into a single arraycalled
quence in which jobs should be processed on machines the sequence array and denoted by 8. In general this will
offen arises. It is hoped that the procedure presented here be as follows:
will help in finding sequences that are either better than a
[411 Miz * + = Mtn |
shop foreman ean make up by hand or, if only as good, } }
may be generated more economically, Often the hand S=|: jap Beg t+ «Bap |
schedules may not anticipate difficulties, and few persons
will try to improve a sequence once one has been decided
eet as ~~" faa)
upon. The procedures to be shown. here for finding. se-
quence are applicable to companies haying stall size where#,; is the order in whieh the ith job will be processed
business computers such as the [BM 1400 series data on the jth machine, For example,if a given three jobs will
processing systems, The procedures and results estah- be processed on four machines, and if each will be se-
lished here are developed from an algorithm by Giffler quenced on four machines in the same order, then
and Thompson (1). The solution closely resembles a
123 4
branch and bound algorithm developed independently
by Little, Murty, Sweeney and Karel (3) to solve the S=/1 234
traveling salesman problem, 1 23 4)
This sequenee array specifies that all three jobs will be
processed first on machine 1, second on machine 2, third
On a shop routing sheet the technological ordering of 4
on machine 3, and fourth on machine 4, All three jobs
job or commodity ix given, One job may be processed first
have the sume ordering in order to sitnplify the calcula-
on machine 1, second on machine 2, and third on machine
tion for illustration. The solution procedure would be
3. Another job may be processed first on machine 3,
exactly the same if different orderings were used for the
second an machine 2, and third on machine 1. In the solu-
different jobs.
tion to be developed, all jobs are assumed to have the
Now that the technological ordering is specified, the
technological order specified completely, although they
process times for each job on each machine must be given.
do not need to have the same ordering, This ordering for
Standard time is a predetermined estimate of the time
’ Now Professor of Industrial and Systems Engineering, Uni- required to process, set up, and clean up a job on a given
versity of Florida. machine, and it is assumed that standard times are given
* This presentation is based in part on research performed for a
doctoral dissertation by Charles K. White, under sponsorship by
as integer values. The assumption is made that the se-
the Purdue Research Foundation and under the direction of quenee of the jobs has no effect on the standard time.
Thomas L. Bartlett, These times are specified into a single array called the

BS
facility array and denoted by F. In general this will be as row 2; hence, the third operation on job 2 is performed on
follows: machine 3.
fin fis +--+ San) The assumption is made that no two joba may be
pan Fe fra Son | processed on a given machine at any common interval of
time. In the example, the problem array indicates that
job | is tentatively scheduled for the time interval (0, 6),
Mf Sais:*'* Som) job 2 is in (0, 1) and job 3 is given the interval (0, 4). By
this assumption, this schedule is not possible. When more
where f,, is the time to process the ith job on the jth
than one job is scheduled for the same machine at some
machine, A simple example will be considered. In this
common interval of time, the jobs are said to be in con-
sample problem the following facility array is assumed:
flict. In order to resolve this conflict the jobs must be
6 8 O 4) placed in a sequence. Finding a procedure for obtaining
i i
Feil 3 9 6} the best sequence over all jobs and machines is the central
problem of scheduling and of this presentation.
45 3 6, One more factor must be considered before attempting
Forexample, job | on machine | takes 6 units of time to to solve this problem. A method must be devised to
complete, and job 3 on machine % takes 3 units of time. evaluate a sequence to determine if it is better than an-
By the cumulation of the rows of F the ordering and other sequence. The criterion used to evaluate the se-
times can be shown in one array called the problem array quences developed in this sample problem will be that of
and denoted by 7’. In general this will be as follows: minimizing the total time to complete all jobs on all ma-
chines. Other criteria will be developed later,
Pur Pure * Pw)
Procedure to solve the sample problem
Pa fe Pa Pe
The first step to solve the sample problem is to set up a
work array or linear array. A work orlinear array is the
‘Pat Puss * * Pow array in which the solution is developed and appears as
follows:

A facility block is the entire set of columns under a given

facility, f,.
where/’,, is the tninimumtime to complete the ith job on
the jth machine, This definition implies that no overlap of
For ease in following the solution, Figure 1 should be
followed throughoutall the steps.
operations is permitted on any job.
Step 0 in the solution is to enter the completion time of
In the example above the problem array appears as
follows: the first operation of each commodity in the work array,
and set 7 equal to the smallest of these times. 7 is the
6H Ba time prior to which the schedule is complete and fixed
Peil 4 13 19 without conflicts and the time at and after which con-
fliets may exist.
498 B
fa
The problem array is the one which will be used in the
ealeulation both by hand and in computersolutions. Cr Oy Cs
From this array the lower bound of the total time to
process the N jobs is easily found. In the example this }elals
lower bound is seen to be 27 units of time. The actual time
to process the jobs will be st least as great as this lower
bound, If the processing time of job 2 on machine 3 is de- As 1} is the smallest of these three, 7 is equal to 1.
sired, one subtracts 4 from 13 to get 9, the desired time. Step 1, Check for conflicts between jobs ending at time
On the other hand, if the order of job 2 on machine 3 is T and those ending later than time 7 within each facility
desired, one notes that 13 ia the third smallest number in block,

January-February, 1965
 In the foregoing there are still conflicts; however,
there is no conflict for the time interval (0, /). Through-
out the solution there will be no conflict for jobs ending
at any time less than 7, where 7 is increased every
iteration until the solution is obtained. The schedule is
completely defined for jobs ending at times less than T
in the solution,
As the conflict is resolved about 6, the lower bound for 6
will be found. Note that the greatest time must be found
on machine 4 as all orders are processed last on machine 4.
The minimum time is 27; however, the lower bound can be
found only by observing that the three commodities on
machine 4 do conflict. The arrival times of the three com-
modities on machine 4 are 23 for commodity 1, 19 for
commodity 2, and 18 for commodity 3. (These times are
the completion times for machine 3.) By arranging these
start times in ascending order, the minimum overall time
can be found. As these are all minimum possible arrival
times, commodity 3 cannot be complete before calendar
time 24 and commodity 2 cannot be completed before
calendar time 244-6= 30, and commodity 1 cannot be
completed before 30+4=34 units of time. This deter-
mination of bounds only applies if the minimum time to
complete all jobs on @ machine is the objective. The num-
ber 34 is larger than 27, and this number will be used as
the lower bound for conflict resolved about C\. If the
Within facility block, 1, commodity 2 ends at time 1,
commodities do not all have the same routing, the same
As shown previously, this is in conflict with commodity|
procedure is followed to find a minimum completion time
and commodity 3, that is, commodities 1, 2, and 3 over-
for each machine, then the greatest of these minimum
lap on the same machine,
times for each machineis the lower bound.
Step 2. Cheek all members of conflict for constraints,
Repeating the same procedure by choosing the second
As no constraints were placed on any commodities, this
member of the conflict set, the original problem array
stop can be omitted for this problem. This step will be

nf eu
would become the following:
explained more fully in the second example.
Step 3. Find an estimate of the lower bound for each 15 24 28
member of the eonfliet, Resolve the conflict about the 43
member having the least lower bound.
There are three members in the conflict that is to be re- 5 WB WW,
solved. This conflict should be resolved to give the best Observe that in finding the lower bound the problem
chance of arriving at an optimal or near optimal sequence array is only temporarily updated. After the least lower
over all jobs and machines. One way is to establish a bound is found, it will be permanently updated, The
lower hound in terms of total time for each member of the minimum time to process all jobs on machine four is
contliet. If this lower bound relates closely to the criteria, 19+6+4=29. Hence, a lower bound about conflict mem-
the least lower bound of the three would give basis for re- ber 2 is 29.
solving the conflict. By resolving the confliet about each Again for the third member of the conflict the problem
member and then finding the minimum possible time to array becomes:
completion, such a lower bound is established. 10 18 2 a
For example, by resolving the conflict under considera- P='5 8 WT B
tion about commodity 1, C, is seheduled for the time in-
terval (0, 8) while Cy must take the interval (6, 7) and (; 4912 8B
has (6, 10). In the problem array all times which follow The minimum time to complete commodity 2 is 24 if
(and Cy on machine | must be increased by 6, The prob- the overall time is to be minimized. The minimum time for
lem array becomes changed as follows: commodity | is 31 and not 244+4=28. Care should be
6 4 3 2 taken to avoid missing such idle time in finding lower
bounds. The lower bound about the third member of the
Pei7 0 19 %) confliet is 31.
10 15 18 By The procedure for finding a lower bound can be sum-

% THE JOURNAL OF INDUSTRIAL ENGINEERING Volume XVI * No. 1

marized in the following way. On a given machine the LeveL

earliest starting time for each commodity should be

i som n : ”
found and ordered from lowest to highest. Using this J eo} “5
ordering, find the minimum time to complete all com- » the ee
modities on that machine without overlap. The lower by ry u try

bound is the greatest of these times taken over all ma- 5 a ‘te bt Nou
4 un ws Nk
chines.
7 wa
As 20 is the least of the three lower bounds, the conflict . » n * [ar] x
should be resolved permanently about this member.
ca 2
The possibility of also resolving the remaining conflict
should be considered, If only one solution is to be found,
the immediate resolution of this remaining conflict
is desirable; however, if an optimal solution is desired or , 1 {
iterative improvements are to be made in the solution,
then it is more convenient for caleulation to resolve only
‘| —
one member at a time,
Now that the resolution is made it is desirable to keep ’ a |
track of this resolution and the lower bounds. This can be
done most easily in a manual solution by representing the le pp —
decision by a node and the three members of the conflict
by branches generating from the node as shown at the top © 3; » # WkRRH HOD BAaAAS AR

of Figure 2. Referring to Figure 2, the numbers on the left Figure 2. Decision tree and Gontt chart for problem one.
hand side are referred to as levels. Across from level 1 the
three lower bounds just calculated are shown.
array and the calculation returns to Step 1.
In the sample calculation T’=4 is the next higher
integer, and T is set equal to 4.
As there ix no conflict involving jobs ending at time
JT =4, the calculation returns to step 4 where the next
facility to process commodity 2 is found to be facility 3.
Under facility 3 and commodity 2, the number 13 is
It takes a minimum of one decision to arrive at the nodes entered.
labeled 34, 29, 31; hence, the conflict level is the minimum At this point the problem will iterate through the same
number of decisions to arrive ata given node in the deci- steps until there is no higher integer T’. When there is no
sion tree, The decision tree is the entire diagramat the higher integer T’, a solution has been found with a total
top of Figure 2, and it represents all the decisions made to time equalto 34. In Figure 2 the decisions to arrive at this
arrive at a solution of the sample problem. The nodes solution can be traced. Note that the lower bounds never
with the square boxes represent final solutions and are decrease as the conflict level down a given path increases.
called terminal nodes, The nodes with the letter z below If all lower bounds down all paths from a given node are
themindicate that no better solution can be found by con- greater than the lower bound of this originating node,
tinuing down this path. A path is a series of branches con- then the lower bound of the originating node can be re-
necting nodes of increasing conflict levels, vised and set equal to the least of the lower bounds over
Step 3 is now complete as there are no more conflicts in- all paths generating from this node. Hence, at conflict
volving jobs ending at time T=1 in the work array, level 2 for the decision between 15 and 16, the lower
Step 4. Find the next facility, if any, to process the bound ean be changed from 29 to 34. By a series of such
commodity with value 7 and enter 7 plus the process revisions all paths in the decision tree are exhausted. The
time for this next operation in the corresponding box of last solution found must be the optimal solution to the
the next facility. problem as the revised lower bounds at conflict level 1
The work array in Figure | indicates that commodity 2 will all be greater than or equal to this solution.
on machine | has a completion time of T=1. Machine 2 Step 6. At the previous conflict level, all lower bounds
processes commodity 2 next, andit has a process time of 3 should be compared to the solution. If any lower bound is
units of time; thus +3 = 4 is entered into the work array leas than the solution, return to Step 3. If all lower bounds
under facility 2 and commodity 2. Note that this value of are greater than or equal to the last solution found, back
4 can be obtained directly from the problem array. up one level and repeat this step. When all levels are ex-
Step 6. Test to see if T is the largest entry in the work hausted the optimal solution has been found.
array.If it is, the calculation goes to step 6. Ifit is not, T Again in Figure 2, the conflict level 4 has two branches.
is set equal to the next higher integer, 7’, in the linear One branch leads to the solution just found, and the other

Janwory-Febreary, 1965 THE JOURNAL OF INDUSTRIAL ENGINEERING 37

has a lower bound of 37. As 37 is greater than #4, no Step 4. For each T in the array find the next facility (if any) to
better solution can lie down this path. Backing to conflict process the commodity and enter T plus the new job time in the
corresponding box of the next job!
level 3, the conflict is between 24 and 23, The least lower Step 6 Uf T is the largest entry-in the array go to step 6;if not,
bound down the path corresponding to 24 is 34, the solu- find the integer 7’ in the linear array that is next larger than 7
tion, and the least lower bound down the path correspond- and set 7T'= 72 Go back to step 1,
ing to 23 is 36 which is greater than 34, The lower bound Step 6. Return to the previous conflict level and compare lower
labeled 34 is the same as 34 at conflict level 3. Following bounds to solution. If any Jower bound is ees than the solution,
return to step 3, If all lower bounds are groater than or équal to the
the reverse path again gives a lower bound for 15 equal to solution, bark up one level and repeat the step. If all levels are
4 and for 16 equal to 37. Again the lower bound at con- exhausted, then the optimal solution i¢ the last solution found.
flict level 2 should be changed from 29 to 34. Upon revers-
ing to conflict level 1, the lower bound corresponding to 5
ix $2. As 32 is less than 34, a better solution might be Using the ideas just developed, it is apparent that
found downthis path. It is necessary to return to step 3 other criteria can be chosen for finding an optimal se-
and repeat the above iterations, When levels are ex- quence. One such criterion is that of minimizing the late-
hausted, the solution corresponding to T'=32 must be ness over all orders. [f 7’, is the calendar time at which
the optimal solution to the problem. Had only # near op- commodity ¢ is completely processed and d; is the due
timal solution been desired, then the calculation would date for commodity i, then commodity is late if T,><¢,
have stopped at the solution corresponding to T= 34. and it is on time otherwise, Summing over all the com-
modities which have (7',—d,)>0 gives the total lateness,
Summary of the assumptions ond computation algorithm The object is to minimize this total lateness. In symbols
In summary the following assumptions were required this can be stated as:
to solve this problem:
x
1, Deterministic times are used, For this value in practice min > (T,-—4d)).
standard times are integer values, By » suitable choice of time 1
seale this asstimption can always be met, Tid,
2. "The processing time is independent of the sequence,
S. Transportation and set-up times are negligible of are in- The same data will be assumed for this problem as were
eluded in the operation times, used in the previous problem. The due date for commed-
4. Each order must have a specified technological ordering, ity 2 will be 30 units of calendar time, and commodity 3
but all orders do not need to follow the same ordering.
5, Lote cannot be split during thy scheduling process, This due date will be 20 units of calendar time. As constraint
amwumption con be relaxed by renumbering of lota, however. commodity 2 will be designated as a rush order, and it will
6. Only one job ean be processed on a given machine at a given always be processed before commodities 1 and 3. As a
time. further constraint a fourth commodity will be introduced
7, No operations may be interrupted unless an entire new
schodtle is generated.
as work in process, and it has priority over all other com-
& Only one machine per type te allowed in this problem, modities, The fourth commodity is proeessed only on
All jobs ean be provessed at any time during the scheduling machine 1, The problem array for this problem: is as fol-
period, lows:
10, No overlap, (A given job may be processed on only one
machine at a time.) 6 i 2B 2
Pin | 458 19
Using these assumptions the following algorithm ean
be used for problems similar to the one solved. 4 9R
Step 0, Enter Completion Tinwof the first operation of each 2 2 2 2
commodity in the array, and set 7 equal to the amallest of these
tines? Before solving the problem, it is necessary to establish
Step 1. Within each forility block cheek for confliets between i procedure to find lower bounds, From the problem array
jobe ending ot time T and those ending later than time T!
Sup, 2 Examine members of the conflict for problem con-
the minimum possible completion time of order i can be
straints, If thore are no constraints on the members go to stop 3, called ¢,. By resolving about each conflict member tem-
For example, if a job ie in process and it is in the conflict, resolve porarily as before, a lower bound can be found for each
whonut thit member, or if it it a rush job, resolve about it, Set member of the conflict. In symbols this appears almost
lower bounds af all other members of the conflict to infinity and the same as the expression for the criteria,
go directly to step 4.
Step 3. Find an estimate of the lower bound for each member of ¥
the conflict. Resolve the conflict about the member having the L= > -d).
least lower bound. If a tie exists, break it by a random selection. tt
If a lower bound is greater than or equal to the best solution found, t>d,
set thie lower bound to infinity.

* This is easily done by storing all conflict levels on tape when

* Restated from Giffler’s algorithm (1). using « computer.

3% THE JOURNAL OF INDUSTRIAL ENGINEERING Volume XVI * Mo. 1

 Daevtity Med
+346=27. This means that the lower bound is (27-20) 7,
VMeateew 1 Velw ! Vetbew 1 Mathew 4
Likewise, if the conflict is resolved about (,, then the lower bound
‘- ': + “Tt T a is (34 —30) + (21 — 20) = 5. Thus, it is resolved about ().
Laven cy) « |! an ce Gs GIiGIQiGa |G te 4. Conflict 4 was resolved by the rush job,
eh 5, Conflict 5 was resolved by the rush job.
' "i zle } |
6. The lateness of all orders not rushed was 11. At conflict
> pel eles | level 3, the lower bound for (, was 7; therefore, this path must be
investigated.

itateet
Ll alee 14 7. The lower bounds for conflict, are L,= (31-20) =11 and
L, = (39 —30) +-(27 —20) = 16. Neither of these is better than the
sivlialrlz taleley 1 ae presentsolution,
- -+— tees a=} neem - liens

ios }7 2M) eles)

i
we ZY} 2F
By reversing, it is found that there are no other paths
® post ste ta palo lez bool ae ie bss ante to follow; therefore, the optimal solution is the one corre-
sponding to a lateness of 11.
») a toate paw 6 | 15 n
Two observations should be made. The first is that if
lateness is not a problem for @ given shop, then many of
the lower bounds may equal zero early in the computa-
tion. An alternate criterion such as minimizing set-up
time would be useful in this case, The second observation
is the constraints tend to reduce the number of paths;
therefore, for an optimal solution, constraints help in the
solution rather than hinder it. For large problems in which
only one solution is desired this is not usually the case.
The time to arrive at a solution may become quite large if
too many constraints are entered into a large problem.

Computationol experience
Even for moderate size problems, the foregoing proce-
Figure 3. The solution and decision tree of problem two, dure is too long to do on existing computers. The proce-
dure does allow one to come as close to the solution as
The assumptions used are the same as before and the time and cost permit. Instead of using the lower bounds to
solution follows the algorithm precisely, The solution of eliminate solutions, it may be desirable to use the lower
the problem is shown in Figure 3. The resolution of the bound as a decision rule to develop a single solution
conflicts can be summarized as follows: which, while it may not be optimal, will usually be at
1, At conflict 1, the resolution was around Cy because it is least a near-optimal solution. In developing this idea,
work in proces and precedes rush jobs, several sample problems were run for both minimizing
2 At conflict 2, the resolution was around (, because it is a lateness and for minimizing total schedule time. The re-
rush job.
4. At conflict 4, lower bounds were caleulated for C, and C;, If
sults showed the initial solution for the minimizing late-
the conflict were resolved about C, then the earliest time C, could ness rule usually involved less total time than the mini-
be completed would be 9484944 =30 and C, would be 1245 mum time rule itself, As it did not appear profitable to

Table 3
Comparison of dispotch rules for various size problems

mn Otyert Vunction Latensa* Machine Idle Time Total Time

wan [aN sor. woNG MIN RAN SHOP, LONG MIN RAN SHOP, LONG
7X0 2258 3412 3004 4386 1735 2867 2038 2228 866 941 828 8u2
6x12 3625 7268 5817 9340 1546 2751 3311 2951 1072 1386 1348 1264
9x10 Sil4 5803 3759 6521 4205 8456 5894 5747 1139 1768 1372 1250
7Xu 5642 7992 R408 8962 1937 3898 4078 2241 1038 13976 ° 1325 1184
10X10 2266 3861 3990 4904 3848 5486 Hu $412 904 1204 1174 1137
10x12 3368 5135 5376 6843 3027 §223 5644 5078 981 1238 1285 1221
8x16 6886 12639 10050 16204 3247 5252 5448 7040 1435 1645 1569 1787
xis 4822 10220 7405 10614 4536 8060 7313 6192 1290 1506 1563 1431
10x18 8o41 15043 12570 — 5030 7342 7976 _ 1006 1961 1939 _-
10x18 7442 12403 10701 17960 4244 6532 8094 7316 1384 1624 1822 1723
Average | 4836.4 8377.6 7108.9] 9531.6, 3335.5, 5586.6 5521.0 4911.7) 1180.5, 1473.9 1422.5, 1323.2
* The criterion for these problems was to minimize the total amount of lateness. MIN refers to the solution the algorithm de-
yoteeen oo ee asta RAN Ne eee eee iiva OH OF. m tos eoetion by tia eet eptetion Oak amine ae
LONG is the solution by the longest remaining time decision rule. For each of these solutions the total machine idle time and
over-all time was computed for comparisons.
S

Sorwory-Febrvory, 1965
 Table2
Compertson of peta née for 510 cise problems
Stentor
»
MIs
Cheers Funrtice Lotenens

Rad OP,
*
|
LONG Min |
Masipne Séle Time

RAN $8. OF.

a
Lona Min
_Tetal

RAK
Tne
om. OF. tore
\ z165
2489
5
3485
2468
759
2°
4250
boo
483
teop
1283
1257
2235
«*
90
0
745
ges
5
817
1006
«°
747
2
3 yarh) ns 3506 36) 4483 108 1746 2212 1003 710 46 1011 T72
4 dow 3233) Wi guia a7 asta zine aso 614 B51 we 7h
& z |

3283 3184 1408 2175 2212 731 064 1034 «°

6 v4us gery oO 1618 esa 2187 BIT 057 1143 984
3350) a723)
2
7
8
my
SHS $803
3847
2049
tu
70)
4100
TW
879 2254
2034
1417
aa)
«1708
1018
TO
#21
$73
1012
au3
5
oan
718
805
2367 2604 4003 1349 1427 ret!
0
|

ee 2a MMI Bil |
1417 100 1833 O74 825 sau

Avernge 2247 7 44091] 4 4752.6 7 1650.3 1995. 1302.8 722.5 91.8 870.4,
782.4

Mid, Dev. |
554.9 ABLE) 674.3 7.0 387.7, 413.4) 40.4} 87.0; 76.7 17 |
288
*
No Kolutinn,
**
Seo fuutaute on Table 1.

study the minimum thue decision rule further, a series of References

problens were solved using the minimum lateness rule (i) Gervien, B., axe Trospnon, G. L., "Algorithms for Solving
2
only. Tables 1 and show the relationship of this rule as Production Scheduling Problems," IBM Research Report
compared to random solution, shortest operation time RC-118, June 15, 1830.
rule, and ougest renmining time nile. Even though the (2) Harras, J., "Combinatorial, Probabilinte, and Statistical
Aspects of an XJ cheduling Problem," Report NYO.
abject was to minimize lateness, the machine idle time 2540, AES Computing and Applied Mathematical Science,
and total ime were conmdderably shorter for the mini- New York Universty, New York, February 1, 1959.
mun ateness rule than for the others, (3) Larrea. J,
Mewry, K., Sweeney, 1), ano Kanes, C., "An
Using the assumption that the solutions are normally Algorithm foe the Traveling aleeman Problem," Journal
distributed (2), a series of 5X10 problems were solved of prrations Research, November~ December, 1963, pp.

using random digits and ordering. The probability of b 072 -OND,

(4) Warra, C. &., "An Algneithm for Finding Optimal or Near
taining a solution by this rule as had as the expected ran- Optimal Solutions to the Production Scheduling Problem,"
dom one is lesa than 0001 while the probability of ob- unpublished =Pbh.D. Dissertation, Purdue University,
taining @ random solution a8 guod as the expected olu- Lafayette, Indiana, 1064
tion by the cule is bee than 02. While those figures are
only rough approximations, they do tend to show that
the minimizing lateness rule produces very good e
uences on the first solution, The computational time is
about the sume as for other rulen.
Other problems were tried using a more complex eri-
erion which minimized a weighted combination of a
chine idle time and lateness, The results were very
favorable as the rocedure did a better job on the first
solution than the siinpler eriteria. In ractice all conflicts rofecsienal sregrom e @ isatished ttitudes e achieve.
mean plus dedicated ateraination
to impreve. With thie tlosephy
ie mind,
xrnerated at one time should be resolved at that time in- wo hove been mest weccemty) in providing ews clients with the ecowery
technics) perenne! te Gl) ssledt jab pertunitios. 1/ yeu ore @ desler, inter:
tead of resolving about only one member of the conflict er Seniee
wth de
when only a single near-optimal solution is desired. owe io ME, Ew
adeteial
m DISSATISFACTION ne © tong liet of
otentions!, Righty

A
procedure has heen developed for finding optimal
solutions for a wide variety of posatble criteria and con-
is An Engineer's
Biggest Ascot ane
pan!
la basic dest, glen.

wbher ant siestica

ctie.
dwuanion
ere vell-
straints. As moet ractical problems sre much larger than laciede let oveivetion, aentiogns, osh, plant
ie
leveut or production ontrat
could be solved economically by this rocedure, the ides Atl laquistes tll eceive iametion end conktentiol nsideration
of using tho lower ound as a erision rule for developing
& single near-optimal solution was developed. The experi-
S ASSOCIATES, INC.
ManagementPerneneel Consultants claws 1968

ments on these ecision rules showed that they might Sr. Aasocigte indvetrial Enginesring of
nee Sidy.,
231 Frith nine meee Feb.,
Pe anner"
ae
profitably be used in place of other rules for ertain appii- te
taro Jeanette Emad,
conte? Chienge Manager, 30 16. Michigan Ave..
cations. Ciage

THE JOURNAL OF INDUSTRIAL ENGINEERING Vebune

XVI ° Ma, 8