Computers and Industrial Engineering Vol. 25, Nos 1-4, pp. 251-254, 1993
0360-8352/9356.00+0.00 Copyright © 1993PergamonPress Ltd
Printed in GreatBritain.All fightsreserved
SCHEDULING IN AN MRP ENVIRONMENT
GiJrsel A. Slier and Rafael A. Lizardi
Industrial Engineering Department University of Puerto Rico - Mayagtiez Mayagiiez, P.R. 00681
on the performance of sequencing rules and MRP. McAreavey et. al. [4] described how MRP Call provide a working solution to job shop scheduling. They suggested a Computer Aided Resource Management (CARM) Model which provided the basis for a comprehensive planning system. Fry et. al. [2] examined the effect of three different product structures on the performance of selected priority dispatching rules. Results indicated that there is a strong relationship between the product structures and the performance of the sequencing rules. Russell and Taylor [5] evaluated 11 sequencing rules and analyzed the sensitivity of job structure on the performance of the rules. They concluded that sequencing rules have a significant impact on the flow time and tardiness of jobs, regardless of the job structures. Goodwin and Goodwin [3] examined the impact of operating policy rules on multistage system performance. The results indicated that the development of more sophisticated rules incorporating additional status information may yield more benefits with respect to system performance. Biggs [1] like other authors recommended the use of other policies of scheduling to further enhance MRP information utilization in Resources Scheduling.
ABSTRACT This paper reports the results found in regard to the performance of three scheduling policies in a material requirements planning environment. The policies included in the study are shortest processing time, earliest due date and Moore's algorithm. Each workcenter is assumed to have only one machine and as a result this study is limited to the single machine scheduling applications.
INTRODUCTION National and international competition in the manufacturingworld forces many companies to use the best tools and techniques available in order to survive. This general strategy also applies to the production planning and inventory control functions of a manufacturing company. Material Requirements Planning (MRP) is a process of scheduling the shop orders and ensuring that materials are available whenever needed in the manufacturing process. Figure 1 shows the functions involved in a typical Manufacturing Planning and Control (MPC) System as given by Vollman et. al. [6]. MRP uses an infinite capacity approach at rough-cut capacity and capacity requirements planning stages. Because of its focus on timing, an MRP system can generate outputs that serve as valid inputs to the shop floor control (SFC), purchasing, dispatching, and the capacity requirements planning areas.
Resource PlannlnQ
"~
ROUGh - Cut Capet I~y Plsnning
:
Capac I t y
,~
j
t~
Companies using MRP software deal with the allocation of resources by using SFC techniques. To perform shop scheduling one must have up-to-date routing files and shop order files. The load and the capacity at each work center must be known. As soon as MRP assigns the due dates for the final products and lower level components, a better shop performance will be guaranteed using finite loading Detailed Scheduling (DS) techniques.
Finite I ~
Load i ng
:
::,,o::ot t . . . . . . . . . . . . . . . . -~ f
#\
Tlwe i
Matt
I o L
LITERATURE REVIEW Figure
Some of the failures reported with regard to the implementation of MRP results from the fact that MRP is necessary but not sufficient to obtain good detailed schedules. There has been extensive research
251
1.
General Model of a Manufacturing Planning and Control System.
252
Proceedings of the 15th Annual Conference on Computers and Industrial Engineering
PROBLEM DESCRIPTION Many shop floor sheduling systems use rules, but why not use scheduling algorithms? The algorithms have their bases on scheduling rules but they consist of a set of steps and include more search toward a better solution. Therefore they are expected to give better solutions than the rules. The objective of this work is to investigate the performance of rules and a single machine scheduling algorithm in an MRP environment.
STAGE I
STAGE 2
STAGE 3
Figure 2. Production System Studied.
LEVEL
SJ-;
i
. . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
EXPERIMENTAL CONDITIONS In this section, the system studied is described in
Figure 3. Simple Product Structures.
details. Lot Sizing Procedures The lot sizing procedures used are; a. Lot-for-Lot (LFL) b. Economic-Order-Quantity (EOQ) Rules and the Algorithm The techniques included in this study are: a. Moore's Algorithm (MA) b. Earliest Due Date (EDD) Rule c. Shortest Processing Time (SPT) Rule All of these policies are used taking non-zero ready times into consideration. Performance Measures T h performance of the scheduling techniques are evaluated with respect to the following measures: a. Number of Tardy Jobs (nr) b. Maximum Tardiness (MT) c. Total Tardiness ('IT) d. Average Flow Time (AFT) e. Average Completion Time (ACT) Demand Patterns The demand is generated from a uniform distribution. There are three different levels related to the work center capacity utilization. These levels are low (80% to 90% utilization), fight (91% to 100%) and overloaded (over 100%). The number of replications used is 2. Production Stages The system studied has a serial structure with three stages as shown in Figure 2. Each stage has only one work center with one server and each work center performs a unique operation. Product Structures There are three principal products (A, B, and C) considered. Two different sets of product structures are included; one set with simple structures as shown in Figure 3 and the other with complicated structures as represented in Figure 4. The parts at levels 2, 1 and 0 are produced at work centers 201, 202 and 203, respectively.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
LEVEL1
LEVEL
2
Figure 4. Complicated Product Structures.
Time Bucket Approach Time Bucket (TB) establishes when a job is considered for assignment. It runs from one to four periods. If TB is equal to one, only the jobs due in period n are considered during the assignment process. If TB is equal to two, the jobs due in period n are considered in period (n-l) if the material availability does not create a problem. If the attempt is not successful, job is scheduled in period n. The approach can easily be extended to other TB values as well. Figure 5 shows the TB approach used in this study. The due date is always the same, what really changes with the TB is the ready time. With the variation (1 to 4) of the TB, the number of jobs available to be processed earlier increases. The TB brings the advantage of an early production start, but there would not be any advantage if there is not enough material to meet the requirements. In this study it is assumed that at work center 201 there is always enough raw material to start the production.
SOER and LIZARDI: MRP Environment
253 RESULTS
PER tO0
cO
RTS
~
2
3
4
I
I
I 2,4o
I
Throughout the different scenarios studied it has been found that no single policy dominated the others with respect to all performance measures.
10
I
I
I
40
80
i
,o
eo ~2o ~,o 200 240 z,o 320 36o- l~e=
160
200
9
8
0
120
I
7
6
5
280
I
I
I
320
3so
400
0
40
8[3
120
160
200
240
280
320
0
O
0
40
90
120
150
~00
240
280
0
0
0
0
40
80
120
ISO
200
240
{30 = d ~
TB = 2 t
TB = 3 TB = 4
oate
RT = ~eetcty t l m e
T8 = t t r m
Table 1 shows the percentage that a given scheduling policy had the best performance regardless of the scenario. It was found that in most of the cases studied, E D D gave the best results in terms of minimization of MT, q'I', ACT and AFT. It is worth to mention that SPT outperformed MA with respect to all of the performance measures.
bucke~
Figure 5. Time Bucket Approach.
SCHEDULING POLICIES
Tables 2-4 show the results for TBs 2, 3 and 4. As the TB increased, the performance of MA improved and conversely the performance of SPT deteriorated. Table
i. R e s u l t s
iPERFORMANCE
Shortest Processing Time As the demand level increased, the number of jobs to be scheduled increased as well. When this happened the jobs with large processing times were assigned at the end. This adversely affected meeting the due dates. In most of the cases, the jobs that were completed at one work center did not wait for other workcenters become available since the queues were empty. In other words, the rule was applied when the queue content was greater than zero. One of the observations made is that there is no guarantee that a job that is finished early at one work center is going to be early at the next work center. This is also true for the jobs that are tardy at one work center. They may be early or at least meet their due dates at the next work center. These two conditions depend on how early or how late the job is finished at the previous work center and also it depends on the length of the queue at the next work center. This delay in the assignment of jobs was more critical when the lot sizing procedure changed from LFL to EOQ. This is because the size of the lots increased with EOQ and led to higher production volume.
MEASURE
Moore's Algorithm The behavior of this algorithm was the same as EDD for the low and almost always for the tight level demand under LFL. The results varied when EOQ was used. Once the demand increased to the high level, the results varied from the EDD rule drastically. When removing a job it is important to check if it would improve the solution. Because there were cases when removing a job with maximum processing time did not improve any tardy job due to non-zero ready times of the jobs. CAIE 25-1/4--R
SCHEDULING
POLICY
SPT
EDD
MA
62.50%
45.83%
54.17%
MT
33.33%
100%
8.33%
TT
33.33%
100%
8.33%
ACT
50%
75%
8.33%
AFT
50%
70.83%
12.50%
nT
Table
2. R e s u l t s
PERFORMANCE MEASURE
f o r T B = 2.
SCHEDULING
POLICY
EDD
SPT
MA
27.50% 91.67%
nT
4.17%
MT
0
100%
25%
TT
0
100%
25%
ACT
25%
75%
16.67%
62.50%
16.67%
AFT
33.33%
Table
3. R e s u l t s
PERFORMANCE
Earliest Due Date There were no problems in meeting the due dates when the demand level was low or tight and the lot sizing procedure was L F L However, when EOQ was used tardy jobs occurred. In the case of high demand, there were more tardy jobs.
for T B = i.
for T B = 3.
SCHEDULING
POLICY
SPT
EDD
MA
n T
0
33.33%
95.83%
MT
0
100%
25%
MEASURE
TT
0
100%
25%
ACT
12%
79.17%
29.17%
AFT
37.50%
62.50%
12.50%
Table
4. R e s u l t s
PERFORMANCE MEASURE
for T B = 4.
SCHEDULING SPT
nT
0
MT
0
TT
0
ACT
25%
AFT
37.50%
POLICY
EDD
MA
37.50% 95.83% 100%
25%
100%
25%
66.67% 20.83% 58.33%
16.67%
Proceedings of the 15th Annual Conference on Computers and Industrial Engineering
254
T a b l e 5. P e r f o r m a n c e Demand Level.
of
the
LOW SPT
% EDD
%
MA
%
SPT %
EDD %
18.75
65.63
75
25
50
84.38
6.25
MT
12.50
i00
25
12.50
I00
37.50
TT
12.50
i00
25
12.50
i00
37.50
ACT
21.88
81.25
21.88
37.50
71.88
AFT
40.63
68.75
15.63
43.75
62.50
6. P e r f o r m a n c e of the S c h e d u l i n g P o l i c y A c c o r d i n g to the Lot Sizing Procedure.
SPT % E D D
ECONOMIC ORDER QUANTITY
% M A % SPT % E D D
~ MA %
n r 2 0 . 8 4 6 8 . 7 5 8 3 . 8 4 10.42 8.34
85.42
MT
16.67
i00
41.67
0
i00
0
TT
16.67
100
41.67
0
i00
0
A C T 35.4: 6 8 . 7 5 3 7 . 5 C
25
79.17
0
AFT!54.17 54.17 27.0S
25
72.92
2.09
7. P e r f o r m a n c e of the S c h e d u l i n g P o l i c y A c c o r d i n g to the P r o d u c t Structure.
% M A % SPT % E D D
% MA %
nI 16.6743.7589.58 16.6733.34 79.17 MT
8.34
i00
20.84 8.34
i00
20.84
TT
8.34
i00
20.84 8 . 3 4
i00
20.84
4 1 . 6 7 4 7 . 9 2 [6.67 6 . 2 5
i00
20.84
70.8429.17
to
the
MA
%
SPT
% EDD %
MA
%
3.13
93.75
0
i00
0
0
i00
0
28.13
25
68.75
6.25
21.88
34.38
58.38
6.25
CONCLUSION MA outperformed E D D and SPT in terms of nr as expected. However, E D D performed very well with respect to all performance measures. Even though SPT is known to perform well at least with respect to AFT and ACT, the results obtained in this study did not support that claim. It is also possible to modify these scheduling policies in order to improve their performance. The modifications to the scheduling policies could not be covered in this paper. ACKNOWLEDGEMENT Authors would like to thank Ayse Siier and graduate student Genaro Brito for their assistance in the preparation of this paper. REFERENCES Biggs, Joseph R. "Priority Rules for SFC in an MRP System Under Various Levels of Capacity." |tit. J. of Prod. Research, V23, No.1(1985): 33-46. Fry, Timothy D.; Michael D. Oliff; Elliot D. Minor and G. Keong Leong. 'q'he Effects of Product Structure and Sequencing Rule on Assembly Shop Performance." Int. J. of Prod. Research, V27, No.4(1989): 671-686. Goodwin, Jack S. and James C. Goodwin. "Operating Policies for Scheduling Assembled Products." Decision Sciences, V13, (1982): 585-602.
COMPLICATED
SIMPLE SPT %EDD
According
HIGH
nr
LOT F O R LOT
Table
Policy
TIGHT
Table 5 presents the results for different demand levels. E D D outperformed SPT with respect to all performance measures at each level. MA was the best when nr was considered. The results about the lot sizing procedure are presented in Table 6. The performance of SPT was better under L F L The similar trend was observed for MA as well. EDD performed consistently well in both cases and even better when E O Q was used (except for nr). Product structure made a significant difference in the performance of E D D and SPT with respect to ACT and AFT as can be seen in Table 7.
Table
Scheduling
8.34
8.34197.92 20.84
McAreavey, D.; J. Hoey and R. Leonard. "Designing the Closed Loop Elements of an MRP System in a Low Volume, Make to Order Company (with case study)." Int. J. of Prod. Research. V26, No.7(1988): 1141-1159. Russell, Roberta S. and Bernard W. Taylor III. "An Evaluation of Sequencing Rules For an Assembly Shop." Decision Sciences. V16, (1985): 196-212. Vollman, Thomas E.; William h Berry and D. Clay Whybark. Manufacturing Planning ~,O Control Systems. 2nd Ed. Illinois: Dow Jones Irwin, 1988.