Mixing static and dynamic flowtime estimates for due‐date assignment

• Published date: 01 March 1993
• Author: Kevin J. Dooley,Michelle M. Vig
• DOI: http://doi.org/10.1016/0272-6963(93)90034-M
• Date: 01 March 1993

Journal of’ Operations Management, 11 (1993) 67-19

Elsevier

Mixing static and dynamic flowtime estimates for due-date

assignment

Michelle M. Vig and Kevin J. Dooley

Division of Industrial Engineering, Department qf Mechanical Engineering, University of Minnesota. 111 Church Street

S.E., Minneapolis, MN 55455, USA

(Received 18 December 1990; accepted in revised form 10 March 1992)

Abstract

Prediction of job flowtimes is important from the perspectives of both providing internal control and customer

satisfaction. More specifically, accurate and precise flowtime prediction can facilitate proper timing for the release of

materials and resources, improved accuracy of delivery dates to customers, and identification of jobs that require

expediting. The objective of this paper is to introduce a new approach to flowtime prediction which improves perfor-

mance quality over that found in existing dynamic flowtime estimation models. This new approach. mixed flowtime

estimation, incorporates both static and dynamic flowtime estimates into a single, flowtime prediction model. In this

investigation, mixed and dynamic models are compared experimentally using computer simulation of a job shop under

various congestion conditions and dispatching heuristics. The results of this investigation reveal that the mixed flowtime

prediction models provide significant improvements in job shop due-date estimation performance. Statistically significant

performance improvements are obtained in both the average lateness and fraction tardy jobs for mixed estimates.

Specifically, average lateness is reduced by 25% in the balanced shop and 40% in the bottleneck shop. This improvement

enables a corresponding increase in the accuracy of job flowtime estimates, and, hence, due-date assignment accuracy.

Due-date performance improvements are observed for each scheduling heuristic investigated.

1. Introduction

Prediction of job flowtimes in production

systems has always been an important element in

production planning and control systems. The sig-

nificance of these predictions is implied by the

variety of tasks which use these numbers. Typical

applications include assignment of due-dates for

internal and external customers, evaluation

criterion of scheduling and other control systems,

release of jobs to the shop floor, evaluation of

alternative production system configurations, and

evaluation of the system’s performance during

uncontrollable or unforseen disturbances.

Correspondence to: K. Dooley, Division of Industrial Engin-

eering, Department of Mechanical Engineering, University

of Minnesota, 111 Church Street SE., Minneapolis, MN

55455, USA.

Flowtime estimates are usually static or

dynamic. Static estimates are time invariant, single

point estimates that are meant to predict the

expected value of flowtime for a single job or group

of jobs. They can be obtained in a variety of ways,

but are normally found via queuing analysis or

steady-state simulation results. Dynamic estimates

are acquired using regression equations which

predict job flowtime from various job and produc-

tion system characteristics. They are dynamic in

that they are time variant - as job characteristics or

shop floor congestion levels change, so do the cor-

responding flowtime estimates. The methodology

for obtaining the appropriate regression data

usually includes simulation.

Historically, due-dates or flow allowances were

approached as an uncontrollable factor in schedul-

ing problems, often used only as a performance

measurement parameter. Among the first due-date

0272-6963/93/$06.00 0 1993 Elsevier Science Publishers B. V. All rights reserved.