Date of Award

9-1991

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Operational Sciences

First Advisor

Paul Auclair, Major, USAF

Abstract

This research investigates scheduling performances of commercial project management software. It is based on difficulties that the Royal Australian Air Force experienced in evaluating microcomputer-based packages in maintenance task scheduling of aircraft undergoing Depot Level Maintenance. A literature review found only one paper quantifying commercial software performances. No standard performance metrics or test data were found. Several packages were evaluated using a common set of 110 single-project resource- constrained networks. The performance measure used was closeness to optimally minimum Total Durations. No package performed best on all problems. On average, the best produced schedules 5.2% longer than optimal, with 95 of 110 problems 0% to 10% above optimal, and values ranging from 0% to 35%. To describe software performance in terms of network characteristics, a linear regression model was proposed. A highly satisfactory fit was obtained using only 2 of the 58 network measures: SADUR (sum of all task duration) and ACONMX (resource constrainedness). Most project scheduling software is written and used assuming task durations have negligible variability, despite contrary industry experience. Rudimentary simulations examined solutions' robustness in a dynamic environment. The results indicate that actual projects will, on average, run longer than optimal deterministic schedules.

AFIT Designator

AFIT-GLM-ENS-91S-3

DTIC Accession Number

ADA246684

Comments

The author's Vita page is omitted

Download

Share

COinS