Eric C. Klein

Date of Award

2-2025

Document Type

Thesis

Degree Name

Master of Science in Engineering Management

Department

Department of Systems Engineering and Management

First Advisor

Daniel J. Weeks, PhD

Abstract

Hurricane Michael struck Tyndall Air Force Base as a Category 5 storm with 160 mph winds, damaging or destroying 90% of its infrastructure (484 buildings). This devastation prompted the Department of Defense to rebuild Tyndall as the “Installation of the Future,” encompassing 44 Military Construction (MILCON) projects and 260 Facility Restoration, Sustainment, and Modernization (FRSM) projects. Since reconstruction began, over 630 Change Request Forms (CRFs) have been closed, adding $67.8 million in costs and 4,880 construction days. The Air Force Civil Engineer Center's Natural Disaster Recovery (NDR) Division developed a programmatic risk model to predict future costs and delays. However, the model lacked academic rigor and data-driven decision-making. This research analyzed historical CRF data using text mining and word frequency analysis to categorize CRFs into nine bins: Wastewater, Stormwater, Fire, Electrical, Fire and Electrical, Concrete, Two, Three+, and Other. CRFs were assigned to categoriesbased on specific keywords in their descriptions. For example, CRFs were placed in the "Fire" bin if they included terms like “fire,” “UFC 3-600,” “alarm,” or “hydrant.” CRFs overlapping categories were grouped into composite bins such as Fire and Electrical, Two, or Three+. CRFs that didn’t fit any category were labeled “Other.” Descriptive and inferential statistics revealed distinct cost patterns across the bins. Categories such as Fire, Electrical, and Other incurred the lowest costs, while CRFs spanning multiple bins, like Fire and Electrical, Two, or Three+, were significantly more expensive. These findings demonstrate that relying on the overall mean of the CRF dataset as a performance metric is inefficient due to the substantial differences in costs across categories. Instead, categorizing CRFs and applying confidence intervals provides a more accurate metric for predicting future costs and time impacts, improving the effectiveness of the programmatic risk model.

AFIT Designator

AFIT-ENV-MS-25-M-089

Comments

An embargo was observed for this posting.

Distribution A: Approved for public release, Distribution Unlimited. PA case number 88ABW-2025-0299

Download

Share

COinS