TITLE:
An LLM-Augmented Framework for Risk-Informed Schedule Planning in Airside Construction: Integrating CPM and Probabilistic Simulation
AUTHORS:
Hossein Taheri, Mana Samadi
KEYWORDS:
Large Language Models, Construction Scheduling, Airfield Construction, Schedule Risk, Critical Path Method, Monte Carlo Simulation, Probabilistic Scheduling, Risk-Informed Planning, Artificial Intelligence
JOURNAL NAME:
World Journal of Engineering and Technology,
Vol.14 No.4,
October
30,
2026
ABSTRACT: Construction scheduling in active airport environments is complicated by restricted work windows, operational access requirements, and dependencies that may not be fully reflected in conventional critical-path analysis. This study examines an LLM-augmented framework for identifying schedule vulnerabilities by combining Critical Path Method (CPM) indicators, a heuristic Baseline Risk Index (BRI), large language model (LLM) interpretation, and probabilistic duration analysis. The framework was examined through an anonymized airside rehabilitation project at a major U.S. airport, using 1591 baseline schedule records and nine retrospective validation cases. Baseline schedule characteristics were analyzed separately from subsequent project outcomes to reduce hindsight bias. Monte Carlo simulation was applied at the activity level using provisional duration distributions and risk-event assumptions. The results revealed substantial differences in schedule flexibility across work packages. Work Package E had approximately 67% of its eligible activities within one to five days of total float and subsequently experienced a 32-day schedule variance. Work Package D recorded a 70-day variance, despite having lower low-float exposure. Meanwhile, the project’s substantial-completion milestone was achieved on its baseline target date. The LLM helped consolidate schedule-risk indicators into structured observations for planner review, although no independent improvement over conventional CPM and BRI analysis was demonstrated. The probabilistic results provided comparative activity-duration exposure rather than project completion forecasts. These findings suggest that combining conventional schedule measures with contextual risk screening and structured interpretation can support earlier recognition of localized vulnerabilities, while further validation is needed to establish measurable improvements in planning performance.