Planning Construction Projects in Deep Uncertainty: A Data-Driven Uncertainty Analysis Approach

Abstract

Construction planning is significantly affected by many uncertain factors derived from construction tasks, the environments, resources, technologies, personnel, and more. Uncertainty analysis approaches are thus critical to supporting the decision making associated with construction planning. However, the precise probability distributions (PDs) of uncertain factors are sometimes inaccessible, especially for construction projects in a novel context with limited previous experiences or similar references. These situations constitute a deep uncertainty problem, and probability-based methods are no longer applicable for construction planning. To address this challenge, an uncertainty analysis approach that integrates Latin hypercube sampling (LHS), discrete-event simulation (DES), and the patient rule induction method (PRIM) is proposed. Specifically, it is progressed by LHS and DES to generate a wide array of uncertainty scenarios represented by possible PDs to quantify the robustness of various construction decisions; then, PRIM is used to identify the vulnerable scenarios that will jeopardize project completion. The approach was implemented on a real-world project, and the results demonstrated that it was able to identify the most robust construction schemes and vulnerable scenarios for construction planning. This research contributes a data-driven technology that provides an uncertainty analysis approach for construction planning without relying on assumed probability distributions from limited, unreliable project references.