CIAM-UTC-REG54
Research Team
Nii Attoh-Okine, University of Delaware
Sue McNeil, University of Delaware
Funding Sources
University of Delaware Federal Share — $62,188
University of Delaware Match Share — $62,188
Total Project Cost — $124,376
Agency ID or Contract Number
69A3551847103
Start and End Dates
03/04/2022 — 10/03/2023
Project Description
Limited resources, aging infrastructure, and an increased focus on performance and life cycle costs rather than simply condition have led to widespread adoption of asset management principles by state DOTs. The practice of asset management is heavily influenced by a broad base of experience with asset specific tools for pavement and bridge management, commercial software, and the federal requirement to develop Transportation Asset Management Plans for NHS bridges and pavements.
While states use deterioration and costs models, identify goals and objectives, and aim to identify optimal preservation and improvement decisions, the state of the practice does reflect widespread application of the complex decision-making models developed by the research community (Chen & Bai, 2019; Chen, Liang, Wu, & Sun, 2019). In fact, there is some skepticism and fear of the loss of transparency, despite demonstrated gains realized in the application of the more complex models (Wang & Pyle, 2019). This project develops a framework that addresses this concern. The project begins with a complex model developed as part of a previous CIAMTIS project “Strategic prioritization and planning for multi-asset transportation infrastructure maintenance, rehabilitation, and improvements: Phase 1 – Prioritization through optimization” (Zhou, Miller-Hooks, Papakonstantinou, Stoffels, & McNeil, 2021). The framework will be developed by defining scenarios, parameterizing scenarios, running the model, comparing the results to simpler tools such as decision trees and thresholds, and translating them into updated tools that are more accessible to state DOTs. The objective is to use hybrid network analysis and machine learning tools to understand when the complex model can be reduced to the simpler tools and when the application of the more complex model are warranted.