Incorporating Project Bundling, Non-Stationary IRI Markov Models, and Equivalence Testing in Transportation Infrastructure Asset Management
Open Access
- Author:
- Hu, Pengsen
- Graduate Program:
- Civil Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 23, 2024
- Committee Members:
- Jay Regan, Professor in Charge/Director of Graduate Studies
Scarlett Miller, Outside Unit & Field Member
Shelley Stoffels, Chair & Dissertation Advisor
S. Ilgin Guler, Major Field Member
Shihui Shen, Major Field Member - Keywords:
- Transportation Infrastructure Asset Management
Non-Stationary Markov Models
Markov Chain Monte Carlo
Project Bundling
Multi-Objective Optimization
Quality Control/Assurance
Equivalence Testing - Abstract:
- Performance modeling, maintenance action optimization and performance monitoring are the key components of pavement management system (PMS). Developing International Roughness Index (IRI) models is a crucial step in pavement maintenance decision-making. Traditionally, the natural deterioration of IRI in large pavement systems has been modeled as a stationary Markov process in large transportation systems. However, factors such as pavement age, traffic levels, and structural strength can vary significantly among pavements and may influence the deterioration process of IRI. IRI models developed using data from a specific site or small region my loose generality and are not applicable to other pavements. In this dissertation, real IRI data from the Long-Term Pavement Performance program (LTPP) database were utilized. To maximize data utilization, pavement age was reset to zero after overlay actions were applied. The non-stationary characteristics of IRI transition probabilities were examined across three age ranges, two levels of Equivalent Single Axle Load (ESAL), and two road functional classes. Employing the Markov chain Monte Carlo (MCMC) method with the Metropolis-Hastings algorithm, the study estimated the mean and confidence intervals of transition probabilities for each scenario. The findings revealed that IRI transition probabilities remained independent of pavement age for the first 15 years. Interstate pavements, characterized by stronger structures, exhibited slower deterioration compared to non-interstate pavements, particularly in worse condition states. Furthermore, the study identified that unexpected ESAL levels led to accelerated deterioration in non-interstate pavements. Conversely, interstate pavements demonstrated a consistent deterioration rate across different ESAL levels, suggesting that these pavements were designed to accommodate the prevailing ESAL levels. Overall, these findings shed light on the dynamic nature of IRI transition probabilities and provide insights into the varying deterioration processes by traffic levels and road functional classes. Project bundling techniques were used to optimize the developed maintenance actions. The study area is located in Newark, Delaware. Three types of assets: pavements, sidewalks and bikeways and the corresponding three types of users: automobile user, bicyclist and pedestrians are considered, totaling 50 maintenance projects. The project bundling problem is formulated as a Multiobjective Multidimensional Knapsack Problem (MOMKP). Seven objectives relating to cost and user travel times are established. Dou to the numerous potential project combinations, the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is utilized as the solution method for computational efficiency. The relationship and trade-offs among the objectives are discussed. It was found that frequent pavement reconstruction and major repair projects can significantly increase pedestrian travel time, especially for ADA users. To improve the pedestrian experience, it's preferable to minimize the occurrence of such disruptive activities. Instead, prioritizing regular routine maintenance or minor repairs for pavements would be more conducive to pedestrian accessibility and comfort. The NSGA-II results converged after 700 iterations. It has shown that NSGA-II could reduce total agency cost of 0.65 million USD and total construction duration of 78 days. Performance monitoring plays a crucial role in providing condition data to support accurate infrastructure deterioration models and feedback on implemented policies. With the rapid development of infrastructure condition inspection equipment and technologies, data sources have become more diversified, and data collection efficiency has significantly improved. However, the accuracy of the new data needs to be examined. Additionally, the mixture of measurements resulting from changes in data sources, equipment, or technologies makes it more challenging to provide consistent and accurate feedback. To address these challenges, Two One-Sided T Tests (TOST) were applied to various scenarios involving three types of pavement data: cracking data, International Roughness Index (IRI) data, and Traffic Speed Deflectometer Device (TSDD) data. The TOST was utilized to examine whether data collected by equipment vendors can produce similar results to ground truth data, thus ensuring data accuracy. Furthermore, a case study was conducted to demonstrate how TOST can be incorporated into the vendor selection process. Additionally, we demonstrated the utility of TOST in investigating whether two different sources of IRI data can be used interchangeably to draw consistent conclusions in the context of pavement maintenance and condition rating. Similarly, for TSDD data, we adopted TOST to examine whether data collected at two sample frequencies can be deemed equivalent. Overall, the application of TOST in these scenarios highlights its effectiveness in assessing data accuracy, facilitating vendor selection processes, and ensuring the reliability of data interchangeability in infrastructure management decision-making.