ORCID Identifier(s)


Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Civil Engineering


Civil Engineering

First Advisor

Himan Jalali Hojat


Failure of sewer mains poses a significant threat to the society, necessitating a robust risk assessment tool that integrates failure likelihood and associated consequences for effective prioritization of mitigation efforts. This dissertation addresses this need through three key objectives: 1. Failure Likelihood Assessment: The study utilizes Monte-Carlo simulation to evaluate the probability of sewer main failures in common agressive environments, considering factors like sulfide and chloride exposures. It highlights that chloride-induced cracks and bond strength loss are more critical than sulfide-induced wall thickness loss. The degradation of concrete and reinforcement properties under chloride attack significantly reduces ductility, emphasizing the importance of factors like rust expansion coefficient, reinforcement size, and cover thickness in controlling service life. 2. Quantifying Consequences: This study establishes a quantitative framework to predict the monetary consequences of sewer main failures. It accurately predicts direct costs associated with two repair methods—Cured in Place Pipe (CIPP) as lining and open-cut replacement—using stepwise regression models. Additionally, it addresses indirect costs, including noise cost, pavement reduction value, traffic delay, and vehicle operating cost, by employing Monte-Carlo simulation to account for uncertainties. This approach offers a precise alternative comparison method, superior to qualitative or scaled quantitative techniques. 3. Adaptive Neuro-Fuzzy Integration: To enhance the versatility of fuzzy inference systems (FIS) in risk assessment, this research integrates FIS with an adaptive neural network, resulting in an adaptive neuro-fuzzy system capable of learning FIS parameters and adapting to evolving decision-maker preferences. The model undergoes training using diverse optimization algorithms and learning rates, with the Adam optimizer demonstrating efficiency in reducing training trials. In conclusion, this dissertation culminates in a comparative analysis of the proposed neuro-fuzzy model with conventional risk assessment approaches, such as the risk matrix and parameter multiplication. Findings reveal that the parameter multiplication method, though sensitive to input uncertainties, does not rely on decision rules, making it suitable when decision rules are absent or precise predictions are available. Conversely, the risk matrix and neuro-fuzzy approaches prioritize actions based on decision rules rather than exact likelihood and consequence values, rendering them more appropriate for informed decision-making. Notably, the adaptive neuro-fuzzy approach enhances interpretability and facilitates smoother transitions between risk classes, promoting superior decision-making and action prioritization, particularly for pipes within the same risk category.


Sewer pipes, Chloride corrosion, Sulfide erosion, Reliability analysis, Risk assessment, Neuro-fuzzy, Machine learning, Monetary consequences


Civil and Environmental Engineering | Civil Engineering | Engineering


Degree granted by The University of Texas at Arlington