Pipeline Risk Management Using Artificial Intelligence-Enabled Modeling and Decision Making

Overview

Fast Facts
Project No.	954
Contract No.	693JK32150001CAAP
Research Award Recipient	Rutgers, The State University 3 Rutgers Plaza New Brunswick, NJ 08907
AOR/TTI	Zhongquan Zhou Nathan Schoenkin Brady Dague Nusnin Akter
Researcher Contact Info	Dr. Hao Wang, Associate Professor Department of Civil & Environmental Engineering Rutgers University, The State University of New Jersey 500 Bartholomew Road, Piscataway, NJ, 08854 Phone: 848-445-2874, Email: hwang.cee@rutgers.edu

Financial and Status Data
Project Status	Active
Start Fiscal Year	2021 (09/30/2021)
End Fiscal Year	2025 (09/30/2025)
PHMSA $$ Budgeted	$349,328.00

Main Objective

The research objective is to develop Artificial Intelligence (AI) - enabled tools to improve accuracy of probabilistic performance modeling and support decision making of inspection and repair actions in pipeline risk management. The educational objective is to engage, inspire, and train undergraduate and graduate students through research and prepare them for future career in pipeline industry.

Public Abstract

Quantitative risk-based management has been widely accepted in pipeline industry for supporting cost-effective decisions while ensuring pipeline safety. However, there is a need for processing pipeline inspection data to minimize data errors and to extract essential information for risk assessment and decision making. This research intends to develop Artificial Intelligence (AI) based solutions that can identify connections between pipeline safety datasets through data analytics, develop data-driven probabilistic prediction models of pipeline degradation using Bayesian Neural Network (BNN), quantify probability of failure with uncertainties, and support decision making of pipeline inspection and repair using an innovative Reinforcement Learning (RL) approach. The outcome from this project will provide tools to pipeline operators for facilitating reliable and autonomous decision making as well as saving pipeline inspection and repair costs.

Anticipated Results: The project is expected to develop artificial intelligence (AI) based solutions that can identify connections between pipeline safety datasets through data analytics, develop data-driven probabilistic prediction models of pipeline degradation using Bayesian neural network (BNN), quantify probability of failure with uncertainties, and support decision making of pipeline inspection and repair using an innovative reinforcement learning (RL) approach.

Potential Impact on Safety: The results of this study will allow operators to utilize AI for quantitative risk-based management and make better informed integrity management decisions.