An Innovative Simulation-Free Approach for Probabilistic Assessment of Buried Pipeline Integrity Under Landslide-Induced Ground Movement

Abstract Buried pipelines serve as reliable conduits for transporting energy products such as oil, gas, and chemical fluids. Their pivotal role in facilitating the transportation of these resources underscores their substantial impact on businesses, economies, and the overall quality of life for people. Landslide-induced ground movement poses a risk to buried pipelines, potentially causing buckling damage or tensile rupture, threatening their structural integrity and operational safety. Hence, it is crucial to investigate the response of buried pipelines to mitigate the potential impacts of landslide-induced ground movement. However, due to the uncertainties in material, geometry, soil properties, and ground movement, it is imperative to conduct reliability-based analyses rather than deterministic analyses. Reliability methods like Monte Carlo, known for their simplicity and effectiveness, are widely used in reliability analyses. However, their significant drawback lies in the need for extensive simulations to generate responses, a challenge that proves impractical for studying pipelines buried through areas prone to ground movement. As such, this study introduces an innovative simulation-free approach based on the Physics-Informed Neural Network (PINN) to predict the response of inelastic pipelines under landslide-induced ground movement. Using the simulation-free approach for response predictions, the Monte Carlo simulation is then employed to determine the probability of failure. PINN offers a comprehensive solution to the underlying physics of the problem by incorporating stochastic variables like ground movement and eliminating the need for an extensive number of simulations to obtain the response. PINN utilizes a deep learning approach that operates without the need for training data by leveraging the underlying physics expressed through differential equations. The applicability of the employed method to pipelines subjected to permanent ground movement is demonstrated through a case study. The strain-based limit state function relies on established equations from the literature for determining strain capacity, while the strain demand is predicted using the PINN-based approach.