Subsidence Analysis in Buried Pipelines Due to Fluid Leakage: Numerical Modeling and Machine Learning

Resumo

Buried pipeline systems are critical components of urban infrastructure, supporting essential services such as sewage, water supply, and the transport of oil and gas. Fluid leaks in these systems can significantly alter the stress distribution in the surrounding soil, leading to vertical deformations, ground surface subsidence, and, in severe cases, the formation of sinkholes. These processes involve the interplay of complex physical phenomena, including soil mechanics, solid mechanics, and unsaturated flow in porous media, making their accurate modeling both challenging and computationally intensive.This study proposes a numerical and data-driven framework for analyzing subsidence caused by fluid leakage in buried pipelines. A high-fidelity database is constructed from coupled finite element simulations of the elastic equilibrium equations (∇·σ = f) and the unsaturated flow equations governed by Richards’ equation, incorporating the Van Genuchten–Mualem constitutive relationships. The coupling is achieved through the effective stress–pore pressure interaction, capturing the mechanical-hydraulic behavior of the system.To identify the most influential parameters—such as pipeline depth, diameter, leak flow rate, and soil elasticity—a machine learning approach based on tree-based symbolic regression is employed. This method enables the discovery of explicit physical-mathematical laws from the simulation data, yielding interpretable analytical expressions that enhance the efficiency of subsequent numerical analyses.The developed models will be applied to a case study in the city of Rio de Janeiro to assess the potential impacts of hypothetical leakage scenarios. The results aim to support the prediction of structural risks and inform preventive maintenance strategies, contributing to the resilience and sustainability of urban buried infrastructure.