When sensing meets computational Intelligence: A comprehensive survey on pipeline damage detection and localization

Intelligent computational methods are increasingly central to pipeline damage detection and localization, enabling a transition from periodic inspection toward continuous, data-driven structural health monitoring. This review provides an integrated and critical synthesis of machine learning, deep learning, and advanced computational approaches for pipeline monitoring, examined in direct relation to pipeline failure mechanisms, sensing data characteristics, and localization requirements. Modern sensing modalities (e.g., acoustic, fiber-optic, pressure, and ultrasonic measurements) are examined primarily as inputs to intelligent computational frameworks for pipeline damage detection and localization under nonstationary conditions. Beyond methodological developments, the review critically examines implementation challenges that limit real-world deployment, including sensitivity to noise, scarcity of labeled fault data, generalization across pipeline assets, computational scalability, and integration with operational constraints. Emerging paradigms such as physics-informed and hybrid modeling, transfer learning, generative data augmentation, edge computing, and autonomous monitoring are discussed as enablers for reliable and scalable deployment. Unlike prior reviews that treat sensing and analytics in isolation, this work unifies physical failure processes, sensing strategies, and intelligent computation within a single engineering framework, highlighting pathways for translating advanced analytics into actionable decision-support tools for inspection planning, risk assessment, and maintenance of large-scale pipeline infrastructure.