Toward Intelligent Fatigue Management of Welded Metal Sheet Structures by Integrating Simulation, Sensing, and Machine Learning

Welded metal sheets are used across automotive, aerospace, marine, and civil infrastructure applications for their high strength and light weight, yet fatigue failure remains a persistent concern in relation to their long-term structural stability. Fatigue crack growth is strongly influenced by the interaction of weld geometry and manufacturing-related defects. This review presents a comprehensive overview of fatigue-life prediction approaches, outlining the welding processes, fatigue mechanisms, computational tools, and progress made in predictive approaches central to this field. Fatigue performance is evaluated across several welding approaches, including laser beam, resistance spot, tungsten inert gas, and hybrid laser–arc welding. Advanced computational approaches for predicting fatigue life, including finite element modeling, cohesive zone modeling, crystal plasticity modeling, and multi-physics/multi-scale simulation, are also examined. Recent innovations in experimental characterization and structural health monitoring are addressed in detail. The review further highlights how machine learning-enabled digital twins and sensor fusion frameworks improve fatigue damage detection and maintenance prediction. Unlike prior reviews that treat conventional and data-driven methods separately, this work integrates welding process selection, fatigue life prediction, experimental validation, and structural health monitoring within a single closed-loop simulation–sensing–machine learning framework specific to thin welded sheets. Despite such advancements, the accurate prediction of environmental degradation and microstructural evolution remains challenging, and future research is required to integrate physics-based and data-driven models for critical welded structures.