Water hammer propagation in pipes under multi-defect coupling and gradient boosting decision tree-based intelligent prediction

Accurate prediction of water hammer pressure is vital for the safety of pressurized pipe networks. Conventional calculation methods often fail to account for the nonlinear coupling of multiple defects, such as erosion, scaling, and leakage, that develop during long-term service. This work presents a unified framework integrating experimental modeling, a physics-based numerical solver, and a gradient boosting decision tree (GBDT) algorithm to address this challenge. Applied to acrylate polymer blended with poly(vinyl chloride) resin pipes, the framework incorporates complex parameters encompassing pipe dimensions, defect geometries, and operational conditions. Experimental validation confirms that the baseline numerical model accurately captures transient responses. Extensive coupled simulations reveal that water hammer pressure amplification is highly scenario dependent, ranging from approximately 1.25 to 2.2 times the initial steady-state pressure depending on the specific combination of scaling severity and erosion levels. A pivotal finding is the dual role of erosion; while its direct effect on peak pressure is modest, the resultant increase in the diameter-to-thickness ratio profoundly intensifies fluid–structure interaction, shifting the failure risk from pure overpressure to coupled pressure–structure instability. In contrast, leakage primarily serves as an energy dissipation mechanism, attenuating wave amplitude. The developed GBDT model demonstrates exceptional capability in learning the complex mappings from multi-defect parameters to extreme pressures, achieving an R2 of 0.99 on a comprehensive high-fidelity numerical dataset. This unified framework bridges hydraulic transient analysis with defect mechanics and data-driven intelligence, providing a robust tool for risk assessment in aging pipe infrastructure.