Fragility analysis of buried continuous pipelines under normal faulting through analytical solution combining with machine learning technique

Crossing major active faults is often unavoidable for pipelines in earthquake-prone regions. This study explores a finite difference analytical solution for continuous pipelines subjected to normal faulting. The effectiveness of the proposed method is evaluated by comparing the calculated results with the data from centrifuge and full-scale laboratory tests. Combining with the Lasso regression machine learning technique, multiparameter probabilistic risk assessment of pipelines can be performed to generate fragility curves. Probability density function curves of normalized location of pipe failure are also computed. Results show that pipelines with a larger pipe wall thickness and a smaller pipe diameter buried at a shallower depth in soils with a lower elastic modulus are less prone to failure. The dominant failure mode of pipelines transits from local buckling to tensile strain, depending on the exceedance of critical diameter-to-thickness ( D/t) ratio and burial depth. Greater attention should be placed on the failure mode of local buckling, and the normalized location of pipe failure moves away from the fault plane, when a shallow burial depth is selected. The fragility relations help to understand the relative vulnerability of pipelines with different parameters, and provide implications for the design and repair of pipelines in seismic-prone regions.