Bucking Reliability Assessment of Offshore Fluid Tanks Subjected to External Pressure With Active Learning Kriging Approaches

Offshore fluid tanks are critical components of offshore oil and gas infrastructure and are often subjected to external pressures that may lead to catastrophic buckling failure. Reliable assessment of their structural integrity is paramount to ensure offshore operations’ safety and environmental sustainability. Traditionally, reliability assessment relies on deterministic methods or experienced formulas, which may not accurately capture uncertainties from the structural and environmental parameters. Furthermore, classical probabilistic methods for reliability assessment also have their limitations. Approximation approaches like the first-order reliability method (FORM) rely on the linearity of the problems. Sampling-based approaches for the reliability assessment, e.g., Monte Carlo simulation (MCS), require many simulations/experiments, which is unrealistic for the real engineer design and optimization process. To address this problem, this study leverages advanced machine learning techniques, named active learning reliability methods, for buckling reliability assessments of offshore fluid tanks under external pressure. One Automatic Tank Cleaning (ATC) container machine from Dynamic Well Solutions AS company is used for the reliability assessment case study. The linear buckling analysis of the ATC tank is performed using finite element methods. The uncertainties encompassing geometric characteristics, material properties, and external loads are considered for the reliability assessment. One typical active learning reliability approach, AK-MCS with U learning function, is applied to estimate the probability of failure considering different minimum buckling factor requirements. The results demonstrate that the active learning reliability approach can significantly reduce the computational cost for the buckling reliability assessment. It enables an accurate and probabilistic assessment of the buckling reliability. Applying active learning reliability techniques could facilitate structural design optimization considering uncertainties.