To aid the design of empty, open-top, unstiffened, ground-supported, steel cylindrical-tanks against wind-induced buckling, this study proposes a fast and innovative artificial neural network (ANN) to predict buckling load-multiplier, assessing the protective effects of geometric aspect ratios of tank and fill level, based on stability analysis.
A multiphysics system coupling has been utilized to perform finite element methodology-based one-way wind-structure interaction analysis by joining computational fluid dynamics and structural mechanics (eigenvalue buckling) solvers. The accuracy of the numerical model is ensured through experimental and theoretical validations. Basic wind speed (Vb), tank diameter (D), filling height to tank height (HF/H), tank height to diameter (H/D) and tank radius to wall thickness (r/t) ratios have been varied as inputs for studying the wind-induced buckling through buckling load multiplier (λ).
Four different stability conditions, namely safe stability (λ>2), low stability (1<λ ≤ 2), critical stability (λ≈1) and instability (λ<1) are observed based on load-multiplier values. An economically safe buckling capacity is observed for H/D ratios of 0.5 and ≥ 0.75 up to 1.0 in 75% filled tanks with diameters of 15m and 20m with r/t ratios of 1,000 and 750, respectively. An empty tank with H/D ≤ 0.25 is completely safe against wind-induced buckling when r/t ratio 750 and 1,000 are ensured, respectively for tank diameter ≤ 15m and 20m.
ANN has been trained efficiently with ≥ 60% data from the multiphysics analyses, which showcased 97.03% accuracy for assessing the buckling load multiplier of unstiffened, open-top, steel tank against wind-induced buckling. The developed ANN model can predict the required fluid level inside the unstiffened tank to maintain its stability against wind-induced buckling, based on the velocity of an impending storm and the tank’s geometrical features.
