A modified analytical model for predicting seismic behaviors of unanchored liquid storage tanks

Large-scale liquid storage tanks, which are generally constructed unanchored, would suffer from severe damages subjected to intensive earthquakes, e.g., uplift, shell buckling and piping connection fracture, or even failure. This paper aims to develop a modified analytical model for performance assessment of large-scale unanchored storage tanks subjected to earthquakes, with a special focus on the effect of the uplift behavior. The stiffness of the rotational spring at the base of the analytical model was derived from an artificial neural network (ANN) on the spring resistance function. For this purpose, finite element (FE) models of unanchored storage tanks for static analysis were first verified by using experimental and numerical results from literature. Subsequently, the static analysis was performed on a large-scale unanchored storage tank to obtain the spring resistance function, i.e., the overturning moment – uplift height relationship, and a parametric analysis was further developed. Accordingly, a database consisted of 324 FE models was generated to establish the ANN and the spring resistance function was consequently derived. Responses of the large-scale storage tank under different earthquakes obtained from the modified model were compared with the numerical simulation result and reasonable accuracy with deviations of −24% and − 14% on the peak uplift height and peak base shear force was witnessed, respectively.