Experimental and numerical analysis of the seismic response of water storage tanks incorporating nonlinear soil-structure-liquid interaction

Abstract

Post-earthquake observations have repeatedly shown the failure of liquid storage tanks. The likelihood of failure depends on, among other variables, the boundary conditions at the base and top of the tank, the aspect ratio (liquid height to tank radius) and the interaction between the liquid content, the tank and the surrounding soil (soil-structure-liquid interaction). The main objective of this doctoral thesis is to evaluate, through physical experiments, the seismic response of storage tanks incorporating soil-structure-liquid interaction. The long-term focus of research in this field is to enable a realistic design of tanks under earthquake loadings. The outcome is to derive an understating of tank response under simplified conditions that will support the long-term goal. A scaled low-density polyethylene tank with six different aspect ratios is tested on a shake table. The tank is placed directly on a steel plate, i.e., representing a rigid supporting medium (rock), and on sand in a laminar box, i.e., representing a flexible supporting medium (soil site). Uplift defined here as the transient partial separation of the base of the tank from the supporting medium is also considered. To evaluate the effects of chaotic sloshing on the tank response, a floating lid is utilised as a barrier. The results are compared to those from a commonly used nonlinear elastic spring-mass model for the seismic analysis of storage tanks. The seismic tank response strongly depends on the coincidence between the dominant frequency of the excitation and the first-free vibration frequency of the soil-tank-liquid system. It is shown that when the coincidence of frequencies occurs, sloshing may become chaotic and the occurrence and magnitude of uplift increases, affecting the concentration of stress in the tank wall. It is also revealed that the aspect ratio plays a significant role, especially when the tank lies on a flexible supporting medium. The comparison between the experimentally obtained response and that from the spring-mass model reveal that the simplifications made in the spring-mass model are not always adequate to represent the tank response. The differences in the results increase when chaotic sloshing occurs.