Simplified Seismic Dynamic Analysis of Sloshing Phenomenon in Rectangular Tanks with Multiple Vertical Baffles

محورهای موضوعی : Journal of Water Sciences Research

M Hosseini ¹ (Associate Professor, International Institute of Earthquake Engineering and Seismology, Tehran, Iran, and Part Time Faculty Memner in Graduate School, Civil Engineering Deptment, Islamic Azad University- South Tehran Branch, Tehran, Iran.)
H.R Vosoughifar ² (Assistant Professor, Department of Civil Engineering, Islamic Azad University- South Tehran Branch, Tehran, Iran)
P Farshadmanesh ³ (Graduate M. Sc., Department of Civil Engineering, Islamic Azad University- South Tehran Branch, Tehran, Iran)

تاریخ دریافت : 1394/02/27 تاریخ پذیرش : 1394/02/27 تاریخ انتشار : 1392/04/10

کلید واژه: Neural network, Time History Analysis, Multiple Vertical Baffle, Sloshing phenomenon,

چکیده مقاله :

Sloshing is a well-known phenomenon in liquid storage tanks subjected to base or body motions. In recent years the use of multiple vertical baffles for reducing the sloshing effects in tanks subjected to earthquake has not been taken into consideration so much. On the other hand, although some of the existing computer programs are capable to model sloshing phenomenon with acceptable accuracy, the full dynamic analysis subjected to random excitations, such as earthquake induced motions, is very time consuming. In this paper a method is presented for this purpose based on conducting several dynamic analysis cases, by using ANSYS-CFX for rectangular tanks with various dimensions, subjected to seismic excitations, and then using neural network to create simple relationships between the dominant frequency and amplitude of the base excitations and the maximum level of liquid in the tank during the sloshing. At first, the numerical modeling has been verified by using some existing experimental data. Then, several cases of time history analysis have been conducted to obtain the required numerical results for teaching the neural network. Finally, the predicted results of the neural network have been compared with those obtained by some other cases of analyses as control values.