Influence of effective stiffness on the performance of RC frame buildings designed using displacement-based method and evaluation of column effective stiffness using ANN

•The use of gross stiffness misinterprets the dynamic response of the buildings.•FEMA356 specified stiffness factors are also found deficient.•Developed an ANN model for estimating effective stiffness ratio of columns.•A model equation is also presented based on the trained weights of the ANN model. In the analysis and design of reinforced concrete (RC) buildings, member stiffness is generally taken as the gross stiffness, or as the effective stiffness as an empirical approximate fraction of gross stiffness. However, actual stiffness is strength dependent. Improper approximation of member stiffness affects the distribution of forces and deformation demands which lead to inaccurate structural response evaluation. Thus, the use of actual effective stiffness based on strength is very important in nonlinear analyses to evaluate the actual performance of the building under seismic conditions. A series of RC frame buildings have been designed using displacement-based method. The designed building have been subjected to nonlinear analyses incorporating three categories of stiffnesses, namely, gross stiffness, effective stiffness as per FEMA-356, and, actual effective stiffness based on strength. The results of nonlinear time history analysis show that the buildings analysed with gross stiffness exhibit very conservative drift and high performance level in comparison with those of buildings with effective stiffness based on strength. FEMA specified effective stiffness ratio yielded drift and performance level in between those with gross stiffness and effective stiffness based on strength. It is noted that, the estimation of actual effective stiffness for column sections is a tedious process. Therefore, to ease the process, a soft computing technique called artificial neural network (ANN), has been used to estimate the effective stiffness of columns sections. The statistical performances of the ANN model show that it can be used as an alternative tool to estimate the effective stiffness of RC column sections with good accuracy. Moreover, a model equation is presented based on the parameters of trained neural network which can be readily used to estimate effective stiffness ratios.