LAUSR

Effective stiffness of reinforced concrete (RC) members has a very important role in the performance evaluation of RC frame buildings through nonlinear dynamic analyses. The beam effective stiffness can be readily computed using mechanics, but the evaluation of column stiffness is a complicated process and the use of support vector regression helps in this regard. Therefore, in this study, an attempt is made to predict the effective stiffness ratio of reinforced concrete columns using support vector regression (SVR) approach. A data set of 208 samples, which are collected through nonlinear dynamic analysis of reinforced concrete buildings using SAP2000 software, is utilized to develop the SVR model. The input parameters considered are reinforcement percentage, axial load and depth of the column section in both the perpendicular directions, and the output parameter is the effective stiffness ratio of columns. Three different kernel parameters are used, namely exponential radial basis function (ERBF), Gaussian radial basis function and polynomial function for SVR modelling, among which ERBF is found to be the most suitable one. The obtained results indicate that the statistical performance of the SVR-ERBF model is better than the models with other two kernels in predicting the effective stiffness ratio of reinforced concrete columns. Performance of the SVR model is compared with the results of multi-variable regression analysis. In addition to that, a sensitivity analysis is also performed to check the influence of each input parameter on output responses.