Predicting natural vibration period of concrete frame structures having masonry infill using machine learning techniques

The natural period of vibration is one of the most significant factors used in the seismic design of buildings. Although the building design codes and previous studies provide some empirical methods to compute the natural period of vibration (T), their marginal accuracy and inability to incorporate the effect of masonry infill on vibration period significantly limits their use. Thus, researchers are constantly trying to find new and accurate methods to calculate T of concrete structures. To this end, this study presents the novel approach to predict T of reinforced concrete framed buildings (RC buildings) using Multi Expression Programming (MEP), Extreme Gradient Boosting (XGB), and Gene Expression Programming (GEP) machine learning algorithms. For this purpose, an extensive dataset of 569 points was gathered from previously published studies and split into training and testing sets for training and testing the algorithms respectively by using several error evaluation metrices like coefficient of determination (R²), Root Mean Squared Error (RMSE), and Objective Function (OF) etc. The results of error evaluation showed that XGB is the most accurate algorithm having the least OF value of 0.012 compared to 0.037 of MEP and 0.048 of GEP. Additionally, several explanatory analyses like sensitivity and shapley analysis were conducted on the XGB model which showed that number of storeys and opening ratio are the most contributing variables to prediction period of vibration. Thus, the models developed in this study can be practically utilized for determining natural vibration period of reinforced concrete frames with masonry infill.