Sharif University of TechnologySharif Journal of Civil Engineering2676-476838.21.220220522INVESTIGATION OF EFFECTIVE PARAMETERS IN SHORT COLUMN FAILURE USING CORRELATION AND MACHINE LEARNING METHODSINVESTIGATION OF EFFECTIVE PARAMETERS IN SHORT COLUMN FAILURE USING CORRELATION AND MACHINE LEARNING METHODS77872266110.24200/j30.2022.59055.3021FAZ. NouriInternational Institute of Earthquake Engineering and SeismologyF. Nateghi AlahiInternational Institute of Earthquake Engineering and SeismologyJournal Article20211009Column shear failure remains one of the most important causes of damage and collapse of reinforced concrete buildings during recent earthquakes, which should be avoided due to its low ductility and brittle failure mechanism. In previous studies, various parameters have been introduced as effective factors in short column mechanism and in each case, the effect of some of them has been studied. However, there is no comprehensive study that investigates the importance of all of these parameters. In this paper, using Monte Carlo algorithm and considering the normal distribution for 10 effective parameters in short column formation, including column cross-section size, column longitudinal reinforcement ratio, column transverse reinforcement ratio, effective column length, concrete compressive strength, reinforcement yield strength, beam length, axial force ratio, infill wall to column height ratio, and wall thickness, a database consisting of 200,000 samples is created. OpenSees software is used to model the concrete moment frame by considering the flexural and shear behavior of the column, and the model is verified by comparison with experimental studies. Then, by using push-over analysis, the type of failure mechanism of the column in a moment frame with infill and opening is determined to be flexural or shear failure. The importance of each parameter is investigated using machine learning methods including Principal Component Analysis (PCA), Decision Tree (DT), and F-Test (FT) as well as Pearson and Spearman correlation methods. DT and FT machine learning methods as well as both Pearson and Spearman correlation methods are well able to identify the importance of each parameter in the formation of the short column. By summarizing the results of all methods, the parameters of the percentage of column shear reinforcement as well as the ratio of wall to column height have been determined as the most important and effective parameters. Also, the least important parameters are fy, L-Beam and axial load ratio. The result of this paper will be useful for designer of RC building and also to develop models and criteria for rapid short column identification in seismic evaluation of existing buildings.Column shear failure remains one of the most important causes of damage and collapse of reinforced concrete buildings during recent earthquakes, which should be avoided due to its low ductility and brittle failure mechanism. In previous studies, various parameters have been introduced as effective factors in short column mechanism and in each case, the effect of some of them has been studied. However, there is no comprehensive study that investigates the importance of all of these parameters. In this paper, using Monte Carlo algorithm and considering the normal distribution for 10 effective parameters in short column formation, including column cross-section size, column longitudinal reinforcement ratio, column transverse reinforcement ratio, effective column length, concrete compressive strength, reinforcement yield strength, beam length, axial force ratio, infill wall to column height ratio, and wall thickness, a database consisting of 200,000 samples is created. OpenSees software is used to model the concrete moment frame by considering the flexural and shear behavior of the column, and the model is verified by comparison with experimental studies. Then, by using push-over analysis, the type of failure mechanism of the column in a moment frame with infill and opening is determined to be flexural or shear failure. The importance of each parameter is investigated using machine learning methods including Principal Component Analysis (PCA), Decision Tree (DT), and F-Test (FT) as well as Pearson and Spearman correlation methods. DT and FT machine learning methods as well as both Pearson and Spearman correlation methods are well able to identify the importance of each parameter in the formation of the short column. By summarizing the results of all methods, the parameters of the percentage of column shear reinforcement as well as the ratio of wall to column height have been determined as the most important and effective parameters. Also, the least important parameters are fy, L-Beam and axial load ratio. The result of this paper will be useful for designer of RC building and also to develop models and criteria for rapid short column identification in seismic evaluation of existing buildings.https://sjce.journals.sharif.edu/article_22661_3c037d79dfa1e608b3acc7ffcafae999.pdf