A comparison of machine learning- and regression-based models for predicting ductility ratio of RC beam-column joints. (March 2022)

• Record Type: Journal Article
• Authors: Dabiri, Hamed; Rahimzadeh, Khashayar; Kheyroddin, Ali
• Is Part Of: Structures. Volume 37(2022)
• Page Start: 69
• Page End: 81
• Publication Date: 2022-03
• Subjects: Beam-column joints -- Ductility ratio -- Machine learning -- ANN -- Regression -- Random Forest; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2021.12.083
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: Seismic response of Reinforced Concrete (RC) beam-column joints has always been evaluated by researchers due to their undeniable influence on the overall behavior of RC building frames under seismic loads. In the present study, an attempt has been made to predict displacement ductility ratio of RC joints using machine learning- (Artificial Neural Network and Random Forest) and regression-based (linear, nonlinear and ridge) methods. Therefore, a dataset including the results of over 170 experimental studies conducted on RC joints was collected from the international peer-reviewed publications. Parameters reflecting beam and column dimensions (length, cross section width and depth), reinforcement detailing (longitudinal and transverse reinforcement of beams and columns), material properties (concrete compressive strength, yield strength of longitudinal and transverse reinforcement) and retrofitting techniques were considered as input variables for predicting the only output parameter, displacement ductility ratio. The predicted and actual values were compared together and the efficiency of the models was assessed by Taylor diagram and performance metrics including RSME, MAE, MAPE and R 2 . The evaluation results proved the high reliability of proposed models for predicting ductility of RC joints.