Prediction of Joint Shear Strain-Stress Envelope Through Generalized Regression Neural Networks

Abstract

In structural engineering practice, it is widely accepted that beam-to-column joints in reinforced concrete frames can be idealized as rigid regions. However, recent studies demonstrated that severe damage can be observed in these regions and neglecting inelastic deformations can lead to misinterpretations in performance-based seismic design and assessment process. Despite the large experimental and analytical efforts in establishing a generalized method for predicting inelastic behavior of exterior and interior beam-to-column joints, literature survey revealed that there has only been little consensus about the factors affecting the shear stress-strain envelope. This study introduces the application of Generalized Regression Neural Networks to joint deformation problem and proposes a prediction model. Accuracy and reliability of the proposed model are demonstrated with statistical measures and comparison to various methods available in the literature.