Identification of Semi-rigid Joints in Steel Frame Structures Using Vibration-Based Technique

Abstract

The connections are the critical locations where damage commonly occurs. However, very limited vibration-based studies using the artificial neural network (ANN) are found to identify the damaged connections. Again, the ANN-based technique requires retraining of network for little variation in geometry of the structure. Moreover, those studies were limited to damage at the beam end connection. In this present study, health monitoring of steel plane frame structures having semi-rigid connections either at the beam side or at the column side using a limited number of sensors is addressed. With that purpose, a single-storey and a two-storey frame are considered. The frames are modeled using plane frame elements, in which two rotational springs are placed at the ends to affect the stiffness of the rotational springs only. The frames are excited using an impact at the right top corner, and strain time responses are collected from the connections only. The strain data are then transformed into frequency spectra. Using the frequency spectra, an objective function is developed and minimized using particle swarm optimization (PSO) to get the updated fixity factors for all the springs. In order to get more accurate values of fixity factors of the two-storey frame, it is divided into sub-structures, and it is found that the technique estimates fixity factors with an acceptable error.