DAMAGE DETECTION IN STEEL RAILWAY BRIDGES

Abstract

Structures are exposed to damages during their service life which can severely affect their safety and functionality. Thus to monitor a structure plays very important role so that it can be repaired before it gets affected due to damage. Finding out location of damage and extent of damage helps us to decide further strategies to repair the structure. Bridges are important structures as they play important role in transportation and rescue operation. There can be visible or invisible damages caused due to various forces such as wind, blast, and earthquake. Damage can also cause due to aging, repetitive loading and unloading, corrosion. There have been some disastrous failures of bridges due to undetected progressive damage in the past. There is therefore considerable interest in continuous monitoring of bridges. In case of truss and girder bridges cracks at joint locations causes rotation of the joints. Rotational stiffness of the members connected to the joints gets reduced. If this effect is not taken care of, it results in widening of cracks and failure of members connected to the damaged joint. Therefore rotational stiffnesses of joints were reduced in truss and girder models. Free vibration analysis was performed by connecting partial fixity spring to the members connected to the damages joint. Frequency, displacement and curvature were obtained at joint locations. Curvature was obtained by using central difference method. Parameters obtained from analysis of damaged cases were compared with undamaged case to obtain particular pattern in variation. Percentage changes in frequency were obtained and displacements were compared using modal assurance criterion. Variation was found to be negligible. Artificial neural network is one of the software which is used in various fields to achieve reasonable results wherever changes occurred in parameters due to damage. Here, application of artificial neural network is presented to detect damage using frequency, displacement and curvature in truss and frequency, displacement in Girder Bridge in India as damage causes changes in these parameters. The technique which was employed to overcome the issues associated with many unknown parameters in a large structural system is the sub structural identification. Trial and error method is used to improve results of neural network training. Required data for training was obtained by analyzing the models of bridge for different damage patterns. Data is trained efficiently to detect damage and accuracy is also examined in terms of percentages.