Identifying Damages in girders of Bridges Using Square Time-Frequency Distribution and Neural Network

Damage detection in structures has received much attention, especially in recent years. Due to the importance of bridges, the identification of damage in this type of structure is very significant. In this research, a new method for identifying damage in concrete girders of bridge decks is presented. Ease of use, high accuracy, and reduction of monitoring costs have been considered defaults for presenting the new method. In this research, using signal processing tools and artificial intelligence, damage-sensitive features have been extracted in such a way that the damage with its severity and location is determined with very high accuracy (about 99%) and error percentage less than 1. In fact, damage can be detected with such accuracy only by using vibration response signals received from a sensor. According to the proposed method, the response signals received from the structure in healthy and damaged conditions are processed using the time-frequency function. Then the processing results are defined as inputs to the neural network, and the appropriate outputs are determined. In other words, the neural network is trained. To evaluate, validate and ensure the performance of the method, the numerical model of concrete beams and the numerical model of Shahid Madani Bridge in Tabriz in normal and noisy conditions have been used. The results of the calculations show the high diagnostic accuracy of this method and the least amount of error in determining the health of the structure as well as identifying the location of the damaged element.