مجتبي حبيبي

The detection of damage to important structures such as bridges is one of the fundamental issues in civil engineering and structural maintenance management. In recent years, the detection of structural damage to prevent their sudden collapse and causing heavy casualties and financial losses has attracted the attention of many researchers. In general, the subject of structural pathology can be divided into local and global detections, and the general detection method can be divided into two types: static and dynamic. In this study, the dynamic general detection method (based on vibration) has been used. This research aims to detect damage in bridges and eva‎luate and compare the sensitivity of two modes of free and forced vibration in detecting structural damage using continuous wavelet transform and deep learning-based intelligent system. For this purpose, vibration data from the KW51 bridge accelerometer in Leuven, Belgium, collected under different loading conditions, were used. Initially, these data were divided into two categories of free vibrations (without moving load) and forced vibrations (with moving load), and through continuous wavelet transform, time-frequency features were extracted and converted into wavelet images. These images were given as input to a convolutional neural network to predict the state of the structure (healthy or damaged), of which 80% of the images were used for training and 20% of the images were used for testing the network. Analysis of the results showed that the accuracy of the convolutional neural network model in classifying and detecting damage in the forced vibration mode was 98.4% and 97.12%, respectively. This finding shows that forced vibration performs better than free vibration for damage detection and that forced vibration provides more accurate and detailed information about structural damage, because controlled external excitation can reveal hidden damage patterns that cannot be detected in the free vibration mode.

تشخيص خرابي , پايش سلامت سازه , شبكه عصبي كانولوشن , ارتعاش آزاد , ارتعاش اجباري , تبديل موجك پيوسته , يادگيري عميق

Damage detection , structural health monitoring , convolutional neural network , free vibration , forced vibration , continuous wavelet transform , deep learning

Mojtaba Habibi

Dr. Hasan Ziyari