نيلوفر احمدي

Structural health monitoring is one of the popular branches in civil engineering. The purpose of structural health monitoring is to detect the presence of damage, determine its location and the severity of the damage. One of the emerging and widely used fields in structural health monitoring is the use of artificial neural networks and deep learning. In deep learning, a model can identify the target subject according to the data it receives. So far, most researches have examined raw data for structural health monitoring, while various types of data transformations and pre-processing can result in other important information and features. In the upcoming research, continuous wavelet transform has been used to highlight the frequency of raw signal matrix data of a bridge for damage detection with neural networks and semantic segmentation of images. Semantic image segmentation is able to eva‎luate the location and frequency value in the group of accelerogram images. Two neural network models FCN and U-Net have been investigated for this purpose. The results of the two mentioned architectures indicate the high accuracy of the above two models in predicting the important features of the data. Finally, by comparing these features, it is possible to separate healthy and damaged structures.

پايش سلامت سازه , يادگيري عميق , شبكه عصبي كانولوشني , تقسيم¬بندي معنايي تصاوير , تبديل موجك

structural health monitoring , convolutional neural network , semantic segmentation of images , wavelet transform , deep learning

niloofar ahmadi

gholamreza ghodrati amiri