Aplicación de un modelo de Deep Learning con redes neuronales convolucionales para evaluar la integridad estructural y predecir fallas en puentes de concreto mediante exploración visual de imágenes

Abstract

In the area of civil engineering, it is very important to review the structural health of concrete bridges. This review aims to maintain the safety and improve the maintenance of these structures. On the other hand, in the area of Artificial Intelligence, Convolutional Neural Networks have been successfully used to analyze and classify images of different types and origins. This thesis aims to create a system that uses a Deep Learning model using Convolutional Neural Networks with the VGG 16 architecture. Its main function is to evaluate the structure of bridges and classify damages through images taken from them, focusing on problems such as cracks, efflorescence and peeling. The VGG 16 model was tested to identify these structural damages demonstrating satisfactory results. The results indicate that the system has a high accuracy, especially in the classification of cracks, with a maximum accuracy of 0.81. This shows that the system is effective in detecting and classifying defects, becoming a useful tool for the inspection and maintenance of bridges.