Implementation of innovative techniques based on machine learning and computer vision for monitoring bridges and viaducts in the States of São Paulo and Pará

Abstract

The evolution of the Structural Health Monitoring (SHM) process has brought new challenges to be overcome. Changes in structural dynamics generated by operational and environmental variation continue to be the greatest technical challenge, as they mask the existence of structural damage, changing the level of severity. Although the number of approaches to attenuate these effects is growing rapidly, most still suffer from problems of unfamiliarity with the initial conditions, such as the high number of input parameters and the instability of the models. In addition, there is a growing need for approaches to autonomous data acquisition without direct contact with the structure. This demand has driven the development of image-based techniques for modal analysis and identification, which can be coupled to machine learning models for real-time dynamic monitoring. This project aims to develop new approaches for data acquisition, modal analysis and damage detection based on computer vision and machine learning. The project aims to improve the technique of digital twins for damage detection in bridges and viaducts. Instrumentation data, executive design and the mechanical and geometric properties of the structural elements will be used to calibrate the digital models. In addition to creating these models, the SP team will share instrumentation information to improve damage detection algorithms in structures developed by the PA team. The systems will be applied in field tests to monitor bridges and viaducts, and validated with reference instruments. The monitoring system will provide scientific and technological innovations to the states of SP and PA, with an impact on the industrial, economic and environmental spheres. (AU)