Banca de DEFESA: Victor Fabrício Magalhães Carvalho

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Victor Fabrício Magalhães Carvalho
DATE: 15/09/2023
TIME: 08:30
LOCAL: Online via plataforma Teams
TITLE:

Structural damage detection based on acceleration time history and Artificial Neural Networks

KEY WORDS:

Structural Health Monitoring (SHM); Damage Detection; Artificial Neutral Networks

PAGES: 100
BIG AREA: Engenharias
AREA: Engenharia Civil
SUBÁREA: Estruturas
SPECIALTY: Mecânica das Estruturas
SUMMARY:

The natural or accidental deterioration of large structures and the growing complexity of projects demand constant monitoring of the health of constructions. This process enables interventions, with the aim of avoiding catastrophic failures and also reducing costs with repairs and replacement of structural components. In this sense, damage assessment methods based on variations in the modal properties of structures were widely studied, due to their global assessment capabilities. However, some challenges were presented to the practical application of these methods, such as the loss of information on the structure's vibration signals during the modal identification process and the small variations recorded in parameters such as natural frequencies, making it difficult to assess the state of a structure exposed to environmental effects. In order to overcome these challenges, new monitoring methodologies that directly apply statistical techniques to the acceleration time histories obtained through monitoring have been developed. This work aims to evaluate the use of statistical indicators obtained from the acceleration time history of structures as an alternative input for training and testing Artificial Neural Networks in structural damage detection methodologies. To this end, the study was carried out in two stages: in the first stage, only numerically modeled beams were studied, considering different cases of structural damage; in the second stage, numerical models and experimental data of a 3D frame were used, considering different cases of structural damage. In both stages, the structures were subjected to impacts and the acceleration signals were extracted for further processing in the form of statistical indicators, which in turn were used to train and test the Artificial Neural Networks proposed for structural damage detection, varying and analyzing factors such as the number of neurons in the hidden layer, the network activation functions, the need to filter the noise of the experimentally obtained time series and the effect of peak amplitudes of accelerations in the training of the networks. The tests showed good performance using statistical indicators as parameters for training networks for damage detection.

COMMITTEE MEMBERS:
Presidente - 1225412 - GRACIELA NORA DOZ DE CARVALHO
Interno - 1111740 - GUILHERME SANTANA ALENCAR
Externa ao Programa - 2213675 - SUZANA MOREIRA AVILA - UnBExterna à Instituição - IVIANE CUNHA E SANTOS