Uso de Autocodificadores Variacionais para Monitoramento de Integridade Estrutural

Detecção e Quantificação de Alterações Estruturais Utilizando Autocodificadores Variacionais combinados com Cartas de Controle T²

Abstract

Structural Health Monitoring (SHM) aims to ensure the safety and reliability of civil infrastructures. Autoencoders, unsupervised machine learning models, have shown promise for SHM by learning data features and reducing dimensionality. However, comprehensive studies comparing autoencoder models in SHM are scarce. This study investigates the effectiveness of the Variational Autoencoder (VAE), combined with T² Control Chart, to detect and quantify structural changes in three civil engineering structures. Acceleration signals collected by accelerometers were used as input for the autoencoder to perform unsupervised classification. The latent layer data generated by the VAE were applied to T² analysis, and the obtained values were compared between signal subgroups to identify structural changes. The study concludes that the VAE model combined with T² is effective in both identifying and quantifying changes in structures. The ongoing development of this technique can contribute to advancements in SHM, promoting greater safety, cost reduction, and long-term durability of civil engineering structures.