Probabilistic Structural Damage Identification with Uncertain Data by Deep Learning Techniques

Abstract

This PhD thesis proposes innovative methods based on deep learning techniques, such as convolutional neural networks and Bayesian neural networks, for structural damage identification with uncertain data. These approaches improve the performance and reliability of structural damage detection and quantification under the effect of uncertainties, such as measurement noise and modelling inaccuracies. Numerical and experimental studies are conducted to validate the accuracy and performance of the proposed approaches.