Search

Structural Health Monitoring of civil infrastructures often suffers from the limited availability of damage labelled data. The work here seeks to overcome this issue by using Transfer Learning approaches, in the form of Domain Adaptation, for leveraging information from a source structure with determined health‐state labels to make inferences on an unlabeled monitored structure. The idea is to exploit source data to train a Machine Learning algorithm and achieve improved early‐stage damage detection capabilities across a population of bridges. To account for differences in the underlying distributions of each structure, Transfer Learning is seen as a strategy enabling population‐level bridge SHM. In this paper, the natural frequencies obtained from multiple vibration measurements are extracted to characterise different domains during pristine and abnormal conditions. Such damage‐sensitive features are aligned via Domain Adaptation and used to train a standard classifier within a shared feature space. The methodology is validated on the heterogeneous population composed of the Z24 and S101 bridges. The results prove the effectiveness to successfully exchange damage labels, thus increasing available information for health‐state inference for SHM applications with sparce datasets. [ABSTRACT FROM AUTHOR]

Over the last decades, the rising number of aging infrastructures has progressively fueled much interest toward the field of structural health monitoring. Following the increasing popularity of artificial intelligence algorithms, an autoencoder‐based damage detection technique within the context of unsupervised learning is proposed in this paper to provide support for practical engineering applications. The developed methodology uses the autoencoder to reconstruct raw acceleration sequences of user‐defined length collected from a healthy structure. To quantify the errors between the original input and the reconstructed output, which may be representative of damage occurrence, two indexes of reconstruction loss are selected as damage‐sensitive features. To support damage detection, a selected number of short‐time sequences are finally grouped into a unique macrosequence. The novel procedure can effectively both work at the single sensor level, as well as combine the predictive models using an ensemble learning strategy. Avoiding system identification, results obtained in the Z24 bridge demonstrate that the proposed method is quite effective for local damage detection with limited computational effort and using a limited number of sensors, thereby suitable to be easily applicable in the context of real‐time bridge assessment. [ABSTRACT FROM AUTHOR]