This paper proposes a novel deep-learning approach for verifying concrete cracks, leveraging a Siamese network with triplet loss. Inspired by FaceNet, the model is trained on an augmented 'Crack dataset' containing rotations and shearing to enhance its ability to distinguish subtle crack variations. This method facilitates the evolution-centric monitoring of cracks, enabling early detection of potential failures and contributing to the development of digital twins for structural health moni-toring. The model achieves high accuracy (97.36 %) and precision (95.77 %) in verifying crack patterns, demonstrating its potential for improving infrastructure maintenance and safety.

Triplet Loss-Based Concrete Crack Verification for Structural Health Monitoring and Digital Twin Applications

Neyestani A.;Picariello F.;Tudosa I.;Daponte P.;De Vito L.
2024-01-01

Abstract

This paper proposes a novel deep-learning approach for verifying concrete cracks, leveraging a Siamese network with triplet loss. Inspired by FaceNet, the model is trained on an augmented 'Crack dataset' containing rotations and shearing to enhance its ability to distinguish subtle crack variations. This method facilitates the evolution-centric monitoring of cracks, enabling early detection of potential failures and contributing to the development of digital twins for structural health moni-toring. The model achieves high accuracy (97.36 %) and precision (95.77 %) in verifying crack patterns, demonstrating its potential for improving infrastructure maintenance and safety.
2024
cracks verification
digital twins. Resnet
measurement
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12070/68766
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact