Damage detection and assessment are key objectives of structural health monitoring. Inspections timing and results accuracy represent crucial aspects to properly assess the integrity of a structural system over time. The goal of this work consists in the development of an automated damage classification methodology based on machine learning, that is able to process sampled measurements of the dynamic structural response as input and provide a quantitative health assessment of the structure as output. The methodology is validated on a broad set of examples, including both numerical simulations and data recorded on real structures, evidencing high accuracy across all instances. Moreover, the computational experiments show consistency in the assessment of location and severity of the damages even in presence of additive white Gaussian noise.
Using supervised learning for damage detection and assessment in structural health monitoring / Pastore, T.; Mariniello, G.; Menna, C.; Festa, P.; Asprone, D.. - (2019), pp. 485-492. (Intervento presentato al convegno 2019 International fib Symposium on Conceptual Design of Structures tenutosi a Eduardo Torroja Institute for Construction Science (IETcc), Serrano Galvache Street, n 4, esp nel 2019).
Using supervised learning for damage detection and assessment in structural health monitoring
Pastore T.;Menna C.;Asprone D.
2019
Abstract
Damage detection and assessment are key objectives of structural health monitoring. Inspections timing and results accuracy represent crucial aspects to properly assess the integrity of a structural system over time. The goal of this work consists in the development of an automated damage classification methodology based on machine learning, that is able to process sampled measurements of the dynamic structural response as input and provide a quantitative health assessment of the structure as output. The methodology is validated on a broad set of examples, including both numerical simulations and data recorded on real structures, evidencing high accuracy across all instances. Moreover, the computational experiments show consistency in the assessment of location and severity of the damages even in presence of additive white Gaussian noise.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
