Multilayer perceptrons are utilized in this work for vibration-based damage detection of multicomponent aerospace structures. A back-propagation algorithm is utilized along with Monte Carlo simulations and advanced structural theories for training Artificial Neural Networks (ANN’s), which are able to detect and classify local damages in structures given the natural frequencies and the associated vibrations modes. The latter ones are feed into the network in terms of Modal Assurance Criterion (MAC), which is a scalar representing the degree of consistency between undamaged and damaged modal vectors. Dataset and ANN training process is carried out by means of Carrera Unified Formulation (CUF), according to which refined finite elements with component-wise capabilities can be implemented in a hierarchical and unified manner. The proposed results demonstrate that CUF-trained ANNs can approximate complete mapping of the damage distribution, even in case of low damage intensities and local defects in localized components (stringers, spar caps, webs, etc.)

Component-wise damage detection by neural networks and refined FEs training / Pagani, A.; Enea, M.; Carrera, E.. - In: JOURNAL OF SOUND AND VIBRATION. - ISSN 0022-460X. - STAMPA. - 509:(2021), p. 116255. [10.1016/j.jsv.2021.116255]

Component-wise damage detection by neural networks and refined FEs training

Pagani, A.;Enea, M.;Carrera, E.
2021

Abstract

Multilayer perceptrons are utilized in this work for vibration-based damage detection of multicomponent aerospace structures. A back-propagation algorithm is utilized along with Monte Carlo simulations and advanced structural theories for training Artificial Neural Networks (ANN’s), which are able to detect and classify local damages in structures given the natural frequencies and the associated vibrations modes. The latter ones are feed into the network in terms of Modal Assurance Criterion (MAC), which is a scalar representing the degree of consistency between undamaged and damaged modal vectors. Dataset and ANN training process is carried out by means of Carrera Unified Formulation (CUF), according to which refined finite elements with component-wise capabilities can be implemented in a hierarchical and unified manner. The proposed results demonstrate that CUF-trained ANNs can approximate complete mapping of the damage distribution, even in case of low damage intensities and local defects in localized components (stringers, spar caps, webs, etc.)
File in questo prodotto:
File Dimensione Formato  
Pagani, Enea, Carrera - Journal of Sound and Vibration - 2021.pdf

accesso aperto

Descrizione: PDF editoriale open access
Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Creative commons
Dimensione 2.42 MB
Formato Adobe PDF
2.42 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

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/11583/2907457