PSI - Issue 41

ScienceDirect Structural Integrity Procedia 00 (2022) 000–000 Structural Integrity Procedia 00 (2022) 000–000 Available online at www.sciencedirect.com Available online at www.sciencedirect.com ScienceD rect Available online at www.sciencedirect.com ScienceDirect

www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia

Procedia Structural Integrity 41 (2022) 145–157

© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the MedFract2Guest Editors. Abstract Understanding the mechanical behavior and properties of works of art, such as panel paintings, is helpful for evaluating the failure mechanisms in play, especially during exhibitions in confined spaces. A panel painting is typically formed by a wooden panel and by layers of gesso, followed by paints and varnishes on the top. Induced by environmental changes, the mismatch in the moisture response of a stiff gesso layer and of the underlying wood panel produces risk of fracture in the pictorial layer. This happens because the gesso layer experiences tension which leads to cracking if the mechanical strain exceeds a critical level. The proposed contribution aims to develop a 3D simplified model for paintings, to detect the environmental conditions which may lead to exceeding the critical strain levels. To this aim, a penny-shaped crack has been simulated inside the gesso layer, centered along the symmetrical planes. To establish if the crack is in critical conditions, the failure criterion of the strain energy density (SED) method has been used: when the critical SED value is reached, the crack is assumed to be in unsafe conditions, otherwise it is in safe conditions. Several combinations of the geometric parameters describing the model have been checked, allowing to define whether the different conditions are likely to lead to a further damage development. Finally, the obtained results have been used to train a preliminary extreme gradient boosting machine (XGBoost) that may be able to classify and predict the two possible outcomes: safe and unsafe. This way, by exploiting the capabilities of machine learning, it could be possible to limit the need of numerical simulations and to introduce new rationales in the framework of work of arts conservation. © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the MedFract2Guest Editors. Keywords: panel paintings; cultural heritage; strain-energy density; crack; Machine learning; XGBoost 2nd Mediterranean Conference on Fracture and Structural Integrity Predicting damage evolution in panel paintings with machine learning America Califano a,b , Pietro Foti b *, Filippo Berto b , Marco Baiesi a , Chiara Bertolin b a University of Padua, Department of Physics and Astronomy “Galileo Galilei”, Via F. Marzolo 8, Padua, 35121 (Italy) b Norwegian University of Science and Technology, Department of Industrial and Mechanical Engineering, Richard Birkelands vei 2B, Trondheim, 749 ( Norway) Abstract Understanding the mechanical behavior and properties of works of art, such as panel paintings, is helpful for evaluating the failure mechanisms in play, specially during exhibiti ns in confined spaces. A panel inting is typically formed by a wooden panel and by layers of gesso, followed b paints and varnishes on the top. Induced by nvironmental changes, the mism tch in the moisture response f a stiff gesso layer and of the underlying wood panel pro uces risk of fracture i the pictorial layer. T is happens because the ge so layer experiences tension which leads to cracking if the mechanical strain exceeds a ritical level. e pro osed contribution aims to develop a 3D simplified model for paintings, to detect the environmental conditions which may lead to exceeding the critical strain levels. To this aim, a penny-shaped crack has be n simulated i side the gesso layer, centered along the symmetrical planes. To establish if the crack is in critical conditions, the failure criterion of the strain nergy density (SED) method has been used: when the critical SED v lue is reached, the crack is assumed to be i unsafe co ditions, otherwise it is in safe conditions. Several combinations of the geometric parameters describing the model have been checked, allowing to d fine whether the ifferent conditions are likely to lead t a further damage velopment. Finally, the obtained results have been used to train a pr liminary extreme gradi nt boosting machine (XGBoost) that may be able to classify and predict the two possible outcomes: safe and unsafe. This way, by expl iting the capabilities f machine learning, it could be possible to limit the ne d of numerical simulations and to introduc new rationales in the framework of work of arts conservation. © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review u der re ponsibility of MedFract2Guest Editors. Keywords: panel paintings; cultural heritage; strain-energy density; crack; Machine learning; XGBoost 2nd Mediterranean Conference on Fracture and Structural Integrity Predicting damage evolution in panel paintings with machine learning America Califano a,b , Pietro Foti b *, Filippo Berto b , Marco Baiesi a , Chiara Bertolin b a University of Padua, Department of Physics and Astronomy “Galileo Galilei”, Via F. Marzolo 8, Padua, 35121 (Italy) b Norwegian University of Science and Technology, Department of Industrial and Mechanical Engineering, Richard Birkelands vei 2B, Trondheim, 749 ( Norway)

* Corresponding author. E-mail address: pietro.foti@ntnu.no * Corresponding author. E-mail address: pietro.foti@ntnu.no

2452-3216 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license ( https://creativecommons.org/licenses/by-nc-nd/4.0 ) Peer-review under responsibility of the MedFract2Guest Editors. 2452-3216 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license ( https://creativecommons.org/licenses/by-nc-nd/4.0 ) Peer-review under responsibility of the MedFract2Guest Editors.

2452-3216 © 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the MedFract2Guest Editors. 10.1016/j.prostr.2022.05.017