Chloride-induced corrosion in reinforced structures is complex, involving simultaneous material aging and diverse uncertainties. To computationally interpret such a process, the time-variant material aging was often ignored to avoid numerical difficulty and the arbitrary chloride threshold was invoked to determine corrosion initiation, which, however, may inevitably lead to false assessments. In this paper, a novel computational architecture integrating a physics-based aging corrosion method and the recently developed extended support vector regression algorithm is proposed. In specific, the physics-based method is featured of a chemo-physical-mechanical model coupling with an electrochemical model, where realistic aging corrosion mechanism can be simulated with considering the associated uncertainty. In addition, the machine learning algorithm is adopted to greatly enhance the computational efficiency in uncertainty quantification. The developed approach is applied to model the reported experiments on both microcell and non-uniform macrocell corrosion under various exposure conditions. It is shown that the proposed method is able to precisely predict the initiation and the onset of steady-state corrosion, while efficiently handling the designed randomness in model, material, and exposure condition. Furthermore, through comparative studies, the significance of adopting physics-based approach for achieving robust stochastic aging corrosion analysis and reliability assessment is discussed.

Physics-based stochastic aging corrosion analysis assisted by machine learning / Yu, Y.; Dong, B.; Gao, W.; Sofi, A.. - In: PROBABILISTIC ENGINEERING MECHANICS. - ISSN 0266-8920. - 69:(2022), p. 103270. [10.1016/j.probengmech.2022.103270]

Physics-based stochastic aging corrosion analysis assisted by machine learning

Sofi A.
2022-01-01

Abstract

Chloride-induced corrosion in reinforced structures is complex, involving simultaneous material aging and diverse uncertainties. To computationally interpret such a process, the time-variant material aging was often ignored to avoid numerical difficulty and the arbitrary chloride threshold was invoked to determine corrosion initiation, which, however, may inevitably lead to false assessments. In this paper, a novel computational architecture integrating a physics-based aging corrosion method and the recently developed extended support vector regression algorithm is proposed. In specific, the physics-based method is featured of a chemo-physical-mechanical model coupling with an electrochemical model, where realistic aging corrosion mechanism can be simulated with considering the associated uncertainty. In addition, the machine learning algorithm is adopted to greatly enhance the computational efficiency in uncertainty quantification. The developed approach is applied to model the reported experiments on both microcell and non-uniform macrocell corrosion under various exposure conditions. It is shown that the proposed method is able to precisely predict the initiation and the onset of steady-state corrosion, while efficiently handling the designed randomness in model, material, and exposure condition. Furthermore, through comparative studies, the significance of adopting physics-based approach for achieving robust stochastic aging corrosion analysis and reliability assessment is discussed.
2022
Chloride attack
Corrosion analysis
Machine learning
Physics-based modelling
Uncertainty quantification
File in questo prodotto:
File Dimensione Formato  
Yu_2022_Probabilistic_Physics-based_editor.pdf

non disponibili

Tipologia: Versione Editoriale (PDF)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 5.72 MB
Formato Adobe PDF
5.72 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

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/20.500.12318/135499
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 13
  • ???jsp.display-item.citation.isi??? 12
social impact