Imprecise abstract argumentation as a support for forensic engineering - Archive ouverte HAL Access content directly
Journal Articles Engineering, Construction and Architectural Management Year : 2021

Imprecise abstract argumentation as a support for forensic engineering

, (1) , , ,
1
Franck Taillandier
  • Function : Author
Claudio Carvajal
  • Function : Author
Benjamin Delhomme
  • Function : Author
Bruno Beullac
  • Function : Author

Abstract

Purpose Civil engineering structures are regularly confronted with failures that can lead to catastrophic consequences. It is important, after a failure, to be able to identify the origin and the sequence of factors that led to it. This failure analysis by experts, called forensic engineering investigation, generally leads to the drafting of an expert report. These reports do not inform on the processes that guided the experts to a conclusion and the uncertainties involved. This paper aims to propose a new methodological approach to formalize the opinions of experts in forensic engineering. Design/methodology/approach The research consists in combining abstract argumentation with the theory of imprecise probabilities to take into account epistemic and stochastic uncertainties to support forensic engineering investigation. Findings A model and a tool to support forensic analysis are presented. An application on the collapse of the Brumadinho dam highlights the interest of the chosen approach. Originality/value This work is the first use of the abstract argument framework in civil engineering, and so in forensic engineering. Furthermore, it provides an innovative model based on imprecise probability for AAF.
Not file

Dates and versions

hal-03738920 , version 1 (26-07-2022)

Identifiers

Cite

Franck Taillandier, Cédric Baudrit, Claudio Carvajal, Benjamin Delhomme, Bruno Beullac. Imprecise abstract argumentation as a support for forensic engineering. Engineering, Construction and Architectural Management, 2021, 29 (1), pp.523-550. ⟨10.1108/ECAM-09-2020-0714⟩. ⟨hal-03738920⟩
5 View
0 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More