Communication Dans Un Congrès Année : 2019

A bayesian Network for the deduction of corrosion level based on NDT inspection data

Résumé

Reinforced concrete structures deteriorate with time, mainly due to corrosion of the steel reinforcement, which leads to shorten their service life. Hence, there has been an interest in an effective and preventive maintenance strategy for assessing the actual level of deterioration with regards to the steel reinforcement corrosion. As a preventive strategy, the deterioration must be detected before serious damage of the structures appears. That is why the Non Destructive Testing (NDT) has become a relevant tool. Some NDT techniques have been used to detect corroded rebars in on-site structures, such as half-cell potential, resistivity and polarization resistance measurements. Other NDT techniques can be used as indirect indicators to locate a problematic zone in a structure such as the capacity measurement, which is sensitive to the water content in concrete. From these NDT techniques, pertinent measurements can be extracted to assess the corrosion level in different parts of a structure. Some of these measurements, called “Rapid observables”, are easy to make and exploit, and give a general view of the current state of the structure. However, they are not very precise and can have an average to low reliability. Other observables can be obtained, called “Complementary observables”, which are more precise and reliable, but more difficult to exploit and which take longer to make. An efficient and reliable strategy would combine rapid and complementary observables to assess the corrosion level of a reinforced concrete structure. For this study, a multi-criteria inspection methodology to determine the corrosion level of an on-site structure has been proposed and tested for existing concrete structures which have been inspected through rapid and complementary observables. One main pitfall that has to be taken into account is that many points on the concrete facing have been inspected through the rapid observables, while only a few ones have been through the complementary observables. Moreover, the points with the available rapid observables were not the same as the points where the complementary observables were provided. Therefore, a Bayesian Network has been set up, so as to infer the corrosion level at each inspected point with whatever the available observables. Based on the database collected through several inspection campaigns, the most accurate architecture of the network has been investigated. The Bayesian Network was also a relevant tool allowing to take into account both the experimental uncertainty concerning the NDT inspection data acquisition and the supposed reliability of the observables. Those aspects can hence be integrated into the evaluation of the probability of active corrosion at each inspected point.

Domaines

Fichier non déposé

Dates et versions

hal-04971437 , version 1 (28-02-2025)

Identifiants

  • HAL Id : hal-04971437 , version 1

Citer

Sam Ang Keo, Thomas de Larrard, Frederic Duprat, Sandrine Geoffroy. A bayesian Network for the deduction of corrosion level based on NDT inspection data. EUROCORR 2019, Sep 2019, Seville (Spain), Spain. ⟨hal-04971437⟩
73 Consultations
0 Téléchargements

Partager

  • More