Influence of the spatial variability of the corrosion parameters on SLS failure probability of marine reinforced concrete structures

Abstract

Existing reinforced concrete marine structures are subject to the penetration of chemical agents, mainly chlorides, which leads to their degradation through pitting corrosion of rebars. Because of our limited knowledge about both their health states and environmental conditions, the definition of the parameters inducing corrosion presents uncertainties as well as spatial and time variabilities, so that neglecting them may imply an early and unexpected failure. In case of the Serviceability Limit State (SLS) associated to pitting corrosion initiation, numerous researchers showed indeed that considering the spatial variability of input parameters increases the failure probability. Thus, it is now generally recommended to model SLS input parameters as random fields.

Nowadays, a major challenge for assessing structure performance is to maximize the added value of uncertainties reduction through an accurate probabilistic modelling of each parameter. Indeed, spatially variable parameters with small sensitivities may be deterministically defined whereas highly sensitive ones have to be geostatistically modelled. This implies to define their marginal lawsﾒ types and parameters as well as their spatial correlation functions and fluctuation scales. These definitions can either be based on measurements data or on a priori knowledge. The former is better fitted to the structure but may induce important inspection or monitoring costs. The latter only requires predefined geostatistical models but may results in a rise of computational costs for a limited decrease of uncertainties.

This work aims to facilitate the choice of the optimal geostatistical definition method by performing all-at-time sensitivity analyses of the probability of failure to both marginal law hyper-parameters and fluctuation scales of pitting corrosion SLS parameters.

Our study-case is a typical reinforced concrete wharf beam subject to constant end-to-end loading, with real data supplied by the APOS project [1]. Chloride diffusion is modelled by a semi-empirical model with both time-dependency and correlated diffusion parameters. We consider maximum available precision on marginal law hyper-parameters and large uncertainty on both fluctuation scales and correlation of diffusion parameters.

Results show that reducing the uncertainties on the probability of failure requires to primarily assess the correlation among the diffusion parameters, and then hyper-parameters of diffusion parameters. The only fluctuation scale characterization which may then have an effect for reducing the uncertainty on the probability of failure is the diffusivity one.

[1] Othmen I., Bonnet S., Schoefs F., ﾓInvestigation of analysis methods for chloride profiles within a real structure in marine environmentﾔ, Ocean Engineering, 157/1 June 2018, 96-107, doi.org/10.1016/j.oceaneng.2018.03.040 - 2018