Probabilistic and sensitivity analysis of analytical models of corrosion onset for reinforced concrete structures

Chloride ingress and carbonation are the main factors leading to corrosion initiation of reinforced concrete structures. Lifetime assessment and maintenance optimisation of these deteriorating structures has been based for several practical applications on analytical predictive models. The literature review shows that there are numerous analytical models of chloride ingress and carbonation. It was also found that comprehensive lifetime assessment requires that the selected models be able to propagate uncertainties in a correct manner. Thus, this article proposes a methodology for probabilistic sensitivity analysis of analytical models for chloride ingress and carbonation. Given that different parameter notations were found for each model, a homogenisation of notations and generic forms are first proposed. The sensitivity analysis is based on the following indicators: elasticity coefficient, Pearson coefficient, bias of the mean and standard deviation of the response of the model. The results highlighted which parameters are influential for each model by considering three concrete types. Overall results indicated a higher effect of the non-linearity of the models on the sensitivity indicators. It was also found that the level of influence of each parameter depends also on the concrete type and the duration of the exposure to chlorides or carbon dioxide.