Influence of concrete strength estimation on the structural safety assessment of existing structures

• Monte Carlo simulations are used to evaluate concrete strength estimation methods.
• Estimation performance can be coupled to structural reliability quantification.
• Characteristics of estimation errors are calculated using numerical statistical methods.
• Performance of estimations should be considered when performing safety evaluations.
• The results serve as guidance for choosing estimation methods and number of samples.

In case of existing concrete structures, the estimation of the characteristic strength values from limited data is a difficult, but important task. There are currently different commonly used estimation methods available, among which the classical coverage method, a Bayesian procedure with vague prior distributions (as mentioned in EN 1990) and the method described in EN 13791. There exists however currently no comprehensive framework in order to quantify the influence of these concrete strength estimation methods on the safety assessment of existing structures. In order to analyse this effect, a previously developed semi-probabilistic partial factor approach for the evaluation of existing structures is considered herein, namely the Adjusted Partial Factor Method. The influence of the different concrete strength estimation methods on the safety level of existing concrete columns is investigated, considering both the application of unchanged partial factors compared to new structures and partial factors adjusted according to the Adjusted Partial Factor Method. The performance of the different estimation methods are evaluated and compared using Monte Carlo simulations and FORM analyses. The relative performance of the estimation methods seem to be rather independent of the partial factor approach applied, however the Adjusted Partial Factor Approach allows to achieve a coherent performance with respect to a target reliability index and alternative reference period. The performance of the classical coverage method and a Bayesian method with vague prior information are comparable and yield a higher safety level when more than 5 test samples are considered. In case only very few concrete samples are used for the assessment (i.e. 3–5), the EN 13791 yields a comparable safety level, mainly due to the reduced variability with respect to the estimation error. Finally, the analysis also showed that for the investigated situation, taking more than about 8 test samples into account does not lead to an increase in assessed safety level.