Assessment of concrete strength combining direct and NDT measures via Bayesian inference

摘要

Assessment of existing reinforced concrete structure entails a series of steps, among which the evaluation of the mechanical properties of concrete can be considered a corner-stone. To this end, direct compression tests on cores extracted directly from a structure provide the most reliable estimation of the strength. Unfortunately, the number of cores usually accepted is often limited because the method is expensive and invasive. For this reason, non-destructive (ND) methods are mostly used, whose results are usually calibrated using a limited number of destructive tests, to provide some preliminary information about the homogeneity of the investigated concrete and possibly to suggest zones where to extract other cores. In addition, non-destructive tests may be used to enlarge the database for the estimation of concrete strength. The main drawback in using this approach is that a correlation formula between the in situ measures and concrete strength is required. In many cases, such formula cannot be easily generalized and must be restricted every time based on pairs of indirect and direct measures of strength. Moreover, the use of different experimental techniques (destructive and non-destructive) provides information with different reliability, and the results are thus difficult to combine. For these reasons, in this paper, a technique based on Bayesian inference is proposed to combine in a rational manner the results of direct and indirect measures, providing the probabilistic distribution of the concrete strength and some significant properties such as the median and characteristic value. In this paper, the proposed methodology is developed by considering the ultrasonic pulse velocity (UPV) technique, one of the most popular non-destructive methods employed for the evaluation of concrete strength, although it may be easily extended to other types of in situ measures. To validate the proposed method, two real cases, for which experimental data are available, are analyzed and discussed.