PSI - Issue 44

Dario De Domenico et al. / Procedia Structural Integrity 44 (2023) 633–640 Dario De Domenico et al. / Structural Integrity Procedia 00 (2022) 000–000

636

4

21 steel bars tests, i.e., extraction of portions of bars and tests in laboratory for tensile strength determination of steel. The experimental outcomes were critically interpreted, in an interconnected manner, to obtain reliable information on the state of corrosion of the bridge piers. In particular, the core strength, ranging from 22.0 MPa to 52.7 MPa, was corrected to account for the disturbance induced during the extraction process (Dewar et al. 1987). The resulting concrete cubic strength ranged from 25.2 MPa to 55.3 MPa, with a mean value of 37.2 MPa, a mode equal to 29.1 MPa and a coefficient of variation (CoV) of around 20%. The carbonation depth was of 40-50 mm for the majority of the extracted cores (see again Fig. 3).

core test

pachometer test for steel bar detection

ultrasonic pulse velocity semi- direct measurement

sclerometric test with Schmidt rebound hammer

carbonation depth

corrosion potential mapping wrt saturated Cu/CuSO4 reference electrode

corrosion potential map

mean -464.6 mV; std 32.5 mV high probability of corrosion

Fig. 3. Experimental tests performed on the corroded piers of the Zappulla viaduct.

In addition to core tests, SonReb tests were executed by measuring the rebound index I r (ranging from 30 to 38) and the ultrasonic pulse velocity V us (ranging from 3850 mm/s to 4580 mm/s), which are then used to estimate the concrete compressive strength through literature calibration formulae (Cristofaro et al. 2020). In particular, the average SonReb (SR) strength calculated from three popular power-law expressions (Rilem NTD4 1993; Gašparik 1992; Di Leo & Pascale 1994) was first considered, based on the following expression:

b c c s r R V I a = ⋅ ⋅ u

(1)

where a , b , c are three regression coefficients. Besides this average SR strength, an ad-hoc calibrated predicting formula was developed to minimize the uncertainties inherent to available literature formulae in predicting strength data without a proper calibration of their empirical coefficients. This formula was based on the results of 21 out of the 59 SonReb tests that were purposely performed close to core tests for calibration purposes. For such 21 tests, a nonlinear least-square fitting procedure was used to determine the most appropriate regression coefficients a , b , c to be used in Eq. (1) in order that the SonReb strength approaches the core strength in the best possible manner, that is:

2

  

   

, c i R R a b c R − , c i ( , , ) exp est exp

i =  ∑

2

, , a b c min

(2)

ε

, c i

where c i R is the corresponding SonReb strength obtained by Eq. (1), depending on the three regression coefficients. The resulting squared error resulting from Eq. (2) was equal to 0.9322 for the average SR strength and decreased to 0.6181 for the proposed calibration formula, thus confirming a significantly higher accuracy; the resulting regression coefficients of the proposed formula are: a = 2.716 10 -3 , b = 0.81168, c = 0.76265. Once the regression coefficients a , b , c are identified, a lognormal distribution is fitted to the entire set of 59 datapoints (21 core strengths plus 38 SonReb tests) as shown in Fig. 4, by , exp c i R is the i -th core strength identified from the experiments, and , est