Study Regarding the Evaluation of Prediction Models for Determining the Concrete Compressive Strength Using Non-Destructive Testing (NDT) Data: Validation Stage

Abstract

In the evaluation of an existing reinforced concrete structure, a fundamental issue is determining the concrete compressive strength as accurately as possible. This process can be conducted by using destructive and non-destructive methods. The destructive method presumes a limited number of cores extracted from the concrete structure. A higher number of cores would affect the structural safety and is time and resource consuming. Therefore, the conclusions drawn using exclusively this method can also generate errors in correctly estimating the load bearing capacity of a structure, thus leading to the possibility of implementing deficient measures in order to ensure a structural safety. Data obtained via non-destructive methods are more comprehensive. Due to their non-destructive nature, there are no limitations regarding the number of elements investigated and are fast in delivering results. One of the main concerns of researchers in this field is developing a direct relationship between the measured indicators through non-destructive testing (NDT) methods and concrete compressive strength. Over the years different equations with different mathematical structure (linear, polynomial, power, exponential and logarithmic) were developed with the main purpose of delivering fast and accurate results concerning concrete compressive strength by the means of NDT. The aim of this paper is to validate some of the most important prediction models using an original set of data. The database consists in a number of 96 concrete cores that were subjected to Ultrasonic Pulse Velocity (UPV) and Schmidt Rebound Hammer (SRH) testing. The accuracy of the results was determined by using two statistical parameters the mean absolute error (MAE) and mean absolute percentage error (MAPE). The proposed equations have been analyzed in terms of prediction and dispersion of values. It was noticed that some of the formulations predict values that are higher than the ones obtained destructively, others provide a larger dissipation of values, while some equations deliver a compact distribution of results with higher rate in terms of accuracy. This study proposes a data validation of some of the most popular empirical equations, used for the estimation of the concrete compressive strength, elaborated through the years, using a new set of data.