Development of Prediction Models for Mechanical Properties and Durability of Concrete Using Combined Nondestructive Tests

Abstract

For several decades, researchers have attempted to develop statistical models through individual and combined use of ultrasonic pulse velocity (UPV) and rebound hammer data to enhance the prediction of concrete compressive strength and durability. This study proposes statistical univariate and multivariable regression models to predict compressive strength, abrasion, and salt scaling of concrete using UPV and rebound hammer measurements. Stepwise regression analysis was undertaken to develop the proposed models that were then validated using independent data. A scaling quality classification table using rebound hammer, and based on a k-means clustering algorithm, is also proposed. The measurements support the combined use of UPV and rebound hammer to predict compressive strength. On the other hand, rebound hammer values are the only statistically significant variables to predict abrasion and salt-scaling resistance of concrete. Concrete properties had a significant impact on the mean and dispersion values of UPV and rebound number (RN). The procedures used in this paper for model development can serve as a general guideline for developing statistically valid univariate and multivariable regression models for other applications when predicting concrete properties.