A penalized regression approach to prediction of triaxial performance of recycled aggregate concrete

The quest for cost-effective development of sustainable infrastructure systems has motivated the utilization of construction waste materials as a main concrete ingredient. As the use of recycled concrete aggregates (RCAs) has gained increasing acceptance, a need has emerged to investigate and quantify the mechanical properties of RCA. While major research efforts have been directed at modeling various engineering properties of RCA under different loading conditions, its performance under triaxial loading has not been sufficiently investigated.This chapter presents a study to investigate the factors affecting the triaxial performance of RCA. Using an existing experimental dataset consisting of axisymmetric triaxial test results on 193 cylindrical specimens, and an initial predictor set consisting of specimen diameter, lateral stress, water-to-cement ratio, aggregate-to-cement ratio, percentage of recycled aggregate, and exposure temperature, a novel empirical model for the peak stress has been developed. The least absolute shrinkage and selection operator approach has been used to rank the predictors in terms of their impact on the peak stress, and to construct a linear predictive model using a selected set of predictors.The results show that the triaxial peak stress of RCA is mainly determined by the lateral stress conditions, diameter of the cylindrical samples, and, to a smaller extent, exposure temperature, while being relatively insensitive to the changes in water-to-cement ratio, aggregate-to-cement ratio, and the percentage of recycled aggregate. The approach presented in this chapter can be adopted to investigate other mechanical properties.