Bond strength prediction of UHPC-NSC interface

Abstract

The application of ultra-high-performance concrete (UHPC) on top of normal-strength concrete (NSC) is a practical rehabilitation approach to maintaining degraded and damaged concrete members. However, a successful repair operation and consequent adequate performance are very much dependent on the ability of the interface between UHPC and NSC to present a superior performance of bonding under various surface conditions. Consequently, predicting the strength of the bond at the interface joining the existing NSC and the newly placed overlaying UHPC - with sufficient certainty - has become a vital and required step in assessing and maintaining UHPC rehabilitated NSC structural elements. In this work, Artificial Neural Network (ANN as well as Gene Expression Programming (GEP methods are utilized to predict the bond strength between the overlaying UHPC and the substrate NSC using a comprehensive database set consisting of 264 experimental data points gathered from the literature. A parametric ANN analysis is performed to examine and assess the effect of each parameter on the interfacial bond strength. The following five factors are identified as key parameters through the GEP and ANN analyses: curing method, age of UHPC, the compressive strength of NSC, interfacial surface treatment, and moisture conditions. The developed ANN and GEP models have good accuracy and closer predictions of the bond strength of the slant shear test and the splitting tensile strength with root mean square error (RMSE) values of 5.0, 4.3, and coefficient of variation (COV) values of 37%, 24%, respectively.