Interfacial bond behavior between ultra high performance concrete and normal concrete substrates

Ultra-high performance concrete (UHPC), possessing excellent mechanical properties, can be used as an excellent material to repair or strengthen the reinforced concrete structural elements. However, the effectiveness of repairing and strengthening by applying the layers of UHPC on the surface of substrate concrete depends on whether the interface between overlay and substrate ensures adequate bonding performance over the entire life span under different loading and exposure conditions. This paper presents an experimental investigation on the effects of key factors on the interfacial bond strength between UHPC overlay and normal concrete substrate (NC). The different surface roughness of the NC substrate, curing conditions, exposure conditions, and test methods (slant shear, bi-surface shear, splitting tension, and third-point flexure), were considered as variable parameters affecting the bond behavior. The results showed that while the curing conditions have an insignificant effect on the bond behavior, the surface roughness of the NC substrate affected the bond strength significantly. The preparation of the substrate's surface using the sandblasting technique achieved the highest bond strength determined using different bond tests, except the bi-surface shear test, which gave the highest bond strength for the surface prepared using drill holes, owing to the interlocking action of the UHPC that filled the holes drilled in the NC substrate. Furthermore, it was found that the highest reduction of the bond strength under the cyclic exposure conditions with the drill holes' specimens. Even though the bond strength of the UHPC-NC composites subjected to the cyclic exposure conditions got significantly deteriorated, all bond strength results met the acceptable limits for applications as a repairing/strengthening, as specified by different design codes. Finally, the comparison between experimental results and predictions using different existing code models (i.e., CEB-FIP 1990, Eurocode 2 2004, CAN CSA A23.3 2014, AASHTO LRFD 2014 and AFGC 2013) revealed that the AASHTO LRFD and AFGC predictions were substantially conservative.