The accurate determination of the tensile strength of concrete is a topic of growing interest, especially driven by the new ultra-high tensile strength fiber-reinforced concretes. However, this is a complex task, as traditional methods provide inconsistent results (both splitting tensile test and 3-point bending test). The most robust method is the direct tensile test. However, in concrete, it poses a technical challenge that has not yet been satisfactorily solved. The aim of this study is the design of a test specimen that maximizes the percentage of successful failure and also allows a viable anchorage solution to the load application equipment, especially for high and ultra-high tensile strength concretes. An optimized geometry of a bone-shaped specimen that maximizes the probability of collapse occurring at the central neck is presented. First, a mathematical function of the generatrix of the bone-shape specimen is presented. The behavior of the solution is also analyzed numerically and statistically, comparing it with other commonly used solutions for direct tensile tests. Additionally, the design, fabrication, and calibration of a tailored cardan joint for load centering, thus minimizing its eccentricity and the dispersion of the results. Finally, the results of a static test campaign carried out on six carbon-fiber reinforced high tensile strength concrete specimens. The proposed bone-shaped specimen shows a much higher percentage of successful failures than the other specimen geometries commonly used for tensile testing, which improves the quality of tensile characterization test campaigns for high and ultra-high tensile strength concretes.
