Ultra-high performance concrete (UHPC) is characterized by its superior strength, ductility, durability, and particularly its unique post-cracking performance in tension. Dog-bone–shaped specimens are widely used for determination of the tensile behavior of UHPC, but there is no standard test method or specimen design for the characterization of tensile behavior. In this study, an evolving strategy on designing a direct tension test (DTT) specimen is first conducted using numerical finite element analysis. Seven series of DTT specimens made of UHPC and with well-designed dimensions to avoid local stress concentration are then tested experimentally. Results indicate that the post-cracking localization within the gauge measurement region is guaranteed, and the DTT specimen is capable of fully capturing tensile stress–strain responses of UHPC. An idealized constitutive model with three linear phases is proposed to fit the experimental data and thus characterize the linear elastic, strain-hardening, and strain-softening behavior of UHPC in tension. Four tensile material parameters extracted from the experimental stress–strain curves are implemented in the idealized constitutive model from which the multi-phase responses of UHPC in tension are reconstructed. It is found that most tensile material parameters extracted from experimental stress–strain curves, including tensile strength, modulus of elasticity, and dissipated energy, increase with the increased volume fraction of steel fibers, curing age, and displacement loading rate, while the strain capacity at the first cracking remains nearly constant. The DTT specimen developed can be used effectively to characterize the tensile behavior of ductile fiber-reinforced cementitious materials.

A computational homogenization model using microstructures obtained from X-ray micro-CT is developed to estimate the porosity-based elastic properties of ultra-high performance concrete under freeze–thaw action. The model is transformed directly from micro-CT which is capable of reflecting realistic distribution of porosity and heterogeneities inside the ultra-high performance concrete. Factors are taken into consideration, including the determination of representative volume element, the position and numbers of representative volume element cubes, fiber orientation, image resolution, applied filter, and pore distribution. The relationship between the material internal structure and freeze–thaw resistance is studied at micro-scale. The volume-averaged homogenization approach is applied to calculate the effective properties of the ultra-high performance concrete which are compared with experimental data. It is demonstrated that the proposed model provides an effective tool to evaluate the elastic properties of the ultra-high performance concrete based on microstructural characterization data.