Ultra high performance concrete (UHPC) is an advanced cementitious material made with low water to binder ratio and high fineness admixtures, and possesses a unique combination of superior strength, durability, corrosion resistance, and impact resistance. However, increased strength of UHPC results in a brittle behavior. To overcome this brittle behavior of UHPC and improve post cracking response of UHPC, steel fibers are often added to UHPC and this concrete type is designated as Ultra High Performance Fiber Reinforced Concrete (UHPFRC). Being a relatively new construction material, there are limited guidelines and specifications in standards and codes for the design of structural members fabricated using UHPFRC. To develop a deeper understanding on the behavior of UHPFRC flexural members, seven beams made of UHPFRC are tested under different loading conditions. The test variables include level of longitudinal reinforcement, type of loading (shear and flexure), and presence of shear reinforcement. Further, a finite element based numerical model for tracing structural behavior of UHPFRC beams is developed in ABAQUS. The developed model can account for the nonlinear material response of UHPFRC and steel reinforcement in both tension and compression, as well as bond between concrete and reinforcing steel, and can trace the detailed response of the beams in the entire range of loading. This model is validated by comparing predicted response parameters including load-deflection, load-strain, and crack propagation against experimental data obtained from tests on UHPFRC beams with different material characteristics and under different loading configurations. The validated model is applied to conduct a set of parametric studies to quantify the effect of different parameters on structural response of UHPFRC beams, including the contribution of stirrups and concrete to shear capacity of beams, to explore feasibility of removing the need for shear reinforcement in UHPFRC beams. Results from experiments and numerical model reveal that UHPFRC beams exhibit distinct cracking pattern characterized by the propagation of multiple micro cracks followed by widening of a single crack leading to failure. Also, UHPFRC beams exhibit high flexural and shear capacity, as well as ductility due to high compressive and tensile strength of UHPFRC and fiber bridging developing at the crack surfaces that leads to strain hardening in UHPFRC after cracking. Thus, absence of shear reinforcement in UHPFRC beams does not result in brittle failure, even under dominant shear loading. Data from the conducted experiments as well as those reported in literature is utilized to develop a machine learning (ML) framework for predicting structural response of UHPFRC beams. On this basis, a comprehensive database on reported tests on UHPFRC beams with different geometric, fiber properties, loading and material characteristics is collected. This database is then analyzed utilizing different ML algorithms, including support vector machine, artificial neural networks, k-nearest neighbor, support vector machine regression, and genetic programing, to develop a data-driven computational framework for predicting failure mode and flexural and shear capacity of UHPFRC beams. Predictions obtained from the proposed framework are compared against the values obtained from design equations in codes, and also results from full-scale tests to demonstrate the reliability of the proposed approach. The results clearly indicate that the proposed ML framework can effectively predict failure mode and flexural and shear capacity of UHPFRC beams with varying reinforcement detailing and configurations. The research presented in this dissertation contributes to the development of preliminary guidance on evaluating capacity of UHPFRC beams under different configurations.
Read
