Date of Graduation

5-2024

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level

Graduate

Department

Civil Engineering

Advisor/Mentor

Prinz, Gary S.

Committee Member

Murray, Cameron D.

Second Committee Member

Hale, W. Micah

Third Committee Member

Zhang, Shengfan

Keywords

Cement paste; EDX; Elastic Modulus; Micropillars; Monte Carlo simulations; Ultra-high performance concrete

Abstract

Current design code equations for estimating the elastic modulus of specialty concretes like ultra-high performance concrete are not all suitable. Developing empirical curves from exhaustive experimental testing is challenging for concretes that may have high levels of cement replacement with mineral admixtures and fibers at varying volume contents. As such, an alternative approach is taken that seeks to understand the constituent microscale stiffness mechanisms and relate the results to the macroscale. Micropillars are fabricated on several cement paste phases using a focused ion beam. A modified pico-indenter is used to determine the linear and non-linear behavior under micro-compression loading. Micropillar phases are determined by their chemical compositions, identified with an energy-dispersive X-ray. The compressive strength and stiffness of C-S-H, C-A-S-H, SF, CH, AFt/AFm, and a CH/C-S-H are provided in this study. With the stiffness of individual phases calculated and statistical curves developed, a novel homogenization scheme is proposed to combine cement paste phases. Considering phase stiffnesses as springs (in amounts consistent with phase volume fractions determined by X-ray powder diffraction) allows for them to be combined in series and parallel. Monte Carlo arrangement simulations are run in MATLAB. The same approach is used to homogenize UHPC (fibers, fine aggregate, and cement paste), and the code output is verified through initial experimental testing of cylinders. Cement paste homogenization through spring arrangement simulations proves to be an adequate approach. The code output falls within a 30% error when compared to two different hardened cement paste mixes, potentially within a 10% error if the upper portion of the bimodal distribution is considered outliers. However, when the UHPC homogenization is evaluated, the code fails to provide a reasonable estimation of the results.

Download

Share

COinS