A Stochastic and Computational Approach for Fracture Modeling of Quasi-Brittle Materials

University Of Tennessee Space Institute, Tullahoma TN

Investigators

Abstract

There is an increasing need to predict materials response and failure behavior at macroscopic scale from its microstructural composition. In brittle and quasi-brittle materials, such as glass, concrete, rocks, and ceramics, failure is particularly sensitive to the microstructure leading to a large scatter in failure loads. Most existing fracture models fail to reliably predict this scatter. This award supports fundamental research in developing theoretical and computational tools for fracture of brittle and quasi-brittle materials that directly link their microstructure to failure loads and the scatter observed. Brittle and quasi-brittle fracture mechanics finds applications in a variety of material and structural designs and plays a central role in many other fields. For example, ceramics are used with metals to develop high-strength and light-weight materials for armor designs and aerospace industry. Rock fracture, whether occurring naturally as in earthquake or manmade for enhanced oil recovery and CO2 sequestration is another example. Finally, obtaining more accurate probabilities of fracture reduces uncertainties in current design practices and can aid in the assessment of the structural integrity of existing infrastructure systems. Educational goals focus on development of short course toolkits on random models and computational tools to attract high school students to STEM fields, and software modules that will be shared with scientific community. The field of stochastic partial differential equations provides systematic approaches for the propagation of randomness in an analysis in general. However, there is currently no means to relate material microstructures to initial random field description needed for these stochastic models. This research fills the knowledge gap by deriving continuum models that directly translate microstructure distribution to the initial material field description. Unlike common homogenization schemes, stochastic representative volume elements still preserve the spatial variability and randomness of material. This enables realistic modeling of brittle and quasi-brittle fracture. To ensure accurate rendering of this theoretical model an advanced finite element model is formulated that can efficiently capture complicated fracture patterns by incorporating both bulk and interfacial failure mechanisms. Moreover, a novel adaptive computational scheme eliminates the sensitivity of the failure load on initial mesh discretization and guarantees the estimation of probability of failure within the user-specified error bounds. The microstructure-based probabilistic fracture model approach aims to explain a variety of phenomena that are not well captured with commonly used deterministic models. Some examples are size effect in brittle and quasi-brittle materials, scatter in failure load, and formation of complex fracture patterns even under uniform loads.