While the advance of experimental and computer modeling techniques has continued to push mechanistic understanding and predictive modeling capabilities forward, the capability to generate fatigue data has been almost stagnant. Fatigue engineering and research efforts often operate in a data starved modality (considering the highly stochastic nature of fatigue failures). This impedes attempts to effectively use modern machine and statistical learning tools for fatigue performance prediction, both within standard prognosis frameworks, and integrated computational materials engineering (ICME) frameworks.
This presentation will report on our exploration for opportunities to improve the throughput of fatigue testing machines utilizing the expanded design space offered by technological advancement, e.g., computer aided drawing and manufacturing, data acquisition and computer modeling, and robotic automation. Following our review, we will present two concepts for uniaxial high throughput fatigue testing, with the goal of improving fatigue throughput by ~100x while conforming to popular test standards. Our progress towards this goal, and ultimately the prospects for achieving it, will be presented by sharing the results of multiple design-build-test iterations.
EXTENDED ABSTRACT

Rather than energy release rate, the proposed framework starts from the energy function itself. Instead of strain energy density, it considers the change in volume-specific free energy density from mechanical deformation. The free energy function must capture strain induced orthotropy, known to be critical for polymers but also important for metals plasticity. To capture strain induced orthotropy, free energy is defined in terms of principal strains and by separating deformation into dilatational and distortional contributions. The separation does not utilize deviatoric strain. Rather, it leverages a new distortional strain definition and the new concept of orthotropic dilataion, enabling clean separation to large strain.

The proposed framework clarifies how a generalized Maxwell model spring-dashpot mechanical analog cleanly interperets the First and Second Laws of Thermodynamics. A transition state theory based nonlinear viscoelastic (NLVE) model is mated to the Maxwell model. Nonlinear Maxwell springs feature an instability in their constitutive law, providing a viscoelastic failure criterion. Embedding the instability into the springs in an NLVE model provides a failure criterion that accommodates complex temperature histories, rate dependence, and self generated heat from cyclic loading.
EXTENDED ABSTRACT

To analyze the integrity of piping components in nuclear power plants (NPPs), the U.S. Nuclear Regulatory Commission (NRC) Office of Nuclear Regulatory Research and the Electric Power Research Institute jointly developed a probabilistic fracture mechanics computer code. The Extremely Low Probability of Rupture (xLPR) code simulates crack initiation and growth from fatigue and stress corrosion cracking (SCC) degradation mechanisms and other aspects of piping component structural integrity. This presentation provides an overview of the NRC staff’s applications of the xLPR code since its public release in 2020 to assist in risk-informed regulatory evaluations of leak-before-break (LBB) analyses for pressurized water reactor piping systems with dissimilar metal welds susceptible to SCC. Potential use of the xLPR code to estimate loss of coolant accident (LOCA) frequencies and to interface with artificial intelligence machine learning (AI/ML) models are also discussed.
EXTENDED ABSTRACT