Modeling Fracture in the High and Very High-Cycle Fatigue Range
It is well known in ultrasonic fatigue testing that non-metallic inclusions in aluminium and steels may cause fracture under cyclic loading, inducing a temperature increase due to crack initiation and propagation in the bulk of specimens. To investigate fracture in the high and very high-cycle fatigue range the surface temperature distribution of the samples is captured during experiments. Observation of hot spots in the surface temperature indicates damage. Since thermography data are collected over time for the full surface, temperature profiles can be associated with the position and point of crack initiation. For calibration, it can be instructive to explore different thermal boundary conditions and the heat source term in the framework of a fully-coupled linear thermoelasticity model, based on experimental information collected in ultrasonic fatigue testing. In the context of the thermography data generated it is not immediately obvious which types of conditions are appropriate to simulate material behavior using finite elements and how the associated quantities in the boundary expres-sions can be obtained. To investigate which types of boundary conditions may be appropriate from a physical and a computational perspective, different kinds of conditions are considered, making extensive use of experimental data.
To mimic heating in the bulk of the specimen a phenomenological approach is employed, introducing an appropriate heat source function. In the numerical computation the temperaturedistribution evolution in the sample is the result of a combination of initial conditions, boundary conditions and the heat source contribution. The heat source function’s geometry and intensity is deduced from the difference of temperature profiles obtained from reference computations, based on experimental initial and boundary condition data, and purely experimental distributions, retrieved at the discrete instances captured in the experiment, using a thermal camera. Cooling of the specimen does not involve any excitation of the sample and the heat source term is set to zero. Thus, temperature loss is modeled using thermal boundary conditions, employing real-world data.