Experiment-based Simulations for Heterogeneous Hierarchical Materials: Application to Metal Additive Manufacturing

Kafka, Orion Landauer

doi:https://doi.org/10.21985/n2-f7r9-rs43

Work

Experiment-based Simulations for Heterogeneous Hierarchical Materials: Application to Metal Additive Manufacturing

Public

Download PDF

Download All Files (.zip)

Modeling the mechanical performance of metal produced with additive manufacturing (AM) has proven to be a challenging task. In the as-built state, these materials have been shown to exhibit strong heterogeneity and anisotropy. Even after post-processing, such as heat treatment or hot isostatic pressing and depending on the alloy, some of these characteristics remain. Contemporary testing and modeling approaches are not particularly adept at identifying, capturing, or predicting these relatively complicated behaviors. This shortcoming in the currently available methods is particularly the case for fatigue and failure. This dissertation begins by presenting an experimental exploration of AM materials, looking predominately at spatial heterogeneity and how that might relate to porosity generated during directed energy deposition (i.e. Laser Engineered Net Shaping) processing. The results from these, and related, experiments indicated that in order to explain and predict performance in fatigue and fracture, the two most important factors would likely be porosity (typically between 2 microns and several hundred microns equivalent diameter) and grains (typically columnar and varying from micron-size to millimeter-size). In order to capture these effects well, a crystal plasticity finite element model was developed to explore the fatigue performance of Inconel 718/718+. This model provided some insight into the relative importance of grain orientations and potential impacts of unusually large grains. However, the substantial computational expense involved made it difficult to explore effects of heterogeneity at a realistic scale (e.g. variations occurring over more than 10 times the average grain size). Further, although this type of model is useful for exploring materials, it is also too expensive to practically model part-scale systems of interest to practical users of AM. Thus, the final two chapters will introduce and demonstrate two key developments. First, the combination of reduced order modeling methods with material models suited to the study of crystalline metals (the crystal plasticity law mentioned above) and its integration within a process-structure-properties-performance prediction framework for AM. Second, a one-way or two-way coupled concurrent multiscale model (depending on the requirements of the mechanics simulated) capable of capturing varying microstructures throughout a part-scale. In the case of AM, this variability is linked both to randomness and more importantly to processing conditions, and a change in processing conditions will change the distribution of, in this case, porosity represented by a database of experimental images at the microscale. This model is shown in both fatigue and fracture initiation prediction settings. No other system can currently produce comparable results, nor can any other approach with equivalent capabilities compete in terms of speed. As a design and optimization tool, this allows for consideration of not only minimum and/or mean properties, but also estimation of the spread of properties that can be expected and more importantly the local properties to any region or feature of interest. In terms of AM, this will allow for more generalized design, for example consideration of functional grading, realistic reinforcement to counter possible defects, and toolpaths to account for the specific impact such choices have on both local and overall performance.

Creator