Researchers developed an innovative approach to correlating microstructure and mechanical properties at meso-scale and high-throughput data analysis methodologies using machine learning algorithms.
Predictive modeling and simulation have great potential to reduce the duration of the nuclear materials development cycle, which can take 10-20 years, but not enough focused experimental work has been done to develop and validate the models that will make this a reality. Demonstration of this potential requires the deliberate design of closely coupled modeling and experimental campaigns that generate results which can be compared on length scales consistent with model length scales. In this project, researchers developed an innovative approach to establish a correlation between microstructure and mechanical
properties at meso-scale and high-throughput data analysis methodologies using machine learning algorithms.Suitable machine learning algorithms were applied to efficiently model energy dispersive X-ray spectroscopy datasets. Machine learning algorithms were also explored for feature extraction using the “complexity-minimizing" traits of reconstruction and implement high efficiency optimization algorithm to large scale X-ray spectroscopy datasets. This research demonstrated an innovative approach to measure the mechanical properties of materials at the meso-scale, the same with the modeling and simulation. The high-throughput mechanical testing shortens the development cycle of nuclear fuels and materials and thus expedite the discovery and design of advanced nuclear fuels and materials.