About this Abstract |
Meeting |
MS&T23: Materials Science & Technology
|
Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
Presentation Title |
Microstructure Statistics for Property Prediction in Multifunctional Electrode Composites Using Random Forests |
Author(s) |
William H. Huddleston, Hugh Smith, Yinghui Wu, Alp Sehirlioglu |
On-Site Speaker (Planned) |
William H. Huddleston |
Abstract Scope |
All-solid-state structural lithium-ion batteries are sought for improved safety and systems level weight savings in next generation aerospace concepts. In this work, the influence of processing conditions on microstructural evolution was evaluated for anode composites of strain-free Li4Ti5O12 and metallic nickel current collector. Beyond size distributions, this study explored quantifying microstructural features to describe changes in the spatial distribution and coalescence of nickel particles as a function of sample composition and sintering conditions. Processing-microstructure-property relationships were described by quantifiers including nickel particle count per area, center-to-center distance, edge-to-edge distance, and shape statistics from particle skeletons. Random forests were trained on feature averages and principal component scores from feature distributions. Random forests were analyzed to determine the relative influence of each feature on prediction of electrical conductivity and mechanical strength. This framework and insights can inform microstructural modeling with useful experimental data to improve property prediction and performance assessment. |