||Materials processing (one of the three pillars in the materials discipline) has been dominated by experiment-intensive, empirical, and repetitive studies to achieve desirable microstructures and phases. An epitome of this trial-and-error practice is high temperature material processing. The transient nature and evolution complexity from starting materials (liquid or solid, bulk or powder) to the drastically different compositions and microstructures of final states (monolithic or porous solids) have challenged the field for decades in correlating processing conditions with the final attributes. Machine learning uses statistical and probabilistic models trained on historical data to make predictions about new observations. When properly introduced, machine learning can fit the available data collected over decades in the literature to adaptive models. The training process is technically sophisticated yet operationally simple, especially considering the parallel experimental study of high temperature processing from different precursors/starting materials and under different processing conditions. As long as with enough data, ML can bypass the analytical equations required to describe the one-to-one relations and is more adeptly suited to tackle the drastic and multi-dimensional changes during material processing, especially when the changes are drastic, complex, and intractable. Data-driven ML is a golden opportunity in our field to extend the predictability at the atomic and molecular levels to microscopic and macroscopic levels to fully explore the immense space of composition-processing parameters. The exciting aspects for ML are not just about finding hidden correlations. It can explore new material processing space that cannot be unearthed by empirical experiments and even traditional computer simulation. More excitingly, ML can reveal unknown material design and processing space.
This symposium will cover machine learning topics related to fundamental and applied sciences in high temperature materials processing from experimental, computation, and large database approaches. It will consider all aspects of high temperature material processing and related studies.
Material processing studies using regression
Machine learning for small datasets
Machine learning in materials processing using large databases
Machine learning based on computer modeling
Machine learning in additive manufacturing
Machine learning in sintering
Machine learning in polymer derived ceramics