Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : September 05, 2021 - September 08, 2021
In this paper, it is reviewed that combinatorial technologies for the fabrication and characterization of a large number of samples at once, the use of MEMS and other microfabrication technologies for the characterization of the samples, and our efforts to use machine learning to analyze and improve the search efficiency of the large number of sample data obtained by the combinatorial technologies. We introduce combinatorial arc plasma deposition, which enables the combinatorial deposition of amorphous alloy materials. The composition-graded films fabricated by this method are separated and labeled into thin-film libraries by using microfabrication methods, and their properties are evaluated using MEMS structures. As an example of material search with the aid of machine learning, we describe the identification of physical properties with high contribution to the current density of electrocatalysts by random forest analysis, the examination of search termination conditions by Bayesian optimization, and the estimation of crystal grain size of magnetic materials from Barkhausen noise by machine learning.
