Automated Data Extraction Techniques for SMAs

The emergence of Machine Learning (ML) as a technique for predicting new alloys has been an important step forward in SMA development. ML models are trained with large databases containing wide ranges of data to model the relationships between features and responses. These databases require large numbers of experiments to be ran, which then need to be analyzed to characterize their behavior. This step represents a major bottleneck in the application of ML to SMAs. By taking steps to automate the data extraction from raw experimental data for a variety of experiments we can significantly speed up the rate at which we can populate search spaces, improving model accuracy and material discovery. In this presentation, we will present work that develops a generalized algorithmic approach to automated data extraction for a range of experiment types used in SMA characterization.