Data-driven Shape Memory Alloy Discovery using Artificial Intelligence Materials Selection (AIMS) Framework

Previous studies have focused on minimizing hysteresis under no stress, but not under applied stress.

W. Trehern, R. Ortiz-Ayala, K. C. Atli, R. Arroyave, I. Karaman

Problems Associated with Shape Memory Alloys (SMA)
• Thermal hysteresis – the difference in temperature for the forward versus reverse phase transformation – limits application.
• Consequences of thermal hysteresis are a larger required temperature excursion for application and reduced efficiency.
• Previous studies have focused on minimizing hysteresis under no stress, but not under applied stress.

New SMA Discovery via Machine Learning
• Materials informatics framework (A) was developed and used to analyze a SMA dataset of ~6,000 experimental data entries, constructed from raw data available.
• Ni32Ti47Cu21 (B, see red dot)was predicted by deep neural network regressors to have a small thermal hysteresis under stress.
• Ni32Ti47Cu21 has the smallest thermal hysteresis reported in the literature, based on experimental measurements (C) and comparison to literature under zero stress (D) and applied stress (E).

Designing Materials to Revolutionize and Engineer our Future (DMREF)