Integrated Tool for Accelerated Materials Design & Development (AMDAD) of New Shape Memory Alloys

Traditional methods, by which new materials are designed, developed & manufactured on a mass scale, take a long time while incurring a huge cost. With the advent of computational materials modeling tools and Artificial Intelligence (AIML) techniques, there have been efforts to gain insights into materials systems, capture complex trends, and identify materials with targeted properties thereby reducing the timelines while saving the cost on experiments. Additionally, Image analysis methods in AIML can also help in capturing the processing conditions that the material undergoes.

We have developed an integrated tool, AMDAD, with graphical user interface that has all the machine learning operations in one place starting from data collection, data cleaning, data preprocessing, model fitting, optimization to reading the microstructures. This enables researchers to not only link the Processing-Structure-Properties of the existing material systems but also design new materials with desired properties.

The tool’s capability is demonstrated through a case study on identifying NiTiCu shape memory alloy compositions with targeted properties with single (maximum transformation temperature-TT) and multi (TT and hysteresis) objective optimization under given processing conditions. The bounds on the compositions & heat treatment conditions were selected based on the phase diagrams. For Multi-Objective optimization, Neural Networks coupled with Genetic Algorithm were used to make the predictions out of which the alloys with targeted properties were selected from the Pareto front and experimentally validated. The measured values found to be within reasonable limits of the predictions. The tool can further be used to predict other mechanical properties of SMA.