Automated high-throughput system developed to generate structural materials databases

A NIMS research team has developed a high-throughput automated system capable of generating data sets from a single sample of a superalloy used in aircraft engines. The system successfully produced an experimental dataset containing several thousand records, each consisting of interconnected processing conditions, microstructural features, and resulting yield strengths (hereinafter referred to as “Process-Structure-Property datasets) – in just 13 days.

Datasets are generated over 200 times faster than conventional methods. The system’s ability to quickly produce large-scale, comprehensive data sets could potentially significantly accelerate data-driven materials design. This research is published in Materials and design.

High-precision experimental data is essential for studying materials mechanisms, formulating theories, building models, performing numerical simulations and machine learning, and driving materials innovation. In particular, large amounts of accurate Process-Structure-Properties datasets are indispensable for optimizing processing methods for heat-resistant superalloys and the complex, multi-element microstructures of these materials. However, developing such databases typically requires years of continuous experimental work and a significant investment of resources. These challenges have long hampered the development of high-performance superalloys.

This NIMS research team recently developed a new high-throughput automated evaluation system capable of generating process-structure-property data sets containing thousands of data points from a single sample of a Ni-Co-based superalloy developed by NIMS for use in aircraft engine turbine disks. These data sets include processing conditions (heat treatment temperatures), microstructural information (e.g. precipitate parameters), and mechanical properties (e.g. yield strength).

The superalloy sample was heat treated using a temperature gradient furnace developed by the team, thereby mapping a wide range of processing temperatures. Precipitate and yield strength measurements were obtained at various coordinates along the temperature gradient using a scanning electron microscope automatically controlled using a Python API and a nanoindenter.

The system then quickly evaluated and processed the collected data. As a result, in just 13 days, the system managed to generate a volume of Process-Structure-Properties data that would have taken conventional methods approximately seven years and three months to produce.

The research team plans to apply this system to building databases for various target superalloys and developing new technologies to acquire yield strength and creep data at high temperatures. Additionally, the team aims to formulate multi-component phase diagrams, essential for materials design, based on the constructed superalloy databases, and explore new superalloys with desirable properties using data-driven techniques.

The ultimate goal is to make new heat-resistant superalloys that could help achieve carbon neutrality.