Machine learning assisted development of high strength iron-based alloy

Abstract

High strength alloys are materials with alloying additions designed to produce a specific combination of mechanical qualities such as strength, toughness, weldability, formability as well as atmospheric corrosion resistance. One such example is the Iron-Aluminium (Fe-Al) based alloy that is highly favorable due to their properties of high specific yield strength, low cost as well as high corrosion resistance. However, initial methods of traditional alloy development have proved to be an extensive, time-consuming process. Studies of superalloy took nearly 40 years to understand the mechanical behavior even before development. The resources and manhours invested in development of alloys has therefore proved to be inefficient. With the use of Machine Learning, the development of alloys can be accelerated through assisted screenings and high throughout experimentation, thus reducing the overall duration as well as improving the experiment's efficiency. In this study, we attempt to demonstrate an alternative approach that utilizes Machine Learning (ML) algorithm that is trained on a given set of composition with given mechanical properties to predict novel alloy composition with high specific yield strength. Predicted novel compositions of Fe-Al will be validated with the developed alloy through a series of experimental test to determine the model’s accuracy.