A digital twin framework for mechanical testing powered by machine learning

Abstract

The aviation industry demands innovation in new materials and processes which can demonstrate high performance with minimum weight. Strength-to-weight ratio (STR) is the key metric that drives the value justification in this demand stream. However, aviation's test and certification procedures are time-consuming, expensive, and heavily regulated. This study proposes a Digital Twin (DT) framework to address the time and high costs of mechanical testing procedures in the aviation industry. The proposed DT utilizes new Machine Learning (ML) techniques such as Transfer Learning (TL). Hence, a proof-of-concept study using TL in the Aluminum material group has been demonstrated. The promising results revealed that it was possible to reduce the test load of new material to 40% without any significant error.