Predicting Material Properties and its Validation Methodology

Machine Learning Algorithms represent a valid methodology to provide material property prediction. ML can support better decision-making in material selection and performance assessments, when there is a data gap. The benefit to aerospace companies is that they can shorten the material selection process by reducing the need for expensive and time-consuming material characterization tests in materials laboratories.

The Paper will present the different ML models used to perform data predictions, the way to train the algorithms with large curated properties data set and the methodology developed by Polytechnic of Turin to validate the algorithms by comparing the prediction with experimental data.