Fatigue design of automotive parts: Application of machine learning methods to design parts according to a stiffness loss criterion

The main objective of this study is to challenge the use of machine learning techniques to model the stiffness loss occurring during durability tests after crack initiation without explicitly simulating the crack, on automotive parts. Extensive studies have been conducted at Vibracoustic to investigate the durability of carbonfilled natural rubber, under various test conditions. This database includes notably relaxing uniaxial tension tests on multiple samples geometries, and a part of them will be our focus in this paper. By using a sub-sample of the total available database, the authors aim to first explain the different steps taken towards the creation of a FAIR data-model. Our first findings confirm that some techniques from the Machine Learning field can be proficiently used in this case study. The next step will be to use an enlarged database with more loading conditions and geometries including real parts.