Title of master thesis: Micromechanics-based artificial neural networks and transfer learning for modeling short fiber reinforced composites in automotive applications
Overview
- Date:Starts 7 June 2023, 14:00Ends 7 June 2023, 15:00
- Location:GU Physics, von Bahr, Soliden 3rd floor
- Language:English
Abstract:
For the automotive industry, improving efficiency is crucial while weight reduction is one key factor for high efficiency. Short fiber reinforced composite (SFRC) provides superior performance with less weight and is suitable for mass production, making it an interesting material for the industry to adopt. However, the mechanical modelling of SFRC is difficult, a full field analysis could take trials to generate a proper realization, and then the analysis could take hours to finish. Moreover, there are countless fiber orientations and volume fraction, both influential to the mechanical response. Therefore, the use of data-driven models for SFRC has gained popularity. Previously, work has been done to use mean field analysis results to train a recurrent neural network, to predict elastoplastic stress response of SFRC with different strain path and properties. In this study, efforts have been made to enhance the MF network with a small number of full-field data, to enable the enhanced network to provide prediction with a full-field level of accuracy.
Supervisor: Mohsen Mirkhalaf
Examiner: Mohsen Mirkhalaf
Opponent: Xinhao Wang
