| Abstract Scope |
The capability to predict the responses of materials with certain microstructures under various loading conditions is crucial for understanding material behaviors and facilitating material design. However, this process can be expensive and challenging, especially for heterogeneous material systems with a large design space, where physics-based repetitive numerical simulations and extensive experiments may be required. Convolutional neural networks (CNNs) has been demonstrated as a computationally feasible way to develop high-fidelity predictive models for heterogenous materials with complex microstructures. However, one key issue in material prediction tasks is limited and unbalanced data caused by the costs associated with different material testing techniques. Such imbalanced data can lead to biased model predictions and poor generalization on unseen material structures. To overcome this challenge, we propose using multi-task learning (MTL) to provide deep learning models with more knowledge of material behaviors, specifically targeting the unbalanced data problem. |