Speaker
Description
The development of new materials and the optimization of their properties are critical challenges in the field of materials science, particularly within the framework of multiscale modeling and homogenization. Traditional data-driven methods, while effective, often fall short in accurately predicting the complex behavior of microstructures under diverse loading conditions, especially when faced with the intricate microstructures of multi-phasic materials. We propose a new approach by embedding essential physics-based constraints, such as Voigt-Reuss bounds, directly into the architecture ensuring that the network not only learns from empirical data but also adheres to established physical laws, thereby enhancing the predictability and interpretability of the model. Importantly, this approach guarantees that the predictions remain within analytical bounds, providing a built-in error control mechanism that significantly enhances the reliability of the model's output. Furthermore, by infusing fundamental physical bounds, our model requires significantly less data, making it particularly advantageous in data-scarce environments where traditional data-driven approaches may falter. We employ high-resolution images of microstructures as the primary input, coupled with detailed information on material properties and loading conditions. This allows for a nuanced understanding of the microstructure response and refining the predictive capabilities of multi-scale models, enabling more accurate and efficient predictions of material responses without the extensive reliance on exhaustive simulation data.
