Publications by authors named "Viet Dung Vu"
Article Synopsis
- A new method is introduced for learning acoustical responses with limited experimental data, using a multiscale-informed encoder to simplify the learning process.
- A neural network is used to link important parameters to latent variables, leveraging transfer learning and multiscale surrogates to improve model accuracy.
- The effectiveness of this methodology is demonstrated by predicting sound absorption coefficients of packed rigid spheres, revealing that it can achieve good results even with a small dataset, thus facilitating acoustical model development in data-limited scenarios.
The transport and sound absorption properties of random close packings of monodisperse spherical particles are explored following a multiscale approach. First, the discrete element method is used to simulate the free fall of the monodisperse particles in a bounded domain to create virtual samples that are representative of real samples. Different particle diameters ranging from 1 to 16 mm are studied.
View Article and Find Full Text PDF