705 - Optimal memory-storage and low-rank strategies for advanced and data-driven modelling in computational mechanics
Organized by: E. Benvenuti , P. Diez , G. Manzini and M. Nale
The MiniSymposium aims to bring together multidisciplinary expertise in advanced
numerical techniques for managing big data in computational mechanics, including
machine learning algorithms and low-rank strategies. Key areas of focus include:
1. Low-Rank Approximations in numerical solutions, such as SVD and POD for
finite element solvers.
2. Tensor-Based Methods for efficient storage and processing of multi-dimensional
data, including tensor decomposition and QTT methods.
3. Reduced-Order Modeling (ROM) using projection-based and data-driven
ROMs in structural dynamics and fluid mechanics.
4. Machine Learning applications in computational mechanics, including
supervised/unsupervised learning, PINNs, and surrogate modeling.
5. Big Data and HPC solutions, with a focus on parallel computing, memoryefficient
algorithms, and distributed data storage.
6. Data-Driven and Physics-Based Models integrating multi-fidelity and multiscale
modeling approaches with adaptive strategies.
7. Uncertainty Quantification and Optimization using probabilistic methods and
stochastic reduced-order models.
8. Real-Time Simulation and Control involving low-rank methods in FEA for
structural health monitoring and control of smart infrastructure.
9. Data Compression and Efficient Storage techniques to reduce I/O bottlenecks
in HPC and manage simulation data efficiently.
The MS welcomes contributions addressing applications in multi-physics and multi-scale
problems, including coupled systems, materials science, geomechanics, and
biomechanics, while also embracing a wide range of computational science and
engineering applications.