Reduced order models, networks and applications to modeling and imaging with waves (Vladimir Druskin, Cambridge)

02.03.2018 10:30

Geophysical seismic exploration, as well as radar&sonar imaging deals with large scale wave propagation problems. Conventional imaging approaches are often computationally extensive and unreliable. In this talk I will show how the model order reduction can address these issues. In the model order reduction, one approximates response (transfer function) of a largescale dynamic system using a smaller system, known as reduced order model (ROM), that retains certain features of the larger problem. We consider ROMs that can be realized via sparse networks, e.g., mass-spring strings. In particular, I will discuss recent application of such ROMs to multiscale modeling and direct imaging in strongly heterogeneous models with multiple echoes.

Lieu

salle 17, Séminaire d'analyse numérique

entrée libre