KEYWORD |
Reduced order model for scattering problems on distributed-memory architectures for parallel computing
Thesis in external company
keywords COMPUTATIONAL ELECTROMAGNETICS, ALGORITHMS, PARALLEL CODES, REDUCED ORDER MODELS
Reference persons FRANCESCA VIPIANA
Research Groups Applied Electromagnetics
Description Numerical simulations are routinely employed to model complex large scale systems, as the electromagnetic waves propagation. In a similar way, maps are used to describe physical locations. When one is interested in studying the different responses when varying the parameters of the systems, the computational burden becomes too large, as many different simulations are requested. Moreover, often just a reduced part of the system output is of interest. In the maps analogy, this can translate as being interested just in the public transportation network, instead of the whole city geography, to plan different routes. Here, a compact representation retaining only the most important aspects would be useful. From a computational point of view, simpler models can be constructed from a set of representative solutions of the complete system, using different technique to combine them together.
This thesis aims at porting existing code for electromagnetic simulation on a distributed architecture, with the specific aim of using it for building reduced order models for Computational Electromagnetics.
See also http://ismb.it/node/3405
Required skills - Experience with main programming languages (Matlab /Fortran/C/C++)
- Knowledge of parallel computing (MPI)
Deadline 27/10/2020
PROPONI LA TUA CANDIDATURA