Nowadays, despite the constant technological growth of the last twenty years, Computational Fluid Dynamics (CFD) problems represent a hard challenge for scientific computing because of their large demand on computational resources. In fact, in the context of aerodynamics optimization and design, CFD applications require to simulate many different possible realizations of a system which can become prohibitive in terms of computational and memory effort. These considerations have produced an intensive development of reduced order methods to provide highfidelity simulations via efficient and lowdimensional models. In this thesis we focus on an efficient and flexible technique, namely the Inverse Distance Weighting (IDW). This strategy can be used to solve mesh motion problems in a FluidStructure Interaction (FSI) framework. IDW is an interpolation strategy which computes the displacement of the grid nodes starting from the movement of some data points, called control points. The original formulation of IDW uses as control points the nodes of the grid belonging to the interface and requires assembling a matrix [N_s × N_c], where N_s is the number of values to be interpolated and N_c indicates the number of control points. Since the number of interface nodes for a typical mesh motion problem can be very large (of the order of hundreds of thousands), the system to solve assumes significantly high dimensions. In order to reduce the computational and the memory load due to these huge dimensions, we implement an ad hoc algorithm which performs a geometrical and a model order reduction of the system: the former reduction is based on an iterative procedure which selects some relevant control points among the initial ones, while the latter reduction is inspired by the Proper Orthogonal Decomposition (POD) strategy to exploit the advantages of an OfflineOnline splitting between ROM construction and evaluation of the solution. The implemented algorithm is validated on some examples, from benchmark cases to more complex three dimensional configurations such as a wing or a hull. All the reduction strategies developed are implemented using an efficient C++ object oriented code and an opensource Finite Element library (libMesh), while the processing of the grids is performed by the opensource visualization software Paraview.
A reducedorder inverse distance weighting technique for the efficient meshmotion of deformable interfaces and moving shapes in computational problems
D'AMARIO, ALESSANDRO
2014/2015
Abstract
Nowadays, despite the constant technological growth of the last twenty years, Computational Fluid Dynamics (CFD) problems represent a hard challenge for scientific computing because of their large demand on computational resources. In fact, in the context of aerodynamics optimization and design, CFD applications require to simulate many different possible realizations of a system which can become prohibitive in terms of computational and memory effort. These considerations have produced an intensive development of reduced order methods to provide highfidelity simulations via efficient and lowdimensional models. In this thesis we focus on an efficient and flexible technique, namely the Inverse Distance Weighting (IDW). This strategy can be used to solve mesh motion problems in a FluidStructure Interaction (FSI) framework. IDW is an interpolation strategy which computes the displacement of the grid nodes starting from the movement of some data points, called control points. The original formulation of IDW uses as control points the nodes of the grid belonging to the interface and requires assembling a matrix [N_s × N_c], where N_s is the number of values to be interpolated and N_c indicates the number of control points. Since the number of interface nodes for a typical mesh motion problem can be very large (of the order of hundreds of thousands), the system to solve assumes significantly high dimensions. In order to reduce the computational and the memory load due to these huge dimensions, we implement an ad hoc algorithm which performs a geometrical and a model order reduction of the system: the former reduction is based on an iterative procedure which selects some relevant control points among the initial ones, while the latter reduction is inspired by the Proper Orthogonal Decomposition (POD) strategy to exploit the advantages of an OfflineOnline splitting between ROM construction and evaluation of the solution. The implemented algorithm is validated on some examples, from benchmark cases to more complex three dimensional configurations such as a wing or a hull. All the reduction strategies developed are implemented using an efficient C++ object oriented code and an opensource Finite Element library (libMesh), while the processing of the grids is performed by the opensource visualization software Paraview.File  Dimensione  Formato  

2016_04_DAmario.pdf
accessibile in internet per tutti
Descrizione: Versione definitiva e completa della tesi
Dimensione
6.86 MB
Formato
Adobe PDF

6.86 MB  Adobe PDF  Visualizza/Apri 
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/118102