Efficient implementation of high-order finite elements for Helmholtz problems

Computational modeling remains key to the acoustic design of various applications, but it is constrained by the cost of solving large Helmholtz problems at high frequencies. This paper presents an efficient implementation of the high-order Finite Element Method for tackling large-scale engineering p...

Full description

Bibliographic Details
Main Authors: Beriot, Hadrien (Author), Prinn, Albert (Author), Gabard, Gwenael (Author)
Format: Article
Language:English
Published: 2016-04-03.
Subjects:
Online Access:Get fulltext
Get fulltext
LEADER 01765 am a22001693u 4500
001 384370
042 |a dc 
100 1 0 |a Beriot, Hadrien  |e author 
700 1 0 |a Prinn, Albert  |e author 
700 1 0 |a Gabard, Gwenael  |e author 
245 0 0 |a Efficient implementation of high-order finite elements for Helmholtz problems 
260 |c 2016-04-03. 
856 |z Get fulltext  |u https://eprints.soton.ac.uk/384370/1/paper_postprint.pdf 
856 |z Get fulltext  |u https://eprints.soton.ac.uk/384370/2/nme5172.pdf 
520 |a Computational modeling remains key to the acoustic design of various applications, but it is constrained by the cost of solving large Helmholtz problems at high frequencies. This paper presents an efficient implementation of the high-order Finite Element Method for tackling large-scale engineering problems arising in acoustics. A key feature of the proposed method is the ability to select automatically the order of interpolation in each element so as to obtain a target accuracy while minimising the cost. This is achieved using a simple local a priori error indicator. For simulations involving several frequencies, the use of hierarchic shape functions leads to an efficient strategy to accelerate the assembly of the finite element model. The intrinsic performance of the high-order FEM for 3D Helmholtz problem is assessed and an error indicator is devised to select the polynomial order in each element. A realistic 3D application is presented in detail to demonstrate the reduction in computational costs and the robustness of the a priori error indicator. For this test case the proposed method accelerates the simulation by an order of magnitude and requires less than a quarter of the memory needed by the standard FEM. 
540 |a other 
655 7 |a Article