Computational Ocean Acoustics: Do Finite Element Models have a future?
Martin van GijzenSenior Scientist in the Parallel Algorithms Project at CERFACS
Wednesday, February 27th, 10.30 a.m., CERFACS Conference Room
Abstract :
Computer models that describe and predict acoustic wave propagation in the ocean are widely used for various reasons:
To complement experiments at sea, which are very expensive;
To understand and explain phenomena observed during experiments;
To obtain physical parameters, such as sediments properties that are otherwise hard or expensive to obtain;
To optimise SONAR performance.
For the above reasons a major effort has been made to develop accurate and fast numerical models. The majority of these models is based on either ray or normal mode theory. These types of models are fast but assume simplifications of reality.
Finite Element based models on the other hand make very few assumptions on physics. Inclusion of surface waves or scattering of sound against an object on the sea bottom can be modelled in a natural way. Yet, Finite Element models have gained little popularity. The reason is that a realistic model may need a prohibitive amount of computer power.
During the talk we will discuss two examples of Finite Element calculations.
The first example addresses the question whether a sound reflecting object that is buried half in the sediment (e.g. a World War II sea mine, a container or an oil pipe) can be detected by a given SONAR. This example illustrates perfectly the enormous computing power that is required for this type of calculations.
The second example concerns the modelling of the acoustic basin at TNO that is used to test and calibrate equipment. This model can be used to explain certain phenomena that are sometimes observed during tests in the basin.
At the end of the talk we will discuss the question: do Finite Element models have a future in Computational Ocean Acoustics? Or should we rephrase this by: do Finite Element models have the future?
The talk is intended for a large audience. The applications, the physics behind it, and the computational algorithms involved will be discussed in a descriptive way.
Cerfacs' Conferences 2002 Home Page
