An advanced numerical method for the vibro-acoustic response of immersed structures

Controlling the acoustic signatures of marine platforms is a major challenge for shipbuilding and marine energy manufacturers, for the following reasons:

• for passenger shipbuilders, being able to justify a comfort brand is a differentiating factor in the competition, and meets a growing demand from shipowners. Vibroacoustic criteria, defined in terms of permissible cabin noise levels for example, are becoming increasingly stringent in specifications. This applies not only to passenger areas, for obvious comfort reasons, but also to crew areas, due to changes in the regulatory context;

• for military shipbuilders, mastering the acoustic signature of a ship is a vital element in guaranteeing its stealth under a wide range of operational conditions, in a context where underwater detection systems are becoming more and more powerful.

In addition to these design constraints, the need to reduce weight and optimize production costs is driving down insulation margins. Increasingly precise control of noise and vibration levels is required.

In this context, numerical simulation is increasingly used in the design phase to demonstrate and justify the expected vibration performance over the desired frequency range.

In this paper, we propose a method for improving the predictive capability of vibro-acoustic calculations and controlling their uncertainties, in order to estimate the frequency response of mixed structures (typically made up of metallic and visco-elastic materials) immersed in a heavy fluid, taking into account the random nature of certain input data (such as material characteristics). The technique developed is based on a model reduction method, coupled with a sampling method and a learning algorithm. After outlining the industrial need for predictive vibro-acoustic calculation tools, we review the mathematical models classically used for this type of problem and present the main methods for calculating frequency response, using both standard approaches and the method we have developed to account for uncertainties.

Finally, we illustrate the method on two application cases with a number of degrees of freedom representative of models used for industrial purposes.

A list of the symbols used is given at the end of the article.