An adaptive multiresolution method for parabolic PDEs

Kai Schneider (Université de Provence), Marie Farge (Ecole Normale Supérieure)

We present a new method to extract coherent events out of turbulent signals and apply it to ion density fluctuations measured in the edge plasma of the Tokamak Tore Supra. The algorithm is based on wavelet denoising and uses a threshold which is iteratively determined without any adjustable parameter. The signal is thus projected onto an orthogonal wavelet basis, a nonlinear thresholding is applied to the wavelet coefficients, and the denoised signal is reconstructed in physical space. This procedure disentangles two orthogonal contributions: the coherent events, which contain most of the variance, have a non-Gaussian probability distribution and are correlated, from a weak background noise, which is structureless, quasi-Gaussian and uncorrelated. We will also give perspectives for computing turbulent flows using the CVS (Coherent Vortex
Simulation) we are developing and which is based on the same approach.

This work has been done in collaboration with Pascal Devynck from CEA-DSM-DRFC, Cadarache, and has been financially supported by the
Association Euratom-CEA.


Presentation (PDF File)
A series of related papers (Zip Archive)

Back to Long Programs