Page 6 - Demo
P. 6

Bayesian inversion of current sources ( Plasma, vessel, magnets)
Current tomography 1,2
All current sources are treated as block coils
Forward model is linear
Current sources I -> Measurements D D = K I
Inverse problem
Measurements D-> Current sources I
Generally non inver@ble (in algebraic sense).
Can we find all solu@ons I compa@ble with the measurements D and their uncertainty, along with their probability ?
Ø  This will be called the posterior distribu@on :
P(I | D) = ??
Ø  Flux obtained from profile with maximum probability. Separatrix will be then the Ψ =0 contour.
1J. Svensson (2008) PPCF, vol 50, no. 8 2J.A. Romero, J. Svensson (2013). Nuclear Fusion, vol. 53,no. 3
Presenter : J.Romero. Sesion T07. Computational and theoretical Techniques. APS 2016.
6




















































































   4   5   6   7   8