Techniques for the automatic optimisation of active shields
The present work is about the development of automatic techniques for the optimisation of active shields for stationary magnetic fields. An active magnetic shield is basically made by a number of coils, fed with suitably chosen currents. In this way a magnetic field as equal as possible to the distu...
Saved in:
Published in | Compel Vol. 20; no. 3; pp. 732 - 739 |
---|---|
Main Authors | , |
Format | Journal Article |
Language | English |
Published |
Bradford
MCB UP Ltd
01.09.2001
Emerald Group Publishing Limited |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The present work is about the development of automatic techniques for the optimisation of active shields for stationary magnetic fields. An active magnetic shield is basically made by a number of coils, fed with suitably chosen currents. In this way a magnetic field as equal as possible to the disturbing one is generated. The resulting effect is the reduction of the disturbing field in the area of interest. To achieve reasonable results a stochastic optimisation procedure has been used to optimise the non linear part of the problem (i.e. the geometry of shielding coils); the linear part can be optimised "on the fly" in a much more straightforward way through the solution of a least squares problem. The stochastic optimiser used is based on the very fast simulated reannaling (VFSR), allowing to get a good optimum with a much reduced sampling of the objective function. It is used combining it with a deterministic optimiser (Nelder-Mead simplex method), to get a faster optimisation process as soon as the valley of the global optimum is located. Basically the VFSR has a different point generating function and a different cooling schedule with respect to the standard Boltzmann annealing, but the concept is clearly the same. |
---|---|
Bibliography: | href:03321640110393734.pdf filenameID:1740200307 original-pdf:1740200307.pdf ark:/67375/4W2-N2XPLG9V-D istex:5CA464F93FE16616ABB05AF486E82AC535BC2BBB ObjectType-Article-2 SourceType-Scholarly Journals-1 ObjectType-Feature-1 content type line 23 |
ISSN: | 0332-1649 2054-5606 |
DOI: | 10.1108/03321640110393734 |