Project code: III.1
Project acronym: TomoModelSupra
Field: Fundamental research on Energy
Topic: Nuclear fusion and fission
Project title
Characterisation and modelisation of superconductor cables:
accurate 3D examination of Superconductor Cables by X-ray microtomography and mutliphysic models upgrade
(electromagnetic, thermo-hydraulic and mechanic)
Partners
|
RO team |
CEA team |
Project leader Name |
Ion TISEANU |
Louis ZANI |
Laboratory Institution Address Tel : Head of laboratory |
Plasma Physics and Nuclear Fusion National Institute for Lasers, Plasma and Radiation Physics (INFLPR) Dr. Ion TISEANU |
Institut de Recherche sur la Fusion par confinement Magnétique (IRFM) Dr. Alain Becoulet |
Project Objectives
- Demonstrate that X-ray micro-tomography (μCT) can be used as a very efficient and reliable tool for retrieving the necessary information for constructing detailed 3D models of the superconductor cable (CICC).
- Upgrade the multi-physic models taking into consideration the experimental results from µCT examinations and validate their new configuration through confrontation with experimental tests campaigns.
Abstract
Operation and data acquisition of an X-ray micro-tomograph developed at INFLPR are optimized to produce stacks of 2-D high-resolution tomographic sections of Cable in Conduit
Conductor (CICC) type superconductors demanded in major fusion projects. High-resolution
images for CCIC samples (486 NbTi&Cu strands of 0.81 mm diameter, jacketed in
rectangular stainless steel pipes of 22x26 mm^2) are obtained by a combination of high
energy/ high intensity and small focus spot X-ray source and high resolution /efficiency
detector array. The stack of reconstructed slices is the input data for quantitative analysis
consisting of accurate strands positioning, determination of the local and global void fraction
and 3D strand trajectory assignment for relevant fragments of cable (~300 mm). Strand
positioning algorithm is based on the application of Gabor Annular filtering followed by local
maxima detection. The local void fraction is extensively mapped by employing local
segmentation methods at a space resolution of about 50 sub-cells sized to be relevant to triplet
of triplet twisting pattern.
For the strand trajectory assignment we implemented two types of
algorithms: i) a simple local algorithm trying to match the strands in adjacent slices. Even
with a strand positioning efficiency over 95% such a simple algorithm cannot assign more
than 50% of strands trajectories; ii) a global algorithm of the linear programming type which
provides dramatically improved number of strand trajectories. For the benchmark CCIC
samples 99% of the trajectories are correctly assigned. For production samples the efficiency
of the algorithm is around 90%. Trajectory assignment of a high proportion of the strands is a
crucial result for the derivation of statistical properties of the cable such as twisting pattern, cos(theta) or void fraction.