Neutron Imaging Processing at HFIR and SNS

Talk presented during the International Imaging Workshop in Varenna (2015)

Transcript

1. ORNL is managed by UT-Battelle for the US Department of Energy Neutron Imaging Processing at HFIR and SNS Jean Bilheux bilheuxjm@ornl.gov
2. 2 Presentation_name Outline •  ORNL Neutron User Facilities •  Use of Neutron for Imaging and Strategy •  Examples of Research Projects •  Data Normalization, Visualization and Analysis •  Algorithms –  Cylindrical Geometry Correction –  Adaptive Growing Region •  “Seeing is NOT believing” (Christina Messa)
3. 3 Presentation_name ORNL is Department Of Energy’s largest science and energy laboratory $1.46B budget World’s most intense neutron source 4,325 employees World- class research reactor 3,200 research guests annually $750M modernization investment Nation’s largest materials research portfolio Forefront scientific computing facilities Nation’s most diverse energy portfolio Managing billion-dollar US ITER project Located in East Tennessee (near Knoxville)
4. 4 Presentation_name High Flux Isotope Reactor (HFIR) Intense steady-state neutron flux and a high-brightness cold neutron source Spallation Neutron Source (SNS) World’s most powerful accelerator-based neutron source DOE BES investment has created 2 powerful neutron sources at ORNL
5. 5 Presentation_name Outline •  ORNL Neutron User Facilities •  Use of Neutron for Imaging and Strategy •  Examples of Research Projects •  Data Normalization, Visualization and Analysis •  Algorithms –  Cylindrical Geometry Correction –  Adaptive Growing Region •  “Seeing is NOT believing” (Christina Messa)
6. 6 Presentation_name Neutrons are highly sensitive to Light Elements [M. Strobl et al., J. Phys. D: Appl. Phys. 42 (2009) 243001] Neutron Radiograph of Rose in Lead Flask X-ray Radiograph of camera Courtesy of E. Lehmann,PSI gure 2. Mass attenuation coefficients for thermal neutrons and 0 keV x-rays for the elements (natural isotopical mixture unless ted differently). (Reprinted with permission from [10]. Copyright 08, University of Oxford Press.) c) or plastics contained within metals; in contrast, x-rays Neutron Radiograph of camera
7. 7 Presentation_name Strategy •  ORNL Neutron Facilities have a broad and diverse user community with different levels of data analysis expertise. –  Research spans from materials science to engineering, biology, archeology, geosciences… •  Provide user-friendly/intuitive standalone 2D data normalization and analysis software. –  Software can be easily adapted to user needs (plugins, new algorithms…) –  Software needs to work easily with other software of the community (ImageJ, Octopus…)
8. 8 Presentation_name Outline •  ORNL Neutron User Facilities •  Use of Neutron for Imaging and Strategy •  Examples of Research Projects •  Data Normalization, Visualization and Analysis •  Algorithms –  Cylindrical Geometry Correction –  Adaptive Growing Region •  “Seeing is NOT believing” (Christina Messa)
9. 9 Presentation_name Using neutron CT to study internal structure of turbine blades made by AM Advanced Materials and Processes, March 2013. •  New manufacturing techniques require advanced characterization capabilities for prediction and validation •  Neutron imaging data provide direct tests for model validation and process optimization
10. 10 Presentation_name Fabrication tolerance studies using CAD drawing and neutron CT + Engineering drawing Neutron CT •  Semi-automated analysis •  Need for software capable of looking for features we don’t know exist =
11. 11 Presentation_name Ancient Craft Skills meet Modern Characterization Ryzewski K., Bilheux H., Herringer S., Bilheux J., Walker L., Sheldon B., " The use and refinement of neutron imaging technologies for archeological artifacts", Advances in Archeological Practice, 2, 91-103, (2014).
12. 12 Presentation_name Outline •  ORNL Neutron User Facilities •  Use of Neutron for Imaging and Strategy •  Examples of Research Projects •  Data Normalization, Visualization and Analysis •  Algorithms –  Cylindrical Geometry Correction –  Adaptive Growing Region •  “Seeing is NOT believing” (Christina Messa)
13. 13 Presentation_name Data Normalization for Imaging •  2D – Radiography –  Normalization = - - Image Dark Field Dark Field Open Beam Normalized Image 1 0 Transmission ( ) ( ) ( ) ( ) ( ) j i, DF - j i, OB j i, DF - j i, I = j i, I N
14. 14 Presentation_name Raw Data: 2048x2048 pixels, 721 projections Normalized Data: 2048x2048 pixels, 721 projections Sinograms: 2048x721 pixels, 2048 files Slices: 2048x2048 pixels, 2048 slices 3D reconstruction: 2048x2048x2048 voxels Computed/Computerized Tomography –  Filtered back projection method ~20700 ~700 Counts 1 0 Transmission 1 0 Transmission ∞ 0 Attenuation
15. 15 Presentation_name Data analysis •  Peformed using –  iMARS (Neutron Sciences software) –  VGStudio (commercial software) –  ImageJ (Open Source)
16. 16 Presentation_name iMARS (ImagingMAnipulation Reduction Software)
17. 17 Presentation_name iMARS – MCP input files Raw image of a simple object Life is good !
18. 18 Presentation_name iMARS – MCP input files To avoid artifacts during CT reconstruction, gaps between chips need to be “filled” Using mean of ‘real’ pixels surrounding gap. But other methods are available (linear interpolation for example).
19. 19 Presentation_name iMARS – MCP dead pixels correction Dead pixels are replaced by average value of neighbors
20. 20 Presentation_name Software workflow Data normalization Raw data Visualization CT Data Analysis
21. 21 Presentation_name Outline •  ORNL Neutron User Facilities •  Use of Neutron for Imaging and Strategy •  Examples of Research Projects •  Data Normalization, Visualization and Analysis •  Algorithms –  Cylindrical Geometry Correction –  Adaptive Growing Region •  “Seeing is NOT believing” (Christina Messa)
22. 22 Presentation_name Cylindrical Geometry correction (challenge) n beam pixels image Attenuation coefficient Pixel position
23. 23 Presentation_name Cylindrical Geometry correction (challenge) Cylinder edges Integration over y-axis not straightforward when sample is titled x y Direction of integration 1 2 3
24. 24 Presentation_name Cylindrical Geometry correction (challenge) Cylinder edges Type of parameters to take into account. Direction of integration 1 2 3 •  Center of pixel deftermine value of correction to apply •  Area of pixel can be considered or not (for pixels on the edge) •  ….
25. 25 Presentation_name Colormap (default does not mean right) -  Default works great with software to install -  But not with imaging ! Default colormap
26. 26 Presentation_name Colormap Evaluation of Artery Visualizations for Heart Disease Diagnosis. Borkin et al. (2011)
27. 27 Presentation_name Colormap (matlab)
28. 28 Presentation_name Colormap (python)
29. 29 Presentation_name Colormap (Python vs Matlab)
30. 30 Presentation_name Colormap (new default for Matlab)
31. 31 Presentation_name Summary •  Neutron Imaging at HFIR is overbooked by a factor 3 •  Research spans from materials science to engineering, biology, archeology, geosciences… •  Venus should be up and running in 3 years •  Software needs for imaging are now •  Collaboration ? •  Common file format (HDF5?) •  Common workbench •  Default colorscale is bad (Python, matplotlib (up to now)) •  Matlab -> Parula •  Python -> Viridis •  Other -> Check ….
32. 32 Presentation_name Takes home •  Default colorscale is bad (Python/matplotlib), matlab) •  Matlab -> Parula •  Python -> Viridis •  Other -> Check ….
33. 33 Presentation_name Acknowledgments •  ORNL contributions from: Hassina Bilheux, Lou Santodonato, Ke An, Barton Bailey, Ken Herwig, Scott Keener, Costas Tsouris, Ryan Dehoff, Mike Kirka, Lindsay Kolbus, Todd Toops, Charles Finney, Eric Nafgizer, Derek Splitter, Sophie Voisin, Jeff Warren •  Anton Tremsin, University of CA - Berkeley •  Ed Perfect, Maria Cekanova, Jen Gregor, University of TN – Knoxville •  Chu-Lin Cheng, University of Texas – Pan America/Rio Grande Valley •  Misun Kang, University of California - Davis •  Fei Ren, Temple University, Philadelphia •  Industry collaborators •  US DOE Scientific User Facilities Division, Office of Basic Energy Sciences, US Department of Energy •  US DOE Energy Efficiency and Renewable Energy, Additive Manufacturing Office, US Department of Energy
34. 34 Presentation_name Thank You