Aluminium Cylinder - Flexible Neutron Tomography Dataset

A flexible neutron tomography dataset, acquired at multiple different exposure times, and with a highly composite number of angles, allowing investigations of optimal ratio between angles and exposure times when total experiment time is limited. The cylinder was scanned at IMAT, ISIS Neutron & Muon Source, experiment number RB2480001. Raw Data Here we release the raw data from the experiment with the following exposure times: 30s 15s 7.5s 3.75s - two acquisitions were made at this exposure time, labelled 'a' and 'b' Each of these acquisitions inlcude 840 projections. The pixel size was 48 microns. This data can be found within the raw_data folder with a subfolder for each exposure time. Within the folder for an exposure time, the raw projections can be found in the folder named 'exp_[exposure_time]s. There are also separate folders for the flat fields and dark fields acquired before and after the projections. Preprocessed Data Data pre-processed in Mantid Imaging have been released in the folders with the `preprocessed` prefix, for the following datasets which were created through addition of scans with different exposure times, and/or removal of projections (equally spaced): Exposure Time (s) Summed Datasets Angles 60 30s + 15s + 7.5s + 3.75s_a + 3.75s_b 840 60 30s + 15s + 7.5s + 3.75s_a + 3.75s_b 420 60 30s + 15s + 7.5s + 3.75s_a + 3.75s_b 210 60 30s + 15s + 7.5s + 3.75s_a + 3.75s_b 105 30 30s 210 15 15s 420 7.5 7.5s 840 The data was preprocessed in Mantid Imaging, and in each folder, as well as the tiff files, is the .json file describing the pre-processing steps run in Mantid Imaging. Additionally a `.csv` file is present, containing the angular positions (degrees) of the projections. The processed data is used in jupyter notebooks demonstrating the impact of reduced experiment time and application of iterative reconstruction methods in the Neutron Tomography showcase: https://github.com/TomographicImaging/Neutron-Tomography-Showcase. Reconstructed Data We only share SIRT reconstructions of the dataset with exposure time 30s and angles 210. This is used in a demo in https://github.com/TomographicImaging/Neutron-Tomography-Showcase. These reconstructions are in the recon_data.zip, and are TIFF files. The SIRT-NN folder contains a reconstruction where a non-negativity constraint was applied. The SIRT folder contains a reconstruction without this constraint. The .npy files contain the lists of objective values and corresponding iteration numbers. The Sample There is an upper and lower cylinder in this sample, both made of Aluminium. The lower cylinder has holes with 1mm, 2mm, 3mm and 4mm diameters, and they are filled with: · ZrO2 spheres (0.5mm diameter), · Trisospheres (1mm diameter), · Mixture of ZrO2 (0.5mm diameter) and trisopheres (1mm diameter), · Steel spheres (2mm dia) in Al foil The trisospheres have a ZrO2 kernel directly coated with a pyrolytic carbon layer and an outer SiC layer. The SiC layer is expected to have been too thin to be resolved in the experiment. The upper cylinder contains four rods, each 2mm in diameter. The material was unknown but a portable element analyzer indicated that the rods are made of ~90%Si alloyed with small amounts of Fe/Al. The rods also extend out of the top of the cylinder. Notes During the acquisition of the data, the beam went down, meaning the acquisition was split with a ~9 hour gap between projection 764 and 765 being acquired. Unfortunately due to this unexpected downtime, the 3.75s exposure times were not recorded for projection number 764. As a result, the projection at index 764 in the 3.75s_a and 3.75s_b datasets are synthetic (as we intended to release a demo dataset which appeared standard). They were created from halving the 7.5s projection at this index. We updated the log files to include these entries as if they were recorded in the experiment, and entered the 'counts' entries to be identical to those for the projection 763. Additionally, the dark and flat fields were only acquired once at the 3.75s exposure time, so these files have been duplicated in the entries for 3.75s a and 3.75s b