Muon imaging for innovative tomography of large volume and heterogeneous cemented waste packages This project has received funding from the Euratom research and training programme 2014-2018 under grant agreement No 755371
Contents Muon tomography Our approach Nuclear waste imaging Finding lumps of material in drums Size measurements Material identification Air bubbles Large(r) structures Summary Request for input 2
Introduction There has been a real surge in cosmic ray tomography over the last 10 years. Muon imaging is a technique to explore closed volumes from a safe distance. Muon imaging was first pioneered/ reinvented by Hiroyuki Tanaka Prof Tanaka is measuring the 3D density structure of volcanoes. Now people are working on homeland security legacy nuclear waste volcano imaging carbon capture and storage monitoring mining 3
Cosmic rays: muons Primary cosmic rays interact in the atmosphere and decay into showers of secondary cosmic rays. On Earth we are constantly bathed in (secondary) fundamental particles from cosmic rays. Naturally occurring at decent rates (~100 Hz/m 2 ) Large angular and energy spread. Flux and energy of muons depend on the zenith angle. Highly penetrating. Measured rate up to 10km water equivalent. All these combined make muons an excellent probe to study enclosed spaces from a safe distance. 4
Muon scattering tomography Muon scattering tomography relies on multiple scattering. The angular distribution is Gaussian, with σ 0 depending on the radiation length Χ 0 and thus on Z. DETECTOR LAYERS INCOMING MUON σ 2 15MeV 0 pcβ 2 T Χ 0 The Z 2 means that the technique is very sensitive for high-z materials! T Δx vertex Every track yields a displacement and an angle We measure an angle and a vertex. DETECTOR LAYERS Δθ 5
Our novel reconstruction method Fit each muon pair as coming from a vertex Divide volume into voxels. Exploiting clusteredness of high scatters Take in each sub-volume the 50 tracks with largest scatter and calculate median of weighted metric distance Steel Uranium 6
Nuclear waste Nuclear reactors have been and are producing waste. This is stored in waste containers in concrete. It is important to monitor what goes on inside a container over time. Requires a measurement from distance with closed container For hazard assessment want to measure the size and location identify the material 7
Monte Carlo simulation of a small drum. Waste drum monitoring Top hat scans, 2 weeks of data, 50% momentum uncertainty 8
Waste drum monitoring Can even see shapes of objects cylindrical U rod (1 cm radius, 10 cm height) thin quadratic U sheet (0.5 x 10 x 10 cm 3 ) two W pennies of 1 cm thickness (2 cm and 4 cm radius). Technique clearly works well for examining waste drum contents. How precise can we measure the outlines? 9
Edge finding Note that the distribution of the metric variable is different deep inside a block and at the edge. This is due to mixing. In bins with mixed material, both materials contribute vertices. Uranium away from the edge Uranium at the edge Concrete at the edge Concrete away from the edge To find the edge, we simply shift the grid. Metric distance calculated for 10 mm bins, but with 10 grid shifts of 1 mm to improve resolution. Fit the logarithm of the metric distance with a Landau distribution convoluted with a Gaussian, in bins close to the edge 10
Edge finding In order to distinguish uranium from concrete at the edge, multivariate analysis (MVA) was performed using variables from the raw data and the fit: Maximum value; Respective bin position; Widths of the Landau and Gaussian from fit; Peak amplitude and position from fit. MVA method: Fisher linear discriminant. Extract probability that voxel contains Uranium Plot probability versus x. Concrete Uranium 5 mm uranium Output from the 30 training mm uranium with the Fisher linear discriminant, using samples of pure concrete and pure uranium (2 cm block). 11
Edge finding results Excellent correspondence between real and reconstructed length. U blocks 4 weeks of data. We can define the edges of uranium blocks in concrete, measuring lengths as small as 2 mm. The resolution obtained was σ = 0.7 ± 0.1 mm. 12
Material identification Want to identify the materials. Focussing on Lead: X 0 0.56 cm Tungsten: X 0 0.35 cm Uranium: X 0 0.32 cm Plutonium: X 0 0.30 cm All materials were embedded in a concrete cylinder, with 13 cm radius and 40 cm length. 10x10x10cm 3 Compare the individual muon scattering tracks for two materials using a Fisher discriminant. Variables: scattering angle, offset, χ 2 of the fit Calculate the mean Fisher output for 1hour of data The ROC curve shows that we can discriminate very well. 13
Material identification Separation U and Pb almost perfect for 10x10x10cm 3 after 1 hour. It takes longer for smaller blocks. Still, it works down to 2x2x2cm 3. Separation U and W almost perfect for 10x10x10cm 3 after 1 hour. It takes longer for smaller blocks. Still, again it works down to 2x2x2cm 3. Can even discriminate U and Pu for 3x3x3cm 3 blocks 14
Air bubbles Several countries used bitumen as immobilisation matrix. Alpha-beta-gamma irradiation results in production of (mainly) H 2 gas Small bubbles can move and agglomerate to a large bubble. Waste was typically stored in 220L drum leaving 10% empty. Overflowing drums have been found. Key to determine whether large amounts of hydrogen will concentrate in the bitumen. 15
Air bubbles We follow the same procedure: divide the drum into voxels calculate the discriminator for each voxel study the mean as a function of the amount of hydrogen inside the matrix Mean clearly good indicator for the size of the bubbles. Can measure the size of the bubbles with 1.3±0.6% relative error from 2 litres. Can even distinguish single bubble from more smaller ones. Presented at IEEE conference and submitted to IEEE TNS 16
Large(r) objects Tomography needs incoming and outgoing track. Only sensitive in overlap region. For large silos this is to expensive and cumbersome. Solution: radiography. In radiography, measure absorption along line of muon track and use two detector stations or locations to produce 3D image. 17
Simulation set up Simulated a building. The walls of the building are modeled as concrete, and the silos are filled with a mixture of pure magnesium (40%), magnesium hydroxide (40%), and water (20%) at a combined density of 1.3 g/cm 3. Each detector unit consists of 6 layers of RPCs, each sized 120 x 120 x 0.6 cm 3. Not feasible to place large enough detectors to do tomography. Need to do radiography from OUTSIDE the building. 20m high 50cm thick concrete walls large concrete silos inside 18
Only 1 month of running. Dotted lines not to scale! Results with 1 month of running Can clearly see the uranium swarf blocks and air enclosure. U swarf 0.5, 1m & 2 m edge ρ=4.75g/cm 3 U swarf ρ=4.75g/cm 3 Air enclosure Iron blocks 0.25, 0.5m & 1 m edge ρ=7.87 g/cm 3 U swarf ρ= 1.9 g/cm 3 19
Results with 1 month of running Results for 1 month after background subtraction for heavy elements Even clearer and can now see iron blocks and air clearly U swarf 0.5, 1m & 2 m edge ρ=4.75g/cm 3 U swarf ρ=4.75g/cm 3 Air enclosure Iron blocks 0.25, 0.5m & 1 m edge ρ=7.87 g/cm 3 Measured from OUTSIDE the building! U swarf ρ= 1.9 g/cm 3 20
Performance summary Cosmic ray tomography is an exciting new technique. It allows to look into a closed box from a safe distance. Nuclear waste characterization We can define the edges of uranium blocks in concrete, measuring lengths as small as 2 mm. The resolution obtained was σ = 0.7 ± 0.1 mm. Can discriminate between blocks of U, Pb & W down to 2x2x2cm 3. Can even discriminate U and Pu down to 3x3x3cm 3 blocks. Can find air pockets from 1 litres upwards. The relative volume uncertainty is 1.3±0.6% for bubbles from 2 litres upwards. Can distinguish between a large air pocket or several small bubbles. 21
Performance summary We are now working on Measuring the size of blocks of other materials than U. Exploiting neutrons for isotope discrimination. Studying the performance for much larger waste drums. Building a new large scale, mobile demonstrator. It is a well established technique that works for all applications where you want to look into a closed box and have a bit of time. 22
CHANCE Within chance will produce a mobile, large size system (2x2m 2 ) detectors Operate it in an industrial type environment Work with dummy drums and improve our algorithms In final phase want to take data using real active waste drums at WMO site. 23
What drums/casks should we focus on? what sizes? tomography or radiography? what type of stored waste? EUG input request What materials should we focus on? What questions should we ask? Can you exclude an Uranium block larger then.? needs little data Can you find lumps, identify the material and tell us the size? needs a lot of data Is location important? Is development over time important? We are building a mobile system. Are looking for a test opportunity with real waste drums towards the end of CHANCE. Looking for partners. Please tell us what you need. 24
25 This project has received funding from the Euratom research and training programme 2014-2018 under grant agreement No 755371