Figures Of Merit for the Application of Tomography to the Characterization of Nuclear Waste Drums

NuSec, 15th-16th April, Surrey, UK

Patrick Stowell, Ahmad Alrheli, Daniel Kikola, Anna Kopp, Holger Tietze-Jaensch, Mohammed Mhaidra, Lee Thompson, Elie Valcke, Jaap Velthuis, Michael Weekes

This project has received funding from the Euratom research and training program 2014-2018 under grant agreement No 755371 Introduction

CHANCE: “Characterization of conditioned nuclear waste for its safe disposal in Europe”

• Funded by Euratom research and training program 2014-2018 under grant agreement N° 755371 • 4.25 M€ project EU funded project. June 1, 2017 - 31 May, 2021 • Consortium: 12 partners / 8 European countries • WP4 - Muon imaging for innovative tomography of large volume and heterogenous cemented waste packages

2 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Work Package 4 : Muon Tomography

• Primary showers in the upper atmosphere produce a high flux of at sea level. • Highly penetrating particles that can probe dense materials.

� Small Scatter Angles - Low Density - Low Z Materials �+ �− - Air, Hydrogen

�0 Large Scatter Angles - High Density � - High Z Materials - , Plutonium

�− � Cement Matrix �� Steel Drum

+ Cosmic Ray Muons � Uranium + �¯� � �−

3 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Detector System

• Muon tracking detector system being commissioned at the University of Bristol in a non-laboratory environment. (A. Kopp, Bristol) • Combination of high precision/low cost detector technologies.

Trigger boxes and drift chambers kindly provided by AWE. Resistive Plate Chambers, Bristol

Scintillator AWE Scintillator AWE/Manchester DriftTrigger Triggers, Sheffield Chambers, Sheffield

4 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Detector Simulations

• Several simulation studies being developed for the CHANCE project alongside detector commissioning within the CRESTA cosmic ray simulation package (CRY+GEANT4 tomography interface).

• Void/Hydrogen localisation in homogenous waste packages. (M. Maihdra, Warsaw)

• Multivariate material discrimination. (M. Weekes, Sheffield)

• Large volume imaging. (A. Alrheli, Sheffield)

• Tools for tomography algorithm benchmarking. (P. Stowell, Sheffield, NuSec Funded)

5 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Tomography Algorithms

• Penetrating power of the muons allows clean track reconstruction entering and exiting the volume of interest. • Can reconstruct the scattering vertices inside a waste drum to identify high density or high Z materials.

Concrete θ Point of Closest Steel Drum Approach Uranium

Point of Closest Approach (PoCA) : Assume single muon scattering vertex. Weight larger scatters higher. 30 Days Simulated Exposure

6 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Tomography Algorithms (2)

• Alternative algorithms available which improve on the assumptions made in the PoCA algorithm. • Many to choose from, but need to understand which is best for each of our considered applications.

Binned Clustering (BC) Angle Statistics Reconstruction (ASR) High density materials leave Weight voxels along muon trajectories concentrated vertex clusters. to remove PoCA assumption.

Low Density ri PoCA ASR

High Density rj

→ → r i − r j mij = θi piθj pj

7 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Tomography Algorithms (2)

• Alternative algorithms available which improve on the assumptions made in the PoCA algorithm. • Many to choose from, but need to understand which is best for each of our considered applications.

Binned Clustering (BC) Angle Statistics Reconstruction (ASR)

30 Days Simulated Exposure

8 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Figures of Merit

• Using an optical “ISO” technique instead to quantify imaging performance of different algorithms under identical simulation conditions. • Number of observable objects quantifies feature resolution. • Figure of merit method that is can also be experimentally verified.

Uranium Block Uranium Foils 10x10x10 cm3

Uranium Block 7.5x10x10 cm3 N0−1 xmin = 10 cm × 0.75 30 Days Simulated Exposure

9 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Figures of Merit

• Using an optical “ISO” technique instead to quantify imaging performance of different algorithms under identical simulation conditions. • Number of observable objects quantifies feature resolution. • Figure of merit method that is can also be experimentally verified.

Signal Background Observable Feature

N0−1 xmin = 10 cm × 0.75 Resolution : 2.4 cm

10 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Smallest Observable Seperation

• Binned Clustering and Angle Statistics both have improved feature separation resolution compared to PoCA. • Smearing in ASR algorithm makes the discriminator lower in smallest pieces of Uranium. Value starts to become comparable to steel.

N0−1 xmin = 10 cm × 0.75 Resolution : 1.8 cm

11 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Smallest Observable Feature

• Similar tests can be performed to identify the smallest observable feature in a flat concrete background. • Difficult to distinguish ~4mm objects from background in PoCA. • Smallest objects have a discriminator comparable to steel.

Uranium Block Uranium Foil 3x10x10 cm3

Uranium Block 1.7x10x10 cm3 N0−1 Resolution : 0.9 cm smin = 3 cm × 0.75 12 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Smallest Observable Feature

• Similar tests can be performed to identify the smallest observable feature in a flat concrete background. • Difficult to distinguish ~4mm objects from background in PoCA. • Smallest objects have a discriminator comparable to steel.

N0−1 Resolution : 0.4 cm smin = 3 cm × 0.75 13 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Exposure/Material Mapping

• Request from CHANCE partners to produce a suitable mapping of figures of merit vs material and exposure time. • Small pieces of /Uranium have a comparable signature to larger pieces of Steel. • Attempting to use edge discrimination techniques to resolve this false negative issue in the future and push resolution lower.

Lead 30 days : 0.7 cm Uranium 4 days Lead 30 days Steel 30 days 0.7 cm 0.7 cm 0.9 cm

14 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019 Summary

• Developing Figure of Merit “ISO” tests that can be used to quantify the imaging performance of muon tomography algorithms/systems.

• ISO plots suggest that BC algorithm can better discriminate high Z materials at smaller sizes (ASR implementation smears small objects)

• Trying to understand the complex correlations between exposure time/ material density/performance for available tomography algorithms.

• Algorithms and Figure of Merit tests will eventually be tested on a real muon tomography system being commissioned in a non-laboratory environment at the University of Bristol, UK.

Thank you for listening and thank you to the NuSec Network for funding this work !

15 P. Stowell et al. | NuSec Detection Workshop, Surrey, 15th-16th April 2019