Shower-Aware Dual-Stream Voxel Networks for Structural Defect Detection in Cosmic-Ray Muon Tomography

2026-04-04

AI Summary

This study addresses the limitations of conventional cosmic-ray muon tomography, which relies solely on scattering angle information and struggles to accurately detect internal defects in reinforced concrete. To overcome this, the authors propose SA-DSVN, a dual-stream 3D convolutional network that, for the first time, incorporates secondary electromagnetic shower multiplicity as a key discriminative feature alongside muon scattering dynamics. The two streams are processed in parallel and fused via a cross-attention mechanism to enable voxel-level semantic segmentation of four defect types: honeycomb voids, shear cracks, corrosion cavities, and delamination. Evaluated on Geant4-based Vega cloud-native simulation data, the method achieves 96.3% voxel-wise accuracy across 60 samples, Dice coefficients ranging from 0.59 to 0.81, 100% volume-level detection sensitivity, and inference times of only 10 milliseconds per volume.

Abstract

We present SA-DSVN, a 3D convolutional architecture for voxel-level segmentation of structural defects in reinforced concrete using cosmic-ray muon tomography. Unlike conventional reconstruction methods (POCA, MLSD) that rely solely on muon scattering angles, our approach jointly processes scattering kinematics (9 channels) and secondary electromagnetic shower multiplicities (40 channels) through independent encoder streams fused via cross-attention. Training data were generated using Vega, a cloud-native Geant4 simulation framework, producing 4.5 million muon events across 900 volumes containing four defect types - honeycombing, shear fracture, corrosion voids, and delamination - embedded within a dense 7x7 rebar cage. A five-variant ablation study demonstrates that the shower multiplicity stream alone accounts for the majority of discriminative power, raising defect-mean Dice from 0.535 (scattering only) to 0.685 (shower only). On 60 independently simulated validation volumes, the model achieves 96.3% voxel accuracy, per-defect Dice scores of 0.59-0.81, and 100% volume-level detection sensitivity at 10 ms inference per volume. These results establish secondary shower multiplicity as a previously unexploited but highly effective feature for learned muon tomographic reconstruction.

Problem

• structural defect detection
• cosmic-ray muon tomography
• reinforced concrete
• secondary electromagnetic shower
• voxel-level segmentation

Innovation

• muon tomography
• secondary shower multiplicity
• dual-stream network
• voxel-level segmentation
• cross-attention fusion