Answers are grounded in this paper and cite their sources.
Detachment of ceramic tiles from high-rise façades threatens public safety, while manual tapping is hazardous and thermography is limited under adverse conditions. This paper presents a wall-climbing robot for hollow-tile inspection using contact acoustic sensing under strong aerodynamic ego-noise. A two-stage framework combines a DCCRN-based enhancement backbone with task-specific loss and a ResNet18-CWT classifier. Enhancement achieved an SI-SNR gain of 10.58 ± 3.24 dB and a peak ratio of 1.06 ± 0.06; both classification accuracy and F1-score reached 99.12 ± 0.17% over five runs. A wide-swath linear scanning mechanism enables multiple tapping points per adhesion cycle. A spatial damage-mapping pipeline projects each diagnosis onto a reconstructed 2D façade map, with mean localization error 7.6 mm and maximum 12.5 mm in laboratory validation. Full-scale laboratory tests achieved 93.2% accuracy and 90.7% F1, while field trials achieved 97.2% and 93.0% under real-world noise.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: