A Novel Deep Neural Network-Based Framework for Subsurface Analysis in Non-Stationary Thermal Wave Imaging

Enhanced depth resolution and deeper depth scanning are the most pivotal challenges in thermal wave imaging-based non-destructive evaluation. In the recent past, sweep frequency stimulation and traditional signal processing have been successful in analyzing non-linearity, which has resulted in a hig...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Sensing and imaging Ročník 26; číslo 1; s. 117
Hlavní autori: Swapna, M. N., Suman, M., Prasanthi, Y. N., Ghali, V. S., Mulaveesala, R.
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: New York Springer US 05.09.2025
Springer Nature B.V
Predmet:
Artificial neural networks
Data compression
Deep learning
Defects
Electrical Engineering
Engineering
Frequency modulation
Imaging
Low carbon steels
Microwaves
Neural networks
Nondestructive testing
Orthogonality
Radiology
RF and Optical Engineering
Signal processing
Signal to noise ratio
Sweep frequency
Thermal imaging
Thermal response
Thermal wave imaging
Thermography
Encoded features
Constrained autoencoder
Weights with unit norm length
Orthogonal weights
ISSN:1557-2072, 1557-2064, 1557-2072
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:Enhanced depth resolution and deeper depth scanning are the most pivotal challenges in thermal wave imaging-based non-destructive evaluation. In the recent past, sweep frequency stimulation and traditional signal processing have been successful in analyzing non-linearity, which has resulted in a high-dimensional, noise-embedded thermal response and restricted detectability. It is difficult to extract high-level features from high-dimensional data that are meant to improve fault signatures in a lower-dimensional space by encoding. This work proposes a constrained shallow convolutional autoencoder to address these shortcomings of conventional autoencoders. This approach enforces correlated encoded data, orthogonality between encoder and decoder weights, and unit norm length of these weights to boost performance. By implementing the constraints, the proposed autoencoder aims to deliver superior performance and enhance defect visibility. Experiments conducted on mild steel specimens using quadratic frequency modulated thermal wave imaging validate the applicability of this modality, and the results clearly demonstrate an edge over the existing state-of-the-art modalities as quantified in terms of defect signal-to-noise ratio (SNR).
Bibliografia:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1557-2072
1557-2064
1557-2072
DOI:10.1007/s11220-025-00640-3