4.5 - Portable NIR Spectroscopy for Reliable Characterization...
- Event
- iCCC2026 - iCampus Cottbus Conference
2026-05-05 - 2026-05-07
Cottbus - Band
- Vorträge
- Chapter
- Energiewirtschaft und Material & Prozesstechnologie
- Author(s)
- M. Nobari, H. Engelke, I. Jabłoński - Fraunhofer IPMS, Cottbus, F. Eckert, M. Nobari, I. Jabłoński - BTU Cottbus-Senftenberg, Cottbus
- Pages
- 76 - 79
- DOI
- 10.5162/iCCC2026/4.5
- ISBN
- 978-3-910600-10-2
- Price
- free
Abstract
Portable near-infrared (NIR) spectroscopy offers a powerful non-contact approach for characterizing high-temperature radiating surfaces, enabling in-situ temperature diagnostics without intrusive sensors. However, achieving measurement reliability under intense thermal emission is challenged by detector nonlinearity, emissivity variation, and calibration instability. This study introduces a portable NIR framework integrating a blackbody calibration device (BBCD) with range. We systematiadvanced data-driven modelling to improve accuracy and reproducibility across the 600–1200 cally compare statistical regression methods with AI-based architectures-including fully connected, convolutional, residual, and transformer networks for temperature prediction from spectral data. Results show that AI models, particularly transformer and ResNet architectures, outperform statistical baselines in both accuracy and generalization, especially under irregular sampling and noisy conditions. The findings highlight the importance of BBCD assisted calibration and robust preprocessing in achieving traceable, field-ready temperature measurements, advancing the application of portable NIR spectroscopy in industrial and scientific diagnostics.