A Double-layered Ultraviolet Flame Sensor Design for High-Resolution Spectral Discrimination Utilizing Machine...

Area of Science:

• Materials Science
• Spectroscopy
• Sensor Technology

Background:

• Ultraviolet (UV) spectral signatures are crucial for identifying flame origins in fire monitoring.
• Existing UV flame sensors with spectral discrimination are often complex and expensive due to optical and material design limitations.

Purpose of the Study:

• To develop a cost-effective, multi-channel semiconductor UV flame sensor.
• To enable high-resolution spectral discrimination for practical flame monitoring applications.

Main Methods:

• Fabrication of a multi-channel sensor array using an Al2O3-TiO2 double-layered nanostructure via sequential anodization.
• Design tuning of nanostructure geometric parameters for distinct wavelength-dependent responses (280-320 nm).
• Application of convolutional neural networks (CNN) for single-peak spectral discrimination and U-Net for dual-peak spectral reconstruction.

Main Results:

• The Al2O3-TiO2 nanostructure sensor array demonstrated distinct wavelength-dependent responses within the 280-320 nm range.
• Simulations indicated high responsivity and internal quantum efficiencies for the proposed structure.
• CNN achieved 91% classification accuracy for single-peak discrimination, while U-Net yielded high-fidelity spectral reconstruction with a 7.4% average peak-intensity deviation ratio.

Conclusions:

• The proposed Al2O3-TiO2 nanostructure sensor combined with machine learning offers a compact and effective UV flame sensing approach.
• This technology enables high-resolution spectral discrimination, overcoming limitations of current UV flame detection systems.
• The developed sensor has significant potential for practical and advanced fire monitoring scenarios.