Uncertainty-Aware Deep Ensembles for Robust and Reliable Chemical Sensor Arrays.

Area of Science:

• Materials Science
• Chemical Sensing
• Artificial Intelligence

Background:

• Selective detection of volatile sulfur compounds (VSCs) is crucial for breath analysis and environmental monitoring.
• Conventional sensors struggle with cross-reactivity to similar sulfur-containing gases, limiting practical selectivity.
• Developing advanced sensing platforms is necessary to overcome these limitations.

Purpose of the Study:

• To develop a highly selective electronic nose for detecting and quantifying specific sulfur-containing gases.
• To address the challenge of cross-reactivity in conventional chemiresistive sensors.
• To create a reliability-aware VSC sensing platform for real-world applications.

Main Methods:

• Fabrication of a 15-channel multi-sensor array using metal oxide nanofibers decorated with diverse metal catalysts (Pt, Pd, Ir, Co).
• Utilized an intense pulsed light photothermal process for catalyst anchoring.
• Developed a deep-ensemble learning framework trained on comprehensive datasets under varying environmental conditions.

Main Results:

• Achieved selective detection of hydrogen sulfide (H₂S), methyl mercaptan (CH₃SH), and dimethyl sulfide ((CH₃)₂S).
• The deep-ensemble model accurately classified gas species and quantified concentrations.
• The system demonstrated predictive uncertainty, enabling reliability-aware sensing.

Conclusions:

• The developed deep-ensemble-assisted electronic nose offers a promising solution for selective VSC detection.
• This technology has significant potential for future applications in halitosis-related breath monitoring and environmental sensing.
• The reliability-aware approach enhances the trustworthiness of the sensing platform.