Accelerating Plasmonic Hydrogen Sensors for Inert Gas Environments by Transformer-Based Deep Learning

Area of Science:

• Materials Science
• Chemical Engineering
• Artificial Intelligence

Background:

• Safe large-scale hydrogen technology implementation requires rapid hydrogen leak detection.
• Current sensor solutions lack the necessary response times under relevant conditions.

Purpose of the Study:

• To develop a method for accelerating the response of optical plasmonic hydrogen sensors.
• To eliminate intrinsic pressure dependence in hydrogen sensing.
• To provide uncertainty quantification for safety-critical applications.

Main Methods:

• Development of a tailored long short-term transformer ensemble model for accelerated sensing (LEMAS).
• Testing the model on an optical plasmonic hydrogen sensor in an environment simulating large-scale hydrogen installations.
• Utilizing deep learning for predictive response acceleration.

Main Results:

• LEMAS accelerated sensor response by up to a factor of 40.
• The model eliminated the sensor's intrinsic pressure dependence.
• LEMAS provided uncertainty measures for predictions, crucial for safety.

Conclusions:

• Deep learning, specifically LEMAS, offers a viable solution for accelerating sensor response times.
• This approach is applicable beyond plasmonic hydrogen detection, advancing sensor technology.
• The method enhances safety for critical hydrogen applications.