Background subtraction in pulsed photoacoustics through neural-network processing

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Pulsed photoacoustics measures gas concentrations via laser-induced sound waves.
• Window heating creates background noise, limiting detection sensitivity.
• Traditional methods struggle with background subtraction at low concentrations.

Purpose of the Study:

• To apply neural networks to pulsed photoacoustics for improved detection limits.
• To develop a method for subtracting window-heating background noise.
• To accurately measure trace ethylene concentrations.

Main Methods:

• Utilized neural-network processing for signal analysis.
• Trained the neural network with experimental data patterns.
• Applied the technique to ethylene detection using a TEA CO(2) laser.

Main Results:

• Successfully subtracted window-heating background noise.
• Reduced the detection limit by 80% compared to previous methods.
• Achieved accurate concentration measurements even at low levels.

Conclusions:

• Neural-network processing is effective for enhancing pulsed photoacoustics.
• The developed method significantly improves sensitivity and detection limits.
• This technique offers a robust solution for trace gas analysis.