Photoacoustic Multigas Sensor via CNN-Based Mode Division Multiplexing.

Area of Science:

• Photoacoustic spectroscopy
• Gas sensing technology
• Artificial intelligence in analytical chemistry

Background:

• Existing multigas detection methods like TDM, FDM, and MDM have limitations in simultaneous detection, cost, and universality.
• Mode division multiplexing (MDM) requires distinct gas signal waveforms and linear superposition, limiting its applicability.
• Accurate multigas detection is crucial for environmental monitoring and power equipment fault diagnosis.

Purpose of the Study:

• To develop a universal mode division multiplexing (MDM) solution for simultaneous, same-frequency multigas detection.
• To overcome the limitations of existing MDM techniques by accommodating diverse waveform characteristics and nonlinear superposition.
• To introduce a novel convolutional neural network-based MDM (CNN-MDM) approach for enhanced gas sensing.

Main Methods:

• Proposed a novel CNN-MDM technology assigning unique waveform characteristics to different gases during modulation.
• Utilized a convolutional neural network (CNN) to identify and separate mixed gas signals.
• Conducted experiments using SF6 decomposition components (H2S and CO) as detection targets.

Main Results:

• The CNN-MDM scheme demonstrated applicability to various waveform characteristics and both linear and nonlinear superposition.
• Achieved high linearity between separated signals and gas concentrations (R² values of 0.996 for CO, 0.995 for H2S).
• Established detection limits of 50 ppb for CO and 426 ppb for H2S.

Conclusions:

• The CNN-MDM scheme provides a universal framework for simultaneous same-frequency multigas detection.
• This approach is adaptable to photoacoustic systems with mixed modulation modes and tunable diode laser absorption spectroscopy (TDLAS).
• The developed method significantly advances multigas detection capabilities for environmental and industrial applications.