Fluid Classification via the Dual Functionality of Moisture-Enabled Electricity Generation Enhanced by Deep Learning.

Area of Science:

• Materials Science and Engineering
• Nanotechnology
• Analytical Chemistry

Background:

• Miniaturized sensors are crucial for fluid classification across various applications.
• Moisture-enabled electricity generation (MEG) devices offer potential for self-powered sensing.
• Integrating advanced materials with microfluidics enhances sensing capabilities.

Purpose of the Study:

• To develop a novel intelligent self-sustained sensing approach by integrating MEG with deep learning in microfluidics.
• To create a dual-purpose device for both power generation and fluid detection.
• To demonstrate high-accuracy, rapid classification of different fluid samples.

Main Methods:

• Fabrication of a multilayer MEG device using nonwoven fabrics, carbon nanotubes, PVA gels, and liquid alloy.
• Utilizing a composite microfluidic design with hydrophobic channels and hydrophilic substrates for flow control.
• Synchronous measurement of voltage, current, and resistance signals, followed by deep learning (WDCNN) analysis.

Main Results:

• The MEG device provided stable power output for over 6 hours.
• Distinct electrical "fingerprints" (V/C/R signals) were generated for different fluids.
• A wide-kernel deep convolutional neural network (WDCNN) achieved 100% accuracy in classifying water and fruit juices within 15 seconds.

Conclusions:

• The integration of MEG, microfluidics, and deep learning presents a new paradigm for sustainable intelligent environmental perception.
• This approach offers innovative prospects for analytical science and the development of smart instruments.
• The developed system demonstrates a highly efficient and accurate method for real-time fluid analysis.