Least Squares Neural Network-Based Wireless E-Nose System Using an SnO₂ Sensor Array.

Area of Science:

• Sensor technology
• Artificial intelligence
• Chemical sensing

Background:

• Electronic noses (E-noses) are crucial for detecting hazardous gases.
• Tin dioxide (SnO₂) sensors show promise but suffer from cross-sensitivity issues.
• Accurate detection of odorless gases like methane (CH₄) and carbon monoxide (CO) remains a challenge.

Purpose of the Study:

• To develop a robust E-nose system for detecting and quantifying CH₄ and CO.
• To address and mitigate sensor cross-sensitivity in SnO₂-based E-nose arrays.
• To integrate artificial intelligence and regression techniques for improved gas sensing performance.

Main Methods:

• Utilized an array of SnO₂ gas sensors.
• Developed a classifier using a feedforward artificial neural network (ANN) for gas recognition.
• Employed least squares regression (LSR) for gas concentration estimation by minimizing mean square error.
• Implemented the developed algorithms on a remote microcontroller for a wireless system.

Main Results:

• Achieved a high recognition probability of 98.7% for different gases.
• Obtained accurate gas concentration estimations with 95.5% accuracy for CH₄ and 94.4% for CO.
• Successfully deployed the classifier and estimator parameters in a wireless E-nose system.

Conclusions:

• The combined ANN and LSR approach effectively overcomes sensor cross-sensitivity in E-nose systems.
• The developed wireless E-nose demonstrates high accuracy and reliability for detecting and quantifying hazardous gases.
• This technology has significant potential for real-time monitoring of dangerous, odorless gases.