Fast and robust mixed gas identification and recognition using tree-based machine learning and sensor array response.

Area of Science:

• Chemical Engineering
• Environmental Science
• Food Safety
• Medical Diagnostics

Background:

• Accurate gas mixture identification is crucial across various scientific and industrial fields.
• Challenges include distinguishing between similar gases and mitigating sensor noise.
• Existing methods may lack efficiency or accuracy in complex mixtures.

Purpose of the Study:

• To develop an efficient and accurate method for identifying gas mixtures.
• To classify ethylene-methane and ethylene-carbon monoxide (CO) mixtures using a sensor array.
• To evaluate the performance of tree-based machine learning models for gas classification.

Main Methods:

• Utilized a four-sensor array for gas detection.
• Analyzed gas mixtures with concentrations up to 600 ppm (CO) and 300 ppm (methane).
• Employed sixteen features, including temporal dynamics and statistical metrics, with mean sensor response for noise reduction.
• Developed and compared Decision Tree (DT), Random Forest (RF), and Extra Trees (ET) models.

Main Results:

• The Extra Trees (ET) model demonstrated superior classification accuracy at 99.15%.
• Random Forest (RF) achieved 95.86% accuracy, and Decision Tree (DT) achieved 93.89%.
• Models were trained on a reduced dataset (60%) with significantly decreased prediction times.

Conclusions:

• The Extra Trees model offers a highly accurate and efficient solution for experimental gas identification.
• The methodology effectively mitigates noise and enhances classification robustness.
• This approach has broad applicability in chemical engineering, environmental monitoring, and safety applications.