Type discrimination and concentration prediction towards ethanol using a machine learning-enhanced gas sensor array...

Area of Science:

• Materials Science
• Chemical Engineering
• Data Science

Background:

• Developing advanced gas sensors is crucial for environmental monitoring and safety.
• Controlling material morphology is key to tailoring sensor response.
• Machine learning offers powerful tools for analyzing complex sensor data.

Purpose of the Study:

• To synthesize zinc oxide (ZnO) with controllable hierarchical structures using a microwave-assisted method.
• To investigate the gas sensing performance of ZnO with varying morphologies.
• To apply machine learning algorithms for gas classification and concentration prediction.

Main Methods:

• Microwave-assisted synthesis of ZnO with tunable hierarchical structures via urea concentration.
• Gas sensing measurements of ZnO materials towards six different gases.
• Data analysis using Principal Component Analysis (PCA), Linear Ridge Classification (LRC), Support Vector Machine (SVM), Backpropagation (BP) neural network, and Extreme Learning Machine (ELM).

Main Results:

• ZnO morphology was effectively controlled by urea concentration in a surfactant-free system.
• SVM achieved >0.99 accuracy in distinguishing six confusing gases.
• ELM demonstrated high accuracy (R² > 0.98) in predicting ethanol concentration, even with saturated sensor responses.

Conclusions:

• Microwave-assisted synthesis provides a route to ZnO with morphology-dependent gas sensing properties.
• Machine learning, particularly SVM and ELM, significantly enhances gas sensor array performance for classification and prediction.
• This approach holds promise for developing sophisticated gas detection systems.