Ai-assisted gas sensor system for greenhouse gas detection and monitoring

Overview

This solution is an intelligent gas detection and monitoring system designed to identify and discriminate greenhouse gases such as nitrous oxide (N2O), methane (CH4), and carbon dioxide (CO2). The system integrates acoustic wave sensors (AWS), including quartz crystal microbalances (QCM) coated with functionalized graphene oxide, alongside an array of commercial gas sensors. A multi-input convolutional neural network (CNN) analyzes sensor time-response data as spectral images, enabling accurate gas classification regardless of relative humidity conditions. Wi-Fi connectivity allows continuous, real-time data capture and cloud-based monitoring around the clock.

The technology builds on a validated volatile organic compound (VOC) detection platform that achieved approximately 95% accuracy in identifying acetone. By extending this proven approach to greenhouse gases, the system addresses critical needs in agriculture, environmental monitoring, and emissions management, where reliable gas discrimination under variable humidity remains a persistent challenge.

Technical specifications

Core technology components:

• Acoustic wave sensors (AWS): QCM devices coated with functionalized graphene oxide that exhibit measurable resonance frequency and impedance changes upon gas exposure
• Commercial gas sensor array: Multiple sensor types providing complementary output parameters for robust gas discrimination
• Deep learning engine: Multi-input CNN that converts sensor time-response data into scalogram images for spectral analysis and gas classification
• IoT integration: Wi-Fi-enabled, waterproof sensor hardware with companion applications for continuous 24/7 data logging to the cloud
• Gas sampling chamber: Custom-designed enclosure facilitating controlled gas exposure experiments with purified nitrogen carrier gas

Key features:

• Multi-parameter sensor fusion enabling gas type discrimination independent of humidity
• Scalogram-based image processing for enhanced pattern recognition accuracy
• Cloud-connected architecture supporting remote monitoring and data analytics
• Modular sensor array allowing adaptation to different target gases

Technology readiness level

The underlying sensor platform and deep learning methodology have been experimentally validated for VOC detection, achieving approximately 95% accuracy across 107 acetone and 80 ethanol samples. A structured three-phase validation program over 18 months will extend the technology to greenhouse gas applications: Phase 1 involves laboratory characterization using N2O, CH4, and CO2 gas cylinders with the existing experimental setup; Phase 2 captures real-time field data from agricultural soil environments through collaborations with industry partners in the Cameroon highlands and University Malaysia; Phase 3 focuses on validation and production of formal data sheets. The technology is currently at a readiness stage suitable for collaborative development and pilot deployment.