Accurate and real-time gas detection is crucial for applications ranging from environmental monitoring to industrial processes. Traditional methods are often limited by low accuracy, slow response times, and high costs. This project introduces a scalable machine learning fusion system that integrates sensor fusion techniques to enhance detection performance. With encoder-decoder architectures and a decision fusion model, our approach significantly improves the accuracy of carbon dioxide sensing, achieving a mean absolute percentage error (MAPE) of 2.97% while reducing response and recovery times from approximately 9 minutes to 2 minutes. This system consolidates the strengths of multiple sensors into a single sensor. The system architecture allows for straightforward adaptation to other gas detection systems and other sensing applications involving time-series sensing data.
Project is currently funded by: Member Fees