A low-cost quantitative hydrogen concentration detection system

Abstract Metal oxide semiconductor (MOS) sensor arrays have traditionally been employed primarily for qualitative gas identification rather than for precise quantitative concentration detection. In this study, we present the development of an innovative and cost-effective hydrogen concentration detection system, which is priced at less than 15 USD. This system utilizes an array of 10 MOS sensors to accurately measure hydrogen concentrations. To enhance the performance of our detection system, we implemented Principal Component Analysis (PCA) for effective dimensionality reduction, followed by training a Deep Neural Network (DNN) to create a robust predictive model. The trained model was successfully deployed onto an embedded system using STM32 CubeMX.AI tools, facilitating real-time detection capabilities. Our comprehensive dataset, which includes voltage readings collected under varying temperature and humidity conditions, was instrumental in achieving an impressive accuracy rate of over 95% within a hydrogen concentration range of 100–300 ppm during rigorous testing. The results of our study underscore the feasibility and effectiveness of utilizing a MOS sensor array combined with advanced data processing techniques for quantitative concentration detection. This approach not only demonstrates high precision but also offers a promising solution for practical applications in hydrogen concentration monitoring.