Deep Learning-Enhanced Hybrid Sensing: Enabling Gas Identification in Extreme Vacuum Environments

Accurately identifying gas types in the vacuum environment holds crucial significance in the fields of aerospace engineering and semiconductor manufacturing. However, current gas identification technologies, particularly in vacuum environments, are constrained by the limited detection sensitivity of conventional single sensors and their inability to effectively distinguish specific gases. To address vacuum gas identification challenges, an innovative strategy integrating hybrid Pirani/piezoresistive sensors with a Gramian Angular Field-Convolutional Neural Network (GAF-CNN) model is proposed in this work, achieving 99.14% classification accuracy and robust performance in data-scarce high-vacuum environments by converting multi-vacuum response time-series into discriminative two-dimensional (2D) images, overcoming single-sensor limitations and filling technical gaps in high-vacuum environments. Our strategy fills the gap existing in current technologies for gas identification within a vacuum environment and offers a reliable reference basis and technical support for future developments.