Ultrafast Hydrogen Detection System Using Vertical Thermal Conduction Structure and Neural Network Prediction Algorithm Based on Sensor Response Process

Hydrogen detection plays a crucial role in various scenes of hydrogen energy such as hydrogen vehicles, hydrogen transportation and hydrogen storage. It is essential to develop a hydrogen detection system with ultrafast response times (<1 s) for the timely detection of hydrogen leaks. Here we report an ultrafast (0.4 s) hydrogen detection system based on a wafer-scale fabrication process. It consists of a low power (20.2 mW) hydrogen sensor based on vertical thermal conduction structure and a signal processing circuit introduced with a neural network prediction algorithm based on sensor response process. The fabricated sensor exhibits rapid response, wide detection range, and wide operating temperature, while showing good long-term stability and excellent selectivity. Meanwhile, the model significantly enhanced the detection speed by enabling hydrogen concentration prediction using only the initial 40 data points (sampling frequency of 100 Hz) from the sensor response before the sensor completes the entire response process. This work introduces a novel approach to achieve an ultrafast hydrogen detection system, which demonstrates significant application promise in the fields of low-power sensors and rapid gas detection.