Transient-Response-Based Steady-State Prediction Method for Hydrogen Sensors

Abstract

Hydrogen sensors used in safety monitoring are commonly limited by slow response speed and the ex-post nature of conventional T90 evaluation. To address this limitation, this study proposes a transient-response-based steady-state prediction method for catalytic combustion hydrogen sensors and develops a sensing-computing integrated detection framework. The proposed method utilizes the dynamic information contained in the early transient response to predict the final steady-state output before complete signal stabilization. A first-order-plus-dead-time (FOPDT) model is employed to describe the dominant dynamic characteristics of the sensor. For each response event, only the early-stage transient data are used to estimate the steady-state output increment, time constant, and lag time through nonlinear least-squares fitting. The predicted steady-state voltage is then converted into hydrogen concentration using the established steady-state voltage–concentration relationship. Experiments were conducted using six catalytic combustion hydrogen sensors under step hydrogen concentration changes ranging from 2000 ppm to 30,000 ppm. The results demonstrate that the proposed method achieves hydrogen concentration prediction within the specified error requirements in 5–9 s, reducing the effective measurement time by approximately 70% compared with conventional stabilization-based measurement methods. The prediction accuracy satisfies the error requirements specified in ISO 26142:2010. The proposed approach provides an effective strategy for accelerating hydrogen concentration estimation by exploiting transient response characteristics, offering potential applications in real-time hydrogen safety monitoring and intelligent sensing systems.