Drift-Like Effect Compensation for the Cyclic Temperature Operation of Semiconductor Gas Sensor
Shouwen Zhang, Zhenyu Yuan, Lu Kong, Zhenxin Wu, Fanli Meng
Abstract
Semiconductor gas sensors utilizing cyclic temperature operation (CTO) have been extensively employed in the gas detection and sample collection. However, a drift-like effect is always present in CTO, observed as an increasement in response amplitude over time during consecutive short-term measurements. This effect can significantly impact sensor calibration and field identification accuracy. To address this issue, a cost-effective drift compensation method based on the temporal data decomposition is developed in this article. The raw response data is decomposed into trend, periodic, and residual components using an additive model of temporal data. The drift-compensated data is then obtained by removing the trend component from the raw response data. The application of this drift compensation method can greatly enhance the accuracy of concentration prediction models, reducing the mean absolute error by 40.21% on average and increasing the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value by 0.0179 averagely compared to the scenarios where no drift compensation was implemented for benzene concentration prediction in this study. In summary, this article presents a cost-effective compensate strategy for drift-like effect in the CTO by performing adaptive correction.