7.3 - Run-In Behaviour of MOS Sensors in Temperature Cycled Operation

Event

17. Dresdner Sensor-Symposium 2024
2024-11-25 - 2024-11-27
Dresden

Band

Vortrag

Chapter

7. Smart Sensors

Author(s)

D. Arendes, J. Amann, A. Schütze, C. Bur - Universität des Saarlandes,Saarbrücken/D

Pages

84 - 89

DOI

10.5162/17dss2024/7.3

ISBN

978-3-910600-04-1

Price

free

Abstract

Metal oxide semiconductor (MOS) gas sensors are used in a wide spectrum of application [1],[2],[3]. Promising fields are especially the Indoor Air Quality (IAQ) [4], as well as the medicine field with drug monitoring [5]. In many cases an immediate evaluation out of the raw sensor data is impossible without a machine learning (ML) model. In order to train those ML models a proper individual calibration is necessary [6]. This calibration can be done by providing unique gas mixtures (UGMs) to the sensors in a gas mixing apparatus (GMA) [7]. When MOS sensors are operated for the first time, they show characteristic changes in their sensor responses, which do not correlate with the sensor environment. If this Run-In phase occurs during the calibration it will lead to invalid ML methods, for example incorrect predictions after the initial Run-In phase. Therefore, it is necessary to examine this phase to achieve stable and robust ML models. An easy method to compensate the Run-In phase is a pre-treatment before the actual calibration. This pretreatment can be done by operating the sensor in a designated environment with a specific duration.