MP2 - Bayesian Inference for Reliable Gas Sensing with Metal-Oxide Sensors
- Event
- EUROSENSORS XXXVII
2025-09-07 - 2025-09-10
Wroclaw - Band
- Poster I
- Chapter
- 3D Printed Sensors and AI for Sensors
- Author(s)
- S. Pültz - Saarland University, Saarbrucken (Germany)
- Pages
- 206 - 207
- DOI
- 10.5162/EUROSENSORS2025/MP2
- ISBN
- 978-3-910600-07-2
- Price
- free
Abstract
In this study, a probabilistic regression approach for the quantification and separation of epistemic and aleatoric uncertainties in gas measurement with temperature-modulated MOS sensors is presented. The uncertainty of feature extraction is analyzed using Monte Carlo simulations. The subsequent re-gression modeling is based on Partial Least Squares Regression (PLSR) combined with Hamiltonian Monte Carlo (HMC) to capture uncertainties in the parameter estimator. The evaluation shows that most of the uncertainty is due to feature extraction, while epistemic uncertainties remain comparatively small.