D1.1 Minimal Model Selection for Calibrating a Hall-Stress- Temperature Multisensor System Using LASSO Regression
- Event
- SMSI 2020
-
(did not take place because of Covid-19 virus pandemic) - Band
- SMSI 2020 - Measurement Science
- Chapter
- D1 Model-based Measurement
- Author(s)
- M. Berger, O. Paul - University of Freiburg, IMTEK (Germany), S. Huber, C. Schott - Melexis Technologies SA., Bevaix (Switzerland)
- Pages
- 257 - 258
- DOI
- 10.5162/SMSI2020/D1.1
- ISBN
- 978-3-9819376-2-6
- Price
- free
Abstract
The proposed method takes advantage of LASSO regression to select a reduced-complexity polynomial model for calibrating nonlinear multisensor systems, while addressing the trade-off between higher accuracy and smaller calibration effort. The method is applied to compensate the nonlinear thermal and mechanical cross-sensitivities in a Hall-stress-temperature multisensor system. It enables to (i) reduce the calibration effort, measured by the number of model parameters, by a factor of 1.5 within the space of 4th-order polynomial models without compromising accuracy or to (ii) improve the accuracy by strategically including higher-order polynomial terms without increasing the number of model parameters.