On-line drift compensation for continuous monitoring with arrays of cross-sensitive chemical sensors

Long-term application of chemical sensor arrays for continuous monitoring is challenging as a result drift. Drift correction often requires periodic recalibration, which may not be feasible sensors deeply embedded and deployed uninterrupted monitoring. In this paper, we propose multi-calibration ensemble approach to compensate drift in such applications. Our method uses past measurements ground-truth available, treats them “pseudo-calibration” samples. With these, it builds regression model predict the concentration target analytes by combining (1) current (2) history prior pseudo-calibration We evaluate efficacy proposed using three different techniques, partial least squares, extreme gradient boosting, neural networks, compare against two baselines: models that do use samples, state-of-the-art drift-correction technique. evaluated these on an experimental dataset from bioprocess control application, characterize function cross-sensitivity array amount through computer simulations. The outperforms both baselines dataset, under all simulation conditions, achieving significantly lower normalized root mean square errors prediction variables. These results hold used, indicates agnostic underlying model.