Long-Term Drift Behavior of Electronic Nose

The goal of the dataset is to capture the long-term drift behavior of a commercial electronic nose based on 62 metal-oxide gas sensors. We measured three chemical analytes at different concentrations over 12 months: diacetyl (0.1 ppm and 1 ppm), 2-phenylethanol (200 ppm and 1000 ppm), and ethanol, resulting in 700 time series recordings for each sensor across 40 days. Each single measurement includes a three-stage cycle: 1) baseline air for 5 min, 2) sample air exposure for 5 min, and 3) recovery stage for 5 min. The cycle stage information is given in the dataset as well. The dataset includes the readings from all 62 gas sensors (‘R’), from a temperature sensor (‘T01’) and from one humidity sensor (‘H01’). All data files are provided in .csv format with a single space character as separator for the values and a period as decimal separator. The main directory contains different folders containing the raw data for every measurement day and additionally one file containing the processed data with pre-extracted features calculated for the all measurements. Within each directory, the individual files are named according to the substance and concentration measured. This dataset can be used, for example, to evaluate and develop methods for feature extraction, sensor drift detection and compensation. Especially due to the long period of time and the controlled experimental conditions, it is valuable for studying drift phenomena. Please refer to the following publication for more details about the dataset and the experimental conditions: https://doi.org/10.1038/s41597-025-05993-8