MOX-Compensation-SGP30

In this repository, we provide the data we have used to compensate sensitivity changes of duty-cycled MOX gas sensors with machine learning. Details about our work can be found in our paper:

@inproceedings{tug-eip2021sgp30, 
  author = {Gherman, Markus-Philipp and Cheng, Yun and Gomez, Andres and Saukh, Olga},
  title = {Compensating Altered Sensitivity of Duty-Cycled MOX Gas Sensors with Machine Learning},
  year = {2021},
}

Datasets

In the folder data, we provide 5 different datasets. Each dataset contains data from 3 SGP30 sensors running simultaneously. Device1 was operated continuously, device2 was operated on-demand with an on-time of 5 seconds and random off-times and device3 was duty-cycled with an on-time of 5 secons and a period of 5 minutes.

Name	Duration [h]	#samples [device1]	#samples [device3]	#samples [device2]
dataset_train_test	144	51485	1728	1700
dataset_test_04d	96	34325	1152	1099
dataset_test_18d	63	22768	764	780
dataset_test_22d	72	27133	911	949
dataset_test_53d	96	326164	1152	1944

Additionally, we have included plots visualizing each dataset, where we plot each sample from the continuous datasets and the last value of each sample from the duty-cycle and on-demand datasets.

Dataset Structure

The dataset csv files have the following header:

time,rel. Time,tVOC,CO2eq,Ethanol,H2,Temperature,Rel. Humidity

• time describes the absolute time-stamp
• rel. Time was used internally as a relative time-stamp
• tVOC describes the tVOC measurement
• CO2eq describes the CO2eq measurement
• Ethanol describes the Ethanol measurement
• H2 describes the H2 measurement
• Temperature describes the temperature measurement
• Rel. Humidity describes the relative humidity measurement

The samples from the datasets can be read as followed:

Continuous (device1) datasets:

Each sample is terminated with a row starting with an end identifier. The row before describes the measurement values. The measurement values are duplicated in the end row for convenience.

Duty-cycle (device3) and on-demand (device2) datasets:

Each sample is terminated with a row starting with an end identifier. The rows before describe the measurements taken during the on-time of the device with 36Hz. In these modes, there are no tVOC and no CO2eq measurements available. Temperature and relative humidity can be extracted from the end row.

Contact

Feel free to contact us for any questions on the data or to present your findings on this topic to us. We are curious to hear about your work!

• philipp.gherman@gmail.com
• chengyu@ethz.ch
• andres.gomez@unisg.ch
• saukh@tugraz.at