Large scale sensor networks form an important part of the Industrial Internet of Things. To maintain the operation of such networks over time, quality of the sensor readings needs to be ensured. This leads to the development of a metrological traceable in-situ calibration method based on a Bayesian framework which leverages local sensor redundancy. Furthermore, automation of such in-situ calibration tasks is a key feature. To this end, an extension of existing sensor-related ontologies is proposed to cover relevant metrological terms. Sensor self-descriptions based on these knowledge representations allow for support of in-situ calibration by finding suitable reference sensors and initialization the mathematical method presented here. The mathematical method is evaluated in simulation studies against a state of the art in-situ calibration. The evaluation results show good estimation performance in cases of time-depending input signals or sensors of comparable uncertainty levels, but also reveal higher computational costs. The developed ontologies are evaluated by a corpus comparison, ontology metrics as well as logical checks of the taxonomic backbone and indicate a good agreement with existing ontology quality standards.
