Application of Uncertainty-Aware Sensor Fusion in Physical Sensor Networks

Modern industrial processes often collect redundant information from multiple sensors. It is of interest to leverage this information to form a more accurate or robust estimate of an observed quantity utilizing a sensor fusion operation. To be able to comply with quality requirements on the fused value, preservation of traceability is required. Moreover, the sensor fusion needs to be implemented within suitable digital architectures for Industrial Internet of Things (IIoT) environments. In a continuation of previous work, this is achieved by the usage of digital twins that represent the entities of two IIoT testbeds and the adoption of a method for uncertainty-aware homogeneous sensor fusion, which is presented in full detail. Metrological traceability of the fused value is established by propagating the measurement uncertainty of the input sensor according to metrological standards. The method is implemented as a modular service connected to an existing IIoT architecture. The flexibility of the method is shown by application to two different scenarios with only minimal adaption efforts. The fused sensor values are (as indicated by earlier work) robust to outliers and perform well in practical scenarios within the chosen IIoT architecture.