Bitvector: Fault Tolerant Aggregation Scheme for Monitoring in Nuclear Power Plants

Industrial wireless sensor networks (IWSNs) have become popular, given the lower cost of installation than wired networks and their flexibility. WSNs are applied in monitoring, measurement, and control. Nuclear power plants (NPPs), in particular, have not been eager to adopt WSNs due to reliability, electromagnetic and radio frequency interference (EMI/RFI) as well as security concerns. On the other hand, the two main benefits from adding WSNs to NPPs are increased reliability from using robust wireless monitoring as a backup network for the primary wired system and reduce costs from installation and maintenance (in comparison to wired networks). We propose a new fault model and a fault tolerant in-network aggregation scheme, bitvector. First, in order to mimic the interference environment in NPP, we define a new fault model with time-dependent faults. We propose the fault duration concept, a time span during which the link error rate is fixed, such as when EMI/RFI appears between nodes. Therefore, the fault duration is an indicator of the network stability. Second, based on the fault model, there are some network links whose quality are not good during a fault duration interval. Bitvector estimates neighbor link quality dynamically and orders primary and backup parents dynamically according to link quality estimation. In addition, since data rate is low in NPPs, bitvector is able to use redundancy to improve accuracy. In particular, for some faults, bitvector defers fault detection and fault recovery by one sensing interval. Experiments show that bitvector significantly improves performance in root mean square (RMS) error and correct message rate, benefits of bitvector are even larger for environments with high fault rates.