Edge analytics for anomaly detection in water networks by an Arduino101-LoRa based WSN.

This paper presents a novel distributed data analytic architecture, and corresponding algorithms that apply to infrastructure anomaly detection. The proposed method mainly focuses on smart water networks, demonstrating that the highest possible sensor rate analytic performs at on-edge nodes without requiring the whole date to send back to the server. This approach saves communication costs and lengthens the lifetime of the battery-powered nodes. A complex set of tasks is developed on a single-core Intel Curie processor, Arduino101 and the raw sensor data is compressed using a customized Lempel-Ziv compression algorithm tailored to resource-constrained embedded systems. The compression rate figures are then analyzed but only the compressed data which is associated with the anomalous condition is sent back to the server by means of a LoRa platform. The developed system is evaluated experimentally and the results verify the high resource utilization.