Tracking Long- Term Drift in Wireless Sensor Networks Using Long - Term Memory and Data Fusion

Abstract

Wireless sensor networks are used to guarantee optimal and safe operation of difficult-to-reach industrial and civil structures. Due to their exposed mounting location, the sensors experience severe environmental influences. This leads to erosion and ageing of components which result in drifting standard values. Therefore, online tracking of standard values is paramount to guarantee optimal performance. An algorithm has been developed by fusing measurement data across several sensors during their steady-state. The system is able to track drifting standard values by using long-term memory. Simulations show that the algorithm successfully differentiates between measured data and drift of standard values. Simulations have been verified by applying the algorithm to real-world data of several months. Results show that the algorithm is able to track the drift of standard values, thereby maintaining full sensitivity.