Predictive Data Reduction in Wireless Sensor Networks Using Selective Filtering for Engine Monitoring

In a wireless sensor network, transmissions consume a large portion of a node’s energy budget. Data reduction is generally acknowledged as an effective means to reduce the number of network transmissions, thereby increasing the overall network lifetime. This paper builds on the Spanish Inquisition Protocol (SIP), to further reduce transmissions in a single-hop wireless sensor system aimed at a gas turbine engine exhaust gas temperature (EGT) monitoring application. A new method for Selective Filtering of sensed data based on state identification has been devised, using a skewed double exponentially weighted moving average filter for accurate state predictions. Low transmission rates are achieved even when significant temperature step changes occur. A simulator was implemented to generate flight temperature profiles similar to those encountered in real-life, which enabled tuning and evaluation of the algorithm. The results, summarised over 280 simulated flights of variable duration (from approximately 58 min to 14 h), show an average reduction in the number of transmissions by 95, 99.8, and 91 % in the take-off, cruise, and landing phases, respectively, compared to transmissions encountered by a sense-and-send system sampling at the same rate. The algorithm generates an average error of 0. 11 ± 0. 04 ∘C over a 927 ∘C range.