D. Searson and M. Tham (UK)
estimation, fault detection, sensor validation.
This article provides an overview of a two-stage algorithm for robustly filtering sensor data in real time. The filtering method builds on previous work carried out in this area and is based on heuristic arguments rather than traditional filtering theory: it does not use explicit models or restrictive mathematical assumptions. The method was assessed by application to a number of historical sensor signals from a commercial water treatment plant. It is concluded that the method can successfully reject signal artefacts that persist for more than one sample, and reconstruct the underlying signal quite accurately for a limited number of further samples. Some limitations of the method are also discussed.
Important Links:
Go Back
