Cointegration for the Removal of Environmental and Operational Effects Using a Single Sensor

Abstract

The cointegration algorithm has been developed recently as a powerful means of removing confounding influences from structural health monitoring (SHM) data. In this context, if one has two or more nonstationary time series, describing vibration data at a specific time period, one can cointegrate them in order to find a stationary linear combination of them. This means that it is essential to have datasets from different sensors in order to apply the specific technique, something not always feasible especially for on-line damage detection, due to the additional instrumentation costs and the amount of memory storage needed for these data. The current paper proposes a method that potentially allows cointegration to remove trends when a single sensor is present.