Validation of long-term data from FBG temperature sensors

Monitoring temperature in structures is essential for understanding their thermal behavior and performing thermal compensation for any other type of sensors used for structural health monitoring (e.g. strain). This is particularly important for long-term monitoring applications. Nevertheless, a universal procedure for the validation of temperature data from a monitoring system has not been created and implemented. In this paper, preliminary research on two methods that aim to validate temperature readings from temperature sensors is presented. The methods differ in the type of data available for validation: (1) data from two independent monitoring systems are available, and (2) data from a monitoring system and a nearby weather tower are available. The methods use linear regression and confidence intervals on the linear regression model parameters to quantify annual changes in readings, which are then used to characterize long-term performance of the sensor such as changes in sensitivity or stability, or other unusual behaviors. The methods were applied to data collected over five years from fiber optic sensors based on Fiber Bragg Gratings (FBG) installed on Streicker Bridge on the Princeton University campus, and similar qualitative conclusions were derived from the two methods at instrumented locations where two independent monitoring systems are installed. This validation of the two methods was used as a confirmation that the second method, for which data is generally available, can be used in lieu of the first in cases where two independent monitoring systems do not exist. Based on this validation, the second method was independently used to test data at other locations on the bridge, where only one type of sensors is installed. The validation procedure showed and quantified minor changes in sensitivity and stability in measurements from some instrumented locations after the first two years.