Systematic method for the validation of long-term temperature measurements

Abstract

Structural health monitoring (SHM) is the process of collecting and analyzing measurements of various structural and environmental parameters on a structure for the purpose of formulating conclusions on the performance and condition of the structure. Accurate long-term temperature data is critical for SHM applications as it is often used to compensate other measurements (e.g., strain), or to understand the thermal behavior of the structure. Despite the need for accurate long-term temperature data, there are currently no validation methods to ensure the accuracy of collected data. This paper researches and presents a novel method for the validation of long-term temperature measurements from any type of sensors. The method relies on modeling the dependence of temperature measurements inside a structure on the ambient temperature measurements collected from a reliable nearby weather tower. The model is then used to predict future measurements and assess whether or not future measurements conform to predictions. The paper presents both the model selection process, as well as the sensor malfunction detection process. To illustrate and validate the method, it is applied to data from a monitoring system installed on a real structure, Streicker Bridge on the Princeton University campus. Application of the method to data collected from about forty sensors over five years showed the potential of the method to categorize normal sensor function, as well as characterize sensor defect and minor drift.