Abstract

It is important to assure the reliability of a structural health monitoring system before interpreting the monitoring data for the detection of structural anomalies, Finding a malfunctioning component such as a sensor is an important step in that direction. Damage detection techniques in civil structures fall in the following two categories: data-driven and structural model-based. The data-driven methods provide a direct approach to damage assessment in a structure without creating any structural model (e.g. finite element model). Existence of damage and its location are interpreted by pattern matching of the data series of strain gauges, and temperature gauges at different time ranges. The objective of this study was to explore such methods, including the autoregressive exogenous (ARX) model, and based on that, develop new techniques to detect defective sensors. As a case study, the structural health monitoring data from the Portage Creek Bridge, located in the British Columbia, Canada was utilized to assess the conditions a set of sensors in of an instrumented pier, using methods developed based on the concepts of the sequential and binary search techniques. Continuous data sets of strain and temperature gauges were filtered and normalized. Defective sensors were detected by pattern matching of simulated and real data, using sensitivity analyses of the developed models.