Establishing threshold values for use in structural health monitoring

Abstract

Bridges are an essential component of our infrastructure systems. A significant portion of bridges around the world are reaching the end of their design service life. As a result, it is increasingly important to find ways to assess bridge condition in a fast, accurate, and economic way. One way to do this is to utilize real-time structural health monitoring (SHM) data, and to compare the measured data to established threshold values. The primary objectives of this study were (1) to develop a methodology for establishing threshold strain values that can be reliably used to detect significant abnormal response; (2) to correlate the threshold values to rating factors and truck loads so that bridge owners can easily understand the level of the threshold values; and (3) to develop a methodology for building Damage Indices (DIs) based on the threshold values that can be used to detect smaller or more gradual long-term abnormal response of a structure. To achieve these objectives, Statistical-based Damage Indices (DIs) and Decision Boundaries (DBs) were used to detect minor and significant simulated abnormal response of an instrumented cable-stayed bridge. Gaussian mixture models were used to fit recorded live-load strain measurements using an AICc criteria. Gaussian mixture models were then used to select threshold values based on 99% Upper Specific Limits (USLs). DIs were defined by analyzing the Outlier Ratios (ORs), and corresponding DBs were determined using t-distributions. The validity and sensitivity of the proposed methodology were demonstrated using simulated data that was created by perturbing actual collected SHM data. The results provide practical guidance for bridge owners when using SHM data in their decision-making process.