Impact of Wind Load Characteristics on Computed Bridge Stay-Cable Forces Used for Bridge Health Monitoring

Monitoring the magnitude of stay-cable forces over time can play a key role in assessing the integrity and safety of a cable-stayed bridge. A variety of methodologies have been proposed to estimate stay-cable forces, with vibration-based methodologies being most widely applied in real-world applications due to their simplicity, speed, and economy. In most cases, periodically performed pluck tests are used to excite a cable, and accelerometers mounted on the cable are used to record its vibrational characteristics. Taut string theory (TST) is then applied to compute the cable forces. Because a pluck test is required to excite the cable, cable force measurements conducted in this manner are only done periodically. In this paper, we are interested in accurately computing stay-cable tensile forces on a more continuous basis using wind-induced cable vibration data collected by permanently installed structural health monitoring (SHM) systems. The specific objectives of this study are to (1) compute bridge stay-cable tensile forces over time using SHM data and use the results to establish confidence intervals (CIs) that can be applied to future cable force computations to detect "abnormal " values that might signal changes in bridge condition, and (2) define the wind characteristics (speed, gust, and direction) that lead to "acceptable " vibration records from which accurate cable forces can be computed. By fully understanding the wind load effects on cable vibrations, bridge owners will better know under what wind conditions to collect cable accelerations data. By knowing which data is useful, the amount of data collected and analyzed can be minimized, and the accuracy of the force calculations can be increased. The impact of wind load characteristics on the accuracy of cable force calculations is a key component of this work. To validate the proposed methodologies, data collected by an SHM system on the Indian River Inlet Bridge, a cable-stayed bridge in southern Delaware, was used. More specifically, the accelerations of eight cables recorded during five "storms " and one "regular " two-week ambient event (all occurring from 2014 and 2021), were evaluated. During the events, ambient weather conditions (i.e., mean wind speed, wind gust, wind direction, temperature, and perception) were collected in 5-min intervals. Cable accelerations were processed using power spectrum density to identify natural frequencies of the cables. TST was then applied to compute cable tensile forces. By doing this over an extended time window, 99% CIs for expected values of stay-cable forces were computed. These CIs can be used to monitor computed stay-cable forces over time as a means for assessing cable and overall bridge condition. While computing cable forces, it was found that some acceleration records did not yield accurate natural frequencies. Decision tree (DT) models were built to identify the wind load characteristics that correspond to cable vibrations that do yield accurate frequencies. The DT methodology was applied to establish a common set of wind load conditions for all cables for accurate frequency and cable force calculations. By establishing a set of wind load conditions for a specific bridge, those monitoring the bridge can minimize the amount of data collected, while improving the accuracy of the stay-cable force calculations.