Minimum diagnosable damage and optimal sensor placement for structural health monitoring

Abstract

Structural health monitoring (SHM) can extend the operation of bridges beyond their original life span, increase the safety between scheduled inspections, and allow for a prioritized inspection after extreme events. One of the major challenges is to assess which damages can or cannot be diagnosed (i.e., detected or localized), which is essential to evaluate the value of a SHM system before it is installed, and to optimize the sensor placement accordingly. This work develops a framework to predict the minimum damage, i.e., the minimum change in local structural design parameters that can reliably be detected based on changes in global damage-sensitive features. The diagnosis is considered reliable if the probability of false alarms is low and the probability of detection is high. Equivalently, a damage is detectable if it is significant under consideration of typical uncertainties related to ambient excitation and measurement noise and empirical safety thresholds. The approach requires vibration data from the undamaged structure in combination with a numerical model, and is universally applicable to a wide range of structures and damage-sensitive features. Secondly, a method is proposed to analyze the minimum localizable damage. The results show that optimal localizability is a compromise between high localization resolution, high detectability, and few false localization alarms. Thirdly, a sensor placement strategy is devised that takes as input the desired minimum diagnosable damage and optimizes the sensor layout and the number of sensors accordingly. The method allows one to focus the global damage diagnosis on local structural components. Ultimately, the monitoring of prestressing forces and support displacements is incorporated into the diagnostic framework, so that they can be analyzed and distinguished from changes in material properties or cross-sectional values. Besides the performance evaluation, the framework is suitable for quality control of existing instrumentation on real structures. Therefore, self-validation strategies are implemented to verify the input parameters, to validate the theoretical assumptions, and to check its effectiveness based on non-invasive tests using extra masses. The proof of concept studies based on a laboratory steel beam and a cable-stayed bridge show promising results regarding the practical application of the theoretical contributions.