Damage Characterization of the Z24 Bridge by Transfer Function Pole Migration

In this study, a novel approach to the characterization of structural damage in civil structures is presented. The proposed method is scalable, automated, and well suited to implementation within a low power, low cost wireless sensing network. Structural damage often results in subtle changes to structural stiffness and damping properties that are manifested by changes in the location of transfer function characteristic equation roots (poles) upon the complex plane. Using structural response time-history data collected from an instrumented structure, transfer function poles can be estimated using traditional system identification methods. Comparing the location of poles corresponding to the structure in an unknown structural state to those of the undamaged structure, damage can be accurately identified. Pattern classification methods that have been increasingly popular in recent years for damage detection are leveraged to quantify damage existence, location, and severity. To validate the proposed methodology, response data from the Z24 Bridge is adopted as a test case. Utilization of forced vibration data corresponding to the bridge in various controlled damaged states allow the damage detection methods to be tested; in this study, they are shown to be capable of accurate and robust damage assessment in such a complex civil structure. TABLE OF NOMENCLATURE: y(k), Y(z) Output discrete time history value, Z-transformed output. u(k), U(z) Input discrete time history value, Z-transformed input. e(k) Output estimation error. ai, bi Output observation weight, Input observation weight. H(z) Transfer function. DI Damage Index. i Pole cluster mean separation distance. i Pole cluster standard deviation. w Perceptron weighting vector. bias Perceptron bias term. c Perceptron correction factor.  Perceptron classification threshold.  Perceptron learning rate factor.