Progressive damage of a steel frame at the bolted node detail
Source: BAM division Buildings and Structures
In the past years, the condition of transport infrastructure has deteriorated significantly due to natural aging processes and increasing loads. Recently, numerous examples of damaged engineering structures show that very regular condition inspections are required to ensure long-term safety and reliability. In this context, automated vibration-based damage detection methods can be used complementary. In these methods long-term vibration measurements are carried out on the structures during dynamically excitation either directly by traffic or by artificial sources. Since damage of the structure leads to changes in the dynamic system response, the measurement data are evaluated with respect to such changes. The stochastic subspace-based damage detection is a promising approach. It compares measurements of a current test condition with a reference model of the undamaged structure. A residual calculated from the covariances of the measured data quantifies the difference between the two system states and is evaluated by statistical testing. Natural variations in the residual affects the sensitivity and reliability of the test significantly. Such variations are due to measurement noise and changes in excitation properties and are referred to as statistical uncertainties.
So far, in the method it is assumed that the reference model itself has no uncertainties. However, in practical applications this assumption is not correct since the reference is usually purely data-based and estimated from the measured data. In the present article, it is shown how the uncertainties related to the reference can be considered in the evaluation of the residual. This significantly improves the method in terms of sensitivity and robustness, which can be confirmed by a numerical study and the evaluation of experimental data from a steel frame progressively damaged in the laboratory.
Statistical subspace-based damage detection with estimated reference
Eva Viefhues, M. Döhler, Falk Hille, L. Mevel
published in Mechanical Systems and Signal Processing, Vol. 164, page 108241
BAM division Buildings and Structures