Exploring the limits and extending the borders of structural health monitoring

Abstract eng:
Structural health monitoring by vibration monitoring has been and still is a popular research topic. Especially the good results of state-of-the-art system identification methods applied to ambient vibration data has encouraged ongoing research in this field. Many - claimed to be new - methods are proposed which are often just variants of existing methods or which only work on noiseless numerically simulated data. The same applies to the many methods that have been developed to identify damage from observed changes in the dynamic signature (e.g. modal properties). This paper will present a number of recent innovations to extend the borders of what is realistically feasible with the current methods. One important evolution is the consideration of the uncertainty on the extracted modal parameters, which can be due to the finite number of data samples, coloured excitation, nonstationary structural behaviour, mass loading, inherent non-linearities… It is crucial to estimate this uncertainty because it propagates in the subsequent damage assessment. An example of a steel transmitter mast will illustrate the determination of uncertainties of eigenfrequencies, damping factors and mode shapes when using the stochastic subspace identification method. Important factors, decisive whether or not a particular (small) damage can be detected, are the choice of sensors (type, sensitivity, location, number) and the covered frequency interval. Often, ambient measurements only deliver quantitatively good information of the lower modes. A way to extend the frequency interval is the use of a (rather small) artificial excitation source, extra to the ambient disturbances. An example of a footbridge will highlight the advantages of combined stochastic-deterministic system identification.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket