Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
INNOVATION IN COMPUTATIONAL STRUCTURES TECHNOLOGY
Edited by: B.H.V. Topping, G. Montero, R. Montenegro
Vibration Monitoring for Damage Assessment: Realistic Expectations
G. De Roeck
Research Unit Structural Mechanics, Department Civil Engineering, K.U. Leuven, Belgium
G. De Roeck, "Vibration Monitoring for Damage Assessment: Realistic Expectations", in B.H.V. Topping, G. Montero, R. Montenegro, (Editors), "Innovation in Computational Structures Technology", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 8, pp 159-179, 2006. doi:10.4203/csets.14.8
Keywords: vibration monitoring, system identification, modal properties, damage assessment, finite element updating.
This lecture gives an overview of the necessary conditions for a successful damage assessment by vibration measurements. With regard to the four levels of damage appraisal, attention will be concentrated on levels 1, 2 and 3.
First a critical overview is given about the advantages and disadvantages of the different excitation sources, that is, forced, ambient and impact excitation. The link is made to the ability to measure modal properties and subsequently to identify damage based on these quantities. Also choice of sensors and signal conditioning are briefly addressed.
For damage level 1 the advantages of ARX-models, which simulate the relation between the histories of eigenfrequencies and the histories of temperatures over classical regression techniques, is highlighted .
Many methods have been proposed for damage identification levels 2 and 3. The paper concentrates on the most versatile of these, finite element updating. At the expense of building a finite element (FE) model of the structure all acquired data can be used, that is, natural frequencies, scaled or unscaled mode shapes and modal strains. Moreover, unlike other methods, FE updating does not require a dense mesh of sensors. Potential damage is simulated in an FE model of the structure by adopting a parametric representation of this damage. A limited number of unknown updating parameters appear in this description, which should resemble, as closely as possible, physical damage. Best values of the -parameters can be found by minimising the differences between measured and calculated modal properties. Mathematically, a constrained optimization problem is solved. The objective function is defined as a sum of squared differences. The residual vector contains the differences in the identified modal data (and possibly some derived quantities), such as the natural frequencies, the mode shapes and the modal curvatures. If the objective function contains multiple local minima, the result of this optimization process might be dependent of the initial choice of -parameters. In this case, a global optimization algorithm should be used, like the newly developed coupled local minimisers (CLM) method . This method is a valuable alternative to other global optimization methods, for example, simulated annealing and genetic algorithms.
Two examples illustrate the use of FE updating for damage assessment. The first application concerns the well-known Z24 Bridge in Switzerland. The test consisted of a long term test for quantifying the degree of variance due to environmental influences, followed by short term loading failure tests to prove that changes in dynamic properties are large enough (that is, statistically relevant) and can be linked to damage in a particular part of the structure. From all the failure tests, settlement of an intermediate support is chosen as the application .
The second example is the Tilff Bridge. New in this project was the use of very accurate optical fibres for strain measurements in the critical sections. Very reliable modal strains and curvatures could be extracted by the system identification algorithm. In the damage identification by FE updating, natural frequencies, mode shapes and modal curvatures were used, resulting in a pattern of stiffness reduction which resembles reasonably well the visually observed fraction of broken or strongly corroded cables.
Finally, a series of conclusions are highlighted which can be considered as guidelines for good practice in vibration monitoring.
purchase the full-text of this chapter (price £20)