Keywords: damage detection, modal analysis, dynamic tests, monitoring, bridges.

The dynamic characteristics of a structure can be used as a sensitive indicator of its physical integrity. Changes in the structural members due to damage and deterioration alter dynamic characteristics. The possibility of using the natural frequencies and mode shapes in identifying the existence of damage has been the subject of extensive researches in recent years with various degrees of success.

The shift in the natural frequencies caused by damage and changes in the structural properties is the most commonly used damage detection indicator. However, this shift in the individual modes cannot provide spatial information about structural changes. Alternatively, examining simultaneously changes in several natural frequencies through a comprehensive screening process may identify the damage location. In this paper, the use of measured changes in natural frequencies for locating damage in structures is investigated. A data base for all possible damage levels and damage locations can be numerically developed for the monitored structure before being set in service. Such data base contains the dynamic characteristics of the undamaged structure tuned to the actual conditions in service as well as several possible damage levels and locations. The sensitivity of the various mode frequencies to damage level and location are identified and included in the data base. The measured frequencies during monitoring process are continuously correlated to the formed database where damage level and location may be identified.

The proposed method is based on the fact that the global stiffness matrix of a structure is in effect a summation of stiffness matrices of its elements. Each element, through its location and characteristics, uniquely affects the global stiffness matrix of the structure. Eliminating an element or changing its physical or mechanical characteristics would affect the overall dynamic behavior of the structure in a certain way. This change affects all modes of vibration with different levels according to the location of the damaged member, the level of damage, and the shape of the mode.

The proposed method depends on forming a database containing the natural frequencies of selected modes of the monitored structure calculated for different damage patterns. The changes in the natural frequencies due to damage in some elements are calculated and arranged in a 3-D matrix form used as a database. This database of changes in natural frequencies is formed using modal analysis of a finely tuned finite element model of the real structure. It is used to detect the damaged element, and hence the damage location, when changes occur in the dynamic behavior of the real structure. The detection process is carried out through a screening technique comparing changes in the measured frequencies to those forming the database. The natural frequencies of the monitored structure can be regularly determined using modal testing techniques and compared to reference values determined for the undamaged structures.

The applicability of the proposed method is verified experimentally using two groups of models. The first group contains six steel cantilever beam models, while the second group contains three grid structures. One of the specimens in each group is used as the reference specimen without damage while the others represent different damage patterns for different locations and levels. Damage was introduced to the specimen by reducing the cross section at selected positions. Only three modes were considered in the cantilever specimens and four modes for the grid specimens. Two methods are proposed in this research as search engines to identify the damaged element causing the detected changes in natural frequencies. The first method, the filtration method, relies on comparing the changes in the measured frequencies of the real structure to those contained in the data base in a sequential manner. The second method uses a frequency modal assurance criteria F-MAC factor calculated for all possible damaged cases to indicate the one with the highest correlation.

The proposed filtration method was able to locate damage and identify its level particularly for those locations with high sensitivity to inflict changes in the natural frequency. For less sensitive elements of the models, damage location could be approximately located by identifying limited number of elements most likely causing the observed changes in frequency. Those elements can by visually inspected to precisely locate the damage. The use of F-MAC factor method could also identify the location of damage in most specimens. The combined use of both the filtration method and the F-MAC method resulted in identifying damage location in all considered specimens. The proposed method could also reasonably identify damage levels in the considered specimens. The method is easy to use for field measurements sense it relies of accurate measurement of the natural frequencies which can be achieved using only one sensor. The accuracy of the method can be improved by using finer mesh for the numerical finite element model, incorporating more levels of damage, and considering more modes of vibration.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £135 +P&P)