Abstract:
Damage identification techniques have been widely investigated and used for structural
damage evaluation. Many researchers have shown good results in detecting, locating and quantifying
damage in structures using various damage identification algorithms and methods. One of the popular
and promising damage identification methods is modified damage index (MDI) which utilises modal
strain energy (MSE) and a statistical approach. However, when using this damage identification
method, numerical techniques used in realising the damage detection algorithm plays an important role
for the final outcome. The use of different techniques in detection of damage has not been widely
investigated. In this paper, a finite element (FE) model of a timber beam was developed as a test
strUcture. Modal responses of the test structure were generated using a FE software package. The
damage index algorithm, utilising modal strain energy as its damage indicator, was computed. In the
computation process, different numerical techniques at different stages were utilised to process the
data. Since in practice, the number of modal data is usually limited, it is recommended that the mode
shape data to be expanded using mode shape reconstruction technique. Thus, the raw data was
reconstructed using two different mode shape reconstruction techniques, namely Shannon's sampling
theorem and cubic spline. The computation of MDI is enabled by numerical integration method. In
this paper, two numerical integration methods were performed viz trapezoidal and rectangular rules.
The manipulated data is subsequently transformed into standard normal space. The mode shape was
mass normalised and the mode shape curvature was normalised with respect to the maximum value of
each considered mode. For practicality purposes, the first two flexural mode shapes were used in the
algorithms computation. Among the two proposed numerical integration methods, the rectangular rule
has shown greater potential. The cubic spline mode shape reconstruction technique shows better
results compared to the Shannon's sampling theorem.
