USING VIBRATION DATA AND STATISTICAL MEASURES TO LOCATE DAMAGE IN STRUCTURES

Structural damage, such as cracks, usually cause a local reduction in stiffness. Visual inspection may not locate the damage and many papers consider the identification of this stiffness reduction from measured vibration data. Model updating may be used to estimate the stiffness at all the potential damage sites simultaneously, the largest reduction in stiffness giving the most likely damage site. Often model updating will spread the effect of the damage around the structure, making the localization of the damage difficult. A few earlier papers have assumed that the damage occurs at one position and found the statistically most likely damage location. Their approach is examined in this paper but an alternative statistical method of identification, using generalized least squares theory, is suggested. The method uses the ratio of natural frequencies from both the measured and analytical data. In this way the most likely damage location is identified for a range of damage levels. The different approaches are compared using simulated and experimental data.