Abstract
An experimental crack detection system is developed based on the natural frequency changes. In the detecting system, we employed a Maharanobis discriminant analysis, which is a useful way of determining similarity of an unknown sample set to a known one. Monitoring the Mahalanobis distance of unknown specimen and comparing the distance of the specimen with that of un-cracked specimens, we can detect a crack position and a crack depth simultaneously. Using both the statistical classification method and Excel macro program, we have accomplished the real-time detecting system.
