Structural Damage Detection Using Vibration Response Through Cross-Correlation Analysis: Experimental Study

A detection scheme to localize and quantify multiple damages in structures is proposed. At first, acceleration response histories at multiple sensor locations are collected experimentally from damaged and undamaged structures. Structures with single as well as multiple damages have been considered. The cross correlation between the acceleration histories at each sensor location is computed. The peak amplitudes of normalized cross-correlation outputs are used to form the damage indicators. The traditional threshold-based method of detecting damages from these damage indicators has been observed to be inadequate for the detection of multiple damages. Here, the drawback of threshold-based damage-detection method is overcome by using an artificial neural network. The accuracy of the damage-detection scheme has been shown for different damage cases.