In a previous we proposed a new non-contacting non-destructive method for detecting cracks in concrete structures utilizing an impact by shock waves and vibration detection by LASER vibrometers. It has been found that a shock wave excites flexural vibration in the concrete plate above shallow cracks and results in an exponentially decaying sinusoid in the vibration signal. In the present article a signal processing procedure for testing the concrete integrity is proposed. The method automatically detects an exponentially decaying sinusoid that is, a flexural vibration of the concrete surface. The velocity signal is first segmented into successive frames. In each frame, the signal is analyzed using the linear prediction model (AR model) and the prediction error normalized by the signal power is calculated. This error indicates the degree to which the signal is modeled in the form of an exponentially decaying sinusoid, and presence of sub-surface flaws in the concrete can be determined from its value. When a flexural vibration (a defect in concrete) is detected, its frequency is estimated by solving a modified Yule-Walker equation or using FFT. The proposed procedure was applied to computer-generated signals and actual vibration signals, and the success of the procedure was verified.
抄録全体を表示