ASA 125th Meeting Ottawa 1993 May

1pSA4. Signal analysis techniques for acoustic monitoring.

Pierre Y. Robin

Peeter M. Akerberg

Ben H. Jansen

Robert D. Finch

College of Eng., Univ. of Houston, Houston, TX 77204-4793

Impacting structures with a hammer and measuring the resonant frequency of the induced vibrations is a commonly used technique in acoustic monitoring. However, the transient and nonstationary characteristics of the vibrations limit the performance of classical spectral analysis. In this paper, a number of alternative techniques specifically designed to deal with the transient nature of impact-induced vibration signals are presented. Among the techniques discussed are the short-time Fourier transform (STFT), the discrete wavelet transform (DWT), and artificial neural networks (ANN). The results obtained with applying these techniques to the acoustic vibrations of intact and slotted beams are presented as an example. It was found with the STFT method, that a relationship exists between slot depth and decay rate. The DWT provided evidence that the decay undulates as a function of time, with a frequency that is related to the integrity of the beam. Early results with ANNs suggest that a net can be trained to differentiate between acoustic vibrations of intact and defective structures. [Work supported by NSF Grant No. MSS-9024224.]