A method for the automated interpretation of complex vibration signals is introduced. The method which combines techniques of pattern recognition and neural networks is illustrated using an example of detecting defects in concrete cylinders from impact test signals. Two neural nets were trained, one to detect defects and the other to predict the extent (or width) of defects. The first returned an accurate verdict 86 percent and 76 percent of the time when specimens were not defective and defective, respectively. For the second net, with the exception of the smallest defect, the actual width was within the range of the average plus or minus the standard deviation of the predicted width. Based on the results and limitations of this experimental test, the method appears to be capable of interpreting vibrations signals with reasonable accuracy.
A Hybrid Analysis Method for Vibration Signals Based on Neural Networks and Pattern Recognition Techniques
Contributed by the Technical Committee on Vibration and Sound for publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received June 1999; revised July 2000. Associate Editor: Kou-Well Wang.
Samman , M. M. (July 1, 2000). "A Hybrid Analysis Method for Vibration Signals Based on Neural Networks and Pattern Recognition Techniques ." ASME. J. Vib. Acoust. January 2001; 123(1): 122–124. https://doi.org/10.1115/1.1320444
Download citation file: