Independent Component Analysis of Excavator Noise
To identify excavator noise sources, an acoustic camera was used to acquire sound signals, and FastICA was applied to separate the signals. For strong background noise and echoic interference, the noise separation model was built based on FastICA algorithm in frequency-domain, then principle frequencies were obtained. To find the corresponding parts of these frequencies, modal analysis of major surface parts of the diesel was run in Ansys, and the modal analysis results were compared with principle frequencies. Research shows that ICA can effectively separate excavator sound signals contaminated by strong background noise and echoic interference; and the surface noise radiation sources such as cylinder block, cylinder head and valve cover were found by comparing component principle frequencies and modal analysis results.
KeywordsExcavator independent component analysis modal analysis convolutive mixture
Unable to display preview. Download preview PDF.
- 5.Long, F., He, J.S., Ye, X.Y.: Discriminant Independent Component Analysis as A Subspace Representation. Journal of Electronics 23(1), 103–106 (2006)Google Scholar
- 10.Back, A.D., Tsoi, A.C.: Blind Deconvolution of Signals using A Complex Recurrent Network. Neural Networks for Signal Processing 4, 565–574 (1994)Google Scholar