Unsupervised noise cancellation for vibration signals: part I—evaluation of adaptive algorithms

Abstract

Vibration analysis of machines gains much in efficiency if periodic vibrations can be separated out from non-deterministic ones. The first part of this paper reviews the existing adaptive algorithms that may be used to achieve this goal. The presentation is carried out from the viewpoint of prediction theory and gives a more solid theoretical basis to a number of recommendations for setting the algorithm parameters, as well as speeding up the computation by fast convolution using FFT processing. The application of Self-Adaptive Noise Cancellation is then illustrated on simulated and actual vibration signals. This first part is an introduction to part II of the same topic where a new method is proposed.