FRACTAL APPROACH FOR SIGNAL PROCESSING AND APPLICATION TO THE DIAGNOSIS OF CAVITATION

Abstract The 1/ f noises [1] are turbulent signals that are found in many physical processes. Their name comes from the decreasing law of the spectral density irrespective of the definition used. One of the most famous 1/ f noise models is the fractional Brownian motion defined by Mandelbrot and Van Ness [2]. Their analysis requires the use of non-classical analysis method as, for example, the fractal theory. In this paper, we first give some results from this theory and discuss the tools that can be used. This analysis is then carried out on signals obtained on a centrifugal pump where cavitation phenomenon occurs. The spectrum analysis of these signals reveals the presence of 1/ f noise. A fractal study is then carried out with the help of the Higuchi method. The results highlight the sensitivity of the fractal dimension parameter to the degree of cavitation in the pump.