Machine Condition Monitoring by a Novel Fractal Analysis

The nonlinearities, induced by structural looseness or wear/fatigue of components, are good indicators of the health condition of a machine or structure. However, most existing condition monitoring techniques were initially designed for dealing with linear systems, thus unable to account for these scenarios properly. A few available nonlinear techniques are tried in condition monitoring. However, they are more or less limited owing to either intensive computation or unsatisfactory sensitivity to incipient abnormalities. In view of this, a new fractal analysis-based condition monitoring technique is researched in this paper. Firstly, a few number of fractal analysis methods with efficient computing algorithms are investigated in order to find an ideal one for condition monitoring application. Subsequently, a detailed investigation was conducted to verify the favored method and understand its instantaneous properties, robust performance against noise, and sensitivity to the abnormalities. Finally, following discussing the window width used in practical calculation, the condition monitoring technique developed based on the favored fractal analysis method is validated experimentally. Experiments show that the proposed technique does provide an efficient and successful nonlinear tool for machine operation condition and structural health condition assessment.