Induction Motor Bearing Fault Detection Using a Fractal Approach

Abstract

Fault detection is an important research area in mechanical engineering. Literature surveys indicate that bearing failures are considered the most common failure modes in motors. Various faults related to bearings can be categorized into single-point defects or generalized roughness defects. In many research studies, monitoring methods based on vibration signals are used to detect single-point bearing failures. Depending on which bearing surface contains the fault, the characteristic vibration frequencies can be calculated from the rotor speed and the bearing geometry. It also has been demonstrated that stator current monitoring can provide the same indication without requiring access to the motor. The combination of phase space reconstruction and fractal theory may provide an effective approach to detect bearing generalized roughness faults in induction motors by assembling the estimation of dynamic invariant properties of a nonlinear system. In mathematics, a delay embedding theorem gives the conditions under which a chaotic dynamical system can be reconstructed from a sequence of observations of the state of a dynamical system by lagging the time series to embed it in more dimensions. One can determine the delay time by calculating mutual information with equality distant space cells. False nearest neighbors provides a robust way to determine necessary embedding dimensions. Almost all chaotic systems have a quantifying measurement known as a fractal dimension which is extracted from the original or reconstructed phase space. There are many specific forms of fractal dimension. In this research, correlation dimension is used to estimate the dimension of attractors in nonlinear dynamical systems. Taking the result of Fourier based analysis as a reference for fault detection, experimental results show that the proposed method is as effective in detecting bearing generalized roughness faults in induction motors.