Rolling bearing fault identification using BAT algorithm for the resonant demodulation parameters optimization

The rolling bearing is a fundamental component of rotating machines and its failure may lead to a catastrophic damage of the system. The incipient and correct identification of faults contribute to an early predictive maintenance plan, which avoids additional costs and sudden breakdowns. The resonant demodulation technique, envelope analysis, is a well-established method widely used to identify failures in rolling bearings. However, this method requires the identification of the frequency region that contains enough information about the faults. Thus, the spectral kurtosis gives the impulsiveness measure of a vibration signal and it is used to identify the frequency region of failure. This paper presents the use of the bat algorithm as an optimization methodology to identify the resonant demodulation parameters using spectral kurtosis as the objective function. Bat algorithm is a recent method based on echolocation behavior becoming a powerful option in face of traditional method as genetic algorithm. The feasibility and performance of this approach is validated and discussed by experimental data to explicit its benefit which is able to optimize the process of identifying the optimal frequency ranges for application of the analysis, avoiding the need to analyze the entire spectrum of frequencies.