Intelligent fault diagnosis in rolling element bearings: Combining envelope spectrum and spectral kurtosis for enhanced detection

• Hybrid envelope spectrum and spectral kurtosis method achieves 95%+ bearing fault detection accuracy. • Log-ratio classifier distinguishes inner, outer, and normal states in noisy, variable conditions. • Kurtogram-guided filtering enables real-time, scalable, and efficient fault detection. • Tests confirm robustness, noise resistance, and predictive maintenance suitability. Rolling-element bearings are vital components in industrial machinery, and their early fault detection is essential to prevent costly failures and downtime. This paper presents a robust and scalable fault diagnosis method that combines Envelope Spectrum Analysis (ESA) and Spectral Kurtosis (SK) to enhance detection accuracy under noisy and dynamic conditions. ESA identifies characteristic fault frequencies such as BPFO (Ball Pass Frequency Outer Race) and BPFI (Ball Pass Frequency Inner Race), while SK detects impulsive features linked to bearing faults. To improve classification, a log-ratio classifier is introduced, which differentiates between inner race faults, outer race faults, and normal conditions based on amplitude ratios. Normal conditions show low amplitudes, while fault conditions exhibit distinctive frequency signatures. The proposed hybrid approach demonstrates high accuracy (over 95%) even in challenging environments with high noise and variable operational speeds. Experimental validation, including histogram comparisons and simulations, confirms the method’s reliability and real-time applicability for predictive maintenance. The approach is computationally efficient and suitable for industrial deployment, offering potential to reduce maintenance costs, avoid unplanned downtime, and extend machine life. Future enhancements may include broader system flexibility to accommodate diverse machine configurations and the integration of multi-sensor data fusion for more comprehensive fault detection. Overall, the proposed method stands as a practical, accurate, and scalable solution for modern industrial fault diagnosis.