abstract
- Bearing faults are a leading cause of failures in industrial rotating machinery, making their early and reliable detection essential to reduce downtime and maintenance costs. A Hilbert-Huang Transform (HHT) based methodology is proposed for early bearing fault detection. The Empirical Mode Decomposition (EMD) isolates fault-related Intrinsic Mode Functions (IMFs), which are selected automatically using the Kurtosis-RMS (KR) criterion. Envelope Spectrum analysis is then applied to the reconstructed signal for diagnosis. The proposed method achieved 94% accurate IMF selection and over 85% fault identification accuracy across 18 signals from the CWRU, TianYau Wu, and MFPT databases. Compared with conventional FFT and Envelope Spectrum techniques, the approach improves early fault detection under noise and variable load conditions while reducing manual analysis time.