Comparative Analysis of Fault Detection in Bearings using Vibration Signal Processing, through Matlab-Based Dynamic Simulation

Abstract

Bearing faults are among the most common causes of rotating machinery failure, responsible for nearly 40–50% of industrial motor breakdowns. This makes accurate and early fault detection essential to avoid unexpected shutdowns and costly downtime. In this study, we developed a MATLAB-based dynamic simulation framework to evaluate three widely used fault detection techniques Fast Fourier Transform (FFT), Wavelet Transform (WT), and Hilbert–Huang Transform (HHT) through vibration signal analysis. Simulated signals were generated under four conditions: healthy operation, inner race fault, outer race fault, and ball defect. The FFT method successfully revealed spectral peaks at characteristic defect frequencies, but its performance declined under non-stationary operating conditions. The Wavelet Transform provided improved time–frequency resolution, effectively capturing transient fault impulses and achieving a detection accuracy of 99.2% compared to 92.6% for FFT. The HHT delivered the most adaptive and detailed analysis by decomposing the signals into intrinsic mode functions (IMFs), resulting in the highest fault detection accuracy (99.8%) with the lowest false alarm rate (0.7%). The study highlights that while FFT remains computationally efficient, WT and HHT are more robust under varying operating conditions. In particular, HHT stands out as the most reliable technique for advanced fault diagnostics in non-stationary environments, offering valuable insights for predictive maintenance in modern rotating machinery.