A smart fault identification system for ball bearing using simulation-driven vibration analysis

Bearings are one of the pivotal parts of rotating machines. The health of a bearing is responsible for the hassle-free operation of a machine. As vibration signatures give intimations of machine failure at an earlier stage, mostly vibration-based condition monitoring is used to monitor bearing’s health for avoiding the risk of failure. In this work, a simulation-based approach is adopted to identify surface defects at ball bearing raceways. The vibration data in time and frequency domain is captured by FFT analyzer from an experimental setup. The time frequency domain conversion of a raw time domain data was carried out by wavelet packet transform, as it takes into account the transients and spectral frequencies. The rotor bearing model is simulated in Ansys. Finally, most influencing statistical features were extracted by employing Principal Component Analysis (PCA), and fed to Multiclass Support Vector Machine (MSVM). To train the algorithm, the simulated data is used whereas the data acquired from FFT analyzer is used for testing. It can be concluded that the defects are characterized by Ball Pass Frequency (BPF) at inner race and outer raceway as indicated in the literature. The developed model is capable to monitor bearing’s health which gives an average accuracy of 99%.