Vibroacoustic Diagnostics of Rolling Bearings - Test Rig for Research Dataset Acquisition and Didactic Applications

Abstract

Rolling element bearings are vital in machines used for transportation. Their failure often leads to breakdowns. Diagnosing faults through vibrations and vibroacoustic signals captured by sensors is essential. Spindle vibrations stem from manufacturing defects, misalignment, and inherent vibrations. Bearing damage can affect races, rolling elements, or cages, and various tests can identify these faults. Diagnostics employ signal processing methods such as power spectrum density estimation, auto-regression moving average (ARMA), and Fast Fourier Transform (FFT), along with noise reduction techniques like minimum entropy deconvolution (MED) and spectral kurtosis. Machine learning is increasingly used to enhancing anomaly detection. Advanced methods like the Teager energy operator improve early fault detection, while deep learning models increase the accuracy of predicting bearing degradation. This paper presents a bearing test rig developed to collect data from faulty bearings, providing datasets for analisys. The setup supports research and educational purposes, allowing the study of bearing damage under various conditions. The test rig design enables quick replacement of tested elements and accommodates a wide range of rotational speeds for comprehensive diagnostics. An FFT for the vibration signal envelope was created, and frequencies and bandwidth were determined using a fast kurtogram algorithm.