Monitoring of ball bearings via vibration analysis and envelope technique for predictive maintenance purposes

Abstract

Bearings are one of the primary and most crucial machine components in the industry and transportation systems, since they are critical components in these systems, their failure can lead to various losses and significant damages. Despite the advanced stages of development and manufacturing processes, approximately 90% of unplanned machine shutdowns are attributed to bearing failures. These failures can occur in various ways, making it essential to monitor the bearing's deterioration to replace it before it reaches a critical condition for the industry, machine or vehicle's operation. To monitor ball bearing failures, a range of vibration monitoring techniques are employed, encompassing envelope analysis, kurtosis, bi-spectrum, and wavelet analysis. In this case, frequency graphs based in the signals of the bearings inner race will be generated, through a new signal acquisition system, using of envelope analysis by frequency filters, Hilbert Transform and Fast Fourier Transform (FFT). However, detecting and analyzing these signals can be challenging due to weak signals and noise masking vibration patterns. Although current techniques are effective, they have limitations, such as requiring expert analysis, difficulty in detecting early-stage faults, and inability to differentiate between different fault types. Current techniques, such as envelope and wavelet analysis, are effective but have limitations. New technologies and methods, are being explored to improve fault detection and classification, providing early detection of faults and differentiation between different fault types, ultimately reducing the impact of ball bearing failures on machines and industries. This paper proposed to present a study of the ball bearing failure through vibration analysis from early-stage to advanced-stage of damage for predictive maintenance purposes, applying the envelope and FFT together with programming, to enable the identification of defects in the bearing, especially in the inner race, through a signal acquisition system that can explain the presence of the defect through frequency graphs. Thus, obtaining results that show the presence of defects in three different bearings, with gradual defect magnitudes, differentiating these data from an ideal bearing. The next step we will intend to explore the new technologies like machine learning and artificial intelligence, to also analyze all variants of defects in a bearing.