AN INVESTIGATION FOR THE PERFORMANCE OF CYLINDRICAL ROLLER BEARING

Abstract

Rolling element bearings (REB)s are commonly utilized in rotating machines. Their performance is extremely significant in various industries like chemical, petrochemical, paper making, automotive, electricity generation, and aerospace turbomachinery. REBs have a major impact on the dynamic behavior of the rotating machinery, which is an origin of vibration and noise. The cylindrical roller bearing is a key component among all REBs and is generally used due to its high load-carrying capacity. It is designed to support a slight axial load along with the predominant radial load. This work emphasizes the development of an experimental setup that attempts to support an axial load over and above a radial load. The experimental investigation has been carried out considering defects on individual bearing components, such as IR, OR, and roller. Moreover, united flaws on an IR-OR, OR-roller, roller-IR, and IR-OR-roller along with healthy bearing have experimented. These experiments have been extended for three different fault severities for individual defects and two defect severities for combined defects. The results are acquired for various radial and axial load combinations at numerous speeds. Moreover, the three different positions for the fault on the OR have been experimented here. The envelope analysis has been performed to extract fault frequencies from all combinations. Bearing failure can cause hazardous effects on rotating machinery. The diagnosis of the fault is very critical for reliable operation. The main steps for the machine learning process involve feature extraction, selection, and classification. Feature selection identifies noble features that perform for better classification accuracy with fewer features along with lesser computational time. A critical study is required to catch the best feature subset for a large feature dimension. So, a novel feature ordering and selection technique called Feature Ranking and Subset Selection based on Euclidean distance named FRSSED is proposed. For the analysis, fifteen statistical features of time domain and frequency domain techniques have been utilized. The proposed technique has been verified over two bearing datasets. One was obtained from the experiment, and the other was collected from Case Western Reserve University (CWRU). Moreover, the noise and vibration produced due to defects can be minimized by applying various loading combinations. MADM techniques are required to pick up a unique and ideal solution or ranking of choices from the available alternatives (loading conditions). Therefore, a novel methodology for the Ranking and Selection of Optimum Alternative based on Linear Scale Transformation (RSOA-LST), having combined benefit and cost criteria, has been proposed. Two kinds of analysis, i.e., Local Defect Analysis (LDA) and Global Defect Analysis (GDA), have been carried out here. The effectiveness of the suggested methodology has been verified over other MADM techniques. The outcomes state that the suggested technique effectively bifurcates the influence of alternatives locally as well as globally over other MADM techniques. The work is not limited to experimentation, diagnosis, and optimization. The defects produced in any part of the bearing result in the additional impulsive force, which shows defect frequency and its harmonics of the vibration signals. The nonlinearity in the bearing is caused by nonlinear stiffness, nonlinear damping, radial internal clearance, and defects in individual or combined components of bearings. However, the nonlinearity due to defect is predominant and reflected in the frequency spectrum. Hence, a novel defect function has been proposed for the IR, OR, and roller, considering roller-race interaction. These defect functions have been applied for defects on the individual as well as combined components of bearing. The behavior of cylindrical roller bearing has been estimated from this analysis. The application of combined radial and axial load on bearing, tilts the rollers, resulting in additional radial deformation of rollers due to tilting. Hence, a novel methodology has been proposed to estimate the tilting angle of rollers for cylindrical roller bearing, considering the difference in deflection due to unmixed radial load and composite radial-axial load. This work concentrates on analyzing the tilting angle of rollers for varying loads, contact angles, and clearances.