Modelling of superfinishing process of ball bearings

Ball bearing production involves forging, turning, heat treatment, grinding, and superfinishing (SF). Superfinishing, the most critical step, significantly impacts surface roughness, load capacity, fatigue resistance, form accuracy, noise level, and wear resistance. It also creates micro-textures that enhance lubrication retention. The process entails a complex interaction between an oscillating abrasive stone and a rotating workpiece, driven by hydraulic and pneumatic pressure, and using oil as a cutting fluid. Proper abrasive cleaning is crucial to prevent workpiece imprints and maintain chip removal efficiency. Noise level is important in classifying bearings, and unclassified bearings are caused by scrap. The production of high-precision bearings requires less noise and lower vibration levels. It has become important to reconcile the noise level and the process. To learn the specifics of the SF process, an experimental design was first created. Using the Design of Experiments (DOE) method, a thorough investigation was carried out to examine how the superfinishing (SF) procedure affected wear and surface topography. To determine the ideal process parameters for attaining a more accurate surface quality, 405 experimental observations were made. Various analyses were performed on the findings, including optical measurements of both the abrasive and the worn surfaces, to evaluate important wear parameters. Machine learning algorithms were used to examine better how surface characteristics affect noise. The association between noise levels and process v parameters was investigated using regression analysis. Deep learning algorithms were used to process numerical and categorical data to estimate noise levels based on SFprocess parameters. This strategy ensured accurate product outputs while lowering processing expenses and noisy scrap. Furthermore, the study has demonstrated that the process parameters have an impact on the surface characteristics Ra and Rz that determine the bearings' noise level by using deep learning tools. The effect of the result being scrap is as follows: 1. Priority stone pressure 2. Priority oscillation speed 3. Priority workpiece rpm. Although Ra is utilized more in industrial applications, Rz is more effective than Ra at reducing noise. The Rz parameter's process values need to be examined while attempting to lessen the noise generated by surface roughness during production. SiC abrasive should be used when investigating wear using both the abrasive and the worn surface to achieve good values throughout the process and a sensitive surface. WA abrasive shows an aggressive structure at the beginning of the process. FA and SiC abrasives give more stable results. Therefore, this information is essential and should be used in process and product improvement studies.