Contribution to the selection of cutting fluid type and its application technique for grinding of bearing steel

Abstract

The elevated heat generation in grinding can develop high temperatures at the contact zone, which can adversely affect the surface integrity of the workpiece, especially when grinding hardened steels with conventional abrasives. Thus, the correct selection of cooling-lubrication condition is essential to avoid or attenuate any possible negative effect to workpiece surface integrity. However, the literature lacks work comparing different cutting fluid application technique (e.g. flood and minimum quantity lubrication – MQL) using the same fluid on both techniques. In this context, this work aims to contribute to the selection of cutting fluid type and its application technique for the grinding of bearing steel. Experimental trials were conducted comparing the use of semisynthetic and synthetic cutting fluids, both applied via conventional (flood) and MQL techniques. Different cutting conditions were also tested. A 24 full factorial design of experiment (DOE) was carried out with the following factors: fluid application technique, type of fluid, workspeed, and radial depth of cut. An analysis of main effects and interactions was performed for surface finish (Ra parameter) results, including a prediction model based on the analysis of variance (ANOVA). The morphology of ground surface and microhardness below machined surface were also analyzed. The results showed that the ground surface finish was strongly dependent on the cutting fluid type and its application technique combination: superior finishing was observed with the combination of semisynthetic fluid delivered via flood technique and with synthetic fluid delivered via MQL technique. From the surface morphology analysis, it was observed that the inferior lubrication capacity of synthetic fluid applied via flood condition deteriorated the surface finish and morphology. The surfaces ground with semisynthetic fluid provided, in general, lower values of Ra and lower microhardness variation. The prediction model for Ra showed a maximum error of 14% in comparison to the measured values.