Uncertainty analysis of cutting parameters during grinding based on RSM optimization and Monte Carlo simulation

KAHRAMAN M. F., Ozturk S.

MATERIALS TESTING, vol.61, no.12, pp.1215-1219, 2019 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 61 Issue: 12
  • Publication Date: 2019
  • Doi Number: 10.3139/120.111443
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.1215-1219
  • Kocaeli University Affiliated: No


Due to the importance of high surface quality of machined parts, regarding its functional requirements, it is necessary to determine an appropriate set of grinding parameters. According to the uncertainty of the machining process, the statistical techniques have recently been used to set up an experimental-based model for estimating the performance of machining parameters and optimizing them. The purpose of this study is to demonstrate the modeling and optimization of the grinding process using three approaches. First, multi non-linear regression (MNLR) based on central composite design (CCD) was used to determine the process model. Then the grinding parameters were optimized considering response surface methodology (RSM). Finally, the probabilistic uncertainty analysis was applied by using Monte Carlo simulation as a function of wheel speed and feed rate.. The surface roughness value, which was named the response variable, was estimated by fitting the MNLR model with a predicted regression coefficient (R-pred(2)) of 84.69 %. Wheel speed of 4205.6 rpm and feed rate of 2.969 mm x min(-1) were calculated as RSM-optimized conditions with a surface roughness of 2.26326 mu m. The verification experiments were performed with three replications to verify the predicted surface roughness value obtained with the derived model, and 2.263 +/- 2 % mu m of surface roughness was calculated using RSM optimized conditions. Monte Carlo simulations were found to be quite effective for identification of the uncertainties in surface roughness that could not be identified by deterministic ways.