OBRABOTKAMETALLOV TECHNOLOGY Vol. 26 No. 4 2024 4 Re 13.142 0.058 3.78 0.18 0.0975 1.3 10 V f N Vf VN − = − − − + − × − 5 2 2 2 4.75 5.43 10 62.27 0.0682 fN V f N − − + × + + (3) Using the analysis of variance (ANOVA) method, the adequacy of the resulting equations was verifi ed. The data point variation proportion is measured by the coeffi cient of multiple determinations, or R-squared. A correlation coeffi cient (R-squared) that is between −1 and +1 is always ideal. If R is really close to +1, the equation is important. Ameasure of how much of the variance around the mean is explained by a model is called adjusted R-squared. A measure of the predictive accuracy of the model for the response value is the predicted R-squared. It is considered reasonable agreement when the adjusted and predicted R-squared values are within around 0.20 of each other. Otherwise, there may be a problem with the model or the data. The signal-to-noise ratio, or the range in the expected response relative to the corresponding error, is what is called adequate precision. Four or more is ideal value. The ANOVA for surface roughness, microhardness and roundness error during roller burnishing under dry condition can be referred to [20], and that under NFMQL cutting condition is given in Table 4. The ANOVA for the investigated responses under dry condition is also mentioned in Table 4 for comparative evaluation. The ANOVA results for surface roughness under dry condition and NFMQL condition show model F-values of 46.91 and 19.51, respectively, which means that the models are signifi cant. The “Prob > F” values less than 0.05 indicate that the model terms are signifi cant. The signifi cant model terms observed for surface roughness under dry cutting conditions are V×f, V×N, f×N, V 2, f 2, N 2, and for NFMQL the signifi cant model terms are V, f, N, V×N, V 2, f 2, N2. Ta b l e 4 ANOVA for investigated responses under dry [20] and NFMQL cutting conditions Factors Surface roughness (Ra) (μm) Microhardness (HV) Roundness error (Re) (μm) Dry NFMQL Dry NFMQL Dry NFMQL R-squared 0.9769 0.9461 0.9152 0.9377 0.9407 0.9609 Adj. R-Squared 0.956 0.89765 0.8389 0.8816 0.8873 0.9258 Pred. R-Squared 0.8472 0.848529 0.855 0.8389 0.8933 0.7421 Adeq. Precision 19.328 12.74978 15.464 16.5655 16.002 18.2847 Model F-value 46.91 19.51 11.99 16.71 17.62 27.35 The ANOVA results for microhardness show that the model F-values are 11.99 and 16.71 for dry and NFMQL conditions, which means that the models are signifi cant. There is only a 0.03 % chance that such a large “model F-value” may be due to noise. In this case, V, V×f, V×N, f×N, V 2, f 2, N 2 for dry conditions and V, f, N, V×f, f×N, V 2 for NFMQL conditions are signifi cant model terms. And the ANOVA results for roundness show that the model F-values are 17.62 and 27.35 for dry and NFMQL conditions, which means that the models are signifi cant. In this case, V, N, V×N, f×N, f 2 for dry conditions and V, f, N, V×f, V 2 for NFMQL conditions are signifi cant model terms. Each model created for dry and NFMQL cutting conditions has an R-squared value above 0.9, indicating the proportion of variation in the data points. Therefore, during the roller burnishing of Al6061-T6 alloy, the microhardness, surface roughness and roundness error can be accurately predicted by the established empirical equations. To improve understanding, two-dimensional graphs are created for NFMQL cutting settings by adjusting the feed, speed, and number of passes using the derived equations 1–3. In order to facilitate comparison and better understanding, curves are also plotted for the studied responses under dry conditions using the
RkJQdWJsaXNoZXIy MTk0ODM1