Issue 23

A. Spaggiari et alii, Frattura ed Integrità Strutturale, 23 (2013) 75-86; DOI: 10.3221/IGF-ESIS.23.08 83 Net Torque (Nmm) τ y ( kPa) P ( Bar) B ( mT) R1 R2 R3 Average Std. Dev. Average Std. Dev. 0 50 172.08 213.56 146.21 177.28 33.97516 9.25 19.16% 0 100 514.47 543.45 507.13 521.68 19.20442 27.22 3.68% 0 200 1204.06 998.46 1168.5 1123.67 109.8859 58.64 9.78% 0 300 1156.95 1020.73 1393.3 1190.33 188.5142 62.12 15.84% 10 50 255.78 225.71 248.84 243.44 15.74466 12.70 6.47% 10 100 915.54 482.43 445.15 614.38 261.4832 32.06 42.56% 10 200 1658.38 1704.4 2001.82 1788.20 186.4258 93.31 10.43% 10 300 2374.80 2683.51 2688.13 2582.15 179.5823 134.75 6.95% 20 50 198.97 229.18 157.53 195.23 35.97138 10.19 18.43% 20 100 605.44 665.52 573.83 614.93 46.57584 32.09 7.57% 20 200 1901.04 2037.91 2084.71 2007.89 95.44484 104.78 4.75% 20 300 2824.34 2523.62 2949.16 2765.70 218.7452 144.32 7.91% 30 50 237.58 249.54 291.72 259.61 28.44097 13.55 10.96% 30 100 931.46 753.29 739.61 808.12 107.0344 42.17 13.24% 30 200 2347.10 1808.65 2181.02 2112.25 275.7325 110.22 13.05% 30 300 2992.63 3084.06 2802.75 2959.81 143.4975 154.45 4.85% Table 3 : Experimental net torque and yield shear stresses τ y retrieved using Eq. (4). Both from the experimental curves in Fig. 6 and the experimental data in Tab. 3 it is possible to qualitatively sense that the pressure does have an important effect on the shear stress and it is a positive effect. Since a design of experiment procedure was applied the data are analyzed applying an Analysis of Variance: this statistical method is able to quantify variable influence and variable interaction on the process under scrutiny. Analysis Of Variance On The Yield Stress An analysis of the variance was applied using Design Expert 8.0 software [18]. ANOVA calculates the variance of a response by considering a specific variable and the global variance in the responses. Among the possible approaches to graphically represent the results, one of the most popular is the normal plot, which is used to estimate whether a certain set of data follows a Gaussian distribution or not. If the data approximates a straight line, the phenomenon is statistically " normal" i.e. follows a stochastic law and can be attributed to background noise. The variables or the interactions affecting the system’s behaviour will then fall outside the normal distribution line, thus their effect cannot be ascribed to a stochastic process. The greater the deviation of the point from the normal line, the larger the confidence interval (i.e. the probability that the variables are significant). The half normal plot used in this paper is interpreted in the same way as the normal plot, but allows absolute values of the effects to be considered. Since the half normal line starts at the origin, this produces a more sensitive scale to detect significant outcomes [15], which are immediately detected at a glance. The half normal plot of the experimental values of yield shear stress is shown in Fig. 7. The straight line is built thanks to replicates ( triangles) and provides an estimation of the normal distribution of the experimental error, which by definition has a stochastic distribution. The triangles are an expression of the sum of errors [15], which is calculated by Design Expert software [18]. Since both points representing the two variables (pressure and magnetic field) are far from the error line, ANOVA demonstrates that these variables influence the process. The interaction between the variables is also important, which means that the increase in shear yield stress due to the combined action of pressure and magnetic field is greater than the sum of the two effects taken separately. The experimental test points have a little dispersion, as demonstrated in Fig. 8, provided by the Design Expert software, where the I-beams bars represent the standard deviation.