Assessment of the effect of the steels structure dispersion on its magnetic and mechanical properties

OBRABOTKAMETALLOV MATERIAL SCIENCE Vol. 23 No. 4 2021 The hypothesis about the in fl uence of the considered parameters on each other is tested by the disper- sion analysis. Several values of the grain size factor are analysed. They are obtained for samples with dif- ferent heat treatment mode. The samples contain an equal number of elements. Figure 12 shows the statistical characteristics of the studied data. Samples 1 and 2 are a set of values of the grain size factor obtained during several processing of microstructure images for different values of ultimate strength. Sample 3, in contrast to 1 and 2, was obtained when determining the grain severity by its area. Fig. 12. A block diagram obtained by analyzing samples of the value of the grain size factor obtained with different estimates of the grain size The choice of the method for calculating the grain size factor does not have a signi fi cant effect on the presence of a relationship between this value and the value of internal stresses. It is shown in the block diagram. However, the difference in the average values of the samples can be unexpected. Therefore, a sta- tistically proved conclusion about the unambiguous in fl uence of the factors on each other cannot be made. The variance amount (data spread) for the samples has approximately the same value. It is the main condition determining the correctness of the method application for the variance analysis. Analysis of test statistics, which in this case has the form of F -distribution or Fisher distribution, is shown in Figure 13. The average value of the F -distribution obtained for the analysed dependence is 1.08. It is characterized by 13 and 28 degrees of freedom. The criterion for rejecting or accepting the null hypothesis is the value of F0 , which is 4.915. The probability p has a value of 5.3358 or more is 0.00034 with a random magnitude having a Fisher distribution for the analysed dependence. When compared with the signi fi cance level of 0.05, it can be seen that p is signi fi cantly less than it, which indicates that the null hypothesis is rejected, and the difference in the average values for the analyzed samples cannot be explained only by chance. The analysis described above is carried out for the dependencies presented in Figures 8 and 9. Figures 14 and 15 show the Fisher distributions obtained by analysing the dependencies of ln( σ ) and H c . It can be concluded that the average values for the samples are statistically signi fi cantly different from each other and the considered model (dependence) is statistically proved. The dependences can be described by a linear (with the internal stresses) and a polynomial curve in the fi gures from 8 to 10. The presence of such dependences suggests that the dispersion of the system (the