In the presentation of creep data on the materials for elevated temperature use, it is most desirable to analyse them by use of statistical procedures and express the results in objective form.
In this paper an attempt has been made to fit statistically regression curves for the creep data actually obtained for 2
1/
4Cr-1Mo, 1Cr-1Mo-1/4V, 18Cr-10Ni-Ti and 16Cr-13Ni-3Mo steels, concerning relations of stress vs. rupture-time, stress vs. minimum creep rate and stress vs. parameters (LARSONMILLER etc.), which contain rupture-time and temperature or creep rate and temperature.
The results obtained are as follows;
(1) The maximum significant degree of polynomials used for the regression equations for curve fitting turns out to be the third (cubic).
(2) The frequency distribution of residuals obtained from the estimated value and the observed value of logarithms of the rupture-time, the minimum creep rate and the parameters shows normality.
(3) There are some cases where a segmentwise regression curve composed of two straight lines fits the data better than a single curve with an increased degree of regression.
(4) The propriety of the general use of twenty as the value of the constant C in LARSON-MILLER parameter is confirmed statistically. It should be noted, however, that there are some cases where the contribution in curve fitting suddenly deteriorates when the constant C increase over twenty.
(5) As the results of the statistical studies regarding the comparison of the three different parameter methods, the MANSON'S parameter method is shown to be better than the LARSON-MILLER'S and DORN'S methods in respect of curve fitting for the creep rupture data.
抄録全体を表示