Abstract
This article describes a statistical analysis of certain stress corrosion endurances, made with a view to identifying the type of frequency distribution to which they belonged and hence to recommending appropriate methods of treatment of stress corrosion results.
It was found that the logarithms of the endurances of stress corrosion specimens of aluminum alloys tested in salt peroxide solution are normally distributed and that the variance of the logarithm of endurance is independent of the mean but not constant. It was concluded that the most meaningful representative endurance is the geometric mean, of which the median is a good approximation.
Other estimates, which may show a saving in testing time or cost, can be found which are only slightly, if at all, less efficient than the geometric mean. These other estimates use linear functions of the logarithms of endurance in a censored (i.e., incompletely tested) sample The relative merits of various procedures are discussed by way of example and the corresponding time savings are quoted using a standard deviation, which is representative of the present results. The method of calculation is given in the Appendix so that it can be applied for other values of the standard deviation.