ABSTRACT

For each set of data points the values of A and n can be calculated using a power regression; however the scatter is such that the correlation coefficients are low, which means that statistically there was only a questionable relationship between the variables. Therefore at this stage it has only been attempted to compare the location of each “cloud” of data and to draw a conclusion as to whether a given data set is “better”, “worse” or “similar to” another. In this respect, data which leads to faster crack growth rates is regarded as “worse” and slower crack growth rates as “better”, for a given stress intensity. This is the approach that is used below, with the effect of changing one parameter at a time being discussed.