| ||Format||Pages||Price|| |
|PDF (404K)||23||$25||  ADD TO CART|
|Complete Source PDF (5.5M)||367||$55||  ADD TO CART|
A simple-minded yet quantitative approach to assessing interlaboratory fatigue crack growth rate data is proposed. Seven sets of da/dN versus ΔK data from six laboratories on nominally the same material and loading conditions in a cooperative test program sponsored by the Society of Automotive Engineers (SAE) are analyzed to illustrate this ad hoc approach. Each set of data is subjected to a standard first-order linear regression analysis based on the method of least squares. Three characteristics of the regression line [namely, the location of the “center” of the data, the slope, and the vertical half-width of the confidence band (for some specified level of confidence)] are used to define a composite measure of the closeness of one regression line to another. To illustrate the benefit of a statistically sound interlaboratory test program, the single-specimen SAE data are supplemented with fictitious replica data for the application of an interlaboratory data analysis procedure first proposed by Mandel (ASTM Standardization News, Vol. 5, No. 3, 1977, p. 17). “New” information based on the ad hoc approach of this paper and Mandel's method of interlaboratory data analysis is discussed in the context of other work on fatigue crack growth rate data analysis and the economic aspect of engineering testing.
applied regression analysis, data analysis, engineering judgment, fatigue, fatigue crack growth, fracture mechanics, interlaboratory data analysis, interlaboratory test program, linear regression analysis, mathematical modeling, statistics, steels
Physicist, Center for Applied Mathematics, National Bureau of Standards, Washington, D.C.
Fellow engineer, Westinghouse Research and Development Center, Westinghouse Electric Corporation, Pittsburgh, Pa.