Research Associate, Nova Scotia Research Foundation Corporation, Industry Services Division-Corrosion, Dartmouth, Nova Scotia
Defence Scientist, Defence Research Establishment (Atlantic), Dartmouth, Nova Scotia
(Received 16 December 1985; accepted 20 June 1986)
This paper analyzes the effects of the data averaging increment (ΔC) on the prediction of mean life cycles in 25-mm. (1-in.)-thick HY-80 steel plate using CT test specimens. The analysis shows that the crack propagation in this steel plate is not uniform; the crack rate randomly oscillates along the crack at variable amplitude. The frequency of crack rate variation was found to have the characteristics of a normal distribution curve and was therefore characterized by mean and variance. The mean crack behavior, including mean crack propagation rates and mean life cycles, was determined for crack propagation data obtained at different ΔC and statistically compared. It was found that the same raw a and N data averaged over different ΔC yielded significantly different mean crack rates and mean life cycles.
The 95% confidence limits that were assigned to the mean life cycles (N ± 1.96 √σ/n) were also different and affected by ΔC. This becomes extremely important in engineering design when trying to determine the expected confidence in a life prediction or a safe inspection interval. To use the scatter in da/dN data to generate the design confidence limit, it is essential that the same ΔC be used in the numerical integration that was used to generate the data.
Paper ID: JTE11548J