Investigation of the Effect of Sample Size on Fatigue Endurance Limit of a Carburized Steel 2006-01-0539

Prediction of fatigue performance of large structures and components is generally done through the use of a fatigue analysis software, FEA stress/strain analysis, load spectra, and materials properties generated from laboratory tests with small specimens. Prior experience and test data has shown that a specimen size effect exists, i.e. the fatigue strength or endurance limit of large members is lower than that of small specimens made of same material. Obviously, the size effect is an important issue in fatigue design of large components. However a precise experimental study of the size effect is very difficult for several reasons. It is difficult to prepare geometrically similar specimens with increased volume which have the same microstructures and residual stress distributions throughout the entire material volume to be tested. Fatigue testing of large samples can also be a problem due to the limitation of load capacity of the test systems available.

In this study, we investigate the sample volume effect on fatigue endurance limit through both experimental and computational approaches. Bending fatigue tests were conducted and stress life (S-N) test results of two specimen sizes (highly stressed material volume ratio 6:1) made of carburized 4320 steel were obtained. In the computational analysis work, first, deterministic fracture mechanics based software package CRACKS was used to generate S-N curves of two specimen sizes. The computationally generated S-N curves were then compared with the test results and they agreed with test results quite well. However, the deterministic fatigue life analysis using the CRACKS software did not show any size effect, i.e. same endurance limits were obtained for the two different sample sizes.

Since the deterministic approach cannot capture the size effect, stochastic fracture mechanics analysis was then performed which is based on the following premises. Larger components have a greater volume of material, and thus a larger population of material related initial defects, such as inclusions and surface oxidation layer. Statistically, a larger population contains more extreme defect sizes, both large and small. The presence of larger defects leads to crack growth and failures at lower stress levels, hence a lower endurance limit. In the stochastic analysis, we used the alternating finite element method and linear elastic fracture mechanics to compute the endurance limits of the two specimen sizes with high stressed material volume ratio of 6:1. Larger sample volume reduced endurance limit by 3% to 30% depending on the input data used. We also performed sensitivity studies to evaluate the effect of crack aspect ratio, initial defect size range, defect location, and material's crack growth threshold value ΔK_th, on the size effect.