Experimental LCF Study of a Probabilistically Optimized Double Notch Specimen Geometry for Validation of Deterministic and Probabilistic Design Concepts

To provide answers to the demand of flexibility in both gas and steam-turbine business, improving the preciseness of lifetime assessment concepts is of key interest. For notched components, evaluating the local notch-root loading with multiaxial stress states as well as stress gradients to assess the fatigue life for crack initiation represents an established and common standard. Besides such classical approaches however, a more accurate crack initiation life assessment for notches is established via the so-called notch support factors which take the stress gradient as well as the size effect into account. In this work we apply a probabilistic model for fatigue that is capable of simultaneously considering gradient and size effects via a surface integral over a Weibull hazard density approach.

To demonstrate the potential of extending classical local concepts in that way, we conduct a systematic and advanced experimental testing campaign with a high-chromium forged steel for elevated application temperatures. Therefore, single and double notched round-bars have been tested under global force controlled low-cycle fatigue (LCF) loading conditions. The double notched specimens represent a mild and a sharp notch, which were arranged one above the other in the direction of load. The notch geometry of all specimens, mild and sharp single notched as well as double-notched, have been designed in order to get at first a smaller fatigue life at the sharp with the conventional deterministic prediction and second a smaller fatigue life at the mild with the local probabilistic approach to LCF. The latter takes the combined size and stress gradient into account. Corresponding to this, a numerical optimization of the notch radius and the cross section of the sharp and mild notch was done based on the difference of conventional deterministic life and advanced probabilistic life predictions. Via geometrically maximizing this difference of different model predictions, the likeliness of observing a statistically significant experimental result is maximized as well.

In-situ time-resolved alternate current potential drop (ACPD) measurements in combination with load-triggered digital image acquisition at both, mild and sharp notches, enable the determination and quantification of crack initiation and growth during LCF cycling. In order to study crack initiation behavior and growth kinetics post-experimental analyses have been performed in terms of light microscopy and correlation of ACPD measurements and notch images. Comparing the results of both notch geometries, it has been found for the double notched specimens, that cracking of the mild notch occurred in the first place as predicted by the probabilistic size effect model for most scenarios. Whereas crack initiation at the sharp notches, which were supposed to show earliest crack initiation based on conventional deterministic predictions, could not be detected at all.