The Significance of Fractography for Investigations of Fatigue Crack Growth under Variable-Amplitude Loading

TLDR: In this article, the authors used fractographic evidence to analyze prediction problems offatigue crack growth under variable-amplitude (VA) loading and found that the predicted growth rate per spectrum block can still agree with the experimental value.

Abstract: Fatigue crack growth tests were carried out on 2024-T3 and 7075-T6 central cracked specimens. Variable-amplitude (VA) load spectra were used with periodic overload (OL) cycles added to constant-amplitude (CA) cycIes. The fatigue fracture surfaces were examined in the SEM to obtain more detailed information on crack growth contributions of different load cycles. The striation pattems could be related to the load histories. SEM observations were associated with delayed retardation, the effect of 10 or a single OL on retardation, crack growth during the OL cycIes, and crack growth arrest after a high peak load. Fractographs exhibited local scatter of crack growth rates and sometimes a rather tortuous 3d geometry of the crack front. Indications of structural sensitive crack grwoth under VA loading were obtained. Fractography appears to be inadmissable for the evaluation of fatigue crack growth prediction models in view of similarities and dissimilarities between crack growth and V A and CA loading. Nomenclature a BL cycles CA-loading daidN OL cyc1es VA loading A~ff ASeff Introduction crack length base line cyc1es constant-amplitude loading crack growth rate overload cycle variable"amplitude cycles effective stress intensity factor effective stress range The problem addressed in this paper is the potential usefulness of fractographic evidence to analyze prediction problems offatigue crack growth under variable-amplitude (VA) loading. Verifications of prediction models can be made on different levels. The most global one is a comparison of predicted and experimental crack growth lives. If they do agree, it is possible that an underprediction in the fust part of the crack growth life is compensated by an overestimation in the second part. In other words, in the first part the crack growth rate daidN is overestimated, whereas daidN is underestimated in the second part. The prediction model is apparently not accurate. A more precise verification of a prediction model requires that the crack growth rate is correctly predicted for the entire crack growth life. Such verifications are made by comparing predicted and experimental da/dNvalues as a function ofthe crack length a. Ifthe predicted relation and the experimental one are similar, the prediction model could be supposed to be accurate. However, da/dN values as obtained from crack growth records are calculated slopes. Such da/dN values still have a kind of a global nature. Fig.l shows the load spectra applied in the present test series. The spectra contain small base line (BL) cycles and intermittent larger overload (OL) cycles. Ifthe crack rate during the large cycles is higher than predicted by a cycleby-cycle prediction model, and the crack rate during the small cycles is lower than predicted, then the predicted growth rate per spectrum block can still agree with the experimental value. Local under-predictions can he canceled by local over-predictions. The prediction model might appear to he acceptabie, but it is physically still not correct. Prediction models should he verified by still more detailed crack growth measurements. Ultimately, the crack rate should he measured in each individual cycle. Experimentally it implies that striation spacings have to be measured. Fractographic observations in the electron microscope are essential for that purpose. Experiments were carried out on 2024-T3 (bare) and 7075-T6 (clad) specimens with simple VA load histories (Fig.l). Simple load spectra were chosen to be sure about striation pattems which might be vi si bie in the scanning electron microscope (SEM). The experimental conditions and load histories are described first, followed by results, a discussion and some conclusions. It is not the the aim as yet to compare the observations with results of aprediction model. Experimental conditions Crack growth fatigue tests were carried out in an electro-hydraulic fatigue machine equipped with MTS TestStar computer controlled load monitoring. The tests occurred in normal lab air. Central cracked tension specimens were used, see Fig.2. The specimens were produced from 4 mm thick plates. A lower thickness might have given shear lips in a too early stage ofthe crack growth. Plates of2024-T3 bare and 7075-T6 Clad material were used as they were available