PSI - Issue 7

Jean-Yves Buffiere / Procedia Structural Integrity 7 (2017) 27 – 32 Jean-Yves Bu ffi ere / Structural Integrity Procedia 00 (2017) 000–000

28

2

of the order of 1 µ m or better. Indirect methods give access to bulk information but the crack shape has to be assumed and resolution can be limited specially for sub surface shallow cracks ( ∼ 50-100 µ m ). X ray attenuation and / or phase contrast tomographic imaging gives direct access to the 3D shape of fatigue cracks. When di ff raction is used, access to local crystallography is also possible (Ludwig et al. (2009)). In the last ten years those tomographic techniques have shed new light on the study of fatigue cracks but some limits also exist. In this paper those limits are first recalled and, in a second part, some results illustrating recent developments are presented, first for cracks initiating from natural defects, and then for artificial notches. X-ray tomography is based on the reconstruction of a 3D image from radiographs. The principle of the technique has been well documented in several books (Stock (2008); Bu ffi ere and Maire (2014)) and is not reminded here, instead we emphasize the limits of the method inherent to the imaging of cracks. In theory, a crack can easily be imaged by X ray tomography as, when present in a material, it produces a strong variation of the attenuation coe ffi cient µ . However, because fatigue cracks have an opening in the sub-micron range, voxel sizes in the micrometer range are necessary to obtain a good description of the crack fronts (Bu ffi ere et al. (2010)). Phase contrast imaging helps to increase crack detection (Cloetens et al. (1997)) but can also produce some artifacts ( Bu ffi ere et al. (2006)), it is mainly available at synchrotron sources. Therefore, up to now, to the best of the author’s knowledge, all tomographic in situ observations of growing fatigue cracks have used synchrotron radiation (see Bu ffi ere and Maire (2014) for a list). A direct and important consequence of the small voxel size is the thickness of fatigue sample which has been restricted, so far, to 1 to 3 mm, depending on the voxel size, unless some local tomography is performed (Stock (2008)). For a metallic material with a grain size of the order of 20 to 100 µ m a stable fatigue crack initiated and propagated in situ can cross a maximum of 10 to 20 grains making it, de facto , a microstructurally small crack (Bu ffi ere et al. (2006)). For the same reason, when defects are responsible for crack initiation their sizes might be appreciably smaller than in the real application e.g. in the case of pores in the bulk of cast materials which can reach millimeters. Beamtime availability at synchrotron sources is scarce, hence, in situ fatigue experiments have to be carried out in a limited amount of time. With the development of the new generation of CCD detector based on CMOS technology, the time necessary to record the set of projections required for the reconstruction of the 3D image (scan) is of the order of a few minutes (typically from 1 to 5 minutes). Hence most of the time is spent cycling the samples. In order to be able to study at least 3 to 4 samples during an experimental session (typically from two to four days), fatigue tests are expected to last from 100 to 400 kcycles (loading frequency isin most fatigue machines is of the order of 10 to 50 Hz). The stress levels required to reach those fatigue lives are generally at the upper limit of the high cycle regime of many metals and this has the drawback of reducing crack interactions with microstructure. Finally, the natural shape of a tomography sample is that of a slender structure making it di ffi cult to apply high compressive loads which can produce buckling. As a result many studies have focused on positive R ratio values, up to now. Some of our first observations on cast Al alloys containing pores have shown that, in this type of material, crack initiation occurs first at the intersection between the pores and the surface (where the stress is higher) or at subsurface pores and that grain boundaries are e ffi cient obstacles during this process. The number of fatigue cycles necessary to form a crack around a pore (a stage from which propagation laws can be used ... if available) represents a large fraction of the fatigue life (Ferrie et al. (2005)). Values reported in this study are however probably underestimated because of the relatively large stress level used, as explained in section 2 . As a matter of fact, ex situ 3D observations of samples cycled at lower stress levels had shown a signification fraction of arrested cracks (Bu ffi ere et al. (2001)) with a 3D shape that could be correlated to the presence of grains (Ludwig et al. (2003)) In a more recent study, again on cast Al alloy, it was found (Serrano-Munoz et al. (2017)) that the probability of initiation from an internal casting defect is extremely low compared to that of initiation from a surface defect even 3. Crack initiation and propagation from natural defects 2. X ray tomography for observing fatigue cracks: current limits