Statistical meta-analysis of additively manufactured (AM) 316L austenitic stainless steel and Ti-6Al-4V have revealed the importance of transition between defect and microstructure driven fatigue. Even in cases where the presence of processing defects is mitigated, they can still detrimentally impact fatigue performance. In order to understand the degree to which remaining processing defects are impacting fatigue performance, a tool is required which can provide estimates of fatigue performance based off of non-destructively identified processing defects. Quantitative fractography techniques developed on Ti-6Al-4V have provided the capability to connect specific features of crack initiating defects to fatigue life in such a way that defects imaged using x-ray computed tomography (XCT) a fatigue life-based ranking of defect severity. This pipeline between quantitative fractography to a large quantity of XCT data acts as a database for statistical survival analysis of new processing defects to be performed.
Validation was in part performed using two sets of fatigue specimens. Specimen chemistry, which was measured using direct current plasma emission spectroscopy and inert gas fusion and combustion, was 6.42 wt.% Al, 3.94 wt.% V, 0.18 wt.% Fe, 0.0172 wt.% C, 0.0015 wt.% S, 0.1374 wt.% O, and 0.0223 wt.% N. All material used to fabricate the test specimens was produced using the L-PBF process on a 3D Systems Model ProX 320 system. All specimens had a layer thickness of 60 Ám and were printed at 90░ to the build plate.
Manufacture and testing of specimens in the first two rows of Table 2 is described elsewhere . All other fatigue specimens were scaled down to have a 5.5 mm gauge diameter after machining. Sub-sized specimens were tested on an MTS 810, servohydraulic load frame with a 100 kN axial load cell. Specimens were tested at a load ratio of 0.1 and frequency of 15 Hz at stress maxima of 400 MPa and 780 MPa.
Prior to testing, each specimen was analyzed using x-ray computed tomography (XCT) to characterize the presence of process-related and intentional defects. A GE v|tom|x L300 nano/microCT system was used with an accelerating voltage of 220 kV, a beam current of 0.045 mA, source-to-detector distance of 800 mm, and an exposure time of 2,000 ms to produce a voxel size of 10 Ám.
After testing, the fracture surface of each fatigue specimen was characterized using an Apreo S Scanning Electron Microscope at an accelerating voltage of 2.0 kV, a beam current of 0.1 nA, and a dwell time of 500 ns.
Confocal white light interferometry was performed on crack initiating defects to obtain height maps. A numerical aperture of 0.5 was used, as calculated using Sparrow’s criterion.
Results and Discussion:
Over half a million processing pores were characterized through XCT to create a large database to compare pore characteristics and estimate severity off of. Pores were first assessed in terms of volume. The use of empirical cumulative distribution functions provided a quick means to show how the collection of smaller pores combine into the overall porosity of the material. By calculating the skewness and kurtosis of pore volume distributions, it was found that they follow a strong log-log relationship with each other. Specimens with larger pores tend to have more of those pores. Statistical modeling demonstrated how this trend can be connected back to the choice of processing parameters.
Pore morphology was chiefly characterized according to the Zingg shape criterion which classifies morphology as spherical, rod, blade, or disc. Although a wide range of defect morphologies are obtained when comparing flatness and elongation, the majority of defects images were classified as either spheres or rods. While illuminating, such a result is not necessarily actionable.
The fractal-topology technique for quantitative fractography was used to evaluate the geometric characteristics of internal and sub-surface initiated fatigue cracks. Through these measurements, a strong quadratic relationship was observed between fisheye area normalized by specimen diameter and fatigue life. Modeling in-vacuum crack growth as an ellipse established the first equation to predict defect severity. Additional details about the defect were incorporated by generating height maps around the defect. Assuming a symmetric fracture surface near initiation, estimates of the radius of curvature were made to include information about defect morphology.
Fractographic relationships were applied to the pores characterized through XCT. Linear regression of these metrics provided the metric of defect severity. Kaplan-Meier type survival curves were produced to visualize the probability of a certain type of processing defect initiating failure. When defect severity was scaled by the ratio of tensile strength to applied stress, experimental fatigue lives matched closely with the bend of the knee of the survival curve where fatiguing past this defect severity becomes less probable. A 90% average value can be applied, and tolerance band created by considering plus or minus one standard deviation to develop a range of tolerable fatigue lives.
XCT techniques provide a means to image a large population of processing defects. While these defects may be characterized in terms of size, location, or morphology, there still is a lack to connect XCT data to experimentally obtained fatigue lives. Quantitative fractography provides a means to do so. Modeling fatigue crack growth and final ligament length of internal and sub-surface fatigue cracks provided a means of estimating the degree of in-vacuum crack propagation. This metric followed a strong quadratic relationship with fatigue life. Refinements to this metric were obtained by estimating defect radius of curvature from height maps. Both of these metrics are readily transferable into a set of equations that when applied on XCT data yield a ranking of defect severity. Subsequent survival statistics provide a means to estimate tolerable fatigue lives given the presence of a defect of certain severity.
Keywords: titanium alloys, additive manufacturing, fatigue, defects, x-ray computed tomography