Mechanical Behavior of Differently Oriented Electron Beam Melting Ti–6Al–4V Components Using Digital Image Correlation

Freedom of design and efficiency are among the multiple advantages that have made metal additive manufacturing (AM) so popular [1]. Titanium has two allotropic forms. At low temperatures, the alpha form is characterized by a hexagonal-close-packed crystalline structure up to 882 °C, above which it transforms to beta with a stronger, less deformable, body-centered-cube structure. The addition of elements can increase or maintain the temperature range of stability of phases. For alpha phase, aluminum is among the most common stabilizers, while vanadium is popular for beta. Alpha-beta titanium alloys contain small amounts of beta stabilizers, which cannot strengthen the alpha phase; hence, alpha stabilizers are added. Thus, alpha-beta alloys display bimodal microstructures with features of both phases. The mechanical properties of these alloys depend on the amount and distribution of the alpha and beta phases. The distribution and quantity of these phases are normally controlled by heat treatment. For example, given its strength, lightweight, material compatibility, and resistance at extreme conditions, Ti–6Al–4V is probably the most attractive selection among titanium alloys for aerospace, biomedical, and other fields demanding high performance metals.

Nowadays, it is known that energy management of electron beam melting (EBM) manufacturing settings [2] and process can promote the formation of convenient allotropes [3]. For example, in alpha + beta titanium alloys, a microstructure richer in alpha phase may be more convenient for components prone to fatigue [4]. However, the additional multiple variables intervening in the formation of new material phases, such as cooling rate and manufacturing practices, convey high uncertainty in producing tailored microstructures [5]. In addition, the layering orientation of components fabricated by EBM may affect their strength; for example, the higher strengths are expected in parts that are loaded parallel to the layering pattern [6].

The concern over the numerous variables and its effects on the performance of EBM builds is continuously addressed by researches [7,8]. From a totally opposed and advantageous point of view, these factors compromising the mechanical properties may represent multiple opportunities in developing components with engineered failure modes, to enhance the safety and survivability of systems and structures.

In this work, EBM Ti–6Al–4V specimens were commercially fabricated at 0 deg, 30 deg, 60 deg, and 90 deg, and tested per ASTM E8 Standard Test Methods for Tension Testing of Metallic Materials [9], followed by failure analysis. Contributing to the extended characterization of AM metals, strain fields were generated by three-dimensional (3D) digital image correlation (DIC) for the investigation of shear effects developing due the manufacturing orientation. Thus, in this work, additionally to the typically reported values of strength, axial strain, and elastic modulus, the Poisson's effect of the specimens was measured. Finally, a study case of a specimen with an artificially created brittle failure mode is presented and discussed.

The fabrication started from drafting 75 mm × 15 mm cylindrical rods in solidworks. The rods were fabricated using Arcam AB's Ti–6Al–4V pre-alloyed powder, with spherical particle size from 40 to 100 μm, in an Arcam A2 system set for 50 μm layers. The bars were fabricated in variants of orientation at 0 deg, 30 deg, 60 deg, and 90 deg, represented by an arrow pointing in the manufacturing direction and stripes rendering the layering pattern (Fig. 1). Only the rods at 30 deg and 60 deg required supports for fabrication (Fig. 2).

Three specimens of each orientation (Fig. 1), for a total of 12, were CNC machined with sharp tool, at low speed and high feed rate, as per Arcam's recommendations; the sample average gauge diameter and length are 5.9 mm and 37 mm, respectively. One specimen of each orientation was subjected to DIC testing, and these were prepared with speckle patterns (Fig. 2). The components were fabricated at the best possible similar conditions, under the best commercial production practices by Keck Center. Furthermore, the orientation and identification of specimens were given special attention and preserved all the time. The specimens are labeled by a first number indicating its build orientation, followed by the term “deg” for degrees, which capitalized indicates measurements by DIC, and a unique ID number for the specimen (e.g., 60deg12).

Noteworthy is that although the constant manufacturing settings, the manufacturing space restrictions and productivity compromises of EBM systems, could prohibit all the specimens from being fabricated in a single batch. This implies that each batch has its characteristic thermal gradient, and so does each component in it, depending on its own geometry, location, and orientation in the manufacturing chamber. As the components are not subjected to the same thermal gradient, the microstructure of the builds is altered in different manners that potentially can compromise their mechanical properties; thus, the importance of standardized fabrication practices to ensure quality components.

Testing was conducted on an Instron 5969 UTM with installed load cell capacity of 50 kN, equipped with threaded grips. An Instron 25.4 mm axial clip extensometer was used for axial strain measuring. Additionally, a 3D digital image correlation system was setup to measure axial strain and strain fields by means of a virtual extensometer and virtual gauges available in vic-3d and vic-gauge software, all from correlated solutions.

Following ASTM E8 installation and recommendations, the specimens were tested in an Instron 5669 UTM with threaded grips. As the specimens were considered potentially sensitive to strain rate, given the investigation on the effects of the layering orientation, the testing speed was set at a controlled crosshead speed equivalent to a strain rate of 0.003 mm/mm·min, until failure.

For the axial strain measurement, an Instron 25.4 mm clip extensometer was carefully installed, avoiding damage to the painted speckle pattern. For one specimen of each orientation, the axial strain was dual-measured by the clip extensometer and the 3D DIC system. This dual setup helped in validating the DIC for strain measurements. The DIC and the UTM systems were synchronized to capture images and record data at time intervals of 0.5 s; thus, the comparison of the strain measured by the DIC and the stress measured by the UTM, at the same instant of time, allowed to obtain stress–strain curves with strains measured by virtual and clip extensometers simultaneously.

Studies have utilized DIC to understand the deformation behavior of components. For example, Anzelotti et al. [10] Daggumati et al. [11] demonstrated DIC suitable to measure strains with high accuracy in fiber composites. Rajan et al. [12] calculated an error of approximately 5% for a virtual DIC extensometer compared with a laser extensometer. The comparison of the stress–strain curves obtained herein with the clip and the virtual (DIC) extensometers, resulted in curves with an estimated error of ±1.5%.

The 12 specimens, regardless of their orientation, displayed stress–strain curves with bilinear behaviors. The average yield strengths were found in the range of 700–1100 MPa, comparable to the 950 MPa reported by Arcam, and the 758 MPa and 860 MPa required in ASTM F1108 and ASTM F1472, for cast and wrought Ti–6Al–4V, respectively, (Fig. 3 and Table 1). Although the specimens were fabricated under the same manufacturing settings and the process presented no interruptions, some samples displayed large variability. This consistency in the fabrication process leaves the precursor as the most probable factor affecting the variability of the mechanical properties in the components. The sieving in the recovery of the precursor powder does not guarantee that only particles with dimensions smaller that the sieve openings are reclaimed. In this manner, larger and elongated particles formed by two sintered spherical particles can be found in reclaimed precursor. The larger particles may require higher energy from the beam to form uniform melting pools, thus, resulting in internal flaws. In this context, the reduced strength and large variability of specimens oriented at 30 deg may have resulted from the presence of a larger number of flaws, compared to the other orientations (Figs. 8(c) and 8(d)). Moreover, comparing the median specimen from each orientation, the yield and ultimate tensile strengths were found to be inversely proportional to the manufacturing angle (Fig. 4 and Table 1), except for the 90 deg orientation, where stress concentrations at internal flaws may not originate shear effects, as suggested by the strain fields (Table 2). The average Young's modulus displayed no significant variation suggested due to the manufacturing orientation and was close to the 120 GPa reported by Arcam. Poisson's ratio was measured by DIC gauges on one specimen of each orientation, recording values of 0.32, 0.30, 0.29, and 0.33 for 0 deg, 30 deg, 60 deg, and 90 deg specimens, respectively, (Table 1).

Fractographs for failure analysis were captured by scanning electron microscopy (SEM) in a Hitachi TM-1000 at an accelerating voltage of 15 kV, the fracture profile images with a Revenece VHX-S50 digital microscope, and microstructural features, etched with Kroll's reagent, were observed using a Leica inverted-light microscope.

Regardless of the orientation, the optical fractographs displayed ductile features common in titanium; dimples with the presence of cracks and voids, some resulting from sintered powder particles (Figs. 5(a) and 5(b)). SEM fractographs displayed a ductile-to-brittle fracture with dimples, voids, small fibrous features, and some transgranular cracking (Figs. 5(c) and 5(d)). These observations were in accordance to what is expected for Ti–6Al–4V with a uniformly distributed bimodal microstructure (Fig. 6). Whereas shear effects were not so evident on surface fractographs, the fracture profiles taken at the same orientation as the DIC, and perpendicular to the layering pattern, displayed features suggesting its presence (Fig. 7). The analysis of fracture profiles, for the 0 deg and 90 deg specimens, displayed defined fracture features, such as slight necking, shear lips, and a failure plane normal to the main axial stress. These are characteristic features of flat-face failure modes; a mode corresponding to a planar strain state from a triaxial tensile-stress system. On the other hand, the features in the fracture profiles from the asymmetric specimens (30 deg, 60 deg) were not defined with clarity, indicating the development of a stress system, different than triaxial, and therefore, the presence of shear stresses potentially influencing the failure mode (Fig. 7). Moreover, flaws found close to the fracture in asymmetric specimens (30 deg, 60 deg), could have developed asymmetric stress concentrators and therefore shear effects (Fig. 8).

Supporting the existence of shear effects suggested by the fracture profiles, strain field recordings developed by DIC showed particular shear strain fields ε_xy (Table 2) for the different orientations. The axial-asymmetric specimens (30 deg, 60 deg) presented scattered shear concentrations ε_xy (Table 2(b)) and Table 2(c)). The shear strain concentrations ε_xy found near the fracture areas of the 30 deg and 60 deg specimens (Fig. 9) suggest that it added up to the strain tensor and likely defined the fracture mode. The axial-symmetric specimens (0 deg, 90 deg) displayed more uniformly distributed shear strain fields ε_xy (Table 2(a) and Table 2(d)). The uniformity and low magnitude of the shear strain ε_xy found in the 0 deg and 90 deg suggest that the fracture was practically dependent on the principal strain ε_yy only, as suggested by their fracture profiles (Figs. 7(a) and 7(d)). Therefore, as expected, the axial-symmetric specimens displayed von Mises strain fields practically identical to the axial strains ε_yy, resulting from the almost null and mostly uniform shear strain fields ε_yy (Table 2-(a) and Table 2-(d)). Furthermore, slightly larger differences were found in the axial-asymmetric specimens, when comparing the von Mises to the axial strains. This is a consequence of the shear strain concentrations presented in the fields. Additionally, for the 30 deg and 60 deg orientations, the shear strain fields displayed maximum values on the failure region; therefore, supporting the argument that shear effects contributed to define the location of the fracture (Fig. 9).

A potential advantage of the performance of AM metal components over those traditionally fabricated is that the proper management of manufacturing settings and process can lead to developing localized behaviors for specific and convenient failure modes. In this context, an experiment with one of the tension test specimens showed that it is possible to eliminate the plastic region from Ti–6Al–4V and emulate a purely brittle material failure (60deg99, Fig. 3(c)), a useful behavior for fail-safe components. The 60deg99 specimen showed a value of 770 MPa, for both ultimate and yield strengths, which are approximately 25% and 30%, respectively, lower than the average of the rest of the 60 deg specimens (Table 1). This reduction in strength should be considered and addressed for potential applications. The artificial brittle failure was possible by interrupting the fabrication process, and a careful restart procedure found to maintain the strength of the component [13]. A proper energy management in the EBM process, at the restart, can sinter the metal powder particles to artificially create the conditions for an oriented intergranular-like failure plane to occur between sintered particles, instead of between metal grains. In this manner, brittle and ductile features could be found in the same fracture surface, suggesting the possibility of localizing and exert control over strains within the same build (Fig. 10).

There are many factors affecting the mechanical properties of printed metals. These variables are interrelated and make it difficult to predict their influence on the material properties. The interdependence of these factors originates increased variability compared to more traditional manufacturing methods. For example, the cooling and solidification rates during the EBM process affect the microstructure of the metal, but the thermal gradient of a given part being fabricated is dependent on other factors such as its morphology. In addition, manufacturing practices and settings such as beam power and speed may have more direct effects [14,15]. In that context, the variability shown in the measured mechanical properties are the result from variations during the manufacturing process affecting the microstructure. Moreover, texture analysis of the microstructure may be proposed for further investigation on the correlation of microstructure morphology and strength variations, as suggested in Ref. [16]. In addition, the variability in the strength of components can be affected by internal flaws and defects that may exist and whose quantification and effect on the material properties is the subject of further study. However, on the investigation of the lower strengths displayed by the builds at 30 deg, the specimens were sectioned perpendicularly to the layering pattern, finding more and multiple flaws (Figs. 8(c) and 8(d)) that could act as stress concentrators, resulting in the lower strengths and increased variability, compared to the other orientations (Table 1).

The layering pattern in 30 deg and 60 deg specimens, that is, oblique to the axial load, inherently grant the same orientation to the internal flaws and defects, and these, in turn, can develop oriented stress fields. Following the von Mises failure criterion, these developing shear stresses add up to the principal stresses, resulting in higher stress states leading to failure. However, the possibility of intentionally introducing oriented defects can be seen advantageous in designing beneficial failure modes (Fig. 10). Recalling the relationship between the shear modulus and the Poisson's ratio, the presence of larger shear effects suggests the reason of the smaller Poisson's effects measured in the 30 deg and 60 deg components compared to the 0 deg and 90 deg (Table 1). Therefore, probably comparable to the fabrication of composite materials, the assurance of consistent manufacturing practices for quality builds, and minimum flaws, is extremely important in consistently predicting materials performance. Quality assurance of additive manufacturing metal components is still an extremely challenging task, given the sensitivity of the components to the many known, and probably still unknown, interdependent factors such as precursor quality, cooling rate, manufacturing process, manufacturing settings, manufacturing orientation, and others [7,16,17]

At first sight, the layering approach of powder bed fusion manufacturing may suggest that shear effects could potentially develop under uniaxial loading. In the investigation of the existence of these shear effects, the micrographs of the specimens showed a uniform metal matrix, reported also in Ref. [5], with neither evident visible layer interface, nor phase discontinuity (Fig. 6) that could potentially introduce shear stresses. However, the failure profiles gave some insight about developed shear effects, probably originated from oriented flaws acting as stress concentrators (Fig. 8).

The observed uniformity of the layering interface and material phase suggests that they have minor influence in producing shear strains, and probably, indicating the flaws as the major contributors of these effects. It is important to be aware of the possibility to design failure modes by the development of strain fields counteracting other threatening strains; this can be achieved by the intentional inclusion of manufacturing defects in the metal matrix that can also bias the component to a less detrimental cracking.

Many interdependent factors threaten the quality of components fabricated by EBM, and some effects may still be unknown. Thus, maintaining the manufacturing settings, practices, and process akin as possible is highly advisable to reduce the variability of the mechanical properties and improve the reliability of the components by statistical approaches.

Digital image correlation was exhibited as a reliable method for strain measurements; the system permitted determining commonly reported mechanical properties such as elastic modulus and yield strength, but also measuring Poisson's ratio and shear strains in real time. The advantages of strain measurements by this method may not be as valuable when testing specimens with simple shapes, as they are when investigating components with complex geometries requiring very specific, even nonexistent, instrumentation.

Although the manufacturing layering process did not produce highly noticeable features, and microstructures were shown uniform, the build angle may influence the performance of components due the introduction of flaws and defects producing stress fields, also oriented. The orientation observed in DIC strain fields was approximately that of the layering patterns of 30 deg and 60 deg, and therefore, evidenced the development of shear stresses capable of influencing the failure mode; which the features on the fracture profiles suggested.

Additive manufacturing empowers to exert influence on the behavior and properties of materials by promoting the formation of diminutive features, previously prohibited by size. Exceling on methods to create these small characteristics will lead to the fabrication of safer systems and structures utilizing AM parts with engineered failure modes.

This work was supported by the W. M. Keck Center for 3D Innovation. The authors are grateful to Israel Segura for the EBM fabrication of the specimens tested.

• The National Science Foundation (Grant No. 1405526).