Influence of defects on axial fatigue strength of maraging steel specimens produced by additive manufacturing

Nowadays many materials such as steels, aluminium and titanium alloys can be realised by powder bed solutions melting subsequently powder layers by means of a laser or electron beam (Laser Beam Melting – LBM and Electron Beam Melting – EBM). The microstructure realised by layer-by-layer solidification having high cooling rate cannot be considered isotropic. Therefore, the mechanical properties could be influenced by the building direction. Regarding maraging steel, the study of the influence of the building direction and the heat treatment on the static and axial fatigue strength has been investigated in a previous contribution. A large scatter of the fatigue test results was found because of the presence of detrimental surface and subsurface defects. The aim of this contribution is to present additional axial fatigue test results of maraging steel characterized by different build orientation and providing an analysis of the defects observed at the crack initiation area of the fracture surface.


Introduction
In PBF technologies, such as Selective Laser Melting SLM (or Direct Metal Laser Sintering DMLS) and Electon Beam Melting (EBM) layers of metal powder are subsequently melted by a laser or an electron beam source in order to obtain the part designed by a CAD software without the typical constraints of traditional manufacturing processes [1][2][3][4].
The main advantage of adopting an AM process is the possibility of producing components having customised and complex geometries. Indeed, from structural point of view, topological optimization is the key task in design for additive manufacturing for maximising the performance of the component for a given set of loading.
As reported in literature, the application of design for AM are currently reserved to aerospace, aeronautical and automotive fields because at the moment this new manufacturing process takes advantage in customized components with low production volumes due to the high process costs.
Conversely, the critical point regarding the design of AM mechanical parts is the presence of defects which degrades the fatigue strength of the material [11,12]. Indeed, main causes of lower fatigue performance in AM parts were imputed to the surface finishing, the post heat treatment and the presence of defects currently unavoidable from the manufacturing process [13][14][15]. Typical defects that can be observed in SLM parts are incomplete fusion holes having irregular 3D shapes also known in literature as Lack Of Fusion (LOF). Other defects are pores with elliptical or spherical shape and crack-like defects. The formation mechanism of these defects is explained in [16].
For a design point of view, an accurate estimation of the fatigue limit, intended as the maximum stress that guarantee a fatigue life of a chosen number of cycles, cannot be done without taking into account the mechanics of the defects.
As well-known from [17], the fatigue strength is controlled by the defects having the maximum size and given the complex geometries of the LOF is appropriate to adopt the √area parameter.
Therefore in some contribution [18], statistics of extreme values has been applied in order to estimate the maximum size of the defect starting from inspection by CT scanning.
The aim of this contribution is to add axial fatigue results of maraging steel produced by SLM to those previously published [19] and quantify the √area of the defects present at the crack initiation zone of the fracture surface of the specimens by using a Scanning Electron Microscope (SEM).
Two batches of cylindrical maraging steel specimens were considered in the contribution. The first batch (E) is produced by a EOSINT M280 system, while the second one was provided by SISMA srl, that uses SISMA MYSINT 100 system to produce components made of maraging steel and other alloys. The material under investigation is maraging steel or also commonly known as 18Ni 300. The chemical composition of powders and the process parameters adopted to produce fatigue testing samples are reported in Table 1 and 2, respectively. The specimen's geometry designed for fatigue testing is shown in Fig. 1. For each batch the specimens were built with their axis oriented both at 0° and 90° with respect to the building direction that correspond to the translation of the platform which allows the deposition of the subsequent layer during the process. A schematic sketch of the orientation of the samples can be seen in Fig. 2. In addition, half of the total number of specimens were subjected to aging heat treatment for 6 hours at a temperature of 490°C and then air cooled. For a clear comprehension of the results in the present contribution the nomenclature of the single test series is summarized in Table 3, where the prefix E and S are referred to EOS and SISMA respectively, 0° and 90° indicate the building direction and NT / T refer to Not heat-Treated and the heat-Treated condition, respectively.

Fig. 2.
Orientation of specimens' axis with respect to the building direction.

Testing protocol
Push pull, constant amplitude fatigue tests were carried out by using on the S batch and results were compared with the E batch previously tested [19]. All the specimens were polished by using progressively finer emery paper from grade 80 up to grade 800.

Fig.3.
Deflection of the specimen's axis f a observed in [19].
Since it is known that in additive components the residuals stress induced by rapid cooling of the melted zone might cause geometrical distortion (as demonstrated in [19]), the gross section of each specimens were turned in order to reduce misalignments. Before machining, deflection of the specimen's axis f a (Fig. 3) was measured for all specimens. The fatigue tests were carried out by using a servohydraulic SCHENCK HYDROPULS PSA 100 machine having a 100 kN load cell and equipped with a TRIO

Roughness
In Fig.4 is reported a histogram of the Ra mean values measured on the surface of 5 specimens per test series along the longitudinal direction. It can be seen that all test series present a Ra value approximately 0.5 μm.

Geometrical distortion
Geometrical distortion of each batch, in terms of deflection of the specimen's axis f a , was measured, and the synthesis of the mean values is reported in Fig.5. In the previous work [19] deflection values were used for evaluating the mean stress that is caused by the machine gripping system. Contrary, in this contribution the gross section of specimens provided by SISMA srl, were turned in order to avoid mean stress state during fatigue testing, therefore the value of f a is reasonable negligible.

Fig. 5.
Mean value of deflection f a for each test series.

Hardness
Micro-hardness (HV) was measured in one specimen per test series and the results are reported in Fig.6. Due to the aging hardening heat treatment the hardness of the is doubled with respect to the NT series. This was not observed in S batch, where it seems that the aging treatment has not been effective.    Fig. 8 in terms of SWT parameter. As a result of turning, the SWT value of the specimens of the batch S is simply the applied nominal stress amplitude. For all details of E batch results the reader is referred to [19].

Fractography
In Fig. 9 the available fracture surfaces of specimens 0° oriented at the crack initiation point are shown. In agreement with results from the literature of the recent past, the cause to failure in AM components are LOFs, and such defects present irregular shapes. As reported in Fig. 9, the effective area adopted to evaluate the √area parameter is highlighted by a red dotted line in each image. To choose the effective areas, suggestions reported in [17] were followed.

Conclusions
In this contribution, an experimental analysis of the defects that affect the fatigue behaviour of two batches of maraging steel specimens produced by additive manufacturing has been reported. Both batches are divided in four test series by taking into account the build orientation and heat treated/as-built condition (see Table 3). As observed on the specimen related to the (E) batch, geometrical distortion caused by residual stresses also occurred on the (S) batch, with more relevance in specimens which were not heat treated (NT). While for the (E) batch tested in the previous work the effect of mean stress caused by geometrical distortion had to be taken into account in this contribution the specimens' gross sections were turned to avoid misalignments. For comparing the two batches, fatigue test results were plotted in terms of SWT parameter. All the fatigue data related to (S) batch show a fatigue strength lower than those related to (E) batch. Focusing on the single test series of the (S) batch, the specimens oriented at 0°, both (NT) and (T), exhibit a fatigue strength from 45% to 60% higher than those oriented at 90° (fatigue strength evaluated at 1 million cycles) in disagreement with the results of the (E) test series [19]. It is worth highlighting that the values of the micro-hardness obtained on (S) specimens heat treated are not in agreement with the (E) ones and those reported in literature.
The inspections of the fracture surfaces have demonstrated that the fatigue failures of the AM components are mainly caused by the defects, which in literature are referred to as Lack Of Fusion (LOF). In agreement with the literature, this kind of defects present an irregular shape and are mainly located both at the surface and at the sub surface area of the component. In the examples reported in Fig. 9, it can be observed that the effective area of the defects related to the (E) batch are about 4 times smaller than those calculated for the (S) batch. In addition, at the crack initiation area, a cluster of small defects is often observed; one possible interpretation is to consider such cluster as a single defect having an effective area that circumscribes all defects, but such hypothesis has not been validated yet.