Effect of Pore Defects on Very High Cycle Fatigue Behavior of TC21 Titanium Alloy Additively Manufactured by Electron Beam Melting

Abstract

Titanium alloys additively manufactured by electron beam melting (EBM) inevitably obtained some pore defects, which significantly reduced the very high cycle fatigue performance. An ultrasonic fatigue test was carried out on an EBM TC21 titanium alloy with hot isostatic pressing (HIP) and non-HIP treatment, and the effect of pore defects on the very high cycle fatigue (VHCF) behavior were investigated for the EBM TC21 titanium alloy. The results showed that the S-N curve of non-HIP specimens clearly had a tendency to decrease in very high cycle regimes, and HIP treatment significantly improved fatigue properties. Fatigue limits increased from 250 MPa for non-HIP specimens to 430 MPa for HIP ones. Very high cycle fatigue crack mainly initiated from the internal pore for EBM specimens, and a fine granular area (FGA) was observed at the crack initiation site in a very high cycle regime for both non-HIP and HIP specimens. ΔK_FGA had a constant trend in the range from 2.7 $\mathrm{MPa}\sqrt{\mathrm{m}}$ to 3.5 $\mathrm{MPa}\sqrt{\mathrm{m}}$, corresponding to the threshold stress intensity factor range for stable crack propagation. The effect of pore defects on the very high cycle fatigue limit was investigated based on the Murakami model. Furthermore, a fatigue indicator parameter (FIP) model based on pore defects was established to predict fatigue life for non-HIP and HIP specimens, which agreed with the experimental data.

1. Introduction

Titanium alloys were widely used in aerospace due to their high specific strength and excellent corrosion resistance [1]. The components fabricated by additive manufacturing (AM) were more efficient, had shorter cycles, and free structural design [2,3,4]. Compared with selective laser melting (SLM), the EBM titanium alloy underwent a special thermal history due to the influence of high preheating temperature, resulting in an α + β dual-phase microstructure and low residual stress [5]. Fatigue fracture of these EBM titanium alloy components in aerospace still occurred when subjected to high frequency and low-stress cyclic loads during ultra-long life service (>10⁷ cycles), i.e., VHCF, where fatigue cracks initiated from the specimen interior, and FGA was observed at the crack initiation site [6,7]. Therefore, the VHCF performance of titanium alloy components fabricated by EBM was a key factor for large-scale applications.

The widespread application of AM components was limited due to their inherent defects, such as pores, lack of fusion (LoF), and mirocracks. Fatigue properties of EBM titanium alloy components were still lower than that of the forgings ones, owing to the promotion of crack initiation by defects [8,9]. The Murakami model was widely used to evaluate the relationship between fatigue strength and defects in AM components [10]. The fatigue scatter and size effect of AM materials with defects were investigated based on the extreme value statistics theory and probability fatigue theory [11]. Furthermore, based on the weakest-link theory, a probability fatigue life model was established to analyze the location and size effects of defects [12]. The FIP model based on the stress intensity factor of defects had been applied to predict fatigue life for the titanium alloy [13].

In a very high cycle regime, fatigue crack initiation and fatigue life were much more sensitive to the defects in AM titanium alloys. The investigation by Günther [14] showed that very high cycle fatigue cracks initiated from pores and LoF for EBM titanium alloy, where FGA characteristics were observed around these defects. Fatigue performance was similar to that of the SLM specimens after stress-release treatment, indicating that the EBM specimens did not require stress-relieving annealing due to their low residual stress [8]. However, much research on very high cycle fatigue was focused on SLM titanium alloys [15,16,17]. The internal defects (pores, lack of fusion, etc.) and surface artificial defects dramatically reduced the very high cycle fatigue properties of the SLM titanium alloy, and fatigue strength models based on defect size were proposed based on the Murakami model [18].

HIP was widely applied to improve fatigue performance due to the reduction of defects in EBM titanium alloy components [19]. However, little attention was paid to the very high cycle fatigue behavior of the EBM titanium alloy treated by HIP. As for SLM titanium alloys, HIP significantly improved the very high cycle fatigue performance, which can be comparable to that of conventional forged ones, although HIP treatment coarsened the microstructure and reduced the tensile strength of the SLM Ti6Al4V titanium alloy [18,20]. A very high cycle fatigue crack initiated at the large α phase of SLM Ti6Al4V alloy was treated by HIP [14], suggesting that the reduction of the α phase size after HIP treatment can further improve very high cycle fatigue performance. Furthermore, it should be noted that the combination of HIP and surface machining can effectively enhance the fatigue performance of EBM titanium alloy components. HIP treatment did not generally improve the fatigue life of the AM titanium alloy without machining on the surface due to the effect of the surface roughness [9], while the specimens obtained excellent fatigue properties after HIP treatment and surface machining.

The very high cycle fatigue behavior of the EBM TC21 titanium alloy, which was developed for aerospace structural components, was investigated. Considering that LoF had a more negative impact on fatigue performance than pore defects [21], the EBM process with high energy input and low forming rate was adopted to restrain the dangerous LoF defects. Then, the effect of the internal pore in the EBM titanium alloy on very high cycle fatigue properties and the crack-initiation mechanism was discussed in comparison with that of HIP ones. Very high cycle fatigue strength and fatigue life of an EBM titanium alloy containing internal pore defects were quantitatively evaluated based on the Murakami model and FIP model, which was beneficial for the application of EBM titanium alloy components in aerospace engineering.

2. Experimental Procedures

2.1. Materials and Manufacturing

The TC21 titanium alloy was manufactured by EBM on an Arcam A2 machine (Arcam, Stockholm, Sweden) under a 5 × 10⁻³ Pa vacuum. The chemical composition (wt.%) of the powder is shown in

Table 1. Chemical composition of TC21 powder.

Al	Zr	Mo	Cr	Nb	Sn	O	H	Ti
6.26	2.13	2.38	0.96	1.84	1.87	0.092	0.0043	Bal.

. The diameter size of the particle was in the range of 50 to 125 μm. Preheating the chamber at 650 °C reduced the residual stress during EBM. To avoid the formation of dangerous LoF defects, an EBM process with high energy input and low forming rate was adopted to promote powder melting. EBM fabrication was operated at a voltage of 60 kV, beam current of 30 mA, spot size of 300 μm, and each layer thickness of 50 µm.

The parts on the substrate were a length of 180 mm, height of 100 mm, and width of 15 mm. Fatigue specimens were cut from the building parts, and their longitudinal axis was parallel to the build direction. In order to eliminate the pore defects, HIP for EBM specimens were treated by a QIH-15 machine (ABB company, Vsters, Sweden) at 920 °C under 1000 bar for 2 h, and then the specimens were cooled using nitrogen gas.

Hardness measurements were carried out by a FM310 (Future-Tech, Tokyo, Japan) machine for both HIP and non-HIP specimens. Hardness values were obtained from five tests under a load of 0.5 kg for 15 s.

2.2. Ultrasonic Fatigue Test

Fatigue tests were carried out using a USF-2000 ultrasonic fatigue test machine (Shimadzu, Kyoto, Japan) with R = −1 and a frequency of 20 kHz [22]. The investigation [23] indicated that the frequency had little influence on ultrasonic fatigue tests under low-stress amplitude. Detailed information about the ultrasonic fatigue test machine was present in the reference [24]. To control the thermal effects of high-frequency fatigue tests, compressed air was used to cool the specimens.

Based on the elastic wave theory, the specimen was designed to work at resonance with the amplifier, and the geometries of the fatigue specimens are shown in Figure 1. According to Figure 1, the ultrasonic fatigue specimens were processed from the EBM parts. Thus, the rough surface of the EBM parts was removed by mechanical processing.

Electro-polishing was treated on fatigue specimens to eliminate the machining influence on fatigue behavior. The electro-polishing process was under −20 °C and 20–25 V with 59% methanol, 35% n-butanol, and 6% perchloric acid. The fracture surfaces and microstructure were observed by CS3400 scanning electron microscopy (CamScan, Cambs, UK).

3. Results

3.1. VHCF Properties

Figure 2 shows the microstructure of the EBM TC21 alloy with non-HIP and HIP treatment, respectively. The microstructure for non-HIP treatment was basketweaved with a lamellar α phase, whereas the HIP treatment increased the size of the α phase. The mean hardness for non-HIP and HIP samples was 435 HV and 418 HV, respectively.

S-N curves of the EBM TC21 titanium alloy with the non-HIP and HIP treatment are shown in Figure 3. It showed that the curve of non-HIP specimens displayed a continuously descending characteristic, and fatigue life gradually increased with the decrease of fatigue stress. As the stress amplitude decreased, fatigue crack initiation characteristics changed from surface initiation to interior pore initiation without FGA and interior pore initiation with FGA. When the fatigue life exceeds 2 × 10⁶ cycles, FGA characteristics can be observed around the pores at the fatigue crack initiation site. If the specimens did not break at 10⁹ cycles, the stress amplitude can be defined as the very high cycle fatigue limit. Thus, EBM TC21 titanium alloy obtained a very high cycle fatigue limit of 250 MPa. Very high cycle fatigue S-N curve characteristics and fatigue limit of the EBM TC21 alloy were similar to that of EBM Ti6Al4V [14], but the fatigue limit was far lower than that of forged ones with a basketweave microstructure [25].

Figure 3 also illustrated that HIP treatment dramatically improved fatigue properties in comparison with the non-HIP condition. The HIP specimens obtained a higher fatigue life at the same cyclic stress. It displayed a two-step curve for the HIP specimens, which was similar to that of the forged ones [25]. The VHCF limit was increased from 250 MPa for the HIP specimens to 430 MPa for the non-HIP ones. There was a transition platform from crack surface initiation under high stress to internal initiation under a 550 MPa stress amplitude. The fatigue life of surface-initiated cracks was lower than that of internally initiated cracks due to the environmental and stress concentration factors. However, the non-HIP specimens exhibited a continuous decease feature due to the presence of defects.

3.2. Fractograph

For EBM materials, insufficient energy input resulted in incomplete melting of the powder and LoF, whereas excessive energy input led to an unstable molten pool and vaporization, forming pore defects. As for the non-HIP specimens, the pore defects had a significant impact on the fatigue crack initiation characteristics for the EBM TC21 titanium alloy. Figure 4a shows that under 600 MPa stress amplitude, a crack initiated from the specimen surface without pore defects at the initiation site, as indicated by the red arrow. When the stress amplitude decreased to 450 MPa, fatigue cracks initiated at a large pore inside the specimen (Figure 4b), where stage I propagation along the facet was observed near the pore. Under lower stress amplitude, fatigue cracks initiated at a small pore inside the specimen, and a fish-eye characteristic and FGA were observed near the pore (Figure 4c), which was a typical characteristic of VHCF of titanium alloys [6,7] and were consistent with that of the EBM Ti6Al4V alloy and SLM Ti6Al4V in the investigation by Günther [14]. It should be noted that even if the equivalent diameter of small pores in the EBM TC21 titanium alloy decreased to 12 µm, these small pores still promoted the VHCF crack initiation, and there were FGA characteristics around the pores (Figure 4d).

As for the HIP specimens, the pores in EBM specimens were eliminated by HIP treatment. Under low stress, cracks initiated at the inner of the HIP specimen in the long-life regime, as indicated by the red ring (Figure 5a). An FGA characteristic without pore defects was present at the crack initiation site (Figure 5b), which was similar to the forged TC21 titanium alloy [25] and SLM Ti6Al4V alloy after HIP treatment [18,20]. In comparison, the investigation by Günther [14] suggested that VHCF cracks initiated at the large α phase of the SLM Ti6Al4V alloy treated by HIP with faceted fracture characteristics, which was attributed to the difference in microstructure.

4. Discussion

4.1. Effect of Defect on Fracture Mechanism

Under high-stress amplitude, small pore defects were irrelevant to fatigue crack initiation, and the non-HIP specimens exhibited the surface crack initiation behavior (Figure 4a), and the slip band activity resulted in crack initiation. Fatigue crack initiated at the pore inside the specimens under low-stress amplitude. Stage I propagation along the facet was presented around the large pore under a 450 MPa stress amplitude (Figure 4b), while FGA was observed around the pore under a lower stress amplitude (Figure 4c,d). Figure 6 showed that the areas of the pore ranged from 100 µm² to 3840 µm², and there was no direct relationship between the area of the pore and fatigue life N_f. However, the area of FGA increased with fatigue life N_f. As FGA was formed by the discontinuous propagation of microcracks [6], it indicated that the consumption life in FGA was a major part of fatigue life.

The stress intensity factor range ∆K_pore at the front of the pore inside the specimens can be expressed as [10]:

$$\mathrm{\Delta }{K}_{pore}=0.5{\sigma }_{\mathrm{a}}\sqrt{\pi \sqrt{{area}_{pore}}}$$

(1)

where σ_a is the stress amplitude and ${area}_{pore}$ is the area of the pore at the crack initiation site. As for FGA, ΔK_FGA can be calculated by substituting ${area}_{FGA}$ into Equation (1).

Figure 7 shows that ∆K_pore was not directly related to fatigue life due to the random distribution of pore size but displayed a decreased trend. Similar to cast alloys, there exists the critical stress intensity factor K_cr with pores [26], and the value of K_cr for EBM TC21 titanium alloy was 0.78 $\mathrm{MPa}\sqrt{\mathrm{m}}$ in a very high cycle regime. The pore size was small enough, especially for HIP treatment, where K_pore was lower than K_cr; thus, the pore had no decisive effect on the cracks’ initiation and fatigue cracks initiated from microstructure inhomogeneity for EBM specimens treated by HIP treatment (Figure 5).

Furthermore, Figure 7 also shows that ΔK_FGA was independent of fatigue life from 10⁶ cycles to 10⁹ cycles and had a constant trend in the range from 2.7 $\mathrm{MPa}\sqrt{\mathrm{m}}$ to 3.5 $\mathrm{MPa}\sqrt{\mathrm{m}}$, which was considered as the stable crack growth threshold ∆K_th for titanium alloys [20,27]. The stable crack growth ∆K_th for forged the TC21 titanium alloy ranged from 3.8 $\mathrm{MPa}\sqrt{\mathrm{m}}$ to 4.5 $\mathrm{MPa}\sqrt{\mathrm{m}}$ [25]. The value of ∆K_FGA for the EBM TC21 titanium alloy was lower than that of the forged ones.

As for non-HIP specimens with 400 MPa/1.37 × 10⁶ cycles and 450 MPa/1.26 × 10⁶ cycles, the value of ∆K_pore reached that of ∆K_FGA, and fatigue cracks, which originated from these pores, could continuously propagate to failure. Therefore, the FGA was not observed around the pores for these two specimens. When ∆K_pore is lower than ∆K_FGA, fatigue cracks cannot propagate continuously, resulting in FGA. For the non-HIP specimen with 300 MPa/5.7 × 10⁷ cycles, the equivalent diameter of FGA was approximately 105 μm, and the average propagation rate was 8.77 × 10⁻¹³ m/cycle, which was much lower than the Burgers vector. Thus, microcracks discontinuously propagated within FGA.

VHCF cracks initiated from the heterogeneous microstructure in the HIP specimens (Figure 5). Many models, such as numerous cyclic pressing (NCP) [28] and FGA [29], were established to reveal the FGA formation. A heterogeneous cyclic strain concentration occurred at the α/β interface due to the highly inhomogeneous dislocation arrangement [30]. A nanocrystalline layer was formed at the stress concentration area under very high cyclic loading, and the nanocrystalline boundaries were separated to form fine-grain characteristics [28,29].

4.2. Effect of Defect on Fatigue Strength

To quantitatively analyze the fatigue strength of the EBM titanium alloy with pore defects, the maximum pore size on the fracture surface was estimated by the statistical extreme value method. The pore sizes were assumed to follow the Gumbel distribution [31], and the pore size of EBM titanium alloys was plotted using the Gumbel extreme value analysis. In Figure 8, i/(N + 1) was the cumulative probability corresponding to the maximum pore size in each field of view and arranged in descending order. The maximum pore diameter of the EBM titanium alloy was evaluated as 96.2 μm.

For EBM titanium alloys with internal defects, fatigue strength can be evaluated using the Murakami model [14]:

$${\sigma }_w=\frac{1.56\times (HV+120)}{{(\sqrt{are{a}_{pore}})}^{1/6}}$$

(2)

The parameter ${area}_{pore}$ was considered as the equivalent area of pores perpendicular to the direction of stress loading. Vickers hardness (HV) was the hardness of the material. The location parameter 1.56 was applicable to the internal defects.

Based on Equation (2), fatigue strengths of the EBM titanium alloy were 197 MPa, 230 MPa, and 394 MPa, respectively, corresponding to the maximum pore diameter of 96.2 μm, the average diameter of 60 μm, and the minimum diameter of 12 μm. The maximum size of the pore defect evaluated by the statistical extreme value method corresponded to the lowest limit of fatigue strength of the EBM titanium alloy. Compared with Figure 3, fatigue strength with the average pore size relatively approached the experimental data with an error of 8%. Fatigue strength corresponding to the minimum pore defect was much higher than the experimental data, while it was closer to the fatigue strength of the EBM titanium alloy after HIP treatment. It revealed that HIP reduced the pore size, thereby improving the fatigue strength of the EBM titanium alloy.

4.3. Effect of Defect on Fatigue Life

As mentioned above, the porosity defects significantly reduced the very high cycle fatigue life of the AM titanium alloy, and the defect factor should be considered for the fatigue life predicting model. The FIP model revealed the promoting effect of material defects and cyclic stress on fatigue damage and can be calculated as [26]:

$$FIP=\frac{\mu {\sigma }_a}{E}\left[1+k\frac{\mathrm{\Delta }{K}_{defect}}{\mathrm{\Delta }{K}_{th}}\right]$$

(3)

where Schmid factor μ was equal to 0.408, and parameter k was equal to 1 [26]. ∆K_defect was estimated by Equation (1), and ∆K_defect was considered as the mean value of ∆K_FGA. As for HIP specimens, the pore defects were eliminated, and the K_defect was equal to zero.

The fatigue life was numerically fitted with FIP parameters based on fatigue data of non-HIP and HIP specimens, and the expression was

$$N_f=7.86\times {10}^{-19}{(FIP)}^{-9.46}$$

(4)

The fitting curve between FIP parameters and fatigue life is shown in Figure 9. It showed that the prediction of fatigue life based on Equation (4) agreed well with the experimental data. The HIP specimens had a lower value of FIP and higher fatigue life than those of the non-HIP ones.

Based on the FIP model, the prediction of S-N curves for both non-HIP and HIP specimens is shown in Figure 10. The average diameter of the pores was calculated as 60 μm, and the prediction of fatigue life generally agreed with the experimental data. Obviously, EBM titanium alloys with a smaller defect obtained a longer fatigue life at the same stress level. For HIP treatment, pores in EBM titanium alloy specimens were considered to be healed with a diameter of zero. HIP treatment significantly improved fatigue life owing to the elimination of pore defects. Thus, the fatigue life of EBM titanium alloys with pore defects can be well predicted based on the FIP model.

5. Conclusions

The VHCF of the EBM TC21 titanium alloy with HIP and non-HIP treatment was carried out by ultrasonic fatigue test, and the effect of pore defects on VHCF behavior was investigated based on the Murakami model and FIP model for the EBM TC21 titanium alloy. The conclusions are summarized as follows:

(1)

The curve of the EBM TC21 titanium alloy displayed a continuously descending characteristic in a very high cycle regime, while HIP treatment significantly improved fatigue properties and illustrated two-step curve characteristics similar to the forged ones;

(2)

The VHCF cracks were mainly initiated from the internal pore of the EBM TC21 titanium alloy, whereas cracks were initiated at the inner heterogeneous microstructure for HIP specimens in the very high cycle regime. FGA was observed at the crack initiation site in a very high cycle regime for both non-HIP and HIP specimens. The value of ΔK_FGA corresponded with the threshold of stable crack propagation;

(3)

Based on the Murakami model, the lower limit of fatigue strength for the EBM TC21 titanium alloy was estimated by the statistical extreme value method. Fatigue strength with the average pore size relatively approached the experimental data with an error of 8%. Furthermore, a FIP model based on material defects was established to predict fatigue life for non-HIP and HIP specimens, which agreed well with the experimental data.