Analysis of the Structure and Properties of As-Built and Heat-Treated Wire-Feed Electron Beam Additively Manufactured (WEBAM) Ti–4Al–3V Spherical Pressure Vessel

Abstract

In the present work, a high-pressure spherical vessel was fabricated from Ti–4Al–3V titanium alloy using wire-feed electron beam additive manufacturing and characterized for tightness at high pressure. Studies have been carried out to characterize the microstructures and properties of the vessel’s material in four states: as-built (BM), annealed at 940 °C with cooling in air (HT₁ treatment), quenched in water from 940 °C (HT₂ treatment), and quenched with subsequent annealing at 540 °C (HT₃ treatment). The microstructure of the as-built (BM) samples was composed of grain boundary α-Ti and α/β lath colonies located within the columnar primary β-Ti grain boundaries. The ultimate tensile strength of the as-built material was in the range of 582 to 632 MPa, i.e., significantly lower than that of the source Ti–4Al–3V alloy wire. The subtransus HT₁ heat treatment allowed β→α″ transformation, while both HT₂ and HT₃ resulted in improved tensile strength due to the transformation of β-Ti into α/α′-Ti and the decomposition of α′ into α/β structures, respectively.

1. Introduction

Additive technologies, also known as 3D printing, have become an integral part of the arsenal of production techniques in a number of industries, including rocket and space, aviation and transportation, and energy [1,2,3]. The specific type of metal additive manufacturing employed is contingent upon the intended use, with the size and complexity of the product in question being the determining factors. For instance, in the production of small-sized products comprising numerous small parts, the preferred route is powder addition techniques, such as selective laser melting, directed energy deposition (blown powder), and electron beam melting [4,5]. However, direct metal deposition techniques, including wire and arc additive manufacturing and wire-feed electron beam additive manufacturing (also known as electron beam freeform fabrication), are frequently employed to produce large-scale components [6,7].

Being used on titanium alloys such as Ti–6Al–4V, these techniques offer high production speeds and facilitate the creation of near-net-shaped products. However, all of them are based on the fusion of powder or wire for layer-by-layer deposition and, therefore, result in forming an as-cast structure of the solidified metal. For Ti–6Al–4V, it means that high-aspect ratio columnar β-grains grow, and after the β→α transus, there are formed coarse α/β laths, which reduce the metal strength and add ductility. One of the methods for the elimination of these columnar grains is related to increasing the cooling rate, which, however, unavoidably provokes the formation of detrimental α′-Ti martensite. The cooling rate is higher with the laser powder-bed and wire arc AM methods compared to that of WEBAM, and therefore many results were reported on obtaining high-strength Ti–6Al–4V samples by means of these additive manufacturing methods. However, the WEBAM Ti–6A–-4V sample possesses low strength and high plasticity. Improvements in strength may be achieved by applying heat treatment.

The majority of studies examining the AM methods concentrate on the structure and characteristics of the so-called witness samples, which are defined as samples of a relatively simple shape, typically in the form of vertical walls [8,9,10,11]. Concurrently, the printing processes of complex-shape samples, particularly those of considerable size, diverge in terms of the thermal history of the sample due to the utilization of a sophisticated trajectory of product construction [12]. For instance, as becomes apparent [12], when printing bodies of rotation, the axis of vertical grain growth is deflected in the direction of heat dissipation, which has the potential to influence the tensile force behavior of the material. This characteristic of grain growth can be attributed to the waveform of the crystallization front, as elucidated [13].

In light of the above, the objective of this study was to examine the structure and mechanical properties of a large-scale sample of complex shape (high-pressure spherical vessel) produced by wire-feed electron beam additive manufacturing.

2. Materials and Methods

To carry out the research, a large-sized hemisphere intended for fabricating a high-pressure vessel of 5 dm³ capacity (Figure 1a) was obtained using a wire electron beam additive manufacturing (WEBAM) machine with a vacuum 8.0 m³ chamber at residual pressure not higher than 10⁻³ Pa (Institute of Strength Physics and Materials Science SB RAS, Tomsk, Russia). Ti–4Al–3V welding wire with a diameter of 2.0 mm is a commercially available, inexpensive product intended for welding Ti–6Al–4V workpieces, and the decision at this stage of the study was to use it as a feed wire for WEBAM fabrication of the vessel. The spherical vessel was fabricated in three main steps, as illustrated in Figure 2a–e.

In the first step, the simplest way was to use WEBAM for building two hemispherical workpieces (Figure 2a,b). The following WEBAM parameters were used: accelerating voltage of 30 kV, electron beam current of 45 mA, and linear printing speed of 440 mm/min. The next step was subtractive machining and polishing the hemispheres (Figure 2c,d) in order to improve the quality of both internal and external surfaces.

The final step was electron beam welding the hemispheres together to obtain a sphere (Figure 2e). The hemisphere edges were prepared to provide interlocking between the hemispheres, which were then joined using single-pass welding. The total number of WEBAM-built hemispheres was four, with two of them used for fabricating the spherical vessel and the other two cut into samples for investigation. The total time consumed for fabricating the spherical vessel was three working shifts, which was about 10 times shorter than the time taken using traditional all-subtractive manufacturing.

The sample cut-off scheme and the spherical vessel appearance are represented in Figure 1c,d and Figure 1e, respectively. The weld was tested for leakage using a leakage tester, Pfeiffer Vacuum ASM 34, at hydraulic pressures up to 400 atmospheres.

The resulting workpiece was cut into quarters using an electrical discharge cutting machine, DK7750 (Suzhou Simos CNC Technology Co., Ltd., Suzhou, China). One quarter part was examined in the original state (as-built), while the remaining three were subjected to three different heat treatments similar to those reported elsewhere [14].

One part of the as-built BM samples was annealed at the subtransus temperature of 940 °C for 1 h with cooling in air to the room temperature (HT₁ treatment); another part was quenched in water (HT₂ treatment) from 940 °C for 1 h; and the third part was quenched from 940 °C and then annealed at 540 °C for 1 h, followed by cooling in air (HT₃ treatment). The samples were cut from from the heat-treated hemispheres as shown in Figure 1b. Structural studies were carried out using an Altami MET 1C optical microscope (Altami Ltd., Saint Petersburg, Russia) and an Olympus LEXT OLS4100 confocal laser scanning microscope (Olympus Corporation, Tokyo, Japan). Static tensile tests were performed using the UTS 110M-100 (Testsystems, Ivanovo, Russia) mechanical testing machine operated at a loading speed of 1 mm/min. Specimen gauge dimensions were 12 × 2.7 × 2.5 mm³.

The structural and phase state of the materials was studied using an X-ray diffractometer DRON-3 operated at 40 kV and 22 mA with CoKα radiation, Bragg–Brentano focusing geometry, a 0.05° sampling interval, 30 s counting time, and within the 2θ scan range of 35–100° (Bourevestnik, JSC, Saint Petersburg, Russia). The identification of XRD peaks and data processing were performed using Crystal Impact’s software “Match!” (version 3.9, Crystal Impact, Bonn, Germany).

The grain orientation functions and EDS spectra were obtained using the EBSD add-on Oxford Instruments Nordlys and Oxford Instruments Ultim Max 40 (Oxford Instruments, High Wycombe, UK) add-ons to an SEM instrument Tescan MIRA 3 LMU (TESCAN ORSAY HOLDING, Brno, Czech Republic). The EBSD and EDS examinations were carried out at an accelerating voltage of 20 kV with the electron beam currents of 20 nA and 3 nA, respectively.

Samples with dimensions of 4 × 2 × 1 mm³ were EDM cut from the hemispheres as illustrated in Figure 1c, pos.5, and then ∅3 mm disks were drilled out using a Dimpling Grinder Model 200 (Fischione Instruments, Export, PA, USA) followed by dimpling at the center down to 10–20 μm. The next step was argon ion beam milling at 7 kV for 9 h in a TEM Mill Model 1051 dual-beam ion polishing system (Fischione Instruments, Export, PA, USA) until perforating a hole with surrounding 100–200 nm thick edges.

A TEM instrument JEOL-2100 (JEOL Ltd., Tokyo, Japan) with a 200 kV accelerating voltage and EDS add-on (Oxford Instruments) was used for detecting phases and examining the microstructures.

3. Results

The helium leakage test results demonstrated complete tightness of the WEBAM-produced spherical vessel (Figure 2f). The next step was leakage testing using the hydraulic pressure method (Figure 2g,h). Since the working pressure of the spherical vessel was 200 atm, proof testing was carried out at pressure levels exceeding 400 atm. As a result of this testing, it was found that the WEBAM-produced sphere vessel fractured at pressures in the range of 410–420 atm. Predictably, the fracture occurred and propagated along the girth weld (Figure 2h), which served as the weak part of the vessel. These pressure testing results led to the conclusion that the WEBAM spherical vessel, fabricated from Ti–4Al–3V welding wire, possessed suitable strength.

3.1. As-Built BM Material

Examination of microstructures formed in the as-built metal samples cut from the high-pressure vessel part BM (Figure 1) allowed observation of microstructures typical of the Ti–6Al–4V obtained via additive manufacturing, characterized by the presence of grain boundaries inherited from large columnar β-Ti primary grains, with their long axes oriented along the direction of heat removal (Figure 3a). It is also worth noting that the samples did not contain any defects caused by the additive process, such as, for example, spherical pores or cracks. The long axes of the columnar grains deviated by 20–30 degrees from the vertical symmetry axis due to the scanning strategy of the electron beam, as the vessel hemisphere was built as a body of rotation. Inside the β-Ti primary grain boundaries, there were typical as-cast microstructures composed of allotriomorphic (grain boundary) α_GB-Ti, coarse α/β lath colonies with a mean size in the range of 100–150 μm, and refined “basket-weave” α/β colonies with sizes in the range of 50–90 μm, which grew inside the former β-grain (Figure 3b,c) [15].

With the exception of the deviation in the growth axis of the columnar grains, no visible differences were detected compared to standard research samples obtained by WEBAM in the form of vertical walls. The results of the mechanical tests showed that the as-built material of the WEBAM high-pressure vessel had a significantly lower ultimate tensile strength (UTS) compared to the reference values for this alloy. In the bottom part of the hemisphere, near the substrate, the ultimate strength values of the material measured along the building direction (sample 3) and along the layer deposition direction (sample 1) were 590 ± 22 MPa and 582 ± 22 MPa, respectively. In this case, the strain-to-fracture (STF) was at the level of about 7–10%. Figure 4 shows the stress–strain curves for the specimens that exhibiting both the highest and lowest tensile strength values to illustrate the scatter of the data obtained when tested in different directions. In the upper part of the hemisphere, the as-built material demonstrates higher UTS values: 632 ± 14 MPa and 623 ± 15 Mpa, as measured along the building (sample 2) and layer deposition (sample 4) directions, respectively. The STF values lie within the 10–12% range.

In comparison with the Ti–6Al–4V ELI alloy, the obtained values are more than 30% lower compared to those reported [15]. However, they may be considered acceptable when compared to the UTS of the Ti–4Al–3V wire, which is 712 ± 18 MPa. The differences in strength levels may be related to the lower content of both aluminum and vanadium in the welding wire. Overall, it is concluded that the strength of the WEBAM spherical vessel is high enough to comply with the best industry practices.

The results of mechanical testing on the as-built samples revealed only weak strength anisotropy of 7% with respect to tensile axis directions. An even lower 1.5% anisotropy was measured on samples cut from different heights along the building direction.

3.2. Heat-Treated Material

The structure of the material after HT₁ annealing is practically no different from that of the as-built metal and is also represented by α_GB, α/β colonies, and basket-weaved α-laths inside the former β-Ti grains (Figure 5).

Quenching in water (HT₂) from the subtransus temperature significantly modified the microstructure of the as-built alloy, which is now represented by large α-laths that form a zigzag pattern (Figure 5b), with the space between them filled with smaller α-Ti laths. A similar pattern was observed after quenching followed by annealing at 540 °C (HT₃), as is clearly visible in Figure 5c.

EBSD grain orientation maps show preferential α-Ti grain orientations for all samples (Figure 6). The as-built material can be characterized by the maximum content of grain boundaries with misorientation angles of 60° (Type 2), 60.8° (Type 3), and 63.3° (Type 4) [16]. Annealing at the subtransus temperature (HT₁), quenching (HT₂), and quenching with annealing at 540 °C (HT₃) reduced the fraction of these boundaries from ~47% to ~30%, with a simultaneous increase in the fractions of both Type 6 (10.53°) and Type 5 (90°) grain boundaries (Figure 7).

The results of mechanical tests of these samples showed that HT₁ annealing did not lead to a significant improvement in the properties of the material (Figure 8). The HT₁ samples demonstrated a tensile strength of 600 ± 26 MPa in the lower part of the hemisphere and 627 ± 38 MPa in its upper part.

However, the obtained values show that this type of heat treatment makes it possible to reduce the degree of anisotropy of mechanical properties as measured along the building direction to 4.5%, which is almost half as much as in the as-built sample. Quenching the material in water from 940 °C allowed increasing the strength of the material to 706 ± 39 MPa in the lower part and 697 ± 18 MPa in the upper part of the hemisphere (Figure 8), which is on average 20% and 11% higher than in the original material, while the degree of anisotropy decreased to 1.3%. The last type of heat treatment—(HT₃), consisting of quenching followed by annealing to relieve the residual stresses—demonstrated similar values for the material’s ultimate strength as the sample after the HT₂ quenching. In the upper part of the hemisphere, such a material demonstrated a tensile strength of 716 ± 65 MPa, and in the lower part of 701 ± 37 MPa, that is, 14% and 19% higher than the original as-built material (Figure 8). The degree of anisotropy in this case was 2.1%. Therefore, it was possible to improve the strength of the as-built sample by means of post-heat treatment.

X-ray diffraction patterns of the Ti–4Al–3V samples demonstrated (Figure 9) that in addition to the strong α-Ti peaks, the as-built sample allowed identifying the retained (110)_β-Ti peak. Annealing at 940 °C (HT₁) resulted in considerable reduction of the (110)_β-Ti peak intensity and fewer intensities of α-Ti peaks such as (100)_α-Ti, (002)_α-Ti, (101)_α-Ti, (200)_α-Ti, and (112)_α-Ti peaks. However, new reflections appeared, which were identified as (103)_α-Ti and (201)_α-Ti.

Quenching the as-built samples from 940 °C (HT₂) resulted in an increase in the intensities of (100)_α-Ti and (103)_α-Ti peaks. It was suggested that retained β-Ti was transformed into α′-Ti and, therefore, the corresponding peaks were denoted as α/α′-Ti ones (Figure 9a).

Quenching and the following annealing at 540 °C (HT₃) provided XRD patterns with the (103)_α reflection height growing compared to that in the HT₁ and HT₂ patterns on the account of reflections (100)_α, (002)_α, (112)α, and (201)_α. Such a reorientation may be explained by the existence of strong texturing. This low-temperature annealing would serve to soften the α′-Ti formed at the previous stage of quenching. Therefore, the lattice parameter ratio c/a was used to characterize partitioning of alloying elements, in particular vanadium with an atomic radius of 0.132 nm, i.e., smaller than those of aluminum and titanium. The reduced value of c/a was inherent with the metal quenched from 940 °C (HT₂), which might be evidence in favor of enriching the α′-Ti with vanadium. The metal annealed at 540 °C had a higher c/a ratio because of α′-Ti depletion by vanadium during its decomposition (tempering). However, both of these values were lower than those obtained from the as-built (BM) and annealed at 940 °C (HT₁) metal, where high concentrations of vanadium might be in β-Ti and, allegedly, in α″-Ti, respectively.

The initial structure of the as-built material was characterized by α-phase laths of 1–1.5 μm in length, which were interspersed with β-phase laths of 100–150 nm in width (Figure 10a,c–e). This is corroborated by corresponding SAED patterns and chemical composition analysis (Figure 10,

Table 1. Chemical composition of phases of the as-built (initial) Ti–4Al–3V alloy phases (wt. %).

Phases	Al	Ti	V
α	5.87 ± 2.09	91.34 ± 3.28	2.78 ± 1.67
β	1.44 ± 0.65	77.74 ± 3.46	20.82 ± 3.18

). It is also noteworthy that no distinct boundaries were observed between the α- and β-phases, which may suggest the onset of β-phase transformation into metastable phases. The dimensions of these boundaries are less than 10 nm, which precludes their identification. The nucleation of β-phase decomposition regions in the material in the as-built state may be attributed to the repeated melting and multiple thermal exposures experienced by each of the additively deposited layers. This may result in establishing thermal conditions that are similar to those during the annealing process.

The spherical vessel samples that underwent annealing at 940 °C (HT₁) also exhibited α-Ti laths with β-Ti interlayers distributed between them (Figure 11). This is because the annealing temperature was below the α→β transformation temperature. It is also noteworthy that since the welding wire contained less β-supporting vanadium compared to that of Ti–6Al–4V alloy, its lack should increase the α→β-transus temperature. However, there are some microstructural differences between the as-built and HT₁ samples. The length of the α-Ti laths decreased to 0.5–1.0 μm, while the β-Ti lath width increased to 50–200 nm. Furthermore, the β-Ti laths demonstrate the existence of martensitic structures (Figure 11b,c), which may be due to the formation of metastable α″-Ti resulting from the β-Ti. It can therefore be surmised that the observed β-Ti might contain insignificant quantities of metastable phases, which would otherwise be undetectable. This is corroborated by the observation that the vanadium content in the β-phase diminished after the HT₁ annealing (

Table 2. Chemical composition of phases in the alloy HT₁-annealed Ti–4Al–3V (wt. %).

Phases	Al	Ti	V
α	4.47 ± 0.41	93.17 ± 0.62	2.09 ± 0.36
α″/β	3.32 ± 0.70	89.89 ± 1.67	6.79 ± 2.26

).

The samples that underwent quenching (HT₂) did not exhibit the presence of β-Ti (Figure 12), while those subjected to subsequent annealing (HT₃) exhibited some β-Ti in the form of twins (Figure 13); however, martensitic α′-Ti structures were identified in them (Figure 14). Fast cooling in water from 940 °C resulted in β→α′ transformation; however, some reflections in the SAED pattern did not belong to α′-Ti and may be better associated with those of Ti₃Al (Figure 12b). The α/α′-Ti lath width ranged from 100 to 500 nm.

After annealing of the quenched samples at 540 °C (HT₃), the α-Ti regions had almost not changed and remained almost intact as after HT₂ treatment (Figure 14). Their length is about 1.5 μm, and the lath width of the α-Ti phase in them is ≈170 nm. At the same time, the original α-phase precipitates lost their lamellar morphology and coarsened. According to XRD, this type of α′-Ti is depleted by vanadium and therefore would be herewith referred to as α_d′-Ti. This vanadium is partitioned to the α-twins and forms β-Ti (Figure 13). Thus, the fine structure of the spherical vessel after printing and quenching followed by annealing is represented by the α-Ti laths with former β-laths that decomposed into α′-Ti during quenching and then into α_d′-Ti during low-temperature annealing. Two other components were α/β twinned structures.

4. Discussion

The results presented indicate the viability of the fabrication of complex-shaped, large-sized products by wire electron beam additive manufacturing (WEBAM). However, the achievement of the desired mechanical properties requires subsequent post-processing. The chosen Ti–4Al–3V material for the high-pressure spherical vessel exhibits an ultimate tensile strength of approximately 600 MPa. Furthermore, the degree of anisotropy of the mechanical properties of the investigated samples along the height was approximately 7%. This is due to the lack of active water cooling during the printing of large-sized products with complex shapes. Furthermore, the complex scanning strategy of the electron beam during WEBAM 3D printing results in intensive repeated heating of the printed material. This phenomenon has also been observed previously in large-sized Ti–4Al–3V alloy walls [15].

Usually, subtransus annealing is used to reduce the amount of α′-Ti that is a common finding in laser powder-bed additive methods [17]. The cooling rate in WEBAM is lower, and therefore, no α′-Ti was formed. Nevertheless, the effect of such a treatment was studied for comparison purposes. Annealing at 940 °C caused β→α″ transformation in the WEBAM as-built samples, which is usually feasible in the β-lath enriched with vanadium that comes from the α-Ti during the transformation of primary β-Ti. The mechanical strength of the titanium alloys with orthorhombic α″-Ti is relatively low in comparison to those with the β-Ti [18]. Only a slight strength improvement of the annealed (HT₁) samples occurred compared to that of the as-built ones, while this annealing should result only in reducing the strength since α″ is a more ductile phase compared to β-Ti. Since the α-lath size did not reduce, this slight strength improvement may be explained using the results obtained elsewhere [19,20,21]; i.e., it might have happened that coherent precipitates of Ti₃Al formed in Ti–6Al–4V along the α/β structures and contributed to this slight improvement of the strength of the alloy.

Solely for the purpose of reducing the degree of anisotropy, annealing at 940 °C can achieve a nearly 50% reduction, while the increase in tensile strength is only slight. The anisotropy of the mechanical properties of the HT₁ annealed t alloy can be explained by the electron beam scanning strategy used and the columnar structure formed. In the bottom part of the hemisphere, due to the small diameters of the product, the material is subjected to a large number of heating/cooling cycles and, consequently, repeating β→α and α→β transformations. In the upper part of the workpiece, the diameter of the rotating body is several times larger, and therefore the frequency of heating-cooling cycles when applying layers is reduced, which ensures a more stable thermal history and higher mechanical characteristics. The low degree of strength anisotropy relative to the direction of testing (along the direction of deposition of layers or the direction of growth of the sample) may be associated with the nature of the growth of grains of the primary beta phase. The deviation of the grain growth axis from the vertical direction leads to the fact that the grain boundaries are not parallel (or perpendicular) to the tension axis. As a result, the deformation of the samples in both directions occurs equally without affecting the obtained values of ultimate strength.

Quenching in water from 940 °C (HT₂) induces martensitic transformation of β-Ti into α′-Ti while twinning occurs in the grain boundary α_GB as well as primary α-colonies. The β→α′ transformation was accompanied by the appearance of high elastic stresses that might cause deformation twinning. The combination of the lowest height anisotropy of mechanical properties (1.3%) and an increase in material strength by almost 20% can be achieved through quenching from 940 °C. The α-Ti structures remained almost intact compared to those obtained from quenching. The tensile strength and degree of anisotropy remained at comparable levels.

The maximum ultimate strength values obtained from the WEBAM as-built samples that underwent quenching from 940 °C and quenching/annealing at 540 °C were somewhat lower than those of the as-received Ti–4Al–3V alloy (860 MPa). Additional low-temperature annealing at 540 °C (HT₃) served to relieve the elastic stress and promote additional precipitation of α₂-Ti₃Al particles, which are also capable of dislocation-pinning and might contribute to strain hardening [22,23]. Decomposition of the α′-Ti occurs by partitioning of vanadium from the α′-Ti grains, until its concentration becomes equal to that of primary α-Ti, and further precipitation of both β-Ti occurs [24]. Decomposition of α/α′ results in the formation of a new structure with α-Ti lath and fine β-particles, which are capable of dispersion strengthening [24]. An interesting fact is that these β-particles precipitate preferentially inside the twins whose boundaries attain extra strengthening. The solid solution strengthening is also important because there are still α′-lamellae with some residual concentration of vanadium. All these strengthening mechanisms provide the highest ultimate strength in the samples quenched and annealed at 540 °C (HT₃).

Alternative solutions may be related to the application of post-WEBAM thermomechanical treatment. Nevertheless, the results already obtained demonstrate that the application of wire-feed electron beam additive manufacturing (WEBAM) allows the production of complex, pore-free, large-sized products from Ti–4Al–3V alloy, with the possibility of improving their properties through heat treatment methods.

5. Conclusions

The investigation undertaken into the structure and mechanical properties of the high-pressure vessel has led to the following conclusions:

• A pressure vessel fabricated by welding together two WEBAM-built hemispheres demonstrated its high resilience at a pressure twice as much as its working level.
• The method of wire-feed electron-beam additive manufacturing (WEBAM) allows for the production of a void-free product. However, the resulting material exhibits significantly reduced mechanical properties compared to those required for the Ti–4Al–3V alloy. This is attributed to the absence of active cooling and the complex printing path, which resulted in a material with an ultimate strength ranging from 582 to 632 MPa and a degree of anisotropy in mechanical properties reaching 7%.
• The application of heat treatment in the form of subtransus annealing at 940 °C (HT₁) resulted in β→α″ transformation and only a slight increase in the UTS. The degree of anisotropy was reduced to 4.5%.
• Quenching in water after holding at 940 °C for 1 h (HT₂) was observed to result in an increase in tensile strength by 11–20% due to β→α′ transformation and deformation twinning because of high transformation-induced stresses. Reduction in the degree of strength anisotropy by 1.3% was due to the formation of a modified microstructure.
• The quenching followed by annealing at 540 °C (HT₃) resulted in the decomposition of α′-Ti into α and fine β particles, preferentially inside the deformation twins, which, on the one side, caused both dispersion and grain boundary strengthening, leading to tensile strength improvement, and on the other side, served to relieve the transformation-induced stresses.