3.1 Microstructural characterization of weld toes
Fig. 2 shows cross-sectional OM images of the as-welded joint. Typical weld toe radius and angle were measured to be about 0.27 mm and 27°, respectively (Fig. 1(b)). The base metal and weld toe consisted of ferrite-pearlite layered structure (Figs. 2(c), 2(d)) and coarse bainitic ferrite grains (Figs. 2(e), 2(f)).
FSP was conducted on the weld toe surfaces to improve fatigue strength of the butt-welded joints. Fig. 3 shows cross-sectional OM and SEM images of the weld toes with FSP at the tool rotational speeds of 400, 600, 800, and 1000 rpm (hereafter, samples FSP400, FSP600, FSP800, and FSP1000, respectively). FSP changed weld toe geometry due to plastic deformation, in addition to formation of the SZ. The weld toe radius was significantly increased to about 3.69, 2.65, 3.53, and 3.79 mm in samples FSP400, FSP600, FSP800, and FSP1000, respectively, but with slightly higher weld toe angles (Figs. 3(a)-(d)) in comparison to that of the as-welded toe (Fig. 2(b)). These macroscopic changes did not seem to depend on tool rotational speed. On the other hand, a surface morphological and microstructural feature of the topmost layer of the weld toes after FSP varied with tool rotational speed: a notched surface was observed in FSP400 (Fig. 3(e)), while smooth surfaces were observed in FSP600, FSP800, and FSP1000 (Figs. 3(f)-3(h)). The notched surface in FSP400 can be attributed to materials having been removed at the front of the tool and subsequently accumulating behind the tool. The FSP condition of 400 rpm was not appropriate for sound surface modification on the weld toe. In microstructure of the topmost layer including the SZ, there was an obvious difference in addition to significant grain refinement in all weld toes with FSP. In the OM images for FSP800 and FSP1000 (Figs. 3(g), 3(h)), discontinuous and continuous areas with white contrast were observed at the outermost surface layer, in comparison to those for FSP400 and FSP600 (Figs. 3(e), 3(f)). The white contrast areas correspond to the portions alloyed with constituent elements of the WC tool, and the thicker portion for FSP1000 was provided by the faster tool rotational speed. This suggests that the tool wear amount increased with increasing tool rotational speed. To characterize microstructure of the topmost layer at the weld toe, the specimen surface was etched with 2 vol% nital solution, and cross-sectional SEM images of the outermost surface layer and a layer around 100 µm below the surface were obtained, as shown in Figs. 3(i)-3(l) and 3(m)-3(p), respectively. Fine equiaxed polygonal ferrite grains were observed in both layers mainly in FSP400 and FSP600 (Figs. 3(i), 3(j), 3(m), 3(n)). The volume fraction of bainitic ferrite grains increased with increasing tool rotational speed in the layer 100 µm below the surface, suggesting that the peak temperature during FSP was beyond the phase transformation temperature and the volume fraction of transformed austenite grains increased. The layer eventually showed a fully bainitic structure in FSP1000 (Fig. 3(p)). This indicates that the peak temperature during FSP exceeded the AC3 point with the fully austenite transformation. The bainitic ferrite grains arose through transformation from the austenite phase during cooling in FSP, while polygonal ferrite grains arose through dynamic recrystallization during FSP. The outermost surface layer having smooth continuity formed in FSP800 and FSP1000 is considered to contain tool constituent elements of W and C, as explained in a previous study [31], as shown in Figs. 3(k) and 3(i). This suggests that the layer consisted of fully martensite structure with ultrafine laths.
To identify the elemental contents and their influence on the hardness in the outermost surface layer, EPMA and Vickers hardness tests were conducted. Fig. 4 shows relationship between increment in W and C contents and hardness with respect to depth from the weld toe surface. The increment was estimated by subtracting the average content obtained in the topmost layer of the as-welded toe from the content measured using EPMA. Increment in the W and C contents was concentrated beneath the surface and the segregated depth was about 50 µm in W and 100 µm in C (Figs. 4(a), 4(b)). The increment of W and C contents increased with increasing tool rotational speed and was estimated to be about 2.36 and 2.17 at%, respectively, in sample FSP1000. The hardness increment in the outermost surface layer with respect to the base metal exhibited similar trend of increment in W content in each specimen (Fig. 4(c)). The increment was small in FSP400 and FSP600, and increased with increasing tool rotational speed in FSP800 and FSP1000. Note that the hardness increment due to W and C solid solution hardening was larger than that due to grain refinement (see FSP400 and FSP600). The average hardness of the outermost surface layer was 552 and 878 HV, respectively, in FSP800 and FSP1000, and was higher than that of the water-quenched base metal (449 HV) consisting of fully martensite structure.
To further characterize the distribution of tool constituent elements in the outermost surface layer, TEM-EDS analyses were conducted for the thin film specimens taken from the weld toe surfaces (Figs. 3(j)-3(l)). Fig. 5 shows TEM-EDS results obtained in samples FSP600, FSP800, and FSP1000. Polygonal ferrite grains about a few µm in average diameter were observed in the outermost surface layer of the weld toe in FSP600, as shown in the TEM-bright field (BF) images (Figs. 5(a), 5(b)). W segregation was hardly observed in any ferrite grains or at their boundaries in the EDS-W map image (Fig. 5(c)). On the other hand, grain structure in the outermost surface layer of the weld toes in FSP800 and FSP1000 (Fig. 5(d), 5(g)) became finer than that in FSP600 (Fig. 5(a)). Enlarged images showed obvious change in grain morphology, and the layer included fine elongated grains (Fig. 5(e), 5(h)). The EDS-W map images (Figs. 5(f), 5(i)) exhibited higher homogeneously distributed W content than that in FSP600 (Fig. 5(c)). In addition, a few W-rich particles were observed in FSP1000 (Fig. 5(i)).
To identify W content in the outermost layer, EDS quantitative analyses were conducted at points A-I in the steel matrix in Figs. 5(b), 5(e), and 5(h) and at points 1-3 in the W-rich particles in Fig. 5(g), indicated by red and yellow arrows, respectively. A summary is presented in Table 3. The average W contents obtained in the steel matrix of samples FSP600, FSP800, and FSP1000 based on the TEM-EDS analyses were estimated to be 0.11, 1.43, and 2.25 at%, in good agreement with those obtained using a 5 µm-diameter electron-beam spot using EPMA (Fig. 4(a)), suggesting that the tool constituent W atoms were homogeneously distributed in the topmost steel matrix layer after FSP. On the other hand, the W-rich particles observed in FSP1000 exhibited average W content of 23.42 at%, suggesting that the particle could be an embedded Fe4W2C fragment that had dropped out of the WC tool tip and was milled at the interface between the tool and steel surface. Fe4W2C particles were reported to form at the tool tip as result of reaction of a WC particle with Fe atoms diffused from the steel matrix [31], together with decomposition of most of the fragment into W and C atoms during milling at the interface under high temperature and pressure due to FSP friction. Most of the tool constituent elements seem to have dissolved in the outermost surface layer of steel matrix austenitized around the peak temperatures during FSP, followed by martensitic transformation during cooling, resulting in fine martensite laths in the outermost surface layer (Fig. 5(e), 5(h)). Note that the solid-soluble W and C contents in the a-Fe phase were almost zero at room temperature. The hardness of the topmost layer was significantly higher than that of the water-quenched base metal. This can be explained by solid solution hardening due to the supersaturated W and C atoms (WC-tool constituent elements). The hardening is associated with large lattice distortion caused by substitutional solid solution of the extra W into the martensite as well as interstitial solid solution of the extra C. The higher solid solute W contents could contribute to the higher increase in hardness in FSP1000. The W-rich particles cannot have effectively increased the hardness, since their size of about 100 nm was too large to impede the movement of dislocations.
Table 3 W contents obtained at points A-I and 1-3 in Fig. 5 indicated by red and yellow arrows in the steel matrixes and particles, respectively, based on EDS quantitative analyses
Code
|
Measuring point
|
W content, (at%)
|
Average
|
FSP600
|
Matrix
|
A
|
0.12
|
0.11
|
B
|
0.12
|
C
|
0.10
|
FSP800
|
Matrix
|
D
|
1.40
|
1.43
|
E
|
1.44
|
F
|
1.44
|
FSP1000
|
Matrix
|
G
|
2.36
|
2.25
|
H
|
2.27
|
I
|
2.13
|
Particle
|
1
|
18.46
|
23.42
|
2
|
27.38
|
3
|
24.42
|
3.2 Fatigue strength of welded joints with FSP depending on tool rotational speed
Fatigue strength of weld toes with the FSP-modified topmost layer was investigated. The FSP-modified topmost layer was alloyed with WC-tool constituent elements together with grain refinement. Their applied nominal stress amplitude dependence on number of cycles to failure (S-N diagram) is shown in Fig. 6, together with those of an as-welded joint and a base metal. Since the specimens did not fail up to 2 × 106 cycles (shown with additional arrows on symbols), this number was defined as the fatigue limit. The other joint specimens failed due to local stress concentration at the weld toe. The nominal stress amplitude increased with increasing tool rotational speed, and the values at 2 × 106 cycles were estimated from the S-N curves to be about 164, 214, 223, and 248 MPa in samples FSP400, FSP600, FSP800, and FSP1000, respectively. The nominal stress amplitude at 2 × 106 cycles in FSP1000 increased by 58% in comparison to that in the as-welded joint.
To understand the difference in nominal stress amplitude between these joints, stress concentration effects related to the weld toe geometry were investigated. The elastic stress concentration factor of kt and fatigue notch factor of kf for the butt-welded joints under the bending load were calculated from Equations (1) and (2), below, respectively [32]:
kt = 1 + 0.165(tanθ)1/6 (t/r)1/2 (1)
kf = 1+ (kt - 1)/(1+a/r) (2)
where t is the plate thickness of 10 mm, and θ and r are the weld toe angle and radius, respectively, measured in Figs. 2(b) and 3(a)-3(d). Peterson’s material parameter of a for steels is obtained from Equation (3) [32]:
a=1.087 × 105σuts-2 (3)
where σuts is ultimate tensile strength of the weld toe, obtained using the small tensile specimens. The values of kt and kf calculated in all the joints with and without FSP are summarized in Table 4, together with t, θ, r, and σuts. The kf values of all butt-welded joints with FSP were similar and reduced by about 10% due to the significant increase in the weld toe radius, compared to that of the as-welded joint.
Table 4 Weld toe parameters for the as-welded joint, FSP400, FSP600, FSP800, and FSP1000
Code
|
Plate thickness, t (mm)
|
Weld toe angle,
θ(°)
|
Weld toe radius,
r(mm)
|
Ultimate tensile strength of the weld toe,
σuts (MPa)
|
Elastic stress concentration factor, kt
|
Fatigue notch factor, kf
|
As-welded
|
10
|
27
|
0.27
|
555
|
1.90
|
1.39
|
FSP400
|
40
|
3.69
|
637
|
1.26
|
1.25
|
FSP600
|
32
|
2.65
|
655
|
1.30
|
1.27
|
FSP800
|
36
|
3.53
|
672
|
1.26
|
1.25
|
FSP1000
|
33
|
3.60
|
707
|
1.26
|
1.24
|
The local stress amplitude [33] as fatigue strength at the weld toe of σw, loc was calculated from Equation (4):
σw, loc = σw,nom . kf (4)
where the nominal stress amplitude of σw,nom was defined as that at 2 × 106 cycles in Fig. 6. Note that the σw,loc is associated with effects of factors other than the weld toe geometry modification, such as microstructure and surface roughness modification. Fig. 7 shows relationship between the calculated σw,loc as well as σw,nom for all the joints with FSP variation with tool rotational speed, together with the σw,nom and σw,loc of the as-welded joint and the σw,nom of the base metal. The σw,loc in sample FSP400 was lower than that in the as-welded joint, even though the weld toe microstructure was significantly refined. This can be associated with the notched surface because of low heat input (Fig. 3(e)). On the other hand, values of the σw,loc in FSP600, FSP800, and FSP1000 with the smooth weld toe surface were obviously higher than that of the as-welded joint. Especially, the value of the σw,loc in FSP1000 was about 41% higher than that of the as-welded joint. This indicates that microstructure modification in the topmost layer of weld toes provided much larger increase in fatigue strength than the weld toe geometry modification. Increase of σw,loc in FSP600 can be explained by grain refinement. Those in FSP800 and FSP1000 were a little higher and much higher, respectively, than that in FSP600. The further increase in FSP800 and FSP1000 is associated with the alloying of the topmost layer of weld toes with the tool constituent elements. The alloying led to W and C solid solution hardening together with the related martensitic transformation.
To clarify the effect of the alloyed topmost layer on the fatigue strength increase observed in samples FSP800 and FSP1000, fracture surface observation and analysis were performed. Fig. 8 shows SEM images of fracture surfaces obtained after fatigue failure at stress amplitude of 300 MPa for all the joints with and without FSP. The observed fracture surfaces can be divided into two regions, consisting of flat surface caused by fatigue crack initiation and propagation, and rough surface corresponding to the fast fracture region. Many ratchet marks (white arrows), forming a boundary between two adjacent failure planes, were observed on the fracture surfaces beneath the weld toe surface. The more crack initiation sites there were, the more ratchet marks were present. The average number of ratchet marks per unit length on the fracture surfaces for all the joints is shown in Fig. 9. The average value in FSP400 was more than that in the as-welded joint. The enlarged crack initiation region beneath the weld toe surface showed rough, discontinuous, brittle fracture planes (Fig. 8(c)). The crack initiation occurred at the notch bottom in the rough surface after FSP (Fig. 3(e)), resulting in the lower in FSP400 than that in the as-welded joint. On the other hand, the fatigue fracture surface in FSP600, FSP800, and FSP1000 was smoother and had much fewer ratchet marks than in the as-welded joint and FSP400. The average number of ratchet marks per unit length on the fracture surfaces in FSP600, FSP800, and FSP1000 decreased with increasing tool rotational speed. This suggests that resistance to crack initiation at the weld toe surface during fatigue tests can be enhanced by increasing hardness due to solid solution hardening, by alloying the topmost layer with W and C elements.
To understand the effect of the alloyed topmost layer on increase in the fatigue strength, the fatigue failure position and distributions of the alloyed tool constituent elements near the failure position were investigated. Fig. 10 shows cross-sectional OM and EPMA-W map images obtained near the fatigue failure at the stress amplitude of 300 MPa in sample FSP1000. The EPMA-W map image indicates that the fatigue failure occurred at the surface position with lower alloyed tool constituent elements. The fatigue failure occurred at a similar position in FSP800. This suggests that alloying the weld toe surface with tool constituent elements increased resistance to fatigue crack initiation, and the crack initiated at the excess weld metal with higher thickness than that in the base metal. Especially, the increment of the alloyed W and C contents near the weld toe surface in FSP1000 was the highest, resulting in the highest hardness, leading to the largest increase in fatigue strength among all the joints with FSP modified weld toes. The increment of the alloyed W and C contents is associated with WC tool wear during FSP, and increment of both the W and C contents and WC tool wear increased with increasing tool rotational speed [31]. The tool wear has been considered as an unavoidable issue in high strength materials during FSP, but these results indicate that alloying the weld toe with tool constituent elements could be an effective technique for increasing fatigue strength of welded steel joints.