Content uploaded by Jack Donoghue
Author content
All content in this area was uploaded by Jack Donoghue on Mar 24, 2016
Content may be subject to copyright.
The effectiveness of combining rolling deformation with Wire–Arc
Additive Manufacture on β-grain refinement and texture modification
in Ti–6Al–4V
J. Donoghue
a,
⁎, A.A. Antonysamy
a,1
,F.Martina
b
, P.A. Colegrove
b
,S.W.Williams
b
, P.B. Prangnell
a
a
School of Materials, University of Manchester, Manchester M13 9PL, UK
b
The Welding Engineering Research Centre, Cranfield University, Bedfordshire MK43 0AL, UK
abstractarticle info
Article history:
Received 15 October 2015
Received in revised form 4 January 2016
Accepted 6 February 2016
Available online 8 February 2016
In Additive Manufacture(AM), with the widely usedtitanium alloy Ti–6Al–4V, the solidification conditions typ-
ically result in undesirable, coarse-columnar, primary βgrain structures. This can result in a strong texture and
mechanical anisotropyin AM components. Here, we haveinvestigated the efficacy of a new approachto promote
βgrain refinement in Wire–Arc Additive Manufacture (WAAM) of large scale parts, which combines a rolling
step sequentially with layer deposition. It has been found that when applied in-process, to each added layer,
only a surprisingly low level of deformation is required to greatly reduce the βgrain size. From EBSD analysis
of the rolling strain distribution in each layer and reconstruction of the prior βgrain structure, it has been dem-
onstrated that the normally coarse centimetre scale columnar βgrain structure could be refined down to
b100 μm. Moreover, in theprocess both the βand αphase textures were substantially weakened to close to ran-
dom. It is postulated that the deformation step causes new βorientations to develop, through local heterogene-
ities in the deformation structure, which act as nuclei during the α→βtransformationthat occurs as each layer is
re-heated by the subsequent deposition pass.
© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Keywords:
Additive Manufacture
Titanium
Grain str ucture
Texture,
1. Introduction
Near-net-shape fabrication of metallic components by Additive
Manufacture (AM) is an important new technological area with many
potential applications in the aerospace industry (e.g. [1–17]). AM in-
volves building parts by sequentially consolidating 2D slices of material
that are fused together by a focused heat source [1–3]. A range of AM
processes are now available, mainly based on laser or electron beam
systems, that use powder or wire feedstock [1–12]. Of these techniques,
powder bed methods allow more geometrically complex components
to be produced, but the part size is restricted by slow build rates and
the limited dimensions of the working chamber [1–4].
Recently, a low cost wire-based AM process that exploits standard
welding technology has become of interest to industry [9–11].In
Wire–Arc Additive Manufacture (WAAM) a consumable wire is fed at
a controlled rate into an adapted electric arc (or plasma) welding
torch that is translated by a robot [9–12]. Material is built up in the
form of a weld bead that is overlaid on previously deposited tracks.
Shielding can be provided by an inert gas flooded hood, or deposition
can take place in an atmospherically controlled chamber. The WAAM
process has a much higher deposition rate than most other metal addi-
tive manufacturing techniques (up to 10 kg/h). It also provides better
material utilization than powder based meth ods [9–12], but is restricted
to wider wall thicknesses and cannot produce as fine scale features. This
low cost process is therefore most suited to producing larger scale parts
with less complex geometries.
The α–βtitanium alloy, Ti–6Al–4V, is the ‘work horse’of the aero-
space industry and widely used in airframe and aeroengine applications,
where the production of near-net shape components by AM can result
in significant cost savings. However, a current concern with AM using
this alloy is that coarse primary columnar βgrain structures are nearly
always observed to be produced in the consolidated material. This un-
desirable grain structure is seen across a wide range of AM platforms
[6, 9–16]. With wire based AM the primary βgrains are often as tall as
the build height and with larger components can be tens of centimetres
long [9–16]. This strong tendency to form coarse-columnar βgrain
structures in AM with Ti–6Al–4V is difficult to avoid because it results
from a combination of the solidification conditions in a small heated
moving melt pool, where there is a steep positive thermal gradient at
the solidification front, and the metallurgical characteristics of the
alloy itself [13–16]. In particular, the Ti–6Al–4V alloy system does not
Materials Characterization 114 (2016) 103–114
⁎Corresponding author.
E-mail addresses: jack.donoghue@manchester.ac.uk (J. Donoghue),
alphons.antonysamy@gknaerospace.com (A.A. Antonysamy), f.martina@cranfield.ac.uk
(F. Martina), p.colegrove@cranfield.ac.uk (P.A. Colegrove), s.williams@cranfield.ac.uk
(S.W. Williams), philip.prangnell@manchester.ac.uk (P.B. Prangnell).
1
Now at GKN Aerospace, PO Box 500, Golf Course Lane, Filton BS34 9AU, UK.
http://dx.doi.org/10.1016/j.matchar.2016.02.001
1044-5803/© 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CCBY license (http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
Materials Characterization
journal homepage: www.elsevier.com/locate/matchar
lend itself to nucleation ahead of the solidification front because of the
high partition coefficients of aluminium and vanadium, which are
close to one, and the lack of suitable grain refining particles in the
melt [17]. These process and metallurgical limitations restrict the de-
gree of constitutional supercooling that can occur so that, when com-
bined with a lack of melt inoculants, nucleation ahead of the
solidification front is difficult to achieve [13–16].
Although in AM the βgrain structure developed during solidification
transforms to a fine αand retained βlamellar structure on cooling
below the β-transus temperature, the microstructural memory of the
coarse, directional, primary-βmicrostructure can still have a significant
impact on mechanical performance. In particular, directional growth of
large primary βgrains generally produces a strong b001Nfibre texture,
which gives rise to a related αtransformation texture [13–16] and this
can potentially result in texture clustering of aligned α-plates within
the βmatrix. Such factors are known to be detrimental to fatigue life
[18, 19] and can contribute to mechanical anisotropy [20–22].Inaddi-
tion, with a coarse primary βgrain structure, grain boundary αcan
cause premature failure in transverse loading [21, 23]. However, to
date, little systematic work has been published on the texture found
in AM titanium components produced by wire-based techniques like
the WAAM process.
Potential methods for refining the poor primary grain structure seen
in AM deposits include; i) modification of the solidification conditionsin
the melt pool, through manipulation of the process variables [10], or ii)
altering thealloy chemistry [17, 24, 25]. However, in AM there is limited
scope for changing the process window, because this is dictated by the
conditions required to obtain stable part dimensions [9, 10]. Further-
more, while trace additions of elements like boron are known to act as
growth restrictors in titanium [25] this can have negative consequences
through the formation of brittle second phase particles (e.g. TiB).
In the present work an alternative approach has been investigated
for improving the large columnar β-grain structures and strong textures
typically seen in wire-based AM processes. This has involved the intro-
duction of a small deformation step sequentially with the deposition of
each layer. The deformation step was applied using a roller integrated
with the AM system, so that each deposited layer could be lightly de-
formed before adding a new layer of material (Fig. 1). Although this
set up limits the technique to simpler geometries, the aim of this ap-
proach wasto see if it was possible to introduce sufficient plastic defor-
mation into each layer so that refinement of the β-grains could occur
during re-heating, when the next layer was deposited. It was also
hoped that this might generate a weaker texture, which would lead to
more isotropic mechanical properties [18, 19, 26]. Although the intro-
duction of a light rolling step in AM will moderately reduce the rate of
build-up of material and causes a slight spreading in the wall width,
this can be controlled in an automated manufacturing system and
would not be a major issue when building relatively simple component
designs. In fact, it has been found that rolling increases the accuracy of
the wall dimensions, by correcting variation in the wall width caused
by the bead profile [27]. There are also other potential methods avail-
able for applying deformation to each layer in AM that are not so re-
stricted by geometry, such as by peening [28]; hence the efficacy of
this novel approach is of more general interest.
It should be noted that in this collaborative study the concept of the
hybrid WAAM deformation process was developed by Cranfield Univer-
sity Welding Engineering Research Centre [27, 29], who have previously
published work investigating the effect rolling has on the residual stress
within the builds [27], and noting the effect on the refinement of the mi-
crostructure [30]. The current work, performed at Manchester Universi-
ty, is complementaryin that it investigates the effect the refinement has
on the primary βand final αtexture, and goes into more detail than the
previously into the mechanism of formation and the distribution of the
refined βgrain structure.
2. Experimental
2.1. WAAM samples
The undeformed and rolled WAAM samples were built using a Ti–
6Al–4V alloy welding wire (1.2 mm diameter) with a titanium base
plate of the same alloy. The substrate material was a conventional hot
rolled and annealed plate that had a recrystallized equiaxed α–βmicro-
structure [18]. The samples studied were produced under identical con-
ditions —as onemeter long, 20 layer high, straight, vertical walls, using
a pulsed GTAW welding system with an average current of 110 A. Each
Fig. 1. Schematic diagram of the combined WAAM rolling process.
104 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
wall was a single track wide and had a width of ~6 mm. Argon shielding
was provided by a trailing hood containing a laminar flow device with a
high gas flow rate. This resulted in an average oxygen content in the de-
posited walls of b1500 ppm. The deposition parameters employed are
given in Table 1 and full details of the WAAM process can be found in
ref [9]. To produce the deformed samples, after deposition of each indi-
vidual layer, a 100 mm diameter roller was run across the top of the
walls using a rigid gantry system on which the welding torch was also
mounted (see Fig. 1). Each layer was rolled after the temperature of
the top layercooled naturally to well below 300° (i.e. close to cold defor-
mation conditions and well into the α–βfield). The roller employed
contained a 3.6 mm radius semi-circular groove, designed to approxi-
mately match the curvature of the bead surface. The compressive load
during rolling was controlled using a load cell and was applied directly
downwards through the roller bearings to the top of each wall.
Five sample conditions were analysed; (i) a control sample built
without rolling, as well as two walls to which rolling was applied after
adding every layer, with a down-force of (ii) 50 or (iii) 75 kN, and two
walls where rolling was only applied to the penultimate layer with
the same loads (i.e. (iv) 50 and (v) 75 kN) and the final layer was not
deformed.
In the results presented, a standard reference frame has been used
for the orientation of all the samples where; zis the direction normal
to the deposited layers (and parallel to the wall height), xis parallel to
the wall length and coincident with both the torch travel and rolling di-
rection, and yis the transverse direction normal to the wall surface (see
Fig. 1). For the samples rolled every layer the net reduction in wall
height is given in Table 2, where it can be seen that the 50 and 75 kN
rolling loads resulted in an average compressive strain (ε
z
)of8and
19% respectively.
2.2. Characterisation techniques
The undeformed and rolled AM samples were characterised in two
cross-sections for microstructure and textural analysis; i. down their
centre line in the vertical x–zplane and ii. in the plane of the layers
(x–y), half way up each wall —to observe any microstructural variation
vertically and through the thickness. Following standard preparation
procedures, etching and optical microscopy was used to reveal the de-
posits' macrostructures. Metallographically prepared samples were
analysed by scanning electron microscopy and electron back scatter dif-
fraction (EBSD) orientation mapping using a CamScam Maxim FEG-
SEM. Orientation maps were collected with an Oxford instruments
EBSD system, with Aztec acquisition software. Large area maps of
~12× 6 mm, witha 5 μm step size, were used to obtain average texture
data and provide comparative macro-views of the coarse β-grain struc-
ture across all samples.
2.3. β-phase reconstruction and texture analysis
In a Ti–6Al–4V alloy with a fine lamellar microstructure it is chal-
lenging to index the residual β-phase directly by conventional EBSD be-
cause of its small scale and low volume fraction (~5–9% [31]). The high
temperature parent βphase grain structures and textures that were
originally present after solidification, prior to transformation on cooling,
were therefore reconstructed from room temperature αorientation
data using a procedure developed by Davies and Wynne [32, 33],
based on earlier work by Humbert et al. [34, 35]. Full details of this ap-
proach can be found in ref. [32]. The reconstruction procedure uses
the Burgers Orientation Relationship (BOR) between the αand β
phase to calculate the six possible βparent orientations for each α
plate [32–35]; where the BOR is given by:
100
fg
β== 0002
fg
αb111Nβ==b1120Nα:
The most probable parent βorientation for each αplate is then se-
lected by comparing the most common solution for the misorientations
between neighbouring data points. Variables within the procedure that
can be altered are the misorientations between neighbouring αpoints,
that can be considered the same αvariant, and the allowable maximum
angular deviation from the ideal BOR [32]. Here, these parameters were
kept at 2° and 3°, respectively.
Texture information was extracted from the original measured α
EBSD maps and reconstructed βorientation data and is presented in
standard pole figures. In all the orientation maps, inverse pole figure
(IPF) colouring has been used with the reference axis aligned with the
main fibre direction, which is close to z. High angle grain boundaries
(HAGBs) N15° in misorientation are depicted by black lines.
2.4. Strain distribution in the AM deposits
In order to relate the plastic deformation to the refined microstruc-
ture, it is necessary to map the distribution of the plastic strain as well
as the distribution of the grain sizes. Two different approaches were ap-
plied to determine the plastic strain distribution in the rolled AM sam-
ples using EBSD maps; i) from the pattern quality and ii) the relative
effect on thedeviation of the misorientation of neighbouring αvariants
from fulfilling the BOR. If the microscope conditions are kept constant,
variation in pattern quality can be attributed to local changes in distor-
tion of the crystal lattice caused by the rolling deformation [36].The
standard measure of pattern quality is Band Contrast (BC), which refers
to the brightness of the diffracted bands above background. However,
here the Band Slope (BS) was used, which is the gradient of the intensity
of signal at the edge of a band (slope of the intensity above back
ground). BS was chosen due to the lower sensitivity of the probed vol-
ume to crystallographic orientation, compared to BC, which is more
strongly affected by how well orientated the crystallographic planes
are for diffraction [37]. For method ii) the deviation of misorientations
between neighbouring αvariants from fulfilling the BOR, within the
EBSD data, wasdetermined using the βreconstruction software.In Ti al-
loys this method is effective because, although there will be local vari-
ability between specificαvariants, on average the degree of rotation
of neighbouring αplates away from the ideal BOR increases with plastic
strain [38]. Although neither of the above techniques gives direct values
for the strain, they both give a reliable qualitative indication of how the
strain is relatively distributed within the wall.
Table 1
WAAM deposition parameters used to build the walls investigated.
Deposition parameter Value
Travel speed 270 mm/min
Average arc voltage 12 V
Average current 110 A
Wire feed speed 1.6 m/min
Frequency 10 Hz
Trailing shield gas flow rate 20 l/min
Table 2
Changein average layer heightand wall width after rollingeach added layer in the Ti6Al4V
builds, along with the estimated average true principle strains.
Sample Layer
height
Rolling
reduction
Wall
width
Change in
width
ε
z
ε
y
ε
x
Control 1.13 –5.71 ––––
50 kN 1.04 0.09 6.17 0.46 −0.083 0.077 0.006
75 kN 0.93 0.20 6.71 1.00 −0.19 0.16 0.03
(mm) (mm) (mm) (mm)
105J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
3. Results
3.1. Overview
Macroscopic optical images of sections through the centre of the
WAAM walls (x–zplane) produced with and without deformation ap-
plied to each added layer, are shown in Fig. 2. The main features that
can be noted are the prior βgrain structure that developed on solidifica-
tion, before transformation to an α–βlamellar microstructure on
cooling to room temperature, and the regularly spaced horizontal
white bands. In the un-rolled wall (Fig. 2a) the presence of very large
prior columnar βgrains can clearly be seen, which run upwards virtual-
ly throughout the entire build height. In comparison, when rolling was
applied (Fig. 2a & b), coarse columnar β-grains are absent and the
new grain structure is hard to distinguish at this magnification. Howev-
er, smaller columnar grains can still be observed in the last addedlayer,
which has not been reheated by subsequent deposition passes.
The spacing of the white bands seen in Fig. 2 corresponds to the
height of each added layer. The bands are formed within the Heat Af-
fected Zone (HAZ) that developed by the moving thermal field below
the heat source, as each new layer is deposited [9]. Similar banding
has been seen in other AM processes [6,9,12,15]and is thought to
occur at a depth where the peak temperature reached was just below
the βtransus temperature. This has been reported to result in local
coarsening of the transformation microstructure [9, 12].InmostAM
processes the thermal field depth that causes this effect corresponds
to three to five added layers, depending on the processing conditions
[6, 9, 11]. With the WAAM process the first white band can be seen to
be at a depth of about 8 mm below the top surface (Fig. 2)andoccurred
at a depth equivalent to approximately four layers. In the last layer de-
posited, the newly added and re-melted material solidifies as the β
phase and the material within the HAZ of the last pass above the top
white band is fully re-heated into the βphase field. This region then
transforms directly to αon cooling, although material lower down in
the HAZ will have been re-heated and re-transformed between one
and three times, depending on the sequential layer number [6].Hence,
the top four layers above the last white band, where deformation was
concentrated when each new layer was rolled (see Section 3.2), had a
similar finelamellar transformation microstructure. This microstructure
was relatively uniform and was mainly comprised of a Widmanstätten
αmorphology, with thin layers of retained βbetween the lath bound-
aries (Fig. 3). Regardless of being rolled or not, all the samples
underwent a very similar cooling rate [30] through the βtransus, and
therefore the αtransformation structure was found to be nearly identi-
cal for all build conditions.
3.2. Strain distribution
In Fig. 2 the application of a rolling pass to each added layer can be
seen to have reduced the overall height of the walls by an amount
that increased with the applied load. The average principal (true)
strains implied by the net wall shape change are shown in Table 2.It
can be seen that the samples rolled with an applied load of 50 and
75 kN had deformed by an average compressive strain in zof 8 and
Fig. 2. Macroscopic views of sections through the WAAM walls, cut along their centre, x–z, plane produced; (a) without deformation and with an increasing applied rolling load, of
(b) 50 kN and (c) 75 kN.
Fig. 3. Example of the typical α–βtr ansformation microstructure see n in the WAAM
deposits.
106 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
19%, respectively. The net strain values indicate a decrease in average
layer height and an increase in wall width after deformation, while
the increase in length of the sample along eachwall in the rolling direc-
tion was relatively small. This is to be expected because, with a thin wall
that is attached to a base plate, there is high constraint alongthe wall in
the rolling direction and less lateral constraint than with a conventional
rolling geometry. The net average shape change experienced by the
WAAM walls was thus close to a plane strain, but with the principal
components rotated 90° about the build direction, z, (or ND), relative
to RD in conventional rolling of a wide plate [39]. However, as will be
seen belowfrom the EBSD strain mapping results, the local strain distri-
bution in each layer was found to be highly non-uniform.
The local strain distribution in an individual layer was investigat-
ed using the techniques described in section 2.4.InFig. 4 the change
in EBSD pattern quality determined from the band slope (BS) is plot-
ted with depth down the centre line of the walls rolled with 50 and
75 kN loads. It should first be noted that the periodicity observed in
the BS plot deeper in each wall (on the right hand side of the plot)
is caused by the microstructural bandin g in the transformation struc-
ture described above (Fig. 2). This occurs because local variations in
the coarseness of the microstructure within a layer systematically
changes the density of αplate boundaries across it, which in turn in-
fluences the average value of the BS due to the poorer quality of over-
lapping diffraction patterns encountered at grain boundaries. In
comparison, towards the left side of the graph (i.e. closer to the top
wall surface), which is from the region above the last white band
seen in each wall, there is little variation in the αmicrostructure.
When this effect is taken into account, the underlying trend in the
BS curves can be attributed to the relative plastic strain within the
α-phase. For both rolling loads the BS value can be seen to decrease
to a minimum below the top wall surface before increasing again to
level out at constant value. This behaviour can be interpreted as the
plastic strain from rolling being low near the top surface and increas-
ing to a maximum at a depth of between 1.5 and 2.5 mm, before then
falling off further down the wall to approach zero at a depth of about
5 and 8 mm, for the 50 and 75 kN loads, respectively. It can be further
seen that for the sample produced with a 75 kN rolling load the min-
imum position in BS value is both deeper and wider than with a
50 kN load; suggesting that, not only the local strain was larger
with a greater rolling load, but that the depth where the greatest
plastic strain occurred also penetrated further below the rolled sur-
face (2 as opposed to 1.5 mm, with the 75 and 50 kN loads,
respectively).
The conclusion that theplastic strain generated by rolling each layer
was low at the top surface, and greatest between 1.5 and 2.5 mm deep
in each wall, can be attributed to the constraint imposed by the profiled
roller. This interpretation can be further corroborated by using the BOR
misorientation mapping technique, the results of which are depicted in
Fig. 5.Aswellasconfirming that the maximum strain was concentrated
at a specific depth below the top of each rolled wall, the map in Fig. 5
also demonstrates that the strain developed by the grooved roller was
not evenly distributed across a wall's width, being focused in its centre
with respect to the wall's width.
3.3. Effect of rolling on refinement of the primary β-grain structure
Fig. 6 compares EBSD orientation maps of both the α-phase and re-
constructed parent βgrain structures seen in the un-rolled control and
rolled WAAM walls. The maps depicted are x–zcentre plane cross-
sections taken from the top of each wall to a depth of ~10 mm (equiva-
lent to about the last 8 layers). In the αphase EBSD maps (Fig. 6a–c) a
memory of the parent βgrain structure is evident from the texture clus-
tering seen in the αvariants, which is particularly obvious in the un-
rolled control sample. However, following reconstruction it becomes
very apparent that a coarse columnar βgrain structure with a strong
texture developed in the un-rolled wall before its transformation to α
(Fig. 6d). The red colouring that dominates the βphase in this IPF map
indicates that the columnar grains have strong preferential alignment
with a mutual b001Nfibre [13–16]. In the undeformed wall it can also
be seen that the columnar βgrains have a width of approximately
2 mm and their lengths' can be measured in centimetres.
In contrast, when the rolling step was applied to each deposited
layer, a much more refined equiaxed β-grain structure was observed,
as well as a weaker texture (Fig. 6e & f). Greater grain refinement was
also seen in the wall that had received a higher rolling load. In Fig. 6e
& f it is further apparent from the grain structure that is formed in the
last layer, how quickly a columnar structure is re-established when it
has not been refined by rolling and re-heating. Finally, a band can be
seen in Fig. 6e and f between a depth of 1–4 mm below the rolled wall's
top surfaces where there are unindexed points in the maps. These miss-
ing points result from data being discounted during βreconstruction
because it had too great a deviation from the BOR and the depth of
this band coincides with the strain distribution generated by rolling
the last layer, as described above (Figs. 4 and 5).
Fig. 7 shows EBSD maps from y–ztransverse wall cross-sections that
have been coloured to highlight the size distribution of the refined β
grains. In (a) and (b) rolling was only applied to the penultimate
layer, and in (c) and (d), as before, every layer was rolled. The images
in Fig. 7c and d, therefore, correspond to the transverse sections from
the walls previously described in Fig. 6. The grain sizes given are the
Fig. 4. EBSD pattern quality, measuredby the relative band slope, for the two rolled walls
plotted with vertical distance down their centre from their top surface.
Fig. 5. The strain distribution seen in the top section of the 75 kN rolled wall, inferred by
plotting the deviation of neighbo uring αlaths from fulfilling the ideal Burger's
orientation relationship.
107J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
equivalent circular diameter of the reconstructed β-grains. From Fig. 7a
and b it is evident that when a rolling pass was only applied to the pen-
ultimate layer, and it was re-heated duringdepositing the next and final
layer, refinement of the βgrain structure predominantly occurred
below the top surface within a core region in each wall. This region of
βrefinement can be seen to increase with rolling load and shows
good qualitative agreement with the plastic strain distribution inferred
from Fig. 6.
When rolling was applied to every added layer (Fig. 7c & d) the
width of the refined core region increased, relative to the samples
where only the penultimate layer was rolled (Fig. 7a & b), and extended
all the way down each wall. The average grain sizes measured below the
top layer across these walls are given in Fig. 8a. With both rolling loads,
the grain size can be seen to increase towards each wall's outer surfaces
and there isa central plateaus region where the average grain size is re-
fined to a minimum level, which reduces with increasing applied load.
In the core region at the centre of the walls the βgrain size was reduced
to around 140 and 90 μm with rolling loads of 50 and 75 kN, respective-
ly. In addition, with the higher 75 kN load the grain size in the wall core
was more uniform and the core region was proportionally wider com-
pared to the wall thickness.
An equivalent plot depicting the variation in grain size with vertical
position down the centreline of the rolled wall samples is given in Fig.
8b. The grain size data for both rolling loads shows an increase in aver-
age diameter at the top of each wall, which corresponds to the re-
establishment of a columnar βstructure in the final layer to solidify
(Fig. 6e & f). Below this transition region, rolling each layer refined the
grain size to a relatively uniform minimum level with the larger 75 kN
rolling load, but when the 50 kN downforce was used there was far
more variability. A histogram of the grain sizes in the fully refined
core region of this cross-section is given in Fig. 9. It should be noted
that grain sizes below the reliable measurement limit for the set up
used have been excluded (25 μm) from the EBSD data. However, the av-
erage grain sizes were found to be 130 μmand94μmforthe50kNand
75 kN rolling loads respectively, which corresponds well with measure-
ments made by the line intercept method on optically acquired micro-
graphs in a previous study [30]. In it is also evident from Fig. 9 that
the βgrain size distribution from the refined core regions of the rolled
walls was more tightly distributed around the smaller mean value
when the larger rolling load was used. For example, with a 50 kN
down force the largest grain measured 590 μm) was almost twice the
size of that observed in the 75 kN sample (350 μm).
3.4. Effect of rolling on texture in the WAAM process
Strong crystallographic texturescan cause anisotropy in a material's
mechanical properties and this can be particularly pronounced in titani-
um alloys that have hcp crystal structures [18, 40, 41]. In the un-
deformed AM wall the reconstructed EBSD map in Fig. 6d clearly indi-
cates the presence of a strong β-texture. However, textural changes
can also arise from deformation, or its influence on any recrystallization
or phase transformation that may occur during subsequent heat treat-
ment [40, 41]. It is thus important to compare the texture seen in an
un-deformed build with that found when a rolling step was utilised
with the WAAM process to refine their coarse primary βgrain
structures.
Pole figures obtained from the large area EBSD maps depicting the
α-textures and the reconstructed parent β-textures that existed prior
to transformation, on cooling down below the β-transus temperature,
are provided in Figs. 10 and 11. The EBSD maps were taken from the
centre, x–z, plane of each wall, and for the rolled walls therefore mainly
represent orientations from their more refined core region (Fig. 7). The
Fig. 6. Measured α-phase(a–c) and reconstructed β-parent phase (d–f) IPF orientation coloured EBSD maps from the undeformed control and rolledWAAM walls, produced with rolling
loads of 50 and 75 kN, obtained from mid plane, x–z,wallsections.
108 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
pole figures are orientated with the build direction and compression
axis (z) near to normal to the plane of projection and the torch travel di-
rection (x) is aligned vertically. Before presentingthese results it should
be noted that the sampling statistics for the solidification texture in the
un-deformed wall were poor, owing to the large prior-βgrain size (only
10 grains were covered in the map area), whereas the statistics for the
textures in the rolled samples were more representative because of
their more refined grain structure.
Of the αtextures, shown in Fig. 10, the texture of the un-deformed
as-deposited wall is difficult to describe at first sight because of the
poor βparent grain sampling statistics. However, it becomes much
clearer when βphase reconstruction is performed. From the 100 β
phase pole figure in Fig. 11a it can be seen that the αphase transformed
from βparent grains that had a common b001Ndirection aligned close
to the build direction. The corresponding α-texture seen in Fig. 10ais,
therefore, an b001 Nβtransformation texture that is poorly defined,
partly because of the poor statistics for the parent βgrain orientations,
but also because it has become weakened by the twelve possible αvar-
iant orientations available through the Burgers relationship [42–44].
In comparison to the strong texture seen in the control sample,
much weaker αand βtextures were found in the rolled walls (Figs.
10b&c;11b & c), which had maximum intensities of less than approx-
imately 3 and 4 times random for the αand βphases, respectively. In
both of the rolled walls it was found that the strongest orientations
present in their reconstructed βpole figures were consistent with the
textures containing residual weakened cube components. As can be
seen from Fig. 11b and c both the rolled wall's 100 pole figures have
one b001 Npole aligned close to the build direction, z, with the other re-
lated b001Npoles rotated by different amounts about zrelative to the
deposition direction (e.g. ~10 and 20° in the 50 kN and 75 kN examples
shown). These residual b001 Nfibre orientations were weaker in the
wall rolled with a greater down force; the maximum intensity dropping
from 3.5 times random for the 50 kN rolling load to 2.8 for the 75 kN
load and this led to a corresponding drop in the strength inherited in
Fig. 7. EBSD reconstructedβgrain size maps from transverse, y–z, cross-sectionsnear the top of rolled WAAM walls; (a) and (b) with onlyone rolling pass appliedto the penultimate layer
and (c) and (d) with a rolling pass applied to every layer, both with rolling loads of 50 kN and 75 kN respectively.
109J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
the αtransformation textures, on applying rolling to each layer, from
2.7 to 2.2 times random.
4. Discussion
With titanium alloys, coarse-directional grain structures are of con-
cern in the industrial application of AM processes because of their po-
tential to cause anisotropic properties in aerospace components [13,
16, 23]. The above results show that the introduction of a rolling defor-
mation stage sequentially within the WAAM build cycle could be a use-
ful technique for refining such undesirable grain structures, as well as
for reducing the intensity of the strong textures normally seen in the
as-deposited material. Interestingly, the strain required to achieve a
high level of βrefinement has been found to be relatively low, which
makes it practical to apply such a technology when producing compo-
nents with relatively simple geometries. However, other methods for
introducing plastic deformation in AM are also being investigated,
such as peening [28], which offers higher compatibility with more com-
plex component designs.
4.1. Grain structure development in the conventional WAAM material
By reconstruction of the parent βphase that forms on solidification,
it has been confirmed that in the conventional WAAM process a very
coarse primary columnar grain structure is developed as material is
built up, by sequentially adding many new layers. This columnar grain
structure has a strong preferential b001Nβtexture. It has been previ-
ously noted that the main reason such a coarse grain structure develops
is because the high thermal gradient in the melt pool and the alloy
chemistry do not allow sufficient constitutional supercooling for nucle-
ation to be possible ahead of the solidification front [17]. Similar to
welding, in AM when a layer is added the temperature rapidly increases
below the fusion boundary and the material high in the heat affected
zone will fully transform back to the βphase, whereupon grain coarsen-
ing will occur where the peak temperature reached is significantly
above the βtransus [11] (Fig. 6d). When solidification of the single β
phase subsequently takes place at therear of the moving melt pool, be-
cause nucleation ahead of the growth front is not possible, the coars-
ened β-grains at the fusion boundary then act as a substrate for
epitaxial re-growth back into the liquid following the maximum ther-
mal gradient, which is normal to the solidification front. A columnar
structure is thus developed as grains grow following the rear melt
pool surface [45].
This process repeats asmore layers are added, with the retained βin
the transformation microstructure re-growing, each time when it is re-
heated in each new cycleto re-create βgrainswith the same orientation
they had in the previous pass. When re-heated above the βtransus, the
reformed βgrains can also potentially coarsen each cycle, before provid-
ing the substrate for epitaxial re-growth at each new fusion boundary.
Hence, in AM without deformation the same grain orientations re-
grow over may layers as each new layer is added. As the β-grains devel-
op, grain growth thus occurs both in the solid state and during solidifi-
cation where orientations are progressively selected that have a
preferred b001 Ncrystallographic direction parallel to the maximum
thermal gradient at the solidification front [45], leading to the strong
b001Ntexture seen in the final wall [13–16] (Figs. 5d, 11a). One of
the reasons a deformation step is so effective in causing βgrain refine-
ment and a weaker texture in AM is thus because it has the potential
to disrupt the accumulative ‘ratcheting’effect of this repeated cyclic be-
haviour that is inherent in an additive ‘layer’manufacturing process.
4.2. Grain refinement in the rolled WAAM deposits
In the samples studied, the rolling loads employed led to relatively
modest average compressive plastic strains of 8 and 19%. Nevertheless,
when each deposited layer was sequentially rolled during the AM pro-
cess, the original coarse columnar grain structure was found to become
greatly refined, giving rise to an equiaxed β-grain structure with an av-
erage diameter of b100 μm in the core of the wall produced with the
75 kN rolling load. Although the refined βgrain size increased towards
the surface of the walls, thisoverall reductionin grain size stillcompares
very favourably to the centimetre-scale columnar grains seen in stan-
dard un-deformed samples (Fig. 2). In addition, both the parent βand
αphase texture strengths was substantially weakened in the rolled
samples.
The refined βgrains were formed at temperatures well above the β
transus temperature. This is cleary apparent from comparison of Fig. 2
with the refined region in Fig. 7b. In Fig. 2 the top white bands occur
Fig. 8. The average βgrain size variation seen in walls produced when every layer was
rolled after deposition, with a 50 or 75 kN applied load; (a) across the walls over 5
layers below the final white band and (b) as a function of depth down their centre lines.
In (a) the dashed lines indicate the outside surfaces of each wall.
Fig. 9. Comparison of the βgrain size distributions measured from reconstructed EBSD
maps obtained at the centre of the walls rolled with a load of 50 kN and 75 kN.
110 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
where the maximum temperature reached in the last pass was just
below the βtranus temperature (as has been widely acknowledged
[6]) and this is at a depth of over 8 mm's below the top surface which
is well below the depth where βrefinement occurs in Fig. 7b. In addi-
tion, refinement in the βphase would not beexpected through conven-
tional recrystallization as the driving force is usually insufficient at such
low strain levels of below 20% [46]. It can therefore be concluded that
the grain refinement seen at these low applied strains is caused by the
growth of new βorientations associated with the influence of deforma-
tion on the α→βphase transformation, which occurs during re-heating
above the βtransus temperature, when the next layer is deposited. This
implies rolling creates new βorientations, other than those retained
from the parentβgrains in the original transformation structure formed
during cooling each solidified layer. These orientations can then act as
new nuclei that grow on re-heating each layer above the βtransus tem-
perature when the next layer is added. This prevents the original β
grains simply re-growing the orientations of the substrate during epi-
taxial solidification of the next layer deposited, which leads to the
coarse columnar structures normally seen in undeformed WAAM
walls, by the ‘ratcheting process’discussed above.
Because the strain applied to each deposited layer was relatively low,
the origin of the new βorientations is most probably related to heteroge-
neities that develop within the deformation of the fine Widmanstätten
transformation microstructure found in AM titanium parts. One possible
mechanism, already discussed by the current authors in Ref [47], has
been observed directly using in-situ heating experiments and involves
new βorientations originating from deformation twinning of αlaths
within the deformed region below the top of a rolled wall. Further relat-
ed mechanisms, by which new βorientations can be generated by the
application of relatively low levels of plastic deformation in AM, also as-
sociated with deformation twinning, and from strain concentration at
colony boundaries, are still under investigation and will be the subject
of a future publication.
EBSD strain mapping has shown that the core wall region where β
grain refinement was found to be closely related to the local strain dis-
tribution generated in each rolling pass (Fig. 5). This is strongly influ-
enced by the shape of the roller which was profiled to match the
curvature of the top layer bead profile. With a grooved roller, the top
of the wall is highly constrained and cannot spread sideways, which
causes a ‘dead-zone’nearthe roll surface and forcesthe plastic deforma-
tion to be concentrated at a greater depth within the wall core. Although
this strain distribution was not pre-planned, it is beneficial when trying
to combine deformation with AM techniques like the WAAM process
that features a large re-melt depth (~1.5 mm) as it means that the ma-
jority of useful deformation is not lost by the deformed material being
re-melted by the subsequent pass. Other techniques, such as peening,
tend to concentrate plastic deformation closer to a surface [48] and
would therefore be less suitable for refining the grain structure unless
they are combined with an AM technique that has a low re-melt
depth [28].
A negative effect of the grooved roller is that the strain introduced is
lower near the wall faces, which leads to a more refined core with
coarse grains towards each wall edge. However, when rolling was ap-
plied to every added layer (Fig. 7c & d) the width of the refined core re-
gion was foundto increase and the grain sizein the less refined skin was
still much smaller than in the undeformed wall. The fact that rolling
each layer was more effective than rolling a single layer, is discussed fur-
ther below and can be explained partly by the fact that repeatedly
Fig. 10. Polefigures obtained from large area EBSD maps depicting αtextures measured from the centre of; (a) the un-rolled wall, (b) the wall rolled with a 50 kN load, and (c) the wall
rolled with a 75 kN load.
111J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
rolling eachlayer does not allow a coarse βstructure to develop, prior to
the application of an individual rolling pass, and partly because defor-
mation occurs to a sufficient depth that there is some overlap with the
previous layer, which will therefore be deformed and re-heated above
the βtransus more than once when rolling is applied in every pass.
Apart from in the last added layer, which is not re-heated following
deformation, it was found that there was very little variation in grain
size with build height for the wall deformed with a 75 kN rolling load
every pass. However, more variation was observed in the 50 kN wall
(Fig. 8b). The cause of this variation can be related to the lower depth
of deformation and greater inconsistency in the deposition conditions
that led to more irregularity in the layer height with this sample, as
can be observed from the greater unevenness in the vertical spacing of
the white bands seen in the 50 kN wall as opposed to the 75 kN sample
(Fig. 2). Furthermore, in Fig. 12 it can be noted that there is a clear cor-
relation between the white band layer-thickness and the average re-
fined grain size. It is apparent that a lower rolling load is likely to
exaggerate such variability, because there is a lower penetration depth
of the plastic strain in each layer; i.e. where there is a narrower, or
wider, layer separation this will lead to a finer, or coarser, βgrain size
because of the greater variation in strain overlap between deformation
passes with a lower rolling load.
4.3. Texture development
The observation of a strong b100Nβfibre texture in the un-rolled
walls has been discussed above (section 4.1) and is consistent with
studies performed with other AM techniques [11, 14]. In the 100 pole
figure in Fig. 11a the main fibre axis can be seen to have an intensity
of 15 times random, although this value should be treated with cautions
as, despite the large area mapped (12 by 6 mm), the prior βgrain size
was so large sufficient data could not be obtained to develop a full tex-
ture description (Figs. 10aand11a). Owing to the poor sampling statis-
tics, an attempt has therefore not been made to evaluate the possibility
of variant selection affecting the α-texture. However, it is interesting
that in a recent paper Sargent et al. [31] have claimed that transforma-
tion strains can cause variant selection within similarly coarse grains
found in Ti–6Al–4V castings. Encouragingly, the application of a defor-
mation step to each layer greatly weakened the βparent textures, to
Fig. 11.Pole figures obtainedfrom large areaEBSD maps depictingthe parent phaseβtextures reconstructed fromthe centre of; (a) the un-rolled wall,(b) the wall rolled witha 50 kN load,
and (c) the wall rolled with a 75 kN load.
Fig. 12. Average grain size plotted against layer height for the two rolling loads.
112 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
3.5 and 2.8 times random with the 50 kN and 75 kN rolling loads, re-
spectively. On transformation this was further diluted leading to ex-
tremely weak αtextures being seen in the rolled samples of 2.7 and
2.2 times random. Although the maximum intensities were greatly re-
duced, both of the rolled walls still retained a memory of the original
βphase b001Nfibre texture that tended to be present as a weakerrotat-
ed cube orientations. Overall this suggests that the presence of these
components is caused by insufficient sampling statistics, which resulted
in a tendency for orientations related to one dominant ‘harder’parent β
grain, that was more resistant to refinement, to be retained from the
original b001Nfibre within each EBSD map area.
Finally, the synergistic advantages of applying a rolling step to each
layer after it is deposited are highlighted in Fig. 13, which compares
the grain size and texture strength down the centre of the wall rolled
with the higher load, from its top surface, with an enlarged EBSD map
of the same region. In the last layer to be added it can be seen that
there is a rapid increase in the βgrain size and texture strength towards
the top surface of the deposit. It can be noted that this is partly caused by
βgrain growth in the HAZ below the melt pool, and then is further en-
hanced by b001Ngrowth selection during the development of a colum-
nar structure by directional solidification. However, in this single layer
the grain size reached and texture strength are nowhere near as high
as that developed in the un-deformed wall, where the grain size and in-
tensity of the b001Nfibre progressively increase through the ‘ratcheting’
of these two effects during the addition of further layers as a wall is built.
This is because when each layer is sequentially rolled the development of
a columnar grain structure and strong fibre texture is disrupted after
every layer is deposited, and does not have the same opportunity to de-
velop during multiple repeated cycles.
Of additional significance to the synergistic benefit of rolling every
layer is that each layer will have a finer grain size and weaker texture
before it is rolled than in an unrolled wall, where the columnarstructure
is much more developed. The single data points shown in Fig. 13,taken
from the rolled refined region of the sample where rolling was applied
only to the penultimate layer, of an otherwise undeformed wall (Fig.
7b), demonstrate that, in this case the level of grain refinement in the
deformed core region is similar, but the resultant texture is much stron-
ger (4.5 as opposed to 2.7 times random) than when the starting micro-
structure has been previously refined; i.e. when a single rolling pass was
applied to a coarse grained undeformed columnar structure, following
re-heating by adding the next layer, the grain size refinement was ap-
proximately the same as when you start with a refined grain structure,
but the reduction in texture strength was still far less. This is because
when you start with a stronger texture, that has been able to develop
over many layers of βgrain growth in an unrolled wall, it will require
a larger rolling strain to reduce it to the same level.
Therefore, the addition of a rolling pass to every layer leads to a weak
texture in the wall by both preventing the formation of a strongly
textured columnar grain structure, which requires many layers of un-
disturbed growth to develop, and by this, in turn,ensuring a far weaker
texture in the next layer before it is deformed. The combination of these
two effects thus greatly decreases the rolling strain required to reduce
the texture strength in a WAAM component.
5. Conclusions
The efficacy of a deformation step, on refining the primary βgrain
structure and texture developed in a Ti–6Al–4V alloy, has been investi-
gated during Wire–Arc Additive Manufacture (WAAM). By reconstruc-
tion of the βparent phase, from αphase EBSD maps, it has been
confirmed that in the conventional WAAM process the deposited mate-
rial develops centimetre-scale, coarse-columnar, βgrains that grow
through the build height with an associated strong b001 Nβfibre tex-
ture. This coarse grain structure results from the retained βin the trans-
formation structure re-growing with the same orientations when
heated above the βtransus, with the next new layer added. Once re-
formed, the βgrains coarsen and act as a substrate for epitaxial colum-
nar growth during solidification of the new layer.
The application of a rolling step sequentially to each added layer was
surprisingly effective, in termsof the low level of strain required, to both
refine the βgrain size and weaken the primary βand final αtextures,
which were reduced to close to random by the application of an only
modest 8–20% rolling reduction. However, the homogeneity of the re-
fined βgrains was found to improve with increased levels of deforma-
tion. A profiled roller has also been shown to be advantageous for
increasing the depth of deformation in each layer —thus helping the de-
formed region to survive re-melting in the next addition cycle.
It is postulated that the deformation step causes βgrain refinement
through promoting twinning, which generates new βorientations that
then grow during the α→βtransformation as each layer is re-heated
by the subsequent deposition pass.
There are synergistic advantages of rolling each layer because this
disrupts the establishment of the coarse columnar grain structure that
only normally develops over many undisturbed repeated addition cy-
cles. Rolling each layer also ensures that the βgrain structure is refined
and the texture weaker in each new added layer, before it is deformed,
and this decreases the deformation required to obtained a weaker
texture.
Acknowledgements
The authors would like to thank Prof. Brad Wynne (University of
Sheffield) for provision of the βreconstruction software. J. Donoghue
is grateful for financial support provided by LATEST2 (EP/H020047/1)
and Airbus, UK.
References
[1] I. Gibson,D.W. Rosen, B. Stucker, Additive Manufacturing Technologies, Springer US,
Boston, MA, 2010.
[2] J. Kruth, M. Leu, T. Na kagawa, Progress in additive manufa cturing and rapid
prototyping, CIRP Ann. Technol. 47 (1998) 525–540.
[3] P. Kobryn, N. Ontko, L. Perkins, J. Tiley, Additive Manufacturing of Aerospace Alloys
for Aircraft Structures, 2006 2–12.
[4] E.O. Ezugwu, Z.M. Wang, Titanium alloys and their machinability—a review, J Mater
Process Technol 68 (1997) 262–274.
[5] X. Wu, J. Liang, J. Mei, C. Mitchell, P.S. Goodwin, W. Voice, Microstructures of laser-
deposited Ti–6Al–4V, Mater. Des. 25 (2004) 137–144.
Fig. 13.Pl ot of βtextural strength(given by the maximum intensity in the 100 pole figure)
and grain size variationwith height down the centre line, for thewall rolled with a loadof
75 kN applied to every added layer. Also indicated in bl ue are the values for the
corresponding region in the sample produced with only one 75 kN rolling pass applied
to the penultimate layer. The magnified insert is of the βgrain structure in top 9 layers
of the wall.
113J. Donoghue et al. / Materials Characterization 114 (2016) 103–114
[6] S.M. Kelly, S.L. Kampe, Microstructural evolution in laser-deposited multilayer Ti–
6Al–4V builds : part I, Microstruct. Character. 35 (2004) 1861–1867.
[7] S. Kelly, S. Kampe, Microstructural evolution in laser-deposited multilayer Ti–6Al–
4V builds: part II. Thermal modeling, Metall. Mater. Trans. A 35 (2004) 1869–1879.
[8] F. Wang, J.Mei, X. Wu, Microstructure study of direct laser fabricated Ti alloys using
powder and wire, Appl. Surf. Sci. 253 (2006) 1424–1430.
[9] F. Martina, J. Mehnen,S.W. Williams, P. Colegrove, F. Wang, Investigation of the ben-
efits of plasma deposition for the additive layer manufacture of Ti–6Al–4V, J. Mater.
Process Technol. 212 (2012) 1377–1386.
[10] F. Wang, S. Williams, M. Rush, Morphology investigation on direct current pulsed
gas tungsten arc welded additive layer manufactured Ti6Al4V alloy, Int. J. Adv.
Manuf. Technol. 57 (2011) 597–603.
[11] F. Wang, S. Williams, P. Colegrove, A.A. Antonysamy, Microstructure andmechanical
properties of wire and arcadditive manufactured Ti–6Al–4V, Metall.Mater. Trans. A
44 (2012) 968–977.
[12] E. Brandl,A. Schoberth, C. Leyens, Morphology, microstructure, and hardness of tita-
nium (Ti–6Al–4V) blocks depos ited by wire-fe ed additive layer manufacturing
(ALM), Mater. Sci. Eng. A 532 (2012) 295–307.
[13] A.A. Antonysamy, P.B. Prangnell, J. Meyer, Effect of wall thickness transitions ontex-
ture and grain structure in additive layer manufacture (ALM) of Ti–6Al–4V, Mater.
Sci. Forum 706-7 09 (2012) 205–210.
[14] S.S. Al-Bermani, M.L. Blackmore, W. Zhang, I. Todd, The origin of microstructural di-
versity, texture, and mechanical properties in electron beam melted Ti–6Al–4V,
Metall. Mater. Trans. A 41 (2010) 3422–3434.
[15] P. Kobryn, S. Semiatin, Microstructure and texture evolution during solidification
processing of Ti–6Al–4V, J. Mater Process. Technol. 135 (2003) 330–339.
[16] A.A. Antonysamy, Microstructure, Texture and Mechanical Property Evolution dur-
ing Additive Manufacturing of Ti6Al4V Alloy for Aerospace Applications, University
of Manchester, 2012.
[17] M.J. Bermingham, S.D. McDonald, M.S. Dargusch, D. H. StJohn, Grain-refinement
mechanisms in titanium alloys, J. Mater. Res. 23 (2011) 97–104.
[18] G. Lütjering, J.C. Williams, Titanium, Springer, Berlin Heidelberg, Berlin, Heidelberg,
2007.
[19] I. Bantounas, D. Dye, T.C. Lindley, The role of microtexture on the faceted fracture
morphology in Ti–6Al–4V subjected to high-cycle fatigue, Acta Mater. 58 (2010)
3908–3918.
[20] T. Vilaro, C. Colin, J.D.Bartout, As-fabricated and heat-treated microstructures of the
Ti–6Al–4V alloy processed by selective laser melting, Metall. Mater. Trans. A 42
(2011) 3190–3199.
[21] H.K. Rafi, N.V. Karthik, H. Gong, T.L. Starr, B.E. Stucker, Microstructures and mechan-
ical properties of Ti6Al4V parts fabricated by selective laser melting and electron
beam melting, J. Mater. Eng. Perform. 22 (2013) 3872–3883.
[22] F. Martina, S.W. Williams, P. Colegrove, Improved microstructure and increased me-
chanical properties of additive manufacture produced Ti–6Al–4V by interpass cold
rolling, SFF Symp. 2013, pp. 490–496.
[23] B.E. Carroll, T.A. Palmer, A.M. Beese, Anisotropic tensile behavior of Ti–6Al–4V com-
ponents fabricated with directed energy deposition additive manufacturing, Acta
Mater. 87 (2015) 309–320.
[24] X. Wu, R. Sharman, J. Mei, W. Voice, Microstructure and properties of a laser fabri-
cated burn-resistant Ti alloy, Mater. Des. 25 (2004) 103–109.
[25] M.J. Bermingham, S.D. McDonald, K. Nogita, D.H.St. John, M.S. Dargusch, Effects of
boron on microstructure in cast titanium alloys, Scr. Mater. 59 (2008) 538–541.
[26] B. Baufeld, O. van der Biest, Mechanical properties of Ti–6Al–4V specimens pro-
duced by shaped metal deposition, Sci. Technol. Adv. Mater. 10 (2009) 015008.
[27] P. Colegrove, H.E. Coules, J. Fairman, F. Martina,T. Kashoob, H. Mamash, et al., Micro-
structure and residual stress improvement in wire and arc additively manufactured
parts through high-pressure rolling, J. Mater. Process. Technol. 213 (2013)
1782–1791.
[28] J. Donoghue, J. Sidhu, A. Wescott,P. Prangnell, Integration of deformation processing
with Additive Manufacture of Ti–6Al–4V components for improved βgrain struc-
ture and texture, 2015 TMS Annu. Meet. Exhib., Orlando, Florida, 2015.
[29] P. Colegrove, S. Williams, Added Layer Manufacture, GB2491472, 2012.
[30] F. Martina, P.A. Colegrove, S.W. Williams, J. Meyer, Microstructure of Interpass
Rolled Wire + Arc Additive Manufacturing Ti–6Al–4V Components, Metall. Mater.
Trans. A (2015) 1.
[31] G.A. Sargent, K.T. Kinsel, A.L. Pilchak, A.A. Salem, S.L. Semiatin, Variant selection dur-
ing cooling after beta annealing of Ti–6Al–4V ingot material, Metall. Mater. Trans. A
43 (2012) 3570–3585.
[32] P.S. Davies, An Investigation of Microstructure and Texture Evolution in the Near-α
Titanium Alloy Timetal 834, University of Sheffield, 2009.
[33] P.S. Davies, B.P. Wynne, W.M. Rainforth, M.J. Thomas, P.L. Threadgill, Development
of microstructure and crystallographic texture during stationary shoulder friction
stir welding of Ti–6Al–4V, Metall. Mater. Trans. A 42 (2011) 2278–2289.
[34] M. Humbert, N. Gey,The calculation of a parent grain orientationfrom inherited var-
iants for approximate (b.c.c.–h.c.p.) orient ation relations, J. Appl. Crystall ogr. 35
(2002) 401–405.
[35] N. Gey, M. Humbert, Specific analysis of EBSD data to study the texture inheritance
due to the β→αphase transformation, J. Mater. Sci. 8 (2003) 1289–1294.
[36] T. Maitland, S. Sitzman, Scanning microscopy for nanotechnology: techniques and
applications, Springer Science & Business Media, 2007.
[37] J.-Y. Kang, S.-J. Park, M.-B. Moon, Phase analysis on dual-phase steel using band
slope of electron backsc atter diffraction pattern, Microsc. Microanal. 19 (2013)
13–16.
[38] T.E. Buchheit, G.W. Wellman, C.C. Battaile, Investigating the limits of polycrystal
plasticity modeling, Int. J. Plast. 21 (2005) 221–249.
[39] G. Deiter, Mechanical Metallurgy, third ed. McGraw-Hill Science, New York, 1986.
[40] U.F. Kocks, C.N. Tomé, H.-R. Wenk, Texture and Anisotropy: Preferred Orientations
in Polycrystals and Their Effect on Materials Properties , Cambridge University
Press, Cambridge, 2000.
[41] I. Dillamore, W. Roberts, Preferred orientation in wrought and annealed metals,
Metall. Rev. (1965) 10.
[42] I. Lonardelli, N. Gey, H.-R. Wenk, M. Humbert, S.C. Vogel, L. Lutterotti, In situ obser-
vation of texture evolution during α→βand β→αphase transformations in tita-
nium alloys investigated by neutron diffraction, Acta Mater. 55 (2007) 5718–5727.
[43] A.J.J. van Ginneken, W.G. Burgers, The habit plane of the zirconium transformation,
Acta Crystallogr. 5 (1952) 548–549.
[44] G.C.Obasi, S. Birosca, J.Quinta da Fonseca, M.Preuss, Effect of βgrain growth on var-
iant selection and texture memory effect during α→β→αphase transformation in
Ti–6Al–4V, Acta Mater. 60 (2012) 1048–1058.
[45] S. David, J. Vitek, Correlation between solidification parameters and weld micro-
structures, Int. Mater. Rev. (1989) 34.
[46] R. Ding, Z. Guo, Microstructural evolution of a Ti–6Al–4V alloy during β-phase pro-
cessing: experimental and simulative investigations, Mater. Sci. Eng. A. 365 (2004)
172–179.
[47] J. Donoghue, A. Gholinia, J. Quinta da Fonseca, P.B. Prangnell, In-situ High Tempera-
ture EBSD Analysis of the Effect of aDeformation Step on the Alpha to Beta Transi-
tion in Additive Manufactured Ti–6Al–4V, TMS Titan, San Diago, 2015.
[48] R.K.Nalla, I. Altenberger, U. Noster,G.Y. Liu, B. Scholtes, R.O. Ritchie, On the influence
of mechanical surface treatments-deep rolling and laser shock peening-on the fa-
tigue behavior of Ti–6Al–4V at ambient and elevated temperatures, Mater. Sci.
Eng. A 355 (2003) 216–230.
114 J. Donoghue et al. / Materials Characterization 114 (2016) 103–114