International Journal of Plasticity 179 (2024) 104022

Contents lists available at ScienceDirect

International Journal of Plasticity

journal homepage: www.elsevier.com/locate/ijplas

A predictive mesoscale model for continuous

dynamic recrystallization
Franz Miller Branco Ferraz a, b, *, Ricardo Henrique Buzolin c, Stefan Ebenbauer d,
Thomas Leitner e, Alfred Krumphals e, Maria Cecilia Poletti a, b
a
Institute of Materials Science, Joining and Forming at Graz University of Technology. Kopernikusgasse 24/I, Graz 8010, Austria
b
Christian Doppler Laboratory for Design of High-Performance Alloys by Thermomechanical Processing, Kopernikusgasse 24/I, Graz 8010, Austria
c
Voestalpine Tubulars GmbH & Co KG, Alpinestrasse 17, Kindberg-Aumuehl A-8652, Austria
d
Voestalpine BÖHLER Bleche GmbH & Co KG, Böhler-Gasse 1, Mürzzuschlag 8680, Austria
e
Voestalpine BÖHLER Edelstahl GmbH & Co KG, Mariazellerstraße 25, Kapfenberg 8605, Austria

A R T I C L E I N F O A B S T R A C T

Keywords: Thermomechanical processing of titanium alloys often requires complex routes to achieve the
Modeling desired final microstructure. Recent advances in modeling and simulation tools have facilitated
FEM the optimization of these processing routes. However, existing models often fail to accurately
Continuous dynamic recrystallization
predict microstructural changes at large deformations. In this study, we refine the physical
Torsion
Misorientation angle
principles of an existing mean-field model and propose a calibration method that uses experi­
Titanium alloys mental results under isothermal conditions, accounting for the actual local deformation within
the workpiece. This new approach improves the predictability of microstructural changes due to
continuous dynamic recrystallization during torsion and compression experiments. Additionally,
we integrate the model into the commercial FEM-based DEFORM™ 2D software to predict the
local microstructure evolution within hot torsion specimens thermomechanically treated by
resistive heating. Validation using non-isothermal deformation tests demonstrates that the model
provides realistic simulations at high strain rates, where adiabatic heat modifies temperature,
flow stress and microstructure. This study demonstrates the intrinsic correlation between
microstructure, flow behavior, and workpiece geometry, considering the impact of deformation
history in thermomechanical processes.

1. Introduction

Titanium alloys are attractive for many applications because of their high specific strength, high-temperature properties, low
density, biocompatibility, and good corrosion resistance. Among the well-known commercial titanium alloys, Ti-5Al-5Mo-5V-3Cr (Ti-
5553) is a near-β titanium alloy that is particularly suited to high-end applications due to its high mechanical strength and wide
processing window (Hua et al., 2014).
The thermomechanical processing of titanium alloy components typically involves multiple steps of hot working and heat treat­
ment to achieve the desired mechanical properties. In this context, our intial investigations focused on the microstructure development
of titanium alloys during deformation at elevated temperatures. The findings demonstrated that dynamic recovery (DRV) and

* Corresponding author.
E-mail address: [email protected] (F. Miller Branco Ferraz).

https://doi.org/10.1016/j.ijplas.2024.104022
Received 24 March 2024; Received in revised form 3 June 2024;
Available online 8 June 2024
0749-6419/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

continuous dynamic recrystallisation (CDRX) are the primary restoration mechanisms in Ti-55531 and Ti-6Al-4 V alloys (Poletti et al.,
2016). The majority of studies in the literature agree with these findings. However, at very low strain rates in a temperature range from
700 ◦ C to 780 ◦ C (Matsumoto et al., 2014) observed superplastic behaviour in Ti-5553, which is associated with grain boundary
sliding.
Several semiempirical modelling techniques attempt to describe the behaviour of metals and alloys during hot forming. Using
empirical equations, phenomenological models correlate the flow stress with strain, strain rate, and deformation temperature (He
et al., 2023b, 2023a; Khan and Liu, 2012). However, the predictions of this type of model can be inaccurate beyond the range of the
experimental data and are not sensitive to the deformation history and the initial microstructures. While artificial neural networks
(ANN) and fuzzy sets (FNN) correlate the flow stresses of metals with the microstructure (Peng et al., 2013; Quan et al., 2013), these
approaches require a large amount of measurement data. Furthermore, they lack a physical description of the metallurgical phe­
nomena and are unable to predict outside the validated range.
In contrast, physically-based models such as crystal plasticity (Wang et al., 2020; Wroński et al., 2022) and cellular automata (Ao
et al., 2020; Liu et al., 2020; Ridgeway et al., 2020; Xiong et al., 2021) can describe physical phenomena in the microstructure during
hot deformation and resolve local heterogeneities formed within the grains. These approaches provide information on the texture
evolution of polycrystalline materials and can be integrated with finite elements (Jeong et al., 2021; Joo et al., 2023; Rezaei et al.,
2023). However, crystal plasticity finite element (CPFE) and cellular automaton finite element (CAFE) simulations require consid­
erable computational resources to simulate large and complex-shaped forged components, making process development with such
models infeasible. Therefore, mean-field models represent an optimal compromise between physical descriptions and computational
resources. One of the key advantages of mean-field models is their ability to predict trends regarding the influence of processing
parameters on the final product properties without the need of significant computational resources or validation data. Unlike crystal
plasticity models, mean-field models provide only average values of the microstructural features.
Many mean-field models have been developed to describe the hot deformation behavior of materials with high-stacking fault
energy (Li et al., 2020), including titanium and its alloys (Babu and Lindgren, 2013a; Bai et al., 2013; Fan et al., 2013; Fan and Yang,
2011; Gao et al., 2011; Mandal et al., 2017; Yang et al., 2015). This kind of model based on dislocation density and constitutive
equations can correlate the microstructure, applied stresses and internal-state-variables during hot deformation (Galindo-Fernández
et al., 2018), creep (Pradeep et al., 2024; Riedlsperger et al., 2023), low-cycle fatigue (León-Cázares et al., 2020) and high-cycle fatigue
(Wei Lee et al., 2022).
In recent years, we have presented a mean-field model that describes the microstructure evolution during isothermal deformation
of high-stacking fault energy alloys. The model combines dislocation density rate equations, constitutive equations correlating the
stresses with the microstructure, strain partitioning to account for the change in load condition between the two phases of a dual-phase
material, α-dynamic globularization and continuous dynamic recrystallization of both phases. The model has already been applied for
some alloys of titanium (Buzolin et al., 2021a, 2021b; R.H. 2020) and aluminium (Buzolin et al., 2022) under isothermal conditions.
We also implemented the model in FEM to account for heterogeneous deformation within a Ti-17 hat-shaped sample (Ferraz et al.,
2023).
Although the previously proposed model works well up to moderate strains, it fails at large deformations. Another limitation of
these previous investigations is that the validation method relied on experimental data of microstructures affected by adiabatic
heating. This methodology is inaccurate, because the resulting temperature increase alters the kinetics of microstructure evolution. To
address these limitations, we:

– Refined the physical concepts of the existing model;

– Calibrated the model considering local state variables;
– Implemented the model in FEM as a material user-defined subroutine;
– Validated the model during plastic deformation under moderate and large strains.

Fig. 1. Measured and simulated volume fraction of α-phase as a function of temperature.

2
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

2. Methodology

2.1. Material

We investigated a disc obtained from a commercial billet of Ti-5553 alloy radially forged (cogged) in the α+β domain. The
microstructure of the as-received material consisted of globular primary α-grains and β-grains elongated in the cogging direction with 1
mm in length and ~150 μm in width. The software JMatPro® v.10 calculated the equilibrium volume fraction of the α-phase as a
function of the temperature, as shown in Fig. 1. We estimated the beta transus temperature Tβ=860 ◦ C using the measured chemical
composition listed in Table 1. After 15 min heat treatment at 920 ◦ C, the material exhibited equiaxed grains with ~160 µm diameter
and a high angle grain boundary fraction of ~0.93, whereas after 15 min heat treatment at 820 ◦ C, the material exhibited globular
primary α-grains and an intensively recovered β-matrix with subgrains of ~3 µm and a high angle grain boundary fraction of ~0.48.

2.2. Thermomechanical treatments

2.2.1. Hot compression tests

In the first set of experiments, we used the mobile Hydrawedge® conversion unit of the Gleeble® 3800 system to investigate the
thermomechanical behaviour of the material in the β-domain. Cylindrical samples 15 mm in height and 10 mm in diameter were
heated at 5 ◦ C/s under an argon atmosphere and then annealed for 15 min at the deformation temperatures of 880 ◦ C, 900 ◦ C and 920
◦
C. After being held at the test temperature, we deformed the specimens to a true strain of 0.85 at strain rates of 0.001 s-1, 0.01 s-1, 0.1 s-
1
, 1 s-1, and 10 s-1. The results of these flow curves and their corresponding microstructures were presented by (Buzolin et al., 2021a).

2.2.2. Hot torsion tests

To investigate the microstructure evolution of Ti-5553 in the β-domain up to large strains, we conducted hot torsion tests on a
Gleeble® 3800 device under an argon atmosphere using the specimen geometry SMT001 described by Semiatin et al. (Semiatin et al.,
2017), as shown in Fig. 2. We have chosen hot torsion tests to validate the model because it results in a wide strain distribution within
the deformed sample, which leads to a final microstructural gradient. These tests are typically applied to simulate procesess that result
in large gradients of strains and temperatures such as extrusion and solid-state joining (Gilmore and Liu, 2022; Lippold and Livingston,
2013).
Samples were machined with its longitudinal direction parallel to the cogging direction of the Ti-5553 ingot. We measured the
temperature using two S-type thermocouples: the control thermocouple TC1, located at the end of the specimen gauge, and the
thermocouple TC2, located 10 mm from the specimen shoulder (Fig. 2c). The heating rate was 5 ◦ C.s-1, and the dwell time before
deformation was 15 min. The manufactured samples underwent torsion with up to 8 revolutions at temperatures of 880 ◦ C and 920 ◦ C,
employing two distinct angular velocities: 6.92800 rad.s-1 and 0.06928 rad.s-1. These angular velocities correspond to strain rates of 1
s-1 and 0.01 s-1 at the edge of the gauge, respectively. After deformation, the in-situ water quenching system was activated for 30 s to
preserve the deformed microstructure for further microscopic analysis in regions P1, P2 and P3 (Fig. 2c).

2.3. Metallography

We cut the cross-section of the gauge of the tested samples, mounted them in resin, ground them with SiC paper and polished them
with colloidal silica suspension (OP-S solution). The samples were etched for approximately 2 min using a modified Kroll’s solution: 75
ml water, 15 ml HNO3 (69 vol.%), and 10 mL HF (40 vol.%). We performed scanning electron microscopy (SEM) using a TESCAN
Mira3 microscope with a Hikari detector and a TSL-OIM Data Collector software package for electron backscatter diffraction (EBSD)
analysis. We performed EBSD measurements of (500 µm)2 with a step size of 0.5 µm in the centre of the compression samples and three
different regions of the torsion samples (P1, P2 and P3), as schematically shown in Fig. 2c. We analyzed the measured data using OIM
Analysis™ v.8.6 software. After normalizing the confidence index of the measured points, we used a minimum confidence index value
of 0.5 to clean the data. Finally, we considered a minimum grain size of 2 µm, an average misorientation angle of newly formed LAGBs
θ0 = 2◦ , and the average transition misorientation angle between high and low-angle grain boundaries θc = 12◦

2.4. Mean-field hot deformation model

The mean-field model used in this work is based on (Buzolin et al., 2021a) and (Buzolin et al., 2021b). It describes the micro­
structure and stress evolution during hot deformation of titanium alloys in the single-phase (β) and dual-phase (α+β) domains. The
microstructural and state variables are represented by their one-dimensional averages, which means that there is no local information
at the crystal scale. The model structure consists of:

Table 1
Chemical composition of the investigated Ti-5553 alloy [wt.%].
Al V Mo Cr Fe O N C H Y Ti

5.62 5.03 4.93 2.99 0.35 0.117 0.0094 0.0075 – – balance

3
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 2. a) 3D Isometric view of the torsion sample; b) 2D cross-sectional view of the workpiece; c) 2D cross-sectional view of the sample gauge
highlighting the regions P1 (0.2 Radius), P2 (0.5 Radius) and P3 (0.85 Radius), and the positions of the welded thermocouples TC1 (for temperature
control) and TC2 (for measurement).

1) A set of internal variables used to describe the representative microstructure (Section 2.4.1).
2) A load partitioning model to describe the change in plastic strain partitioning between α and β phases during hot deformation
(Section 2.4.2).
3) Constitutive equations correlating the microstructure-based variables with the flow stress (Section 2.4.3)
4) Velocities that govern the kinetics of climb and glide of dislocations and the high-angle grain boundary motion (Section 2.4.4).
5) A set of differential equations that describe the kinetics of microstructure evolution (Section 2.4.5).
6) The representative distribution of misorientation angle (Section 2.4.5).
7) Auxiliary equations directly or indirectly dependent on temperature and strain rate that should be updated at every simulation step
(Table 4).
8) Fitting parameters (Table 5), which are highlighted in bold in the equations below.

In this paper, we propose some modifications to the model presented in (Buzolin et al., 2021a). These changes aim at improving the
following physical concepts:

I) The dynamic α-globularization and its relation to plastic strain partitioning (Section 2.4.2).
II) The glide velocity, which was previously based on a phenomenological Arrhenius-type equation and is now an output of the
Orowan’s equation (Section 2.4.4).
III) The role of dislocations in long-range stress, taking into account their respective strain fields (Section 2.4.3).
IV) The relationship between high-angle grain boundary velocity and the stored energy due to accumulation of dislocations in the
material (Section 2.4.4).
V) The axial deformation of prior β-grains, which was not previously considered (Section 2.4.5).
VI) The description of the misorientation angle distribution (Section 2.4.5).

2.4.1. Representative microstructure

The microstructure consists of three different populations of dislocations and two types of boundaries distinguished by their low
and high misorientation angles. The mobile dislocations have a density ρm and contribute to plastic deformation by sliding along the
active slip systems (Shanthraj and Zikry, 2011). The immobile dislocations at the cell interior have a density ρi and are static because of
their interaction with material defects such as other dislocations, grain boundaries and phase interfaces (Hansen et al., 2013).
Therefore, they do not contribute to plastic deformation (Ma et al., 2006). The immobile dislocations at the cell walls align to form the
low-angle grain boundaries (LAGBs) and have a density ρw (Castelluccio and McDowell, 2017). Finally, the grains are surrounded
solely by high-angle grain boundaries (HAGBs), and the subgrains are the representative microstructure entity. The initial values for
the subgrain (Φ0sg ) and grain (G0s ) sizes, and of the mobile and immobile dislocation densities (ρm0 and ρi0 ) are given in Table 3 in the
Appendix section 7.1.

4
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

2.4.2. Plastic strain partitioning

During deformation in the α+β domain, the difference in strength between α and β phases leads to different plastic strain rates of
both phases, as reported by (Fan et al., 2018) and (Ji et al., 2015). Furthermore, the morphology, distribution and amount of allotropic
phases play a major role in the plastic flow behaviour (Canelo-Yubero et al., 2018). (Buzolin et al., 2021a) considered a progressive
change from iso-power to iso-stress regime during deformation in the α+β domain, but they neglected the effect of the α-morphology
on the strain partitioning. In the present work, we attribute the change in load partitioning to the dynamic globularization of the
α-phase, as described in (Buzolin et al., 2021b). Initially, the α- phase acts as a reinforcement to the softer β-matrix and the material
flows with a unique strain rate ε̇ in the iso-strain regime. This means that the average strain rate ε̇ε in the material equals the strain rate
in the α-phase (ε̇εα ) and the strain rate in the β-phase (ε̇εβ ):

ε̇ε = ε̇εα = ε̇εβ (1)

Upon deformation, a fraction fglob of the α-phase globularises and this morphological change impacts the flow behaviour, changing
the load partitioning regime of this portion of the material from iso-strain to iso-stress regime. The strain rate of the α-phase in the iso-
stress regime is calculated according to (Souza et al., 2018):
( )
( ( )) Q
ε̇σα = exp nYSα ln sinh αYSα σβ − YSα + ln(AYSα ) (2)
RT
The strain rate of the β-phase is calculated using a simple law of mixture based on the volumetric fraction Fv of each phase:
Fv ε̇σα − ε̇
ε̇σβ = (3)
(1 − Fv )
fglob is physically described as the ratio of boundaries formed during globularisation (Svglob ) to the total amount of α-boundaries (Svα +
Svglob ):

Svglob
fglob = (4)
Svα + Svglob

The evolution of the strain is calculated in each phase using a simple rule of mixtures that is a function of fglob :
( )
ε̇x = 1 − fglob ε̇εx + fglob ε̇σx , x = α, β (5)

A comparison of the flow curves simulated using the iso-power → iso-stress and iso-strain → iso-stress load partitioning models is
shown in Fig. 12 in Section 7.4 of the Appendix. Overall, both predictions are similar and capture the general flow behavior of the
material, which exhibits a more pronounced flow softening at lower temperatures due to the higher fractions of α-phase undergoing
dynamic globularization. However, the physical description of the change in load partitioning is refined with the new proposed model.

2.4.3. Constitutive equations

During plastic deformation, the flow stress (σ ) consists of a sum of the short-range (σsr ) and long-range stresses (σ lr ) and is
calculated independently for each phase according to (Babu and Lindgren, 2013b):
σ = σ lr + σsr (6)
The long-range stress is related to the dislocation-dislocation interactions occurring in the material. (Buzolin et al., 2021a) and
(Buzolin et al., 2021b) assumed that the mobile dislocations contribute to the long-range stress (σlr ). In this work, we consider that they
travel a mean free path λ and only contribute to plastic deformation before their annihilation and/or immobilization, due to their
negligible strain-field compared to the other types of dislocation. Furthernore, we assume and that the immobile dislocations in the cell
interior and at the cell walls have independent contributions to the total stress because of the different dislocation spacing in these two
locations, as already reported by (Roters et al., 2000) and (Gourdet and Montheillet, 2003):
( √̅̅̅̅ √̅̅̅̅̅ )
σlr = MαGb f i ρi + f w ρw (7)

The model does not consider any effect of texture evolution, change in the Taylor factor M, or any variation in the Burgers vector b.
The shear modulus was determined by jMatPro software as a function of temperature and alloy chemical composition. The factors fi
and fw are the volume fractions of cell interior and cell walls, respectively, and α is a constant of the order of unity which depends on
the geometry and strength of the dislocation-dislocation interactions (Zerilli, 2004).
We neglected the Hall-Petch hardening due to the small Hall-Petch coefficient reported for titanium alloys at high temperatures
(Fan and Yang, 2011). The short-range stress is related to the effect of short-range obstacles, such as forest dislocations, point defects,
alloying elements and impurities, on the dislocation motion. The short-range interactions can be modelled using phenomenological
(Roy and Acharya, 2006) and statistical (Groma et al., 2003) approaches based on dislocation dynamics. However, attempting to
describe the role of short- and long-range dislocation interactions in a continuum model requires consideration of the fact that the
accumulation of short-range interactions can cause significant long-range effects. In this work, we have applied the common
simplification in materials science which considers σ sr as the difference between the yield stress (σYS ) and the initial long-range stress

5
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

(σ0lr ):

σsr = σ YS − σ 0lr (8)

The physical meaning of this assumption is that the short-range stress is constat for a given temperature and strain rate and any
change in the total stress during isothermal deformation is attributed to the long-range dislocation-dislocation interactions. The yield
stress is calculated based on a modified Arrhenius approach (Souza et al., 2018) and it is assumed to be a function of temperature and
strain rate:
⎧⎛ ( )⎞ 1 ⎡⎛ ( )⎞ 1 ⎤1/2 ⎫
⎪ nYS 2nYS ⎪
⎪ ε̇exp QRTYS
⎪ ε̇exp QRTYS ⎪
⎪
1 ⎨ ⎜ ⎟ ⎢⎜ ⎟ ⎥ ⎬
σYS = ln ⎝⎜ ⎟ ⎢⎜
+ ⎣⎝ ⎟ + 1⎦
⎥ (9)
αYS ⎪⎪ AYS ⎠ AYS ⎠ ⎪
⎪
⎪
⎩ ⎪
⎭

2.4.4. Velocities
The dislocation glide, dislocation climb and HAGB movement govern the kinetics of microstructure evolution during plastic
deformation at high temperatures. (Buzolin et al., 2021a) and (Buzolin et al., 2021b) assumed that the glide velocity follows a
phenomenological relationship which was a function of temperature and short-range stress:
( )
− Wg Ωσsr
vg = a1 exp (10)
kB T kB T
a1 parameter was recalculated and fitted at each step of simulation to adjust the glide velocity and obey the Orowan’s relationship
(Orowan, 1940). In the present work, the glide velocity is a model output that follows the well-established Orowan’s relationship and
no phenomenological equations are used:
Mε̇
vg = (11)
bρm
The climbing velocity due to dipole annihilation for mobile and immobile dislocations is calculated according to (Ghoniem et al.,
1990):
2πηDs (Gb/2π(1 − ν)λ)Ω
vc = ( ( )) (12)
kB bT 1 − ηln 6π(1+ ν)GΩ
(1− ν)kB Tλ

λ is the dislocation interspacing, Ds is the self-diffusion coefficient, ν is the Poisson coefficient, Ω is the atomic volume, and η is a
fitting parameter related to the transfer of defects into jogs on the dislocation. Finally, the HAGB motion is a function of the HAGB
( )nv
mobility MHAGB , the dislocations accumulated in the material and a short-range contribution ε̇ref ε̇
directly proportional to the

applied strain rate and related to the mechanical response of the HAGB:
( )
( ) ε̇ nv
vHAGB = MHAGB Gb2 ρm + ρi + fw fLAGB ρw (13)
ε̇ref
The wall dislocation density is considered to have a lower contribution to the total stored energy due to their smaller strain-field
size at the cell walls than the immobile and mobile dislocations create at the cell interior (R.H. Buzolin et al., 2020).

2.4.5. Differential rate equations

The rate equations related to the microstructure evolution describe the change in the mobile (ρm ) and immobile (ρi ) dislocation
densities, the boundary density of β (Svβ ) and α ( Svα and Svglob ) phases, the three-dimensional deformation of prior β-grains (li ) in x, y and
z directions and the average misorientation angle of the deformed material (θx ). The differential equations are solved in each simu­
lation step using the Dormand-Prince (RKDP) numerical method (Dormand and Prince, 1986).
[ ]+
(Buzolin et al., 2021a) considered that the subgrains act as a source of mobile dislocations that are produced in a rate ddtρm =
δsg ρm Φsg vg
λm 2
. In this work, we assume that the mobile dislocations immobilize within the subgrains at a much higher rate than the pro­
duction of new mobile dislocations at the subgrain boundaries. Therefore, the Frank-Read is considered the only source of mobile
dislocations and the mobile dislocation density rate reads as follows:
dρm ρm vg ρ vc
m vg −
= ρ3/2 − 8 m − δDRV ρm (ρm + ρi )vg − ρm SvHAGB vHAGB − f int
CDRX Svα/β ρm vg (14)
dt δsg λ

δsg is a critical distance which value is between the dislocation interspace and the subgrain diameter Φsg . − 8 ρmλvc , − δDRV ρm (ρm +ρi )vg
and − ρm SvHAGB vHAGB correspond to mobile dislocation annihilation rates through climb, glide and HAGB motion, respectively.

6
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Additionally, − f int
CDRX Svα/β ρm vg corresponds to the annihilation rate of mobile dislocations at α/β interfaces, which was missing in
(Buzolin et al., 2021a).
Similarly to the mobile dislocations, the annihilation of immobilte dislocations via climb, glide and HAGB motion counteracts the
[ ]+
ρ v
immobilization rate ddtρi = Φmsgg :

dρi ρm vg ρ vc
= − 8 i − δDRV ρm ρi vg − ρi SvHAGB vHAGB (15)
dt Φsg λ

A portion fCDRX of the immobile and mobile dislocations annihilated via DRV and SRV and the dislocations annihilated at the α/β
interfaces will contribute either to the formation of new LAGBs or to the increase in misorientation angle of existing boundaries:
( )
dρCDRX vc (ρi + ρm )
= fCDRX 8 + δDRV ρm (ρm + 2ρi )vg + f int
CDRX Svα/β ρm vg (16)
dt λ
The boundary density corresponds to the total area of boundaries per unit volume, including HAGBs and LAGBs. The current work
[ ]+
dS
considers that a fraction αCDRX of the recovered dislocations will form new LAGBs in the α and β phases at a rate dtSv = αCDRX nθb0 dρCDRX
dt
[ ]−
dSvβ
and β-boundaries are swepted by HAGB motion) at a rate dt = αCDRX − Svβ SvHAGBβ vHAGBβ :

dSvβ b dρCDRXβ
= αCDRX − Svβ SvHAGBβ vHAGBβ (17)
dt nθ0 dt
α/β interfaces with density Svglob are formed within the non-globularized α-platelets of density Svα during dynamic globularization:
dSvglob fBG Svα vHAGBα
= (18)
dt tα
The model considers that the formed α/β interfaces instantaneously migrate, decreasing the boundary density Svα of the non-
globularized α-platelets:

dSvα b dρCDRXα dSvglob

= αCDRX − (19)
dt nθ0 dt dt
The rest (1 − αCDRX ) of the recovered dislocations will increase the misorientation of part of the already existing boundaries at the
following rate:

dθx b dρcDRX
= (1 − αCDRX ) , x = Rayleigh or Exponential (20)
dt n dt
θx corresponds to the average misorientation angle of a distribution related to the fraction of the material that increases its
misorientation. (Buzolin et al., 2021a) and (Buzolin et al., 2021b) considered that the total misorientation angle distribution is a sum of
a Rayleigh and a Mackenzie distribution. The Rayleigh distribution ΘR (θ) accounted for a material in its deformed state with an
average misorientation angle θR :
( )
πθ π θ2
ΘR (θ) = 2
exp − 2
(21)
2θR 4θR
The Mackenzie distribution (Mason and Schuh, 2009) represented a recrystallised microstructure with a random texture. Their
overall misorientation angle distribution is given by:
( )
Θ(θ) = ACDRX fR ΘR (θ) + 1 − ACDRX fR ΘM (θ) (22)

ACDRX refers to the ratio of the new formed boundaries to the total amount of boundaries. Since the Mackenzie distribution only
( √̅̅̅)
exists in a misorientation angle range of [0, 2 tan− 1 2 − 2 ] for a cubic crystal structure, only the fraction fx of the Θx (θ) distribution
belonging to this range is considered for the calculation of the total misorientation angle distribution. Thus, fR is the fraction of the
( √̅̅̅)
Rayleigh distribution within the [0, 2 tan− 1 2 − 2 ] range.
The Rayleigh distribution considers that the LAGBs undergo a homogeneous increase in misorientation angle. However, it does not
represent the evolution of boundary misorientation angle observed in the material. Therefore, we propose in this paper to represent the
deformed material in its deformed state with an exponential distribution ΘE (θ):
√̅̅̅̅̅̅̅̅ ( √̅̅̅̅̅̅̅̅)
π/2 θ π/2
ΘE (θ) = exp − (23)
θE θE

The overall misorientation angle distribution is given by:

7
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

( )
Θ(θ) = fLAGB fE ΘE (θ) + 1 − fLAGB fE ΘM (θ) (24)

The physical meaning of this change is that the formation of LAGBs is a gradual process that takes place continuously in two
simultaneous stages:

– A progressive increase in the misorientation angle of existing LAGBs;

– A continuous formation of new LAGBs, which is not considered when the Rayleigh distribution is used to represent the material in
its deformed state.

The model presented here also describes the three-dimensional evolution of the prior β-grain size due to triaxial deformation in x, y
and z directions, as done by (Buzolin et al., 2022) for AA6082 aluminium alloy:

dli ( )(l − Φ ) (l − Φ ) (l − Φ )
x sg y sg z sg
= fE 1 − fHAGBCDRX εii li ε̇i = x, y, z (25)
dt lx ly lz

εii refers to the strain tensor and li refers to the ellipsoid axes in x, y and z directions, while fHAGBCDRX corresponds to the fraction of
high-angle grain boundaries that were formed during CDRX.

2.5. Model calibration and validation

We used the EBSD measurements as described in Section 2.3 to quantify some microstructural features of Ti-5553 samples after
torsion and compression at high temperatures. The equations of these microstructural features are given in Table 2. Based on the
stereological relationship from Eq. 26 that considers a perfect spherical substructure, the subgrain size (ϕSG ) is inversely proportional
to the boundary density. The surface fraction of LAGBs (Eq. 27) is the corresponding area below the overall misorientation angle
distribution between [θ0 -θC ]. θ0 is average misorientation angle of the newly formed LAGBs and θc is a transition misorientation angle
from LAGBs to HAGBs. The grain size (Gs ) is determined as the ratio between the subgrain size and the fraction of high angle grain
boundaries, as shown in Eq. 29, while the average misorientation angle of the LAGBs reads (θLAGB ) is obtained from EBSD measure­
ments according to Eq. 30. Finally, the wall dislocation density is calculated as a function of Sv , θLAGB and fLAGB according to the work
of (Gourdet and Montheillet, 2003), as shown in Eq. 31, where b is the Burgers vector and nρ is the mean number of dislocation sets in
the cell walls.
We used the microstructural features from torsion and compression samples deformed at strain rates less than 0.1 s-1 to calibrate the
β-microstructure evolution because the local temperature in the region of interest remained approximately constant in these cases.
Samples deformed at strain rates equal to or greater than 1 s-1 were used for validation of the material model as a simulation tool to
predict non-isothermal processes, as they experienced an increase in temperature due to adiabatic heating. Due to friction and/or
temperature gradient, local state variables εeff and ε̇eff in the region of interest are different from the global deformation parameters.
For the compression samples, we estimated the local state variables using the finite element method (FEM) according to Section 2.6.
For the torsion samples, we estimated the final effective strain along the gauge radius (R), with Eq. (33) where N is the number of
revolutions and L is the gauge length:
2π NR
εeff = √̅̅̅ (33)
L 3
After estimating the local state variables, we identified the fitting parameters from Table 6 using the deterministic Nelder-Mead
simplex search method. For that, we employed the experimental and simulated microstructural features calculated using the

Table 2
Auxiliary equations used to quantify microstructural
features from EBSD measurements.
Microstructural features Eq.

2 26
ϕSG =
Sv
∫θC 27
fLAGB = Θ(θ)dθ
θ0
fHAGB = 1 − fLAGB 28
ϕ 29
Gs = SG
fHAGB
∫θC 30
1
θLAGB = Θ(θ)θdθ
fLAGB
θ0

nρ Sv θLAGB fLAGB 31
ρw =
b
ρw b 32
= Sv θLAGB fLAGB
nρ

8
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

equations from Table 2 to compute their root mean square error (RMSEs). As the previous publication considered nρ = 2 and for this
proposed model we have identified a different value of nρ = 1.11, we have compared the results in this paper using a wall dislocation
density normalized by the Burgers vector (b) and the average number of dislocations to form a wall (nρ ), as shown in Eq. 32. The
average RMSE was established as the objective function to be minimised during parameter identification.

2.6. FE simulations

The mean-field hot deformation model described in Section 2.4 was implemented in the commercial FEM-based software
DEFORM™ 2D as a material user-defined subroutine to perform simulations in the prefect-plastic regime. We used the deformation
tensor and temperature numerically determined at each Gaussian point by the FEM engine as input to the material model subroutine,
while the elemental values of stresses and their numerical partial derivatives to the strain and the strain rate obtained by the material
model were used to calculate the von Mises stresses at each Gaussian point with the help of the FEM engine. Using the following setup,
we simulated isothermal hot compression tests, and the ohmic heat treatment followed by hot torsion experiments of Ti-5553.

2.6.1. Isothermal compression simulations

In the pre-processor of the FEM software, we created the Ti-5553 standard axisymmetric cylindrical compression sample as a
perfect-plastic object with 1000 quadrilateral elements with 4 nodes each. The compression anvils made of a tungsten carbide-based
material were created as rigid objects with 1000 quadrilateral elements with 4 nodes each. Then, we simulated an isothermal
compression test at 900 ◦ C and 0.001s-1 until a true strain of 0.85 with various friction coefficients between the workpiece and the
anvils, to determine the value that better aligns with the barrelling observed in the experimental final shape.

2.6.2. Torsion simulations

In the pre-processor of the FEM software, we created the Ti-5553 axisymetric torsion workpiece (Fig. 2c) as a perfect-plastic object
with 3000 quadrilateral elements with 4 nodes each and the torsion grips of AISI-316 L stainless steel as rigid objects with 1000
quadrilateral elements homogeneously distributed.
During the Gleeble experiments, we measured the power angle over time during heating, holding and deformation. We assumed
that the current density is directly proportional to the power angle to fit the evolution of the current density using Eq. (34), Eq. (35) and
Eq. (36) and to minimise the difference between measured and simulated temperatures at the thermocouples TC1 and TC2.
{
0.114 + 4.64x10− 4 t, if TC1 = 920∘ C
Heating i(t) = (34)
0.114 + 4.28x10− 4 t, if TC1 = 880∘ C
{
0.170, if TC1 = 920∘ C
Holding i(t) = (35)
0.161, if TC1 = 880∘ C
⎧
⎪
⎪ 0.170, if TC1 = 920∘ C and ε̇ = 0.01 s− 1
⎨
Deformation i(t) = 0.161, if TC1 = 880∘ C and ε̇ = 0.01 s− 1
(36)
⎪
⎪
⎩
0.170 − 0.0175t, if TC1 = 920∘ C and ε̇ = 1 s− 1

We assumed that during deformation the interface between the workpiece and the grips had an average coefficient of friction of 0.3
and a resistivity of 0.045 Ω.mm2. The ambient temperature was 20 ◦ C, and the heat transfer coefficient between the workpiece/grips
and the environment was assumed to be 0.02 W.m-2.K-1. We obtained the elastic and thermal properties of the Ti-5553 workpiece from
the JMatPro® v.10 software and the steel AISI-316 L grips from the DEFORM™ 2D materials library and the literature (Pichler et al.,
2022; P. 2020).

Fig. 3. Comparison between the gauge deformed at 920 ◦ C at: a) 0.01 s-1; b) and 1 s-1. RD is the radial direction of the hot torsion sample, and LD is
the longitudinal direction.

9
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

3. Results

Firstly, this section demonstrates that hot torsion tests conducted in the Gleeble thermomechanical simulator result in different
macro and microstructure gradients along the torsional specimen depending on the deformation rate (3.1). Subsequently, we estimate
the actual effective strain and effective strain rate (3.2) as well as the temperature gradient (3.3). Finally, we calibrate the mean-field
model at low strain rates with an approach that takes into account the local deformation history of the analysed regions (3.4).

3.1. Macro and microstructure gradient obtained after hot torsion

Fig. 3 shows the macroscopic view of the gauges after hot torsion testing at 920 ◦ C with different rotational speeds. While the
sample deformed at 0.01 s-1 underwent only torsional deformation, the sample deformed at 1 s-1 shows barreling and three types of
microstructures (equiaxed non-deformed grains in region I, elongated grains in region II, and recrystallized refined grains in region III),
indicating that the sample was also compressed in the longitudinal direction during torsion.

Fig. 4. Inverse pole figure (IPF) maps after hot torsion at 920 ◦ C and 880 ◦ C for the overall strain rates of 0.01 s-1 and 1 s-1. The regions P1, P2 and
P3, indicate 0.2R, 0.5R and 0.85R, respectively. Black lines indicate HAGBs and white lines indicate LAGBs. RD is the radial direction of the hot
torsion sample, and LD is the longitudinal direction.

10
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 4 shows the Inverse Pole Figures (IPF) maps measured after 8 revolutions as a function of the position along the gauge and the
tested condition. We observed an equiaxed-like refined substructure in all regions. A significantly finer substructure forms at higher
strain rates and/or lower temperatures. The microstructure contains LAGBs highlighted by the white lines and spreading misorien­
tation within the deformed grains, indicating partial recrystallization in all the investigated conditions. Additionally, Fig. 4g shows a
prior β-grains elongated in the radial direction of the torsion sample, which matches with the macroscopic view in region II (Fig. 3b).

3.2. Actual effective strain and effective strain rate within samples

Section 7.5 in the Appendix compares experimental results with simulation predictions for the evolution of β-microstructural
features during hot deformation. The experimental results were obtained from EBSD maps measured at the centre of hot compressed
samples and at the edge of the gauge of torsion samples (region P3 from Fig. 2) deformed at 920 ◦ C at the nominal strain rate of 0.01 s-1.
The β-microstructural features were simulated using the model presented in Section 2.4 and the model presented in (Buzolin et al.,
2021a) and they takes into account the nominal strain rate and temperature (0.01 s-1 and 920 ◦ C). Thus, these simulations neglected
the existence of temperature and strain gradients within the workpieces. Despite of slightly better results from the current presented
model, both models fail to predict the microstructures at moderate and large strains. This happens because the global strain, tem­
perature and strain rate do not correspond to the local values. Therefore, this simulation approach ignores the effects of sample ge­
ometry, barreling and temperature gradient across the sample on the local deformation.
To improve the predictability of the β-microstructure evolution, we estimated the deformed samples’ local effective strain and
strain rate using FE simulations according to the methodology explained in Section 2.6. For the uniaxial compression tests, we esti­
mated the effective strain distribution along the deformed sample radius of the workpiece of an isothermal hot compression test at 900
◦
C and 0.001 s-1. We adjusted the shear friction coefficient to match the simulated and experimental geometries (Fig. 5a). Fig. 5b shows
the final effective strain distributions along the radial direction "x" of the compression sample for different shear friction coefficients.
The shear friction coefficient that better adjusted the final shape of the compression sample was f = 0.4. Therefore, the actual final
effective strain in the centre of the sample is 1.23, approximately 45 % higher than the global strain of 0.85.
Fig. 5c shows the final effective strain distribution along the radial direction of the torsion samples determined using Eq. (33). The
simulated effective strain distribution of the torsion sample is linear over the gauge radius. The average effective strain rate in the

Fig. 5. a) Barreling of a sample compressed at 900 ◦ C and 0.001 s-1; (b) final effective strain distribution in the horizontal direction x of sample in
Fig. 6a simulated in FEM with different shear friction coefficients (FC); c) calculated final effective strain distribution in the radial direction of the
torsion sample deformed up to 8 revolutions.

11
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

regions of interest of both compression and torsion tests was calculated as the ratio of the final effective strain to the total deformation
time.

3.3. Estimation of temperature gradient in the torsion samples

The thermomechanical system developed during a torsion test in a Gleeble device involves ohmic heating, heat conduction within
the sample, heat exchange between the contact parts and the sample, and heat exchange between the sample and the environment. A
homogeneous temperature distribution can be assumed during slow compression if the measured temperature gradient in the lon­
gitudinal direction of the cylindrical sample is less than 10 ◦ C. On the other hand, the temperature gradient measured in torsion tests is
significantly high due to the long sample. Therefore, we simulated the temperature evolution during torsion in the region of interest
with FE simulations to account for all these heat exchanges within the system.
Fig. 6 shows a comparison between the measured and simulated temperatures at thermocouples TC1 and TC2 during heating and
holding at 920 ◦ C (a) and 880 ◦ C (c) followed by torsion at 0.01 s-1 at the same temperatures (b,d). The simulated temperatures at P1,
P2 and P3 are also shown. The simulations reproduce for both cases the higher temperatures at TC1 than TC2 during heat treatment (a,
c) and the constant temperature during deformation (b,d). Additionally, the simulated temperatures in the region of interest (P1, P2
and P3) are about 40–50 ◦ C higher than the temperature measured at the control thermocouple (TC1), similar to the ~30 ◦ C gradient
measured by (Hogrefe, 2019) in Ti-6Al-4V torsion samples. This difference in the temperature is one of the reasons why the previous
model and parameters (Buzolin et al., 2021a) fails to predict the microstructure of torsion samples.
The difference between the experimental and simulated temperatures could be mainly related to:

Fig. 6. Temperature evolution at the controlling thermocouple TC1, measuring thermocouple TC2, P1, P2 and P3 during heat treatment (a,c) and
deformation (b,d) at 0.01 s-1 for the samples tested up to 8 revolutions at: (a,b) 920 ◦ C and (c,d) 880 ◦ C.

12
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

• Wrong temperatures measured by the thermocouples spot welded to the surface. (Semiatin et al., 2017) investigated the radial
temperature gradient in the Gleeble hot torsion test and its effect on the interpretation of the plastic flow. They used microstructural
observations in a nickel-base superalloy and an analytical heat-conduction model to infer the temperature difference between the
temperature in the outer diameter of the sample and that indicated by K-type thermocouples welded to the sample surface. They
found that there was a 20 ◦ C difference between the real surface temperature and the temperature measured by the thermocouple.
• Wrongly assumed boundary conditions such as the of average friction coefficient (0.3) and the electrical resistivity (0.045 Ω.mm2)
at the interface between workpiece and grips, and the heat transfer coefficient between the workpiece/grips and the environment
(0.02 W.m-2.K-1), which were considered constant for every temperature and were not experimentally determined.
• Wrong assumption that the current density passing through the sample is directly proportional to the power angle and that the
power angle behaved linearly during the experiments.

3.4. Model calibration considering the local deformation history at low strain rates

As explained in Sections 3.2 and 3.3, the local strain and temperatures experienced by the regions of interest differ from those
assumed in the previous publication, leading to incorrect microstructure predictions at large strains, as shown in Section 3.2. Fig. 7
shows microstructural predictions considering the local and global deformation parameters in the center of the workpiece during hot
compression at 920 ◦ C and 0.01 s-1. There is no major difference in the subgrain size predictions. At the same time, the evolution of the
high-angle grain boundary fraction is more accurate when the deformation history is considered, together with the newly modified
model equations. As a result, we could improve the predictions of the grain size and the normalized wall dislocation density. It is
worthy to mention that due to statistical reasons, the fraction of high-angle grain boundaries and subgrain size are more accurately
determined through EBSD measurements than the grain size. Fig. 8 shows the microstructure after 15 min annealing at 920 ◦ C (a) and
after the same heat treatment followed by isothermal deformation at 0.01 s-1 (b). The red squares highlight a representative area of the
EBSD measurements. Both microstructures show that there are not enough grains in the EBSD measurement area that can statistically
represent the average grain size of the sample.
Fig. 9 compares the microstructure predictions during torsion tests at 0.01 s-1 for the three regions of interest (P1, P2 and P3),

Fig. 7. Comparison between measured and simulated β-microstructural features during compression at a strain rate of 0.01 s-1 and temperature of
920 ◦ C with (proposed model) and without (Buzolin et al., 2021a) the deformation history: a) fraction of high angle grain boundaries; b) subgrain
size; c) grain size; d) normalized wall dislocation density.

13
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 8. Optical micrographs in the center of the compression sample a) after 15 min heat treatment at 920 ◦ C and b) after 15 min heat treatment at
920 ◦ C followed by deformation to ε = 0.85 at 0.01 s-1 at the same temperature. The red squares show a representative area of the EBSD
measurements.

considering the local thermomechanical history with additional experimental data. The arrows highlight a clear tendency observed at
both tested temperatures. A continuous decrease in grain and subgrain sizes occurs during deformation. The fraction of high-angle
grain boundaries decreases to a minimal value, which is followed by a continuous increase in the amount of HAGBs due to CDRX.
Finally, the wall dislocations are rapidly produced and a continuos annihilation of those dislocations occur after a peak is achieved. For
both tested temperatures, the final microstructure in P3 achieved smaller grain and subgrain sizes and a higher fraction of high-angle
grain boundaries than P1 (close to the center of the gauge). The different average strain rates experienced in regions P1 (~0.002 s-1),
P2 (~0.006 s-1) and P3 (~0.01 s-1) led to different kinetics of microstructure evolution. The calibrated model captured the material
behaviour for the different strain rates along the gauge.

4. Discussion

Within this section, we want to discuss two different topics. First, how the adiabatic heat generated by plastic deformation affects
the microstructure (Section 4.1). Second, how the misorientation angle distribution function influences the predictions of the
deformed state (Section 4.2).

4.1. Microstructure validation at high strain rates with adiabatic heating

The developed mean-field model integrated into FEM software can be used as a simulation tool to determine the microstructure
developed under non-isothermal conditions. To demonstrate the performance of the calibrated model under these conditions, we
simulated the microstructure evolution at P1 and P3 during torsion at 1s-1 with the control thermocouple set at 920 ◦ C using the FEM.
Fig. 10 compares two simulation conditions, ideal isothermal and non-isothermal with a temperature increase of about 65 ◦ C in the
centre of the gauge (Fig. 10a) due to adiabatic heating. This temperature increase caused a flow softening (Fig. 10b) and accelerated
the deformation in the specimen gauge (Fig. 10c). As a result, the temperature increase leads to a higher final effective strain, as shown
in Fig. 10d. The flow softening due to adiabatic heating is the main reason why the sample deformed at 920 ◦ C, and 1s-1 underwent
compression and barreling during the torsion test (Fig. 3). Furthermore, when considering the non-isothermal condition, the model
predictability of the subgrain size and fraction of high-angle boundaries improves (Fig. 10e-f). However, the simulated fraction of high-
angle boundaries at P3 is far from the experimental results. The Appendix section 7.5 discusses the uncertainties in the modelling and
experimental results.

4.2. Validation of the misorientation angle distribution

The model can also describe the evolution of the misorientation angle distribution, which can be correlated with the degree of
recrystallization and is a necessary step to determine all the microstructural features previously discussed in this paper (ϕSG , GS , fHAGB
and ρw ). Before deformation, a typical Mackenzie distribution represents the fully recrystallized microstructure (Mason and Schuh,
2009). Our previous work considered that the misorientation angle distribution of a partially recrystallized microstructure is a
weighted sum of the Rayleigh and Mackenzie distributions (Buzolin et al., 2021a). The Rayleigh distribution represented the material
in its deformed state, as proposed by (Gourdet and Montheillet, 2003).
Fig. 11a-c compares the misorientation distributions obtained from EBSD measurements after torsion at 920 ◦ C and 1 s-1 in P1, P2
and P3 with those modelled using the Rayleigh distribution to represent the deformed state. The model assumes a continuous increase
in the average misorientation angle, resulting in a change in the shape of the distribution. The regions of higher effective strain (εP1
< εP2 < εP3 ) exhibit higher misorientation angles, as indicated by the red arrows. However, this model does not capture measured
distributions, especially at high misorientation angles. This happens because the Rayleigh distribution considers that the LAGBs

14
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 9. Measured and FEM-simulated β-microstructural features tracked during torsion at a strain rate of 0.01 s-1 and temperatures of 880 ◦ C (a,c,e)
and 920 ◦ C (b,d,f): a,-b) subgrain size; c-d) fraction of high angle grain boundaries; e-f) grain size; g-h) normalized wall dislocation density.

15
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 10. Influence of the adiabatic heat on the evolution of the: a) Temperature; b) Effective stress; c) Effective strain rate; d) Effective strain; e)
Subgrain size; f) Fraction of high angle grain boundaries. The tracking points P1 and P3 are related to regions 1 mm from the centre and 4.25 mm
from the centre of the deformed gauge, respectively (Fig. 2).

undergo a homogeneous increase in misorientation angle, while the experimental results show a bimodal misorientation angle dis­
tribution with the constant formation of new LAGBs of low misorientation angle and the continuous increase in misorientation angle of
the existing boundaries.
In contrast, the use of the exponential distribution is an attempt to predict a constant formation of new LAGBs of low angles, as

16
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 11. Comparison between experimental and simulated misorientation angle distributions in the tracking points P1 (a,d), P2 (b,e) and P3 (c,f)
after 8 revolutions at 1 s-1 and 920 ◦ C: a-c) Rayleigh distribution used to describe the deformed state; D-f) Exponential distribution used to describe
the deformed state The black dashed vertical lines correspond to the transition misorientation angle from LAGBs to HAGBs (θc = 12◦ ).

observed in experimental results. Fig. 11d-f compares the same experimental results with a second modeling approach that uses the
exponential distribution to describe the material in its deformed state. In this case, the overall misorientation follows Eq. 20** and a
more accurate prediction of the misorientation angle distribution is achieved. The red arrows indicate that at higher strains, such as at
P3, the shape of a Mackenzie distribution becomes more pronounced. The physical meaning of this improvement is that the use of the
exponential distribution considers simulataneously a continuous formation of new LAGBs of low misorientation (first peak of the
distributions) and a progressive increase in the misorientation angle of existing LAGBs, which results in a bimodal misorientation angle
distribution, as observed in the experimental results.
Fig. 12a shows a comparison between the experimental and simulated average misorientation angles (θ) from the distributions
shown in Fig. 11. A clear increase in θ along the radial direction of the sample gauge is observed in both experimental and simulated
data. Moreover, the average misorientation angles θ modelled using the exponential distribution are much closer to the experimental
values than those calculated using the Rayleigh distribution, as a result of the better prediction of the misorientation angle distribution.
The measured and simulated average misorientation angles of the LAGBs (θLAGB ), which is conventionally assumed to be in the
range of 2◦ to 12◦ , are shown in Fig. 12b for the same studied conditions. An accurate prediction of this variable is highly desirable
since it is used to calculate the wall dislocation density according to Eq. 31. The results show that the use of the exponential distribution
also improves the predictability of the average misorientation angle in the LAGB domain. However, while θ is underestimated by the
model, θLAGB is overestimated by the model regardless of the distribution chosen to represent the material in its deformed state.

5. Summary and conclusions

A mean-field model based on dislocation reactions was developed to describe the continuous dynamic recrystallization process. The
calibrated model was implemented into the commercial FEM-based software DEFORM™ 2D as a user-defined elemental subroutine to
investigate the microstructure evolution of a near-β Ti-5553 alloy during deformation up to moderate and large strains. Although there
are deviations in the microstructure predictions, this computational strategy delivers a reasonable accuracy for a relatively low
computational effort, enabling the simulation of complex thermomechanical processes in industrial scale.
The simulation and experimental results of the investigated material lead to the following conclusions:

• The newly proposed modifications to the hot deformation model (such as the misorientation distribution), together with the novel
calibration method provide a more realistic simulation approach that considers the local thermomechanical history and improves
the model predictability of microstructural features at lower strain rates.

17
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 12. Comparison between experimental and simulated average misorientation angles at P1, P2 and P3 after 8 revolutions at 1s-1 and 920 ◦ C: a)
from the distributions shown in Fig. 11 in the range from 5◦ to 65◦ ; b) from the LAGB domain (from 2◦ to 12◦ ). Rayleigh and Exponential stands for
the distribution chosen to represent the material in its deformed state.

• The adiabatic heat generated by the plastic deformation at high strain rates increases the local effective strain within the gauge of
the torsional specimen due to the flow softening in this region. Apart from barreling, the subgrain structure is refined and the
evolution of the fraction of HAGBs is faster.
• The new proposed description of the misorientation angle distribution to describe the deformed material improves the predict­
ability of the bimodal behavior of these distributions observed during hot torsion and their respective mean misorientation angles.

CRediT authorship contribution statement

Franz Miller Branco Ferraz: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Method­
ology, Investigation, Formal analysis, Data curation, Conceptualization. Ricardo Henrique Buzolin: Writing – review & editing,
Writing – original draft, Visualization, Validation, Supervision, Software, Methodology, Investigation, Formal analysis, Data curation,
Conceptualization. Stefan Ebenbauer: Writing – review & editing, Validation, Resources, Project administration, Methodology,
Funding acquisition, Formal analysis, Conceptualization. Thomas Leitner: Writing – review & editing, Writing – original draft,
Visualization, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Conceptualization.
Alfred Krumphals: Writing – review & editing, Supervision, Resources, Project administration, Funding acquisition, Formal analysis,
Conceptualization. Maria Cecilia Poletti: Writing – review & editing, Writing – original draft, Supervision, Resources, Project
administration, Methodology, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Data availability

Data will be made available on request.

Acknowledgments

The authors have carried out this work as part the following projects: FWF Project F37729 from FWF Der Wissenschafsfonds and D-
1303000107/CD-Laboratory for Design of High-Performance Alloys by Thermomechanical Processing of the Christian Doppler For­
schungsgesellschaft. We would also like to thank Dr. Katharina Hogrefe from Graz University of Technology for carrying out the torsion
thermomechanical experiments.

Appendix

Initial microstructure

The initial grain and subgrain sizes were obtained by the EBSD measurements of the microstructure after annealing for 15 min at
920 ◦ C followed by water quenching. The initial dislocations densities ρ0i = 109 m− 2 and ρ0m = 1012 m− 2 were considered to represent a
recrystallized microstructure.

18
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Table 3
Initial microstructure used for all the simulations.

Microstructural feature Value determined from EBSDs

ϕ0SG 150 µm
0
fHAGB 0.93
G0S ~160 µm

Auxiliary equations

Table 4
Auxiliary equations used in the model.

Auxiliary variables Eq. Source

( )
btHAGB QHAGB 37 Addapted from (Humphreys and Hatherly, 2004)
MHAGB = M0 exp −
k T RT
( B )
Qdiff 38 (R.H. Buzolin et al., 2020)
Ds = D0 exp −
RT
1 39 (Ghoniem et al., 1990)
λ = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ρm + ρi
√̅̅
2tan− 1 (2− 2)
∫ 40
fx = Θx (θ)dθ, x = R or E
0
Sv − S0vHAGB 41
ACDRX =
Sv
24 42
(1 − cos(θ)), 0∘ < θ ≤ 45∘
π
[ ( ) (θ) ]
24 √̅̅̅
ΘM (θ) = { (1 − cos(θ)) 3 2 − 1 cot − 2 , 45∘ < θ ≤ 60∘
π 2
{[ ( ] ( ) }
24 √̅̅̅ ) 4 θ (Buzolin et al., 2021b)
(1 − cos(θ)) 3 2 − 1 + √̅̅̅ cot − 6 , 60∘ < θ ≤ 60.7∘
π 3 2
α-dynamic globularization
dtα 43
= Bglob fglob vcβ
dt
dlα dtα 44
= − AR2 tα3
dt dt
2lα 2 45
ϕα =
Svglob l2α + 2lα
ϕ 46
ARα = α ≥ 1
tα

Model parameters

Table 5
Parameters with a range established in the literature.

Parameters Unit Description α-phase β-phase Source

θ0 ◦
Average misorientation angle of the newly formed LAGBs 0.2 2 (Gourdet and Montheillet, 2003)
θc ◦
Transition average misorientation angle from LAGB to HAGB range 12 12
ε̇ref s− 1 Reference strain rate 106 106
Qdiff J.mol− 1 Diffusion exponent 250,000 303,000
M – Taylor factor 5 3.05 (Shen et al., 2013)
D0 m2 .s− 1 Diffusion coefficient 1.4 x 10–4 1.0 x 10–4 (Köhler and Herzig, 1987)

Table 6
Comparison between the fitting parameters calibrated for the developed model and the fitting parameters used in (Buzolin et al., 2021a).

Parameters Unit Description α-phase β-phase α-phase (Buzolin β-phase (Buzolin

et al., 2021a) et al., 2021a)

α – Taylor constant 0.25 0.25 0.1 0.1

(continued on next page)

19
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Table 6 (continued )
Parameters Unit Description α-phase β-phase α-phase (Buzolin β-phase (Buzolin
et al., 2021a) et al., 2021a)

αCDRX – Fraction of the dynamic recovery that contributes to CDRX 0.40 0.30 0.45 0.4
x
fCDRX – Fraction of mobile dislocations that contribute to CDRX 1.0 x 10–3 8.34 x 1 x 10–3 1 x 10–3
10–4
nHAGB – Exponent related to the stored energy to move HAGBs 0.65 0.2 – 0.4
δDRV m Critical distance for dynamic recovery 3.33 x 3.63 x 1.75 x 10–8 2.15 x 10–8
10–7 10–7
QHAGB J Activation energy for the mobility of HAGBs during CDRX 3.32 x 1.92 x 3.79 x 10–19 3.79 x 10–19
10–19 10–19
M0 – Pre-factor related to the diffusion process at the HAGBs 7890 3.34E x 60 350.0
10–7
BGLOB – Ratio between the coarsening velocity of the α-phase and 2.00 – – –
the climb velocity of the β-phase
fi – Size of strain field caused by the arrangement of immobile 0.9998 0.993 – –
dislocations
fw – Size of strain field caused by the arrangement of wall 1.78 x 0.007 0.003 0.003
dislocations 10–4
nρ – Average number of dislocations considered to form a LAGB 2.50 1.11 2.00 2.00
int
fCDRX – Fraction of boundary α/β interfaces that contribute to – 0.0159 – –
β-subgrain rotation

Flow behavior in the α+β domain

Fig. 13

Fig. 13. Flow curves in the α+β domain at 840 ◦ C (a) and 800 ◦ C (b). “E” refers to the experimental results, while “M1” and “M2” refer to the
previously published results considering the iso-power to iso-stress change in load partitioning (Buzolin et al., 2021a) and the newly proposed model
considering the iso-strain to iso-stress change in load partitioning, respectively.

Predictions at large strains using the model presented in (Buzolin et al., 2021a) and the new proposed model

20
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 14. Measured (scattered data) and simulated β-microstructural features at constant strain rate and temperature using the model presented in
(Buzolin et al., 2021a) and the current presented model (dashed lines): a) subgrain size; b) grain size; c) wall dislocation density normalized by the
Burgers vector and the average number of dislocations to form a wall; d) fraction of high angle grain boundaries.

Uncertainties of the EBSD measurements

All the comparisons between the model and experimental results shown in the previous sections considered that the newly formed
LAGBs had an average misorientation angle θ0 = 2◦ , as in the previous publication (Buzolin et al., 2021a). (Gourdet and Montheillet,
2003) conventionally considered θ0 = 1◦ for commercial pure aluminium. Fig. 14 shows the effect of θ0 on the modelling and
experimental results during torsion at 920 ◦ C and 1 s-1. The model is more sensitive to a slight change in θ0 than the EBSD mea­
surement. However, the experimental results also show a clear dependence on this value. The variable most affected by θ0 is the high
angle grain boundary fraction. Except for grain size, all other variables shown are affected by a change in θ0 . Higher θ0 results in higher
high angle grain boundary fractions, larger subgrain sizes and lower dislocation densities, as the boundaries with lower mis­
orientations than this threshold are neglected in the calculation of these variables.
Fig. 15

21
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Fig. 15. Effect of the misorientation angle of newly formed LAGBs (θ0 ) on the modelling and experimental results in region P3 during torsion at 920
◦
C and 1 s-1: a) Fraction of high angle grain boundaries; b) Subgrain size; c) Grain size; d) Normalized wall dislocation density.

References

Ao, X., Xia, H., Liu, J., He, Q., 2020. Simulations of microstructure coupling with moving molten pool by selective laser melting using a cellular automaton. Mater.
Des. 185, 108230 https://doi.org/10.1016/j.matdes.2019.108230.
Babu, B., Lindgren, L.E., 2013a. Dislocation density based model for plastic deformation and globularization of Ti-6Al-4V. Int. J. Plast. 50, 94–108. https://doi.org/
10.1016/j.ijplas.2013.04.003.
Babu, B., Lindgren, L.E., 2013b. Dislocation density based model for plastic deformation and globularization of Ti-6Al-4V. Int. J. Plast. 50, 94–108. https://doi.org/
10.1016/j.ijplas.2013.04.003.
Bai, Q., Lin, J., Dean, T.A., Balint, D.S., Gao, T., Zhang, Z., 2013. Modelling of dominant softening mechanisms for Ti-6Al-4V in steady state hot forming conditions.
Mater. Sci. Eng. A 559, 352–358. https://doi.org/10.1016/j.msea.2012.08.110.
Buzolin, R.H., Ferraz, F.M.B., Lasnik, M., Krumphals, A., Poletti, M.C., 2020. Improved predictability of microstructure evolution during hot deformation of titanium
alloys. Materials. (Basel) 13, 1–30. https://doi.org/10.3390/ma13245678.
Buzolin, R.H., Krumphals, F., Poletti, M.C., 2022. Topological aspects in the microstructural evolution of AA6082 during hot plastic deformation. Eur. J. Mater. 0,
1–24. https://doi.org/10.1080/26889277.2022.2125350.
Buzolin, R.H., Lasnik, M., Krumphals, A., Poletti, M.C., 2021a. A dislocation-based model for the microstructure evolution and the flow stress of a Ti5553 alloy. Int. J.
Plast. 136, 102862 https://doi.org/10.1016/j.ijplas.2020.102862.
Buzolin, R.H., Lasnik, M., Krumphals, A., Poletti, M.C., 2021b. Hot deformation and dynamic α-globularization of a Ti-17 alloy: consistent physical model. Mater. Des.
197, 109266 https://doi.org/10.1016/j.matdes.2020.109266.
Canelo-Yubero, D., Poletti, C., Warchomicka, F., Daniels, J., Requena, G., 2018. Load partition and microstructural evolution during hot deformation of Ti-6Al-6V-2Sn
matrix composites, and possible strengthening mechanisms. J. Alloys Compd. 764, 937–946. https://doi.org/10.1016/j.jallcom.2018.06.097.
Castelluccio, G.M., McDowell, D.L., 2017. Mesoscale cyclic crystal plasticity with dislocation substructures. Int. J. Plast. 98, 1–26. https://doi.org/10.1016/j.
ijplas.2017.06.002.

22
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Dormand, J.R., Prince, P.J., 1986. A reconsideration of some embedded Runge-Kutta formulae. J. Comput. Appl. Math. 15, 203–211. https://doi.org/10.1016/0377-
0427(86)90027-0.
Fan, X.G., Jiang, X.Q., Zeng, X., Shi, Y.G., Gao, P.F., Zhan, M., 2018. Modeling the anisotropy of hot plastic deformation of two-phase titanium alloys with a colony
microstructure. Int. J. Plast. 104, 173–195. https://doi.org/10.1016/j.ijplas.2018.02.010.
Fan, X.G., Yang, H., 2011. Internal-state-variable based self-consistent constitutive modeling for hot working of two-phase titanium alloys coupling microstructure
evolution. Int. J. Plast. 27, 1833–1852. https://doi.org/10.1016/j.ijplas.2011.05.008.
Fan, X.G., Yang, H., Gao, P.F., 2013. Prediction of constitutive behavior and microstructure evolution in hot deformation of TA15 titanium alloy. Mater. Des. 51,
34–42. https://doi.org/10.1016/j.matdes.2013.03.103.
Ferraz, F.M.B., Sztangret, Ł., Carazo, F., Buzolin, R.H., Wang, P., Szeliga, D., dos Santos Effertz, P., Macioł, P., Krumphals, A., Poletti, M.C., 2023. Metamodelling the
hot deformation behaviour of titanium alloys using a mean-field approach. Mater. Today Commun. 35 https://doi.org/10.1016/j.mtcomm.2023.106148.
Galindo-Fernández, M.A., Mumtaz, K., Rivera-Díaz-del-Castillo, P.E.J., Galindo-Nava, E.I., Ghadbeigi, H., 2018. A microstructure sensitive model for deformation of
Ti-6Al-4V describing cast-and-wrought and additive manufacturing morphologies. Mater. Des. 160, 350–362. https://doi.org/10.1016/j.matdes.2018.09.028.
Gao, C.Y., Zhang, L.C., Yan, H.X., 2011. A new constitutive model for HCP metals. Mater. Sci. Eng.: A 528, 4445–4452. https://doi.org/10.1016/j.msea.2011.02.053.
Ghoniem, N.M., Matthews, J.R., Amodeo, R.J., 1990. A dislocation model for creep in engineering materials.
Gilmore, A., Liu, X., 2022. Mechanical behavior and microstructure evolution during hot torsion deformation of aluminum alloy AA2219. Adv. Eng. Mater. 24 https://
doi.org/10.1002/adem.202200048.
Gourdet, S., Montheillet, F., 2003. A model of continuous dynamic recrystallization. Acta Mater. 51, 2685–2699. https://doi.org/10.1016/S1359-6454(03)00078-8.
Groma, I., Csikor, F.F., Zaiser, M., 2003. Spatial correlations and higher-order gradient terms in a continuum description of dislocation dynamics. Acta Mater. 51,
1271–1281. https://doi.org/10.1016/S1359-6454(02)00517-7.
Hansen, B.L., Beyerlein, I.J., Bronkhorst, C.A., Cerreta, E.K., Dennis-Koller, D., 2013. A dislocation-based multi-rate single crystal plasticity model. Int. J. Plast. 44,
129–146. https://doi.org/10.1016/j.ijplas.2012.12.006.
He, D., Chen, S.B., Lin, Y.C., Xie, H., Li, C., 2023a. Hot tensile behavior of a 7046-aluminum alloy: fracture mechanisms and constitutive models. Mater. Today
Commun. 34, 105209 https://doi.org/10.1016/j.mtcomm.2022.105209.
He, D., Xie, H., Lin, Y., Xu, Z., Tan, X., Xiao, G., 2023b. High-temperature compression behaviors and constitutive models of a 7046-aluminum alloy.
Hogrefe, K., 2019. Flow modelling of Ti-6Al-4V under moderate and large strains.
Hua, K., Xue, X., Kou, H., Fan, J., Tang, B., Li, J., 2014. Characterization of hot deformation microstructure of a near beta titanium alloy Ti-5553. J. Alloys. Compd.
615, 531–537. https://doi.org/10.1016/j.jallcom.2014.07.056.
Humphreys, F.J., Hatherly, M., 2004. The mobility and migration of boundaries. Recrystall. Related Anneal. Phenom. 121–167. https://doi.org/10.1016/b978-
008044164-1/50009-8.
Jeong, Y., Jeon, B., Tomé, C.N., 2021. Finite element analysis using an incremental elasto-visco-plastic self-consistent polycrystal model: FE simulations on Zr and
low-carbon steel subjected to bending, stress-relaxation, and unloading. Int. J. Plast. 147 https://doi.org/10.1016/j.ijplas.2021.103110.
Ji, Z., Yang, H., Li, H., 2015. Predicting the effects of microstructural features on strain localization of a two-phase titanium alloy. Mater. Des. 87, 171–180. https://
doi.org/10.1016/j.matdes.2015.07.128.
Joo, M., Wi, M.S., Yoon, S.Y., Lee, S.Y., Barlat, F., Tomé, C.N., Jeon, B., Jeong, Y., 2023. A crystal plasticity finite element analysis on the effect of prestrain on
springback. Int. J. Mech. Sci. 237 https://doi.org/10.1016/j.ijmecsci.2022.107796.
Khan, A.S., Liu, H., 2012. Variable strain rate sensitivity in an aluminum alloy: response and constitutive modeling. Int. J. Plast. 36, 1–14. https://doi.org/10.1016/j.
ijplas.2012.02.001.
Köhler, U., Herzig, Ch., 1987. On the anomalous self -diffusion in BCC titanium.
León-Cázares, F.D., Monni, F., Jackson, T., Galindo-Nava, E.I., Rae, C.M.F., 2020. Stress response and microstructural evolution of nickel-based superalloys during low
cycle fatigue: physics-based modelling of cyclic hardening and softening. Int. J. Plast. 128, 102682 https://doi.org/10.1016/j.ijplas.2020.102682.
Lippold, J.C., Livingston, J.J., 2013. Microstructure evolution during friction stir processing and hot torsion simulation of Ti-6Al-4V. Metall. Mater. Trans. a Phys.
Metall. Mater. Sci. 44, 3815–3825. https://doi.org/10.1007/s11661-013-1764-1.
Li, Y., Gu, B., Jiang, S., Liu, Y., Shi, Z., Lin, J., 2020. A CDRX-based material model for hot deformation of aluminium alloys. Int. J. Plast. 134, 102844 https://doi.org/
10.1016/j.ijplas.2020.102844.
Liu, H., Zhang, J., Xu, B., Xu, X., Zhao, W., 2020. Prediction of microstructure gradient distribution in machined surface induced by high speed machining through a
coupled FE and CA approach. Mater. Des. 196, 109133 https://doi.org/10.1016/j.matdes.2020.109133.
Ma, A., Roters, F., Raabe, D., 2006. A dislocation density based constitutive model for crystal plasticity FEM including geometrically necessary dislocations. Acta
Mater. 54, 2169–2179. https://doi.org/10.1016/j.actamat.2006.01.005.
Mandal, S., Gockel, B.T., Balachandran, S., Banerjee, D., Rollett, A.D., 2017. Simulation of plastic deformation in Ti-5553 alloy using a self-consistent viscoplastic
model. Int. J. Plast. 94, 57–73. https://doi.org/10.1016/j.ijplas.2017.02.008.
Mason, J.K., Schuh, C.A., 2009. The generalized Mackenzie distribution: disorientation angle distributions for arbitrary textures. Acta Mater. 57, 4186–4197. https://
doi.org/10.1016/J.ACTAMAT.2009.05.016.
Matsumoto, H., Kitamura, M., Li, Y., Koizumi, Y., Chiba, A., 2014. Hot forging characteristic of Ti–5Al–5V–5Mo–3Cr alloy with single metastable β microstructure.
Mater. Sci. Eng. A 611, 337–344. https://doi.org/10.1016/j.msea.2014.06.006.
Orowan, E., 1940. Problems of plastic gliding. Proc. Phys. Soc. 52, 8–22. https://doi.org/10.1088/0959-5309/52/1/303.
Peng, W., Zeng, W., Wang, Q., Yu, H., 2013. Comparative study on constitutive relationship of as-cast Ti60 titanium alloy during hot deformation based on Arrhenius-
type and artificial neural network models. Mater. Des. 51, 95–104. https://doi.org/10.1016/j.matdes.2013.04.009.
Pichler, P., Leitner, T., Kaschnitz, E., Rattenberger, J., Pottlacher, G., 2022. Surface tension and thermal conductivity of NIST SRM 1155a (AISI 316L stainless steel).
Int. J. Thermophys. 43, 1–17. https://doi.org/10.1007/s10765-022-02991-5.
Pichler, P., Simonds, B.J., Sowards, J.W., Pottlacher, G., 2020. Measurements of thermophysical properties of solid and liquid NIST SRM 316L stainless steel. J. Mater.
Sci. 55, 4081–4093. https://doi.org/10.1007/s10853-019-04261-6.
Poletti, C., Germain, L., Warchomicka, F., Dikovits, M., Mitsche, S., 2016. Unified description of the softening behavior of beta-metastable and alpha+beta titanium
alloys during hot deformation. Mater. Sci. Eng. A 651, 280–290. https://doi.org/10.1016/j.msea.2015.10.109.
Pradeep, K., Buzolin, R.H., Domankova, M., Godor, F., Stanojevic, A., Poletti, M.C., 2024. Dynamic recrystallisation in Inconel®718 at creep conditions. Mater. Sci.
Eng.: A 893, 146146. https://doi.org/10.1016/j.msea.2024.146146.
Quan, G.Z., Lv, W.Q., Mao, Y.P., Zhang, Y.W., Zhou, J., 2013. Prediction of flow stress in a wide temperature range involving phase transformation for as-cast Ti-6Al-
2Zr-1Mo-1V alloy by artificial neural network. Mater. Des. 50, 51–61. https://doi.org/10.1016/j.matdes.2013.02.033.
Rezaei, M.J., Sedighi, M., Pourbashiri, M., 2023. Developing a new method to represent the low and high angle grain boundaries by using multi-scale modeling of
crystal plasticity. J. Alloys. Compd. 939, 168844 https://doi.org/10.1016/j.jallcom.2023.168844.
Ridgeway, C.D., Gu, C., Ripplinger, K., Detwiler, D., Ji, M., Soghrati, S., Luo, A.A., 2020. Prediction of location specific mechanical properties of aluminum casting
using a new CA-FEA (cellular automaton-finite element analysis) approach. Mater. Des. 194, 108929 https://doi.org/10.1016/j.matdes.2020.108929.
Riedlsperger, F., Wojcik, T., Buzolin, R., Zuderstorfer, G., Speicher, M., Sommitsch, C., Sonderegger, B., 2023. Microstructural insights into creep of Ni-based alloy 617
at 700 ◦ C provided by electron microscopy and modelling. Mater. Charact. 198, 112720 https://doi.org/10.1016/j.matchar.2023.112720.
Roters, F., Raabe, D., Gottstein, G., 2000. Work hardening in heterogeneous alloys - a microstructural approach based on three internal state variables. Acta Mater. 48,
4181–4189.
Roy, A., Acharya, A., 2006. Size effects and idealized dislocation microstructure at small scales: predictions of a Phenomenological model of Mesoscopic Field
Dislocation Mechanics: part II. J. Mech. Phys. Solids 54, 1711–1743. https://doi.org/10.1016/j.jmps.2006.01.012.
Semiatin, S.L., Mahaffey, D.W., Levkulich, N.C., Senkov, O.N., 2017. The radial temperature gradient in the gleeble® hot-torsion test and its effect on the
interpretation of plastic-flow behavior. Metall. Mater. Trans. A: Physi. Metall. Mater. Sci. 48, 5357–5367. https://doi.org/10.1007/s11661-017-4296-2.

23
F. Miller Branco Ferraz et al. International Journal of Plasticity 179 (2024) 104022

Shanthraj, P., Zikry, M.A., 2011. Dislocation density evolution and interactions in crystalline materials. Acta Mater. 59, 7695–7702. https://doi.org/10.1016/j.
actamat.2011.08.041.
Shen, J.H., Li, Y.L., Wei, Q., 2013. Statistic derivation of Taylor factors for polycrystalline metals with application to pure magnesium. Mater. Sci. Eng.: A 582,
270–275. https://doi.org/10.1016/j.msea.2013.06.025.
Souza, P.M., Beladi, H., Singh, R.P., Hodgson, P.D., Rolfe, B., 2018. An analysis on the constitutive models for forging of Ti6Al4V alloy considering the softening
behavior. J. Mater. Eng. Perform. 27, 3545–3558. https://doi.org/10.1007/s11665-018-3402-y.
Wang, L., Fan, X.G., Zhan, M., Jiang, X.Q., Zeng, X., Liang, Y.F., Zheng, H.J., Zhao, A.M., 2020. The heterogeneous globularization related to crystal and geometrical
orientation of two-phase titanium alloys with a colony microstructure. Mater. Des. 186 https://doi.org/10.1016/j.matdes.2019.108338.
Wei Lee, H., Fakhri, H., Ranade, R., Basaran, C., Egner, H., Lipski, A., Piotrowski, M., Mroziński, S., 2022. Modeling fatigue of pre-corroded body-centered cubic
metals with unified mechanics theory. Mater. Des. 224 https://doi.org/10.1016/j.matdes.2022.111383.
Wroński, M., Kumar, M.A., McCabe, R.J., Wierzbanowski, K., Tomé, C.N., 2022. Deformation behavior of CP-titanium under strain path changes: experiment and
crystal plasticity modeling. Int. J. Plast. 148 https://doi.org/10.1016/j.ijplas.2021.103129.
Xiong, F., Huang, C., Kafka, O.L., Lian, Y., Yan, W., Chen, M., Fang, D., 2021. Grain growth prediction in selective electron beam melting of Ti-6Al-4V with a cellular
automaton method. Mater. Des. 199, 109410 https://doi.org/10.1016/j.matdes.2020.109410.
Yang, L., Wang, B., Liu, G., Zhao, H., Xiao, W., 2015. Behavior and modeling of flow softening and ductile damage evolution in hot forming of TA15 alloy sheets.
Mater. Des. 85, 135–148. https://doi.org/10.1016/j.matdes.2015.06.096.
Zerilli, F.J., 2004. Dislocation mechanics–based constitutive equations. Metall. Mater. Trans. A 35, 215–218.

24