characterization and modeling of the high strain rate
TRANSCRIPT
Clemson UniversityTigerPrints
All Dissertations Dissertations
12-2018
Characterization and Modeling of the High StrainRate Response of Advanced High Strength SteelsRakan AlturkClemson University, [email protected]
Follow this and additional works at: https://tigerprints.clemson.edu/all_dissertations
This Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations byan authorized administrator of TigerPrints. For more information, please contact [email protected].
Recommended CitationAlturk, Rakan, "Characterization and Modeling of the High Strain Rate Response of Advanced High Strength Steels" (2018). AllDissertations. 2238.https://tigerprints.clemson.edu/all_dissertations/2238
CHARACTERIZATION AND MODELING OF THE HIGH STRAIN RATE
RESPONSE OF ADVANCED HIGH STRENGTH STEELS
A Dissertation
Presented to
the Graduate School of
Clemson University
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy
Automotive Engineering
by
Rakan Alturk
December 2018
Accepted by:
Fadi Abu-Farha, Committee Chair
Dr. Nicole Coutris
Dr. Gang Li
Dr. Garrett Pataky
Dr. David Schmuser
ii
ABSTRACT
The push for lightweighting in the automotive industry has motivated metallurgist
and steel manufacturers to produce new generations of steel that provide significant
improvements over the conventional steels to allow them to compete with the
introduction of low-density materials into the industry. To achieve this goal, metallurgist
introduced different (stronger and more ductile) phases into the ferrite-dominant
microstructure of conventional steels. This has led for generations of AHSSs with
significantly improved properties. However, the complex microstructure led to increased
complexities and unpredictability in the behavior of these materials, especially in their
response to variations in the strain rate. This is particularly important, as the materials in
the automotive industry exhibit different strain rates during their lifetime, and the
performance should be predictable especially during high strain rates as these are
encountered during a crash event where performance dictate safety.
This research work aimed at investigating and developing a methodology to allow
for the accurate modeling of multi-phased AHSSs at muli-strain rates. First, a set of
experimental tests were performed on a selected set of AHSSs having a range of
combination of phases in their microstructure at different strain rates. This allowed for
the investigation of the effect of the different phases in the microstructure on the response
of the materials at multi-strain rates, and it gave an insight into the combination of phases
that would result in a material with a favorable response at high strain rates. Then, the
focus was shifted into a particular third-generation AHSS (Medium Mn. steel), the
iii
selection of which was based on its complexity and importance to the future of AHSSs.
Further experiments were performed on this material to characterize its anisotropy at
multi-strain rates. The experiments were used to calibrate an anisotropic yield function at
different strain rates, and the shape and size of the yield locus obtained were observed to
be dependent on the strain rate.
Furthermore, the experimental results of Medium Mn. steel was used to develop a
constitutive relation predicting the strain sensitivity of the anisotropy of the material at
different strain rates. A modification on the Yld2000-2d yield function allowed to
develop a unique strain rate dependent anisotropic yield function that captures the
yielding and anisotropic behavior of the material at multi-strain rates. The model
developed was validated using finite element simulations of the experimental tests.
Finally, the developed model was used to perform crash simulation on a railroad
tank car. The simulations accounted for the strain rate sensitivity of the hardening and
anisotropy of the material. The results highlighted the impact of the current work by
extracting the load-displacement and the energy absorbed from three different
simulations with variations in accounting for the strain rate sensititvity of the material.
The difference in the results emphasizes the impact and importance of utilizing the
methodology proposed in this dissertation for the crashworhiness analysis of AHSSs.
iv
DEDICATION
My PhD journey has been a tough one and wouldn’t have finished without
the unconditional support of my amazing parents who sacrificed everything to get me
where I am, my three wonderful brothers who show me unconditional love, my
wonderful grandmothers who took care of me and pushed me through my life, my
dear friend Nancy who has been my companion and motivator for the past three
years, and the rest of my family and friends. To all, I say thank you, and I love
you
v
ACKNOWLEDGMENTS
First, I would like to express all my gratitude to Dr. Fadi Abu-Farha for
his wise guidance and continued patience during my study and research over
the past four years. He was always willing to help whenever my research
encountered bottlenecks. It has been a great pleasure to work with him, and I
believe that I have learned a lot from him. I also thank my committee members,
Dr. Nicole Coutris, Dr. Garett Pataky, Dr. Gang Li, and Dr. David Schmuser,
for their patient and inspiring suggestions during my studying experience in
their classes and this dissertation writing.
I would particularly take this opportunity to sincerely acknowledge Dr.
Louis Hector as well as the entire General Motors for offering me a great
internship opportunity to work in the automotive industry in 2015 and 2016.
During this internship, I accumulated precious working experience, and that
was extremely helpful to my research work.
Also, I cherished the collaboration experience and help of Dr. Jun Hu,
Akshat Agha, Hakan Kazan, Dr. Farbod Niaki, Saeed Farahani, Abdelrahim
Khal, Dr. Bashar elzuwayer and all the other group mates. I also appreciate the
technical support from the Clemson University library and Palmetto cluster
groups. Furthermore, many thanks to the U.S. Department of Energy for
sponsoring my research project, and AK Steel, ThyssenKrupp Steel, General
Motors, Fiat-Chrysler Automobiles, and JSOL Corporation for providing the
research materials and educational support.
vi
TABLE OF CONTENTS
Page
ABSTRACT ........................................................................................................... i
DEDICATION ...................................................................................................... iv
ACKNOWLEDGMENTS ..................................................................................... v
TABLE OF CONTENTS ..................................................................................... vi
LIST OF TABLES ................................................................................................ ix
LIST OF FIGURES ............................................................................................... x
CHAPTER
1. INTRODUCTION ........................................................................................ 1
1.1 Lightweighting ................................................................................... 1
1.2 AHSSs ................................................................................................ 3
1.3 DP steels............................................................................................. 8
1.4 Q&P Steel ........................................................................................ 10
1.5 Medium Mn. Steel: .......................................................................... 12
1.6 Motivation ........................................................................................ 13
1.7 Dissertation Overview ..................................................................... 15
1.7 References ........................................................................................ 16
2. EFFECTS OF MICROSTRUCTURE ON THE STRAIN RATE
SENSITIVITY OF ADVANCED STEELS ........................................................ 24
2.1 Introduction ................................................................................... 24
2.2 Experiments ................................................................................... 28
vii
2.3 Results .............................................................................................. 40
2.4 Discussion ........................................................................................ 57
2.5 Summary and Conclusions .............................................................. 65
2.6 References ........................................................................................ 69
3. STRAIN RATE EFFECT ON TENSILE FLOW BEHAVIOR AND
ANISOTROPY OF A MEDIUM MANGANES TRIP STEEL .......................... 80
`3.1 Introduction .................................................................................. 81
3.2 Medium Manganese TRIP Steel ................................................... 85
3.3 Test Specimens .............................................................................. 86
3.4 Instrumentation.............................................................................. 88
3.5 Strain Measurement....................................................................... 89
3.6 Results and Discussion .................................................................. 91
3.7 Conclusions .................................................................................... 101
3.8 References ...................................................................................... 105
4. A RATE DEPENDENT ANISOTROPIC YIELD FUNCTION .............. 111
`4.1 Introduction ................................................................................ 112
4.2 Overview of yield functions........................................................... 113
4.3 Barlat Yld2000-2d ......................................................................... 117
4.4 Calibration of Yld2000-2d ............................................................. 119
4.5 Rate Dependent Yld2000-2d .......................................................... 121
4.6 Numerical Procedure ..................................................................... 126
4.7 FEA Validation .............................................................................. 135
Table of Contents (Continued) Page
viii
4.8 Conclusions and Summary ............................................................ 139
4.9 References ................................................................................... 140
5. CRASH SIMULATIONS OF A RAILROAD TANK CAR .................... 149
5.1 Introduction .................................................................................... 150
5.2 Finite Element Model .................................................................... 153
5.3 Crash simulations using TC128-B ................................................. 159
5.4 Crash simulation using medium Mn (10% wt.) steel ..................... 160
5.5 Summary ........................................................................................ 167
5.6 References ...................................................................................... 167
6. SUMMARY AND CONCLUSIONS ....................................................... 171
6.1 Contributions.................................................................................. 172
6.2 Future Work ................................................................................... 173
Table of Contents (Continued) Page
ix
LIST OF TABLES
Table Page
Table 1.1: Assumed tension properties of steel phases [28] ............................................... 7
Table 2.1 Steel Chemistries (wt. %) ................................................................................. 30
Table 2.2: 0.2% Offset Yield Strength (MPa) at the different strain rates. ...................... 43
Table 2.3: Engineering strain at uniform elongation at the different strain rates. ............ 43
Table 2.4: Ultimate Tensile Strength (MPa) at the different strain rates. ......................... 43
Table 2.5: Engineering Strain at Fracture at the different strain rates. ............................. 44
Table 3.1: Chemical composition (wt.%) of Med. Mn steel. ............................................ 87
Table 3.2: Tensile Properties Associated with RD Flow Curves in Fig. 4.5a. Engineering
stress and strain are reported. ............................................................................................ 92
Table 3.3: Tensile Properties Associated with DD Flow Curves in Fig. 4.5b. Engineering
stress and strain are reported. ............................................................................................ 92
Table 3.4: Tensile Properties Associated with TD Flow Curves in Fig. 4.5c. Engineering
stress and strain are reported. ............................................................................................ 93
Table 4.1: Experimental results used to calibrate Yld2000-2d ....................................... 120
Table 4.2: The calibrated 1 8
for each strain rate ..................................................... 121
Table 5.1: the parameters of the simulations .................................................................. 157
x
LIST OF FIGURES
Figure 1.1: The trend in the CO2 and the fuel economy for vehicles in the USA between
1975-2017 as published by the EPA. .................................................................................. 2
Figure 1.2: The “banana” diagram showing the elongation vs. tensile strengths for
different alloys of steels including the three generations of AHSSs [27]. .......................... 6
Figure 1.3: A micrograph showing the typical microstructure of DP steels ....................... 8
Figure 1.4: The iron carbide phase diagram [38] ................................................................ 9
Figure 1.5: Stress-Strain curve showing the effect of martensite on the strength of DP
steels, where the martensite content of CAS>CHCL>CAL [37] ...................................... 10
Figure 1.6: The heat treating processes used to produce (a) DP, (b) TRIP and (c) Q&P
steels [46] .......................................................................................................................... 12
Figure 1.7: Microstructure of QP980 showing the three phases ....................................... 12
Figure 2.1: Micrographs showing the microstructure of the as-received material for (a)
CR5 (b) SS 201 (c) DP590 and (d) DP980 (e) Medium Mn. (10wt%) (f) QP980. .......... 31
Figure 2.2: Experimental setups and test specimens used for (a) quasi-static testing and
(b) high rate testing (Servo-hydraulic) (c) detailed view of the components of the Instron
VHS unit ........................................................................................................................... 34
Figure 2.3: Photron SA1.1 high-speed cameras ................................................................ 36
Figure 2.4: (a) Engineering stress-strain curve of a test ran at a speed of 20m/s showing a
dominant ringing signature and (b) Engineering stress-strain curve of a test ran at a speed
of 20m/s showing the elimination of ringing .................................................................... 37
Page
xi
Figure 2.5: High rate tensile specimen geometry with (a) 20mm gauge length (b) 10mm
gauge length ...................................................................................................................... 39
Figure 2.6: (a) Evolution of strain along the centerline (using a virtual extensometer) of a
10 mm gauge length specimen tested at 500s-1. Zero position denotes the center of gauge
section. (b) Evolution of strain along the centerline (using a virtual extensometer) ........ 40
Figure 2.7: Engineering stress/strain curves at 0.005s-1 and 500s-1 for (a) CR5 (b) SS
201 (c) DP590 (d) DP980 (e) Medium Mn. (10wt%) and (d) QP980. ............................. 45
Figure 2.8: (a) Kolsky bar at NIST (b) Specimen Geometry used for Kolsky testing at
NIST (c) Engineering stress-strain curves with the 500s-1 obtained from the Kolsky
testing. ............................................................................................................................... 46
Figure 2.9: (a) Yield strength (MPa) vs log strain rate (b) Tensile strength (MPa) vs log
strain rate ........................................................................................................................... 52
Figure 2.10: n-value vs log strain rate .............................................................................. 53
Figure 2.11: Percent Change between 0.005s-1 and 500s-1 of (a) Yield Stress (b) Tensile
Strength (c) Total Elongation and (d) Toughness. ............................................................ 56
Figure 2.12: Effect of martensite on the percentage increase of tensile strength ............. 59
Figure 2.13: Volume fraction of martensite measured at different plastic using: (a)
magnetic induction (b) XRD. ............................................................................................ 64
Figure 3.1: Deformation mechanisms in multiphase AHSSs with Mn alloying content .. 82
Figure 3.2: Field emission SEM image of the as-received, medium Mn TRIP steel (two
phases, ferrite (F) and retained austenite (A)) with ~2 μm grain size. Annealing twins
(AT) appear in austenite grains which are indicative of a low stacking fault energy. ...... 87
List of Figures (Continued) Page
xii
Figure 3.3: Tensile specimen geometry for high rate testing. All dimensions are in mm 88
Figure 3.4: INSTRON VHS 65/80-20 (VHS=very high speed) servo-hydraulic test unit 90
Figure 3.5: Engineering stress-strain curves at the five testing strain rates for: (a) 0º (RD
– rolling direction), (b) 45º (DD – diagonal direction), and (c) 90º (TD – transverse
direction). .......................................................................................................................... 94
Figure 3.6: (a) Tensile (2D DIC) strain contours at the ninth image showing the onset of
Lüdering (red contour patch) in the 500 s-1 curve (see Fig. 3.5). (b) Tensile (2D DIC)
strain contours showing Lüders strain at image twelve in the 500 s-1 curve (c) Tensile (2D
DIC) strain contours showing Lüders strain at image twelve in 500 s-1 curve. ................ 95
Figure 3.7: Normalized flow stresses from Eq. (3.1) along the (a) diagonal direction (DD)
(b) transverse direction (TD). ........................................................................................... 97
Figure 3.8: r-value evolution with strain at different strain rates for specimens tested
along the (a) rolling direction (RD) (b) diagonal direction (DD), and (c) transverse
direction (TD). .................................................................................................................. 99
Figure 3.9: Average r-value and normal anisotropy coefficient (rn) variation with the
logarithm of nominal strain rate for the RD, DD and TD. .............................................. 102
Figure 4.1: Yield locus calibrated using data from experiments at different strain rates 122
Figure 4.2: Variation of the Yld2000-2d constants ( 1 8 ) with the strain rate ......... 123
Figure 4.3: Comparison of yield locus plotted using the experimental data at 0.005s-1
(pink), at 0.5s-1 (blue) and using the equation (green), showing an agreement between the
actual and the predicted yield locus. ............................................................................... 124
List of Figures (Continued) Page
xiii
Figure 4.4: Comparison of yield locus plotted using the experimental data at 0.005s-1
(pink), at 5s-1 (blue) and using the equation (green), showing an agreement between the
actual and the predicted yield locus. ............................................................................... 125
Figure 4.5: Comparison of yield locus plotted using the experimental data at 0.005s-1
(pink), at 50s-1 (blue) and using the equation (green), showing an agreement between the
actual and the predicted yield locus. ............................................................................... 126
Figure 4.6: Comparison of yield locus plotted using the experimental data at 0.005s-1
(pink), at 500s-1 (blue) and using the equation (green), showing an agreement between the
actual and the predicted yield locus. ............................................................................... 127
Figure 4.7:A diagram explaining the process of the cutting-plane algorithm ................ 131
Figure 4.8: Flow chart showing the steps followed in the cutting plane algorithm ........ 136
Figure 4.9: Finite element model used to verify the developed equation ....................... 137
Figure 4.10: Comparison of the experimental results and simulations at 0.005s-1......... 138
Figure 4.11: Load vs. displacement comparison between the experiments and simulations
at 50s-1 using the Yld2000-2d calibration at 0.005s-1 (red curve) and using the developed
equation (green curve) .................................................................................................... 138
Figure 4.12: Load vs. displacement comparison between the experiments and simulations
at 500s-1 using the Yld2000-2d calibration at 0.005s-1 (red curve) and using the developed
equation (green curve) .................................................................................................... 139
Figure 5.1: A schematic showing the collisions associated with a derailment accident [3].
......................................................................................................................................... 151
Figure 5.2: A schematic of a typical North American railroad tank car [18]. ................ 154
List of Figures (Continued) Page
xiv
Figure 5.3: The finite element model of the (a) tank (b) impactor ................................. 155
Figure 5.4: The meshed finite element model of the tank .............................................. 155
Figure 5.5: (a) A tank car used for an actual crash test [8] (b) the model developed to
replicate the boundary conditions encountered during a crash test. ............................... 156
Figure 5.6: The model showing the initialization of the fluids inside the tank. Green
(liquid), blue (vapor). ...................................................................................................... 158
Figure 5.7: Comparison of the acceleration profiles of the impactor between (a) the
simulation without lading and (b) Kirkpatrick results [6]. ............................................. 161
Figure 5.8: Comparison of the force-displacement curves of the impactor between (a) the
simulation without lading and (b) Kirkpatrick results [6]. ............................................. 162
Figure 5.9: Comparison of the force-displacement curves of the impactor between (a) the
simulation with lading and (b) Kirkpatrick results [6]. ................................................... 163
Figure 5.10: Comparison of the three load-displacement profiles of simulations ran using:
1) quasi-static hardening and yielding (green), 2) multi-rate hardening and quasi-static
yielding (blue), and 3) the developed constitutive relation, i.e. multi-rate hardening and
yielding (red) ................................................................................................................... 165
Figure 5.11: Energy absorbed to fracture during a crash simulation of the three different
scenarios described. ........................................................................................................ 166
List of Figures (Continued) Page
1
CHAPTER ONE
1. INTRODUCTION
1.1 Lightweighting
In the past decade, research into lightweighting techniques in the automotive
industry has risen significantly. This trend can be mainly attributed to the desire of the
industry to 1) reduce CO2 emissions 2) increase the fuel efficiency of the vehicles 3)
meet the requirement of recent government environmental standards, and 4) reduce the
depletion of energy sources [1, 2]. Assuredly, there are other means by which the goals
above can be achieved, such as: optimizing current technology, or using cleaner fuels.
However, the former has a shallow impact on the overall weight of the vehicle while the
latter has a high initial cost [3]. Whereas, studies have shown that a 10% reduction in
vehicle weight can result in 5-8% improvement in fuel economy [4]. For example, Audi
A2 showed that 60% of fuel consumption is mass-dependent; 10% reduction in vehicle
weight leads to ~6% reduction in fuel consumption [5]. Additionally, lightweighting
leads to overall positive effects on the vehicle such as: mass decompounding and
improving the vehicle handling [6, 7]. Therefore, lightweighting is the most efficient
method to achieve the goals set by the OEMs and the regulations set by the governments.
The first government standard regulating fuel consumption in the USA for
vehicles dates back to the 1970s [8]. Consequently, OEMs began their lightweighting
efforts, cultivating in the shift from body-on-frame (BOF) structure into body-in-white
(BIW) structures, adopting smaller engines and shifting from rear-wheel drive to front-
2
wheel drive systems [9]. These efforts have cultivated in the significant reduction of the
average weight of the vehicles in the USA between 1976 to 1986. However, at the
beginning of the 1990s, there was an opposite trend observed in the average weight of the
vehicle due to an increase in safety and luxury features in vehicles, and the increased
popularity of SUVs and pickup trucks. Nonetheless, the lightweighting drive was again
on the rise in the early 2000s mainly because of the significant increase in fuel prices and
the stringent regulations by governments. Figure 1.1 shows how the lightweighting trend
affected the CO2 emission and the fuel economy of the vehicles in the USA as published
by the EPA in 2018 [10]. Moreover, until today, lightweighting is one of the top priorities
for vehicle manufacturers around the World, although the methods for lightweighting
now has changed.
Figure 1.1: The trend in the CO2 and the fuel economy for vehicles in the USA between
1975-2017 as published by the EPA.
3
The main methods of lightweighting today are material substitution, parts
consolidation and design optimization [1]. Up till today, steel remains the most dominant
material in an automotive BIW [11], with a density of about 7830 kg/m3. Lightweighting
can be achieved with the replacement of steel in the structure with materials with lower
densities such as aluminum (2700 kg/m3) or magnesium (1738 kg/m3). Examples of
material replacements in the automotive industry include the 1994 Audi A8 which had a
BIW entirely made of aluminum [12] and the 2013 BMW i3 which had carbon-fiber
composite structures [13]. Yet, steels are still dominant in the automotive industry body
structures until today, and that status is projected to remain unchanged in the near future
[14]. This dominance of steel is mainly attributed to three points: 1) Mass production: the
automotive manufacturing has been designed to cater to the properties of steels, and the
introduction of a new material (especially magnesium and polymer composites) will
require substantial changes to the infrastructure, 2) Cost: steel is the cheapest metal on
Earth, all of these replacement materials have a much higher cost, and 3) Advanced High
Strength Steels (AHSSs): the steel industry responded to the lightweighting drive with the
introduction of new grades of steels that have improved properties over the conventional
steels and can achieve lightweighting [14].
1.2 AHSSs
In 1975 after the worldwide oil crisis, a sudden increase in the interest of
developing high strength steels for the automotive applications was observed [15, 16].
The first alloys of steel to be introduced were the High Strength Low Alloy (HSLA)
steels that showed a slight improvement in the yield and tensile strengths over the
4
conventional mild steels accompanied by a reduction in the ductility [17]. The unique
combination of strength and ductility required by the industry was found in dual phase
(DP) steels [18], which showed significant improvements on the strength [19, 20].
However, due to the difficulties in adopting the industry to these new steels, the interest
in these materials died out except for some of the conventional High Strength Steels
(HSSs) such as Interstitial Free (IF) steels, Bake Hardenable Interstitial Free (BH-IF) and
HSLA steels [16, 21]. This interest was then revived with the increased competition from
the low-density materials cultivating in the development of the aluminum Audi A8 in
1994, and the demand for improved passenger safety and environment regulations [22].
The response of the steel industry launched a project named Ultra-Light Steel Auto Body
(USLAB) through an international consortium of 35 companies from 18 countries. The
project aimed at developing a lightweight steel BIW comprised mainly of HSS. The
result of this project was the development of a BIW comprised of 90% HSS with a 25%
reduction in weight and improved structural properties [14, 16]. Consequently, there was
an increase in the research for developing new grades of steels to further improve the
lightweighting capabilities, and the interest in DP steels was revived. This led to the
commercialization of DP steels internationally in 1995.
With the more stringent government standards, the increase in fuel prices, and the
infiltration of the low-density materials in the automotive industry provided a push on the
steel industry to seek better combinations of strength and ductility in their steels. These
combinations were achieved through the development of the Advanced High Strength
Steels (AHSSs). The exact definition of AHSSs can be vague sometimes; they can be
5
defined as steels having tensile strengths higher than 600 MPa with “improved”
formability [16]. However, the main feature that differentiates these materials from HSSs
and conventional steels is their microstructure, which contains more than one phase. DP
steels can be considered the first AHSS and belong to what is called the first generation
of AHSSs. The first generation included other materials such as Transformation Induced
Plasticity (TRIP) steels, Ferrite-Bainite (FB) steels, Martensitic Steels (MS) and
Complex-Phase (CP) steels. DP steels, as the name suggests, consist of two phases;
ferrite and martensite, martensite being the phase attributed with the increase in strength.
The TRIP steels show improvement in the strain hardenability and therefore toughness
which can be attributed to the retained austenite in its microstructure [23]. Moreover,
martensitic steels contain more than 80% martensite in their microstructure leading to the
development of very strong yet brittle material. The first generation of AHSSs showed an
increase in the strength with a general decrease of ductility following the trend shown by
the conventional steels in the banana chart shown in Figure 1.2. The first generation of
AHSSs has already been used in the automotive industry. DP and TRIP steels are used
prominently in energy absorbing areas. While the high strength DP and MS steels are
used in areas that require high stiffness and resistance to intrusion [11, 16].
The second generation of AHSSs can be shown to break the trend observed with
the conventional steels and the first generation of AHSSs by improving both the ductility
and strength (Figure 1.2). The main characteristic of the second generation of AHSSs is
that their microstructure is austenite-based and they have high alloy content. The most
known of the second generation AHSSs is the Twinning Induced Plasticity (TWIP) steels
6
which show strengths reaching 1000MPa and elongations of about 60% [24, 25]. The
properties of the second generation are desirable; however, these materials have a high
cost (alloying), and poor manufacturability and welding. Consequently, their applications
in the automotive industry are limited as compared to the first generation [26].
The difficulties shown with the second generation of AHSSs have led the industry
to focus the research on developing the new third-generation AHSSs that can fill the gap
between the first and second generations on the banana chart. Studying the
microstructure of the first and second generation of AHSSs, it was suggested that to
Figure 1.2: The “banana” diagram showing the elongation vs. tensile strengths for
different alloys of steels including the three generations of AHSSs [27].
7
achieve the desired properties, a microstructure that dominantly contains austenite and
martensite with low alloying is required [28]. The reasons for this combination can be
explained by looking at table 1.1, which shows the different properties associated with
the three main phases of steels; ferrite, martensite, and austenite. Martensite is the phase
that provides the highest strength while austenite provides the highest ductility, therefore,
a combination of these two phases should provide the desired properties. The research
conducted for the development of the third generation AHSSs was summarized by
Fonstein [16] in the following materials. 1) Bainitic ferrite TRIP steel which improves the
strength and ductility of the conventional TRIP steels [29], 2) Quenched and Partitioned
(Q&P) steels, which are produced through a modification on the production process of
TRIP steels [30, 31] and 3) Medium Mn. steels (4-10% wt.) which contain a combination
of ferrite and austenite with a medium Mn. content, that improves the ductility and
strength significantly [32], these steels are discussed in detail in this thesis.
Table 1.1: Assumed tension properties of steel phases [28]
Phase Tensile Strength (MPa) Uniform True Strain
Ferrite 300 0.3
Austenite 640 0.6
Martensite 2000 0.08
8
1.3 DP steels
As mentioned earlier, DP steels were among the first steels to be introduced and
classified as AHSSs. They provide a right combination of strength and ductility and have
been widely used in the automotive industry. They can provide these properties because
of their microstructure, shown in Figure 1.3, which consists of a ferrite matrix with
martensitic islands dispersed in it. DP steels are typically produced by first heating the
material to the inter-critical temperature region, soaking to allow austenite to nucleate
and grow followed by slow cooling to the quench temperature and finally rapid cooling to
allow the austenite to transform into martensite [33, 34]. The different phases and the
temperatures they exist at, are shown in the iron carbide phase diagram shown in Figure
1.4. Note that martensite cannot be seen on the diagram, and that is because martensite
can only be achieved through the transformation of austenite. Austenite is a very unstable
phase at room temperatures and will transform to martensite. It is only stable at high
temperatures and through specific alloying techniques.
Figure 1.3: A micrograph showing the typical microstructure of DP steels
9
The austenite to martensite transformation which occurs during cooling results in
an increase in the density of mobile dislocations within the ferrite phase, which
accommodates the majority of the plastic deformation and causes the excellent
continuous yielding observed in the material [35, 36]. On the other hand, the presence of
the martensite phase causes the high strength shown by the DP steels. As can be seen
with the different DP steels produced, the higher the martensite content, the higher the
yield and the ultimate strength of the material (Figure 1.5) [35, 37]. The volume fraction
of martensite in DP steels is controlled by varying the annealing temperature and the
cooling rate in the quenching process.
Figure 1.4: The iron carbide phase diagram [38]
10
Figure 1.5: Stress-Strain curve showing the effect of martensite on the strength of DP
steels, where the martensite content of CAS>CHCL>CAL [37]
1.4 Q&P Steel
As discussed earlier, austenite is an unstable phase that will transform into
martensite at room temperature (see Figure 1.3). However, austenite has desirable
properties, and its presence in the microstructure will provide improvements on the
material. Therefore, research was conducted to find methods to stabilize austenite at room
temperatures. Two methods are generally used by the industry; the first method is mainly
using specifying alloying element (Mn, Ni, Al, and Cu) to increase the stacking fault
energy of the austenite and thus stabilizing it [39-43]. This method is used to produce
most austenitic stainless steels and TWIP steels. However, the increased allying content
causes an increase in the cost and difficulties in welding. The second method uses a
modification on the heat treatment process used to produce the steels with a little help of
11
Al, Si, P, and Mo to suppress the growth of cementite [44]. This method gave birth to
TRIP steels and Q&P steels.
Q&P steels were first developed at the Colorado School of Mines [30, 45], with a
modification to the process used to produce TRIP steels. TRIP steels are produced [46],
by heating to the inter-critical temperature, similar to DP steels, then quenching the
material to a temperature above the that at which martensite begins forming (~357 oC),
followed by a tempering step at which some austenite is stabilized by the carbon atoms
which are suppressed to form cementite by the alloys. Finally, the material is quenched to
room temperature, during which the unstable austenite transforms to martensite. The
process is visualized in Figure 1.6, which shows a comparison between the three different
processes for forming DP, TRIP and Q&P steels. The figure shows that for Q&P, the first
quenching step does not stop before the martensite starts temperature but instead stops at
a temperature between the martensite start and the finish temperatures. This means that
some of the austenite transforms to martensite during that step. Then the material is
reheated again above the temperature martensite starts forming and held at that
temperature allowing carbon atoms to partition from martensite and work to stabilize the
remaining austenite, preventing its transformation back to martensite during the final
quenching phase [47]. This method allows for a better control over the amount of
martensite the material has and thus can produce materials with higher strengths than
TRIP steels [48]. Figure 1.7 shows a micrograph of the microstructure of QP980,
showing the three phases: ferrite, austenite, and martensite.
12
Figure 1.6: The heat-treating processes used to produce (a) DP, (b) TRIP and (c) Q&P
steels [46]
Figure 1.7: Microstructure of QP980 showing the three phases
1.5 Medium Mn. Steel
One of the methods to stabilize austenite mentioned earlier was using an alloying
element with Mn being one of the main alloying elements that can achieve that. TWIP
steels have high Mn content (~20% wt.) to accomplish that. However, that high content
13
causes an increase in the cost and problems with the manufacturing processes and
welding [49]. Therefore, research was done to develop a steel with an austenite content
higher than TRIP steels and lower than TWIP steel to maintain the cost and weldability
[50]. Aydin et al. [32] developed a steel alloy with medium Mn. content based on having
a stacking fault energy between that of TRIP and TWIP steels. The study showed the
importance of having the Mn content in the microstructure to about 10% to obtain the
desired austenite content and stability to improve over TRIP steels. A medium Mn. steel
is the main steel focused on in this thesis and will be described in more detail in chapters
2 and 3.
1.6 Motivation
The enhancement of the mechanical properties of AHSSs over the conventional
steels comes at the cost of increased complexities in the microstructure. The complex
multi-phased microstructure of AHSSs contains various micro-elements that can change
the macro behavior of the material significantly. This will translate into increased
complexities and difficulties in modeling and predicting the behavior of these materials.
Taking austenite as one of those micro-elements. It remains a metastable phase
even when the material contains enough alloys to stabilize it at room temperature, loading
the material will cause it to transform into martensite (TRIP effect) [51, 52]. While that is
a preferred property that improves on the strength and ductility of the material, it is
highly unpredictable due to its dependence on multiple variables, such as temperature
[53, 54] and strain rate [55, 56].
14
Strain rate, in particular, is a very important factor especially for the automotive
industry, as the materials used in the automotive industry undergo loading at strain rates
ranging from 0.001s-1 to 1000s-1. It is well established in the literature that the mechanical
properties of materials are highly dependent on the strain rate [57-60]. It is, therefore,
crucial to understanding how the strain rate affects the behavior of AHSSs if they were to
be used in an automobile. As mentioned earlier, AHSSs are made of multi-phase
microstructure, each having its own independent properties and mechanisms involved
with deformation. Therefore, each will have its own response to changes in strain rate and
that will provide a challenge to model and predict.
Yet, the aforementioned properties do not represent the full spectrum of
challenges involved with modeling the new AHSSs. Anisotropy is another property that
can influence the accuracy of these model if it is not considered. Anisotropy is the
dependence of material properties on the crystallographic direction at which they are
measured. The anisotropy in a sheet metal can be caused by the manufacturing processes
that impart a crystallographic direction or texture on the material and due to the variation
of atomic spacing in different crystallographic directions [38]. This will cause further
complications in modeling AHSSs because as is the case with strain sensitivity and other
properties, the different phases in the microstructure will have its own degree of
anisotropy.
All these complications will manifest themselves when trying to model the
behavior and performance of these AHSSs in crash simulations. During these
simulations, the strain sensitivity, the anisotropy and the combination of both will
15
significantly affect the behavior of these materials. Therefore, despite the number of
attractive properties AHSSs offer, there is a need for these challenges to be addressed and
studied and that what motivated this work. The focus will be on improving the
understanding of the AHSSs, their microstructure, and its effect on the various properties
of interest for the crashworthiness of these materials. Different grades of DP, QP980 and
Medium Mn. steels will be studied and presented. However, the focus will be on Medium
Mn. steel as it is still a development third-generation AHSS that present great
complexities in its modeling, especially regarding its anisotropy. The thesis aims to
answer three questions:
1) What is the effect of the phases in the microstructure of the material on its macro
properties at multiple strain rates?
2) How does the anisotropy of Medium Mn. steel gets affected at high strain rates?
3) How to develop a phenomenological model to capture unique behavior exhibited
by Medium Mn. steel, with regards to its strain rate sensitivity and anisotropy at
multiple strain rates?
1.7 Dissertation Overview
Chapter 2 in this dissertation will aim to explore the answer to the first question
by presenting a series of experimental techniques and results on multiple AHSSs with
varying microstructures. The unique experimental setups of multi-strain rate testing are
presented and discussed, followed by a detailed discussion of the contribution of each of
the three main phases in the different microstructure on the experimental observations.
16
Then, in chapter 3 the focus will be drawn to Medium Mn. steel, when its
anisotropy at multiple strain rates is explored and discussed. The anisotropy is studied
through different properties at different strain rates to show how strain rate sensitive is
the anisotropy. Following that, chapter 4 will be a detailed presentation and derivation of
a constitutive model for Medium Mn. steel aiming to capture the unique strain rate
sensitive anisotropy of the material. Furthermore, its compilation and the numerical
procedure of incorporating it for use in a finite element software will be discussed. The
model developed will be verified through simulations of experimental test and
comparison with the results. Finally, in chapter 5, the model developed will be used to
perform simulations on a railroad tank car to quantify the effect of incorporating the
different variables in a crash simulation.
1.7 References
[1] D. Rosato, "Exclusive content. Automotive lightweighting trend here to stay,
2011," ed, 2017.
[2] N. McGuire, "Automotive industry lightens up," Tribology & Lubrication
Technology, vol. 73, pp. 30-41, 2017.
[3] R. Tucker, "Trends in automotive lightweighting," Metal Finishing, vol. 111, pp.
23-25, 2013.
[4] L. Brooke and H. Evans, "Lighten up!," Automotive Engineering International,
vol. 117, 2009.
17
[5] M. Barrett, R. Lowe, T. Oreszczyn, and P. Steadman, "How to support growth
with less energy," Energy Policy, vol. 36, pp. 4592-4599, 2008.
[6] M. Verbrugge, T. Lee, P. E. Krajewski, A. K. Sachdev, C. Bjelkengren, R. Roth,
et al., "Mass decompounding and vehicle lightweighting," in Materials Science Forum,
2009, pp. 411-418.
[7] P. K. Mallick, Materials, design and manufacturing for lightweight vehicles:
Elsevier, 2010.
[8] N. H. T. S. Administration. (May, 13). CAFE Overview: "What is the origin of
CAFE?"". Available: https://www.nhtsa.gov/laws-regulations/corporate-average-fuel-
economy
[9] M. Mueller, "Timeline: A Path to Lightweight Materials in Cars and Trucks," O.
o. E. E. R. Energy, Ed., ed, 2016.
[10] E. P. Agency, "Light-Duty Automotive Technology, Carbon Dioxide Emissions,
and Fuel Economy Trends: 1975 Through 2017 ", Rockville, MD: U.S. Environmental
Protection Agency2018.
[11] S. Keeler and M. Kimchi, Advanced high-strength steels application guidelines
V6: WorldAutoSteel, 2017.
[12] L. ULRICH. (2010, September 16, 2017). Speaking of Understatements.
[13] G. Davies, Materials for automobile bodies: Butterworth-Heinemann, 2012.
[14] S. Keeler and M. Kimchi, Advanced high-strength steels application guidelines
V5: WorldAutoSteel, 2015.
18
[15] C. Horvath, "Advanced steels for lightweight automotive structures," in Materials,
design and manufacturing for lightweight vehicles, ed: Elsevier, 2010, pp. 35-78.
[16] N. Fonstein, "Advanced High Strength Sheet Steels," Physical Metallurgy,
Design, Processing, and Properties, Springer, 2015.
[17] L. Meyer, F. Heisterkamp, and W. Mueschenborn, "Columbium, Titanium, and
Vanadium in normalized, thermo-mechanically treated and cold-rolled steels," Micro
Alloying, vol. 75, p. 130, 1975.
[18] S. Hayami and T. Furukawa, "Family of high-strength, cold-rolled steels," in
Proc. Conf. on Microalloying 75, 1977, 311-321, 1977.
[19] R. Kot, "Structure and properties of dual-phase steels," TMS/AIME, p. 362, 1979.
[20] J. Bucher, E. Hamburg, and J. Butler, "Property Characterization of VAN--QN
Dual-Phase Steels," Structure and Properties of Dual-Phase Steels, pp. 346-359, 1979.
[21] J. N. Hall, "Evolution of advanced high strength steels in automotive
applications," Presentation at Joint Policy Council, Auto/Steel Partnership, vol. 18, 2011.
[22] N. Fonstein, H. Jun, O. Yakubovsky, R. Song, and N. Pottore, "Evolution of
advanced high strength steels (AHSS) to meet automotive challenges," Vail, CO, USA.
Google Scholar, 2013.
[23] M. Takahashi, "Development of high strength steels for automobiles,"
Shinnittetsu Giho, pp. 2-6, 2003.
[24] S. Allain, J. Chateau, O. Bouaziz, M. Legros, and X. Garat, "Characterization of
the mechanical twinning microstructure in a high manganese content austenitic steel," in
19
Proceedings of the International Conference On TRIP Aided High Strength Ferrous
Alloys, Ghent, 2002, pp. 75-78.
[25] O. Kwon, K. Y. Lee, G. S. Kim, and K. G. Chin, "New trends in advanced high
strength steel developments for automotive application," in Materials Science Forum,
2010, pp. 136-141.
[26] E. De Moor, P. Gibbs, J. Speer, D. Matlock, and J. Schroth, "AIST Transactions
Strategies for Third-Generation Advanced High-Strength Steel Development," Iron &
steel technology, vol. 7, p. 132, 2010.
[27] G. Thomas, D. Matlock, R. Rana, L. Hector, and F. Abu-Farha, "Integrated
Computational Materials Engineering Lab Heat Results Supporting DOE Targets," Great
Designs in Steel, 2015.
[28] D. Matlock and J. G. Speer, "Design considerations for the next generation of
advanced high strength sheet steels," in Proceedings of 3rd International Conference on
Structural Steels, 2006, pp. 774-781.
[29] K.-i. Sugimoto, A. Kanda, R. Kikuchi, S.-i. Hashimoto, T. Kashima, and S. Ikeda,
"Ductility and formability of newly developed high strength low alloy TRIP-aided sheet
steels with annealed martensite matrix," ISIJ international, vol. 42, pp. 910-915, 2002.
[30] J. Speer, D. Matlock, B. De Cooman, and J. Schroth, "Carbon partitioning into
austenite after martensite transformation," Acta materialia, vol. 51, pp. 2611-2622, 2003.
[31] J. Speer, E. De Moor, and A. Clarke, "Critical assessment 7: quenching and
partitioning," Materials Science and Technology, vol. 31, pp. 3-9, 2015.
20
[32] H. Aydin, E. Essadiqi, I.-H. Jung, and S. Yue, "Development of 3rd generation
AHSS with medium Mn content alloying compositions," Materials Science and
Engineering: A, vol. 564, pp. 501-508, 2013.
[33] R. Rocha, T. Melo, E. Pereloma, and D. Santos, "Microstructural evolution at the
initial stages of continuous annealing of cold rolled dual-phase steel," Materials Science
and Engineering: A, vol. 391, pp. 296-304, 2005.
[34] G. Huppi, D. Matlock, and G. Krauss, "An evaluation of the importance of
epitaxial ferrite in dual-phase steel microstructures," Scripta Metallurgica, vol. 14, pp.
1239-1243, 1980.
[35] U. Liedl, S. Traint, and E. Werner, "An unexpected feature of the stress–strain
diagram of dual-phase steel," Computational Materials Science, vol. 25, pp. 122-128,
2002.
[36] M. T. Rahmaan, "Low to high strain rate characterization of DP600, TRIP780,
AA5182-O," University of Waterloo, 2015.
[37] M. Singh, A. Das, T. Venugopalan, K. Mukherjee, M. Walunj, T. Nanda, et al.,
"Impact of Martensite Spatial Distribution on Quasi-Static and Dynamic Deformation
Behavior of Dual-Phase Steel," Metallurgical and Materials Transactions A, vol. 49, pp.
463-475, 2018.
[38] W. D. Callister and D. G. Rethwisch, Materials science and engineering vol. 5:
John Wiley & Sons NY, 2011.
21
[39] S.-G. Hong and S.-B. Lee, "The tensile and low-cycle fatigue behavior of cold
worked 316L stainless steel: influence of dynamic strain aging," International Journal of
Fatigue, vol. 26, pp. 899-910, 2004.
[40] S. Allain, J.-P. Chateau, O. Bouaziz, S. Migot, and N. Guelton, "Correlations
between the calculated stacking fault energy and the plasticity mechanisms in Fe–Mn–C
alloys," Materials Science and Engineering: A, vol. 387, pp. 158-162, 2004.
[41] X. Peng, D. Zhu, Z. Hu, W. Yi, H. Liu, and M. Wang, "Stacking fault energy and
tensile deformation behavior of high-carbon twinning-induced plasticity steels: Effect of
Cu addition," Materials & Design, vol. 45, pp. 518-523, 2013.
[42] T. Shun, C. Wan, and J. Byrne, "A study of work hardening in austenitic Fe
Mn C and Fe Mn Al C alloys," Acta metallurgica et materialia, vol. 40, pp. 3407-
3412, 1992.
[43] J. Hu, "Characterization and Modeling of Deformation, Springback, and Failure
in Advanced High Strength Steels (AHSSs)," 2016.
[44] E. De Moor, J. Speer, D. Matlock, C. Föjer, and J. Penning, "Effect of Si, Al and
Mo alloying on tensile properties obtained by quenching and partitioning," MS&T 2009,
pp. 1554-1563, 2009.
[45] J. G. Speer, F. C. R. Assunção, D. K. Matlock, and D. V. Edmonds, "The"
quenching and partitioning" process: background and recent progress," Materials
Research, vol. 8, pp. 417-423, 2005.
22
[46] E. De Moor, D. K. Matlock, J. G. Speer, C. Fojer, and J. Penning, "Comparison of
Hole Expansion Properties of Quench & Partitioned, Quench & Tempered and
Austempered Steels," SAE Technical Paper 0148-7191, 2012.
[47] T. D. Bigg, D. V. Edmonds, and E. S. Eardley, "Real-time structural analysis of
quenching and partitioning (Q&P) in an experimental martensitic steel," Journal of
Alloys and Compounds, vol. 577, pp. S695-S698, 2013.
[48] B. C. De Cooman and J. G. Speer, "Quench and Partitioning Steel: A New AHSS
Concept for Automotive Anti‐Intrusion Applications," steel research international, vol.
77, pp. 634-640, 2006.
[49] 김동완, "A study on phase transformation and deformation behaviors considering
transformation plasticity in steels," 서울대학교 대학원, 2016.
[50] D. K. Matlock and J. G. Speer, "Third generation of AHSS: microstructure design
concepts," in Microstructure and texture in steels, ed: Springer, 2009, pp. 185-205.
[51] Z. Nishiyama, Martensitic transformation: Elsevier, 2012.
[52] H. Y. Yu, "Strain-hardening behaviors of TRIP-assisted steels during plastic
deformation," Materials Science and Engineering: A, vol. 479, pp. 333-338, 2008.
[53] I. Tamura, "Deformation-induced martensitic transformation and transformation-
induced plasticity in steels," Metal Science, vol. 16, pp. 245-253, 1982.
[54] M. Berrahmoune, S. Berveiller, K. Inal, A. Moulin, and E. Patoor, "Analysis of
the martensitic transformation at various scales in TRIP steel," Materials Science and
Engineering: A, vol. 378, pp. 304-307, 2004.
23
[55] Z. He, Y. He, Y. Ling, Q. Wu, Y. Gao, and L. Li, "Effect of strain rate on
deformation behavior of TRIP steels," Journal of Materials Processing Technology, vol.
212, pp. 2141-2147, 2012.
[56] K. S. Choi, X. Hu, X. Sun, M. Taylor, E. De Moor, J. Speer, et al., "Effects of
constituent properties on performance improvement of a quenching and partitioning
steel," SAE Technical Paper 0148-7191, 2014.
[57] Z. He, J. Ma, H. Wang, G. Tan, D. Shu, and J. Zheng, "Dynamic fracture behavior
of layered alumina ceramics characterized by a split Hopkinson bar," Materials Letters,
vol. 59, pp. 901-904, 2005.
[58] S. Dong, Y. Wang, and Y. Xia, "A finite element analysis for using Brazilian disk
in split Hopkinson pressure bar to investigate dynamic fracture behavior of brittle
polymer materials," Polymer testing, vol. 25, pp. 943-952, 2006.
[59] B. L. Boyce and M. F. Dilmore, "The dynamic tensile behavior of tough,
ultrahigh-strength steels at strain-rates from 0.0002s−1 to 200s−1," International Journal
of Impact Engineering, vol. 36, pp. 263-271, 2009.
[60] J. Bressanelli and A. Moskowitz, "Effects of strain rate, temperature, and
composition on tensile properties of metastable austenitic stainless steels," ASM Trans
Quart, vol. 59, pp. 223-239, 1966.
24
CHAPTER TWO
2. EFFECTS OF MICROSTRUCTURE ON THE STRAIN RATE SENSITIVITY
OF ADVANCED STEELS
ABSTRACT
The dependence of the strain rate sensitivity of advanced ~1GPa tensile strength
steels on the phases present in their microstructures was studied by testing different steels
at multi-strain rates, namely 0.005s-1, 0.5s-1, 5s-1, 50s-1, and 500s-1. The intermediate and
high strain rate tests were performed using a servo-hydraulic machine, while the quasi-
static tests were performed using a universal testing machine. The four main steels of
interest were the Ferrite-Martensite DP590 and DP980, the Ferrite-Austenite Medium
Manganese steel (10% wt.), and the Ferrite-Martensite-Austenite QP980; the latter being
a transformation induced plasticity (TRIP) assisted steel. For comparison, ferritic CR5
mild steel and austenitic stainless steel 201 were also tested under the same conditions.
Though the differences in the steel chemistries were not taken into account, the results
obtained here suggest a strong relationship between the phase-content of the steel and its
response to the changes in the loading rate. The relationships between the observed
mechanical behavior and the phases present in the microstructure are discussed.
2.1 Introduction
With the ever-increasing push for lightweighting in the automotive sector,
Advanced High Strength Steels (AHSSs) continue to infiltrate the industry due to their
unique properties and merits [1]. The uniqueness of AHSSs can be primarily attributed to
25
their complex multi-phase microstructures, as compared to conventional single-phase
steels, which enable achieving particular blends of strength and ductility. While strength
drives the lightweighting potential of the material, ductility defines the practical limits of
its stamping. However, beyond forming materials into the various automotive
components, materials need to pass specific tests and meet certain criteria to be used in
vehicles; some of the most critical ones are crashworthiness-related [2] Finite element
(FE) analyses are typically performed to simulate a crash event; the accuracy of these
simulations relies heavily on how well the constitutive equations describe the material
behavior.
Automotive materials undergo loading at different strain rates during their
lifetime. From quasi-static (< 0.1 s-1) to intermediate (around 1 s-1) during stamping, to
high rates (>100 s-1) during a crash event. It is well-established in the literature that the
deformation behavior and the properties of steels are highly influenced by strain rate.
Thus, the governing equations used in the various analyses should be able to incorporate
rate effects on the behavior of the material. In order for the equations to achieve that,
accurate experimental material data must be obtained at different strain rates.
Special test instrumentation must be used to reliably generate material data over
the nominal high strain rate range since screw-driven load frames commonly used for
measurement of material data at quasi-static rates (e.g. <0.5s-1) are not applicable. Servo-
hydraulic load frames, in which an actuator, accelerated to the desired speed, engages a
slack adaptor assembly to dynamically deform a test specimen in tension, are the most
commonly employed to test automotive body structure steels at high strain rates [3-6]. A
26
common upper testing limit on strain rate for servo-hydraulic test machines is 500s-1. The
split-Hopkinson bar [7-10], also known as the Kolsky bar [11, 12], is commonly applied
in dynamic material tests at strain rates in the vicinity of (or above) 1000s-1 (there are
some notable exceptions where tests have been conducted near 500s-1). Here, the
specimen is plastically deformed by an elastic wave pulse initiated by a striker through
the load train consisting of incident and transmission bars and the specimen. Unlike a
servo-hydraulic load frame, the test specimen is often not deformed to fracture in a split-
Hopkinson bar test. Another test used to measure the deformation response of automotive
materials at high strain rates is the instrumented falling weight impact tester [13, 14]. In
general, high rate tests are more involved and therefore often more costly than quasi-
static tests for a variety of reasons. Among the more significant of these are the
application of strain gauges to address load cell ringing; tensile specimen design to
eliminate non-uniform strain evolution during elongation; and test system artifacts that
produce spurious specimen bending during elongation. Load cell ringing, which is
common in servo-hydraulic units, appears as a spurious oscillatory signal superimposed
onto the “true” material response (e.g. load vs. time or load vs. displacement) [15]. This
effect results from stress wave propagation, the inertia of the load train components, and
the frequency response of the load cell. Ringing is most severe near the resonant
frequency of the load train (~9kHz for many servo-hydraulic testers) which obscures the
desired material response leading to inaccurate flow data. Most test frame manufacturers
provide the means for reducing ringing through data filtering or fitting routines.
However, uncertainties remain regarding the accuracy of the material data since filtering
27
is in and of itself an intentional distortion. Other techniques for reducing ringing, such as
tuning a test specimen geometry via load train modal analysis for a specific test
instrument, have also appeared in the literature [3, 16]. Strain gauges are commonly
applied to one of the test specimen grip regions. These essentially replace the signal from
the test unit load cell and provide voltage signals that can be converted to force assuming
that Young’s modulus of the material is known within reasonable accuracy and that a
uniform force is transmitted through the test specimen during loading. Strain gauge
measurement of plastic deformation in the test specimen gauge region is often
problematic and digital image correlation (DIC), a non-contact whole field optical strain
mapping technique using high-speed photography is now the preferred method for both
quasi-static and high strain rate testing [17, 18]. For example, issues with adhesion of the
strain gauge at large strains can arise, maximum measurable strains are dependent upon
the strain gauge design, and test artifacts such as specimen bending cannot be easily
revealed. At present, only two European standards for dynamic tensile testing are
available, viz., ISO 26203-1,2 [19] and SEP1230 [20].
Strain rate effects on the mechanical properties of steels have been studied by
many researchers over the past decade; examples include these references [5, 21-28].
However, the majority of these studies did not focus on the microstructure and the role it
plays in dictating the strain rate sensitivity of the steel grade. Few studies considered the
microstructure and its correlation to rate sensitivity, particularly in advanced steel grades.
Hwang et. al [29] studied the effects of ferrite grain size on the strain rate sensitivity in
DP steels. Oliver et al. [6] tested DP and TRIP steels at 0.001 s-1 and 200 s-1 to compare
28
the changes in the microstructure at different strain rates. Tarigopula et. al [17] studied
the effect of strain rate on the mechanical properties and strain localization of DP800. Li
et al. [30] studied the effects of strain rate on the deformation-induced transformation of
austenite in quenched and partitioned steels, up to strain rates of 0.1s-1.
This chapter investigates the response of selected advanced steels with complex
multiphase microstructures to external loading at multi-rates ranging from quasi-static to
rates close to those encountered during a crash event. The responses at the intermediate
and high rates are compared to that at quasi-static rates to evaluate the rate sensitivity of
each material. Moreover, the differences in their responses are correlated to the types of
constituent phases present in their microstructures. This would be the first step to develop
models that can incorporate rate effects in each of the various phases within the
microstructure of advanced high strength steels. The main four materials of interest are
the Ferrite-Martensite DP590 and DP980, the Ferrite-Austenite Medium Manganese steel
(10% wt.), and the Ferrite-Martensite-Austenite QP980. Nevertheless, the deformation of
two single-phase steels, ferritic CR5 mild steel and austenitic stainless steel 201, are also
investigated to provide some basis for comparison and establish the rate dependency of
the individual phases.
2.2 Experiments
2.2.1 Materials and Initial Microstructures
Table 2.1 summarizes the chemical compositions of each of the six materials
tested. The initial microstructures of the six materials were examined by optical
29
microscopy and scanning electron microscopy; the obtained micrographs are shown in
Figure 2.1. The CR5, a very ductile grade of mild steel which is most commonly used for
high formability applications including exterior panels and closures. Its microstructure
(Figure 2.1a) primarily consists of moderate-grained ferrite (~5-25 µm equivalent grain
size), and the removal of interstitials such as C and N from the ferritic matrix is achieved
via microalloying additions. The SS 201 is an austenitic stainless steel, the microstructure
of which exhibits a similar heterogeneous distribution of grains, though larger than the
CR5. The microstructure of this chromium-nickel-manganese alloy exhibits complex
deformation behavior [31]. The DP590 and DP980 microstructures, shown in Figures
2.1c and d, consist of martensite islands/particles dispersed in a matrix of ferrite grains
[32-34]. The area fractions of the two phases (both are very fine) were measured to be
~60% ferrite and ~40% martensite (DP590) and ~30% ferrite and ~70% martensite
(DP980). As part of an ongoing DOE project aimed at the development of an ICME
model for third generation advanced high strength steels (3GAHSS), an experimental
medium Mn. steel was produced through inter-critical annealing in the ferrite-austenite
region [44]. Austenite is stabilized through enrichment with Mn and C, and the initial
microstructure contains no martensite. Its initial microstructure, shown in Figure 2.1e,
consists of ferrite and austenite. Annealing twins appear on the austenite grains and are
suggestive of a low stacking fault energy. Plastic deformation and phase transformation
in tension are highly inhomogeneous and occur in propagating bands as confirmed by
synchrotron X-ray measurements [35]. The QP980 microstructure is a three-phased
TRIP steel subject to a two-step quenching and partitioning heat treatment [36]. Xiong et
30
al. [37] identified both film-like and blocky forms of carbon-enriched austenite in QP980.
The result is a microstructure composed of ~40% ferrite, ~50% martensite and ~10%
retained austenite. The ferrite and martensite grains are fine (similar to the DP980), while
the austenite islands are particularly small (~0.5-1 µm); an example is indicated in the
micrograph in Figure 2.1f. As noted, phase distinction in this material is challenging;
thus, the reported phase fractions were obtained from both optical microscopy and EBSD
maps (not presented here).
Table 2.1 Steel Chemistries (wt. %)
Steel C Mn Si Al Cr Ni Mo P N S Ti CR5
0.0018 0.14 0.005 0.038 0.029 0.035 0.005 0.008 0.006 0.007 0.077
SS 201 0.045 6.5 0.48 0.002 16.7 4.5 0.17 0.036 0.13 0.001 0.002
DP 590 0.06 1.5 0.32 0.023 0.78 0.039 0.10 0.010 0.009 0.002 0.003
DP980 0.09 2.3 0.65 0.040 0.022 0.007 0.11 0.007 0.004 0.001 0.029
Medium
Mn. 0.16 10.56 0.22 1.56 <0.002 - - 0.0023 0.0027 0.0016 -
QP980 0.20 1.79 1.52 0.039 0.03 - - 0.014 0.004 0.003 -
31
(a) (b)
(c) (d)
(e) (f)
Figure 2.1: Micrographs showing the microstructure of the as-received material for (a)
CR5 (b) SS 201 (c) DP590 and (d) DP980 (e) Medium Mn. (10wt%) (f) QP980.
32
2.2.2 Experimental Setups and Procedure
The six materials were tested in tension under both quasi-static, intermediate and
high rate conditions using the setups shown in Figure 2.2. The quasi-static tension tests
were performed on a universal electromechanical load frame, at a constant crosshead
speed of 18mm/min. A standard ASTM E8 specimen geometry was used, and the
corresponding strain rate of deformation was ~0.005s-1. For accurate strain
measurements, a 3D digital image correlation (DIC) system was used to monitor material
deformation during testing. High strain rate testing, on the other hand, was performed at
the General Motors R&D department using the servo-hydraulic machine shown in Figure
2.2b. The Instron VHS (very high speed) unit, shown in Figure 2.2b, is a two column
servo-hydraulic tester with static and dynamic load ratings of 80 kN and 50 kN,
respectively. The load frame has a 250 kN rating. Deformation of a test specimen occurs
via the motion of an actuator, which can achieve a maximum speed of 20m/s, connected
to the lower grip of the gripping mechanism or slack adaptor. The actuator retracts into
the bottom portion of the tester and is hidden from view in Figure 2.2b by the black base.
Actuator speed is controlled using oil pressure which is regulated using a remote pump
and servo-valves on the back side of the unit (not shown). The VHS unit runs in two
modes: a low-pressure mode (pressure = 185bar), which is used to move the actuator
manually when gripping or un-gripping a test specimen, and a high-pressure mode
(pressure = 210bar), which is needed to perform the high-speed testing. The upper grip is
connected to a Kistler 9071A piezoelectric load washer (400 kN rating, 5 MHz sampling)
which is connected to a movable yoke (shown in Figure 2.2c). Also shown in Figure 2.2c
33
are two high-speed digital cameras for strain measurement. These are attached to a
custom tripod and placed on the backside of the tester to capture images of a test
specimen as it is deformed.
As mentioned earlier, the lower grip of the VHS unit is connected to the slack
adaptor which is very crucial for the operation of the servo-hydraulic machine at high
speeds. A slack adaptor is a mechanism for gripping a test specimen that consists of a
stationary upper grip, a lower grip that is connected to the actuator via an aluminum
collar and damping washers. It allows the actuator to attain a set constant speed before
the load is applied to the specimen, and it can be used in both open (high speeds) and
closed-loop control (low speeds). The slack adaptor unit shown in Figure 2.2c is a
redesigned version of the original Instron unit which was found to be unsuitable for high
strain rate testing of AHSS. To minimize inertia and hence load train ringing, the grips
and slack rod were designed with Ti-6Al-4V. To further reduce load train ringing, two
polyurethanes and two aluminum impact washers were placed between the lower grip and
the aluminum collar that screws into the top the actuator (the lower grip screwing into the
collar). The aluminum collar was redesigned to enable a reciprocating fit for the slack rod
that runs through the collar into the actuator. The slack rod (also Ti-6Al-4V) is a
cylindrical rod with one end as a conical (i.e., tapered) section that is engaged by the
aluminum collar. The gripping of the specimens to the grips was done using smooth M2
tool steel pins (see Figure 2.2c) as opposed to the original design of the unit which
involved bolt/nut combinations which resulted in spurious twisting of a test specimen
(detected by DIC) since the specimen would ride up and down the bolt threads during
34
(a) (b)
(c)
Figure 2.2: Experimental setups and test specimens used for (a) quasi-static testing and
(b) high rate testing (Servo-hydraulic) (c) detailed view of the components of the Instron
VHS unit
35
loading. To reduce vibrations that result from the engagement of the slack adaptor rod
with the collar/actuator combination, a damping material was wrapped around the conical
section of the slack rod.
Strain measurement was conducted with a high-speed 3D DIC system. A
significant advantage of 3D DIC is that it provides full-field strain maps of the entire
gauge section while eliminating any load train compliance effects. This is especially
useful in quantifying the inhomogeneous deformation at high strain rates such as local
plastic deformation in propagative instabilities (e.g., these were noted at quasi-static rates
during tensile testing of the Med. Mn. (10wt.%) steel). The 3D DIC system also provides
a careful check on instrumentation artifacts (e.g., spurious bending and twisting of the
test specimen) during high strain rate deformations that otherwise go undetected with
other means. The 3D DIC system consisted of two Photron SA1.1 black and white
cameras (Figure 2.3), multi-channel data acquisition (DAQ) box for image capture and
data logging (strain gage voltages and load washer loads).
The Photron SA1.1 cameras have a maximum frame rate of 675,000 frames/s at a
64x16 pixel resolution. To capture the gauge area of the specimens until failure, a
minimum resolution of 512x512 pixels is needed which limits the maximum camera
frame rate to 20,000 frames/s. At this frame rate, the optimal exposure time is 1/95000 s.
If the longer exposure time is used, then the images will contain motion blur which
prevents strain measurement. This problem occurs when there is a significant motion of
the specimen between the period at which the camera aperture opens and closes.
36
The maximum strain rate required was 500s-1, which was achievable with the
maximum frame rate of 20,000 frames/s. The other nominal strain rates used in the tests
were: 0.5s-1, 5s-1, and 50s-1. These rates were deemed reliable for adequate sampling of
high strain rate behavior of the various steels tested. The captured images are then loaded
in the post-processing software to obtain a point-by-point mapping, on a user-chosen
grid, of displacement and strain field components as well as velocities and strain rates.
A significant challenge with high-speed testing using servo-hydraulic units is load
signal ringing due to vibrations in the load train. The piezoelectric load cell used in the
VHS Instron tester, Kistler 9071A (Fig. 3.2b), provides a ringing-free load signal up to
actuator speeds near 1.0m/s. Beyond this speed, ringing obscures the material behavior
Figure 2.3: Photron SA1.1 high-speed cameras
37
since the load signal swamps that from the actual material response, as shown in Figure
2.4a. To reduce ringing and make sure any minor specimen bending was eliminated, one
linear strain gauge was adhered to both sides of the two gripper ends on each specimen
tested at 500s-1 and specimens tested at 50s-1. Linear strain gauges, with a resistance of
350 Ω and gage factor of 2.04 were used since elastic strains are measured in the grip
region in one direction only. Each strain gauge can measure up to 5% strain. The two
terminals of each strain gauge are connected to a signal conditioning box. The signal
conditioner completes the quarter bridge circuit using in-built resistances and outputs the
voltage signal to the DAQ box. Figure 2.4b shows a flow curve in which ringing has been
substantially minimized.
(a) (b)
Figure 2.4: (a) Engineering stress-strain curve of a test ran at a speed of 20m/s showing a
dominant ringing signature and (b) Engineering stress-strain curve of a test run at a speed
of 20m/s showing the elimination of ringing
38
2.2.3 Specimen Geometry
An extensive search of the literature revealed a wide variety of different servo-
hydraulic test specimen geometries. Yan et al. [15] showed four examples used in their
round-robin study on dynamic testing while ten different geometries are shown in
Appendix IV of Ref. [38]. It was determined that the most important considerations for
the test specimen geometry are that it must take into account the gripping mechanism,
load train ringing [3], strain gauge application, and it must not promote non-uniform
strain evolution. A careful review of literature references [39], resulted in the two test
specimen geometries (Figure 2.5). Each is a dogbone geometry. A fundamental
difference between the two is the gauge length, i.e. one has a gauge length of 20mm
(Figure 2.5a), and the other has a gauge length of 10mm (Figure 2.5b). One gripper end
was designed to be longer than the other to accommodate linear strain gauges. Shorter
gauge lengths allow for higher average strain rates at lower actuator speeds with reduced
ringing. However, inferences about the effect of gauge section length of strain evolution
uniformity from ISO 26203-1,2 [19], SEP1230 [20], and the modal analysis of servo-
hydraulic load train ringing by Yang et al. [3] suggest that shorter gauge lengths can
adversely affect the strain field homogeneity in the gauge area.
The DIC methodology was used to examine the effect of the gauge length on the
strain homogeneity. Repeated tests of AHSS specimens that neck using both geometries
showed anomalous strain accumulation artifacts during plastic deformation. Figure 2.6a
provides the evolution of strains along the centerline of a 10mm gauge length specimen
which is distinctly non-uniform. Figure 2.6b shows uniform strain accumulation in a 20
39
mm gauge length specimen. While Figure 2.6c shows the strain countors after yielding.
Based upon the results in Figure 2.6 and the inferences regarding specimen selection in
ISO 26203-1,2 [19], SEP1230 [20], the specimen with 20 mm gauge length was selected
to perform the test with (Figure 2.5b).
(a)
(b)
Figure 2.5: High rate tensile specimen geometry with (a) 20mm gauge length (b) 10mm
gauge length
40
(a) (b) (c)
Figure 2.6: (a) Evolution of strain along the centerline (using a virtual extensometer) of a
10 mm gauge length specimen tested at 500s-1. Zero position denotes the center of gauge
section. (b) Evolution of strain along the centerline (using a virtual extensometer)
2.3 Results
2.3.1 Flow Behavior
Yield Strength, strain at uniform elongation, tensile strength, and strain at fracture
per steel at each strain rate are listed in Tables 2.2-2.5. Engineering stress-strain curves
obtained at the five strain rates are shown for all six materials in Figure 2.7. Part (a)
displays typical tensile flow curves generated for the CR5 steel. These show a monotonic
increase in flow stress, both yield, and tensile stress, as strain rate is increased. A
significant increase in the yield strength and the ultimate strength is noted for the higher
strain rate, accompanied by a reduction in the uniform and total elongation of the
material; this is indicative of strong strain rate sensitivity. This is typical behavior of mild
steels, and several researchers have reported the same behavior in previous studies [22,
41
26, 40]. The red curve at 500s-1 shows a single hump just beyond 0.1 engineering strain
which is unlikely to be related to ringing since it lacks the periodic waviness
characteristic of ringing. Instead, this results from a stiffness issue pertaining to the 0.75
mm initial thickness of the CR5 steel. Repeated tests of the CR5 at 500s-1 showed the
same behavior which for some tests was more extreme, and for others, it was less
extreme. This was the only material that exhibited this behavior even with dual strain
gauges that functioned as the load cell. A search through the literature revealed very little
information on dynamic tensile testing results for sheet steels with initial thicknesses at or
close to the 0.75 mm thickness of the CR5 material using a servo-hydraulic tester. In the
course of searching for alternatives to servo-hydraulic testing of a thin sheet, only two
notable references were located, and both describe custom high strain testing
instrumentation to deform metallic foils in tension. Specifically, Paul and Kimberly [41]
describe a desktop Kolsky bar apparatus for testing metallic foils (99% Magnesium, 0.2
mm thick) at strain rates up to 1000s-1. Also, Kwon et al. [42] developed a “High-Speed
Micro Material Testing Machine” and tested 0.096 mm thick AA1100 foils over the 1s-1 -
100s-1 range. As a check on the servo-hydraulic instrumentation and test methodology
detailed previously, the CR5 material was tested at 500s-1 on a Kolsky bar at NIST,
Gaithersburg. The NIST Kolsky bar is shown in Figure 2.8a, and the associated specimen
geometry is shown in Figure 2.8b. Here, the specimen is plastically deformed by an
elastic wave pulse initiated by a striker through the load train consisting of incident and
transmission bars and the test specimen. Mates and Abu-Farha [12] provide additional
details on the NIST Kolsky bar. Using strain fields computed from stereo DIC (see the
42
two cameras called out in Figure 2.8a), the 500s-1 flow curve (red) in Figure 2.8c does not
have the hump in the red curve at 500s-1 in Figure 2.7a. Repeated tests of the CR5
material with the Kolsky bar produced results essentially identical to the 500s-1 curve in
Figure 2.7a. An overlay of the 500s-1 curves in Figures 2.7a and 2.8c provided confidence
that a “fit” through the red curve at 500s-1 in Figure 2.7a, so as to eliminate the hump
from the servo-hydraulic test, is reasonable. Servo-hydraulic testing can, therefore, be
used in future tests with these thin sheet materials provided that any filtering or alteration
of the data is validated by other means, such as a Kolsky bar test.
Figure 2.7b shows flow curves for the 201LN (fully austenitic) stainless steel.
Here, there is a reduction in both strain hardening rate and tensile strength at the 0.5s-1
and 5s-1 strain rates relative to 0.005s-1 (see Table 2.3). This is presumably due to
reductions in the transformation hardening (TRIP) effect. Tensile strengths for the 201LN
steel, however, begin to increase with strain rate after an initial decrease and exceed that
of the quasi-static rate at the highest tested rates (see Table 2.4). Inspection of Figures
2.7a, b and Table 2.3 reveals that both the CR5 and 201LN steels exhibit significant
reductions in uniform elongation with increased strain rate. The significant reduction in
work hardening rates between quasi-static and high rate testing is likely an effect of the
reduction of strain hardening due to reduced austenite transformation at increased strain
rates in the 201LN steel. This is attributed to increased adiabatic heating as discussed by
Isakov et al. [43] but a definitive demonstration would require the addition of an infrared
camera that is coupled with the DIC image capture process. The increased yield strength
for CR5, which is most probably due to the dislocation slip in ferrite with little increase
43
in strain hardening rate, is in-line with behavior reported in the literature [44]. Note that
the 500s-1 (red) flow curve in Figure 2.7a shows a very modest ringing artifact (i.e., small
amplitude oscillations) which for all practical purposes can be ignored.
Table 2.2: 0.2% Offset Yield Strength (MPa) at the different strain rates.
Table 2.3: Engineering strain at uniform elongation at the different strain rates.
Table 2.4: Ultimate Tensile Strength (MPa) at the different strain rates.
Steel 0.2% Offset Yield Strength (MPa)
0.005s-1 0.5s-1 5s-1 50s-1 500s-1
CR5 160 226 265 307 322
201LN 320 370 395 416 580
DP590 Steel 365 380 417 447 480
DP980 715 763 776 832 900
Med. Mn TRIP 690 743 765 782 815
QP980 680 720 760 780 870
Steel Engineering strain at uniform elongation
0.005s-1 0.5s-1 5s-1 50s-1 500s-1
CR5 0.27 0.22 0.21 0.122 0.11
201LN 0.59 0.45 0.454 0.422 0.34
DP590 Steel 0.17 0.15 0.15 0.154 0.19
DP980 0.09 0.08 0.077 0.082 0.097
Med. Mn TRIP 0.33 0.30 0.31 0.29 0.237
QP980 0.111 0.113 0.114 0.11 0.136
Steel Ultimate Tensile Strength (MPa)
0.005s-1 0.5s-1 5s-1 50s-1 500s-1
CR5 287 311 335 372 440
201LN 743 700 740 755 870
DP590 Steel 630 640 680 714 800
DP980 1052 1080 1090 1110 1158
Med. Mn TRIP 1115 821 826 867 967
QP980 1017 1020 1050 1065 1130
44
Table 2.5: Engineering Strain at Fracture at the different strain rates.
(a) (b)
Steel Engineering Strain at Fracture
0.005s-1 0.5s-1 5s-1 50s-1 500s-1
CR5 0.495 0.41 0.433 0.48 0.33
201LN 0.63 0.59 0.623 0.63 0.59
DP590 Steel 0.26 0.28 0.292 0.315 0.34
DP980 0.143 0.147 0.17 0.188 0.211
Med. Mn TRIP 0.37 0.39 0.40 0.407 0.44
QP980 0.158 0.185 0.20 0.23 0.245
45
(c) (d)
(e) (f)
Figure 2.7: Engineering stress/strain curves at 0.005s-1 and 500s-1 for (a) CR5 (b) SS 201 (c)
DP590 (d) DP980 (e) Medium Mn. (10wt%) and (d) QP980.
46
(a) (b)
(c)
Figure 2.8: (a) Kolsky bar at NIST (b) Specimen Geometry used for Kolsky testing at
NIST (c) Engineering stress-strain curves with the 500s-1 obtained from the Kolsky
testing.
47
A different behavior is observed in the dual-phase steels; the DP590 and DP980
(Figure 2.7c and d); overall, there is a monotonic increase in flow stress, both yield, and
tensile stress, as strain rate is increased. Total elongation is notably larger at high strain
rates in both cases; the change in the uniform elongation, however, is not as notable and
stays almost the same for both materials. The stress-strain curves do not show the change
in shape observed in CR5 and SS 201; i.e., a significant drop in strain hardenability at
higher rates. On the contrary, the DP980, in particular, seems to strain to harden more
during high rate deformation, while DP590 does not show any notable change between
the different strain rates.
Flow curves in Figure 2.7e for the Medium Mn (10 wt. %) TRIP steel, like the
201LN steel, also exhibit significant differences when the strain rate is increased from the
quasi-static rate. Here, the hardening rate associated with quasi-static deformation is
significantly higher than that at the dynamic strain rates. As strain rate is increased, the
ultimate tensile strength of the Medium Mn (10 wt. %) TRIP steel decreases significantly
between 0.005s-1 and 0.5s-1 and 5s-1 (see Table 2.4). Yield point elongation is clearly
indicated in flow curves at all strain rates except that for 500s-1. An issue that was noted
pertained to results with and without dual strain gauges for strain rates ≥50s-1. If strain
gauges were not used, ringing with a frequency close to that of the resonant frequency of
the load train (~9.5 kHz) was observed only during yield point elongation at the dynamic
rates. This effect vanished when strain gauges were applied to the Medium Mn (10 wt.
%) TRIP test specimens. It is currently thought that the ringing effect within the yield
point elongation regime (which is minimal to nonexistent during subsequent work
48
hardening) results from the reduced damping of the tensile specimen during lateral
deflection of the tensile specimen during high rate testing. That is, the lack of stress
increase with specimen deflection (strain) during yield point elongation results in reduced
specimen damping and notable ringing at the resonant frequency of the load train. Again,
an expansive look through the literature on dynamic materials testing revealed no
dynamic tensile testing results for sheet steels (or any other metal alloys) that exhibit a
protracted yield point elongation up to 500s-1. Plastic instability, denoted by a single
Lüders band during yield point elongation [45] followed by one or more bands that
propagate along the gauge length during hardening, was observed at all strain rates at
which the Medium Mn (10 wt. %) TRIP steel was tested. However, the number of bands
and band propagation speed decreased significantly with increasing strain rate as
indicated above for the 500s-1 rate. The propagative instabilities resembled Type A
Portevin-Le Châtelier bands that are associated with dynamic strain aging and a negative
strain rate sensitivity of the flow stress. This behavior has been observed in other AHSS
such as fully austenitic TWIP steels [46]. Highly inhomogeneous deformation in the
Medium Mn (10 wt. %) TRIP steel has been associated with austenite transformation
localized to the propagating bands [47] using synchrotron X-ray diffraction. Not only is
plastic deformation inhomogeneous, but austenite transformation is inhomogeneous as
well.
Figure 2.7f shows the flow curves for the triple-phased QP980. The response of
the flow curves to strain rate in QP980 is similar to that of DP980. The notable difference
is the drop in the strain hardenability at higher strain rates of QP980. This can be mainly
49
attributed to the presence of austenite in the microstructure and the effect higher strain
rates has on its deformation-induced transformation.
2.3.2 Strain Rate Sensitivity
Figure 2.9a shows the yield strength variation with the strain rate for all the
materials tested. An approximately equal and logarithmic increase in yield strength is
shown by CR5 and Medium Mn (10 wt. %) in Figure 2.9 as measured by the slope of the
linear plots. That is, the relationship between yield strength and logarithm of the strain
rate is linear and of equal slope. The same is not true for other materials; the three other
materials show a variation in the slope of the yield strength increase. 201LN for instance
exhibit an increase in the slope at 500s-1, while DP590 exhibits an increase in the slope at
0.5s-1 and finally DP980 shows the increase of slope at 50s-1.
The same trend of variation in yield strength is observed in all the materials
except 201LN and Medium Mn. (10 wt. %) steels, in other words, materials without any
significant presence of austenite in the microstructure (Figure 2.9b). For 201LN and
Medium Mn. (10 wt. %) steels, there is an initial drop in the ultimate tensile strength
when the strain rate increases from 0.005s-1 to 0.5s-1 and then the tensile strength
increases with strain rate. The initial drop in the tensile strength is significantly larger for
Medium Mn. (10 wt. %) steel as compared to 201LN and the reason for the drop and the
variation between the two materials is discussed in section 2.4.
A differential (or instantaneous) strain hardening coefficient, pn n , which is
independent of any hardening law, and can be defined by [48]:
50
ln
(ln )p
dn
d
(2.1)
where is true plastic stress and p is a true plastic strain. n can be computed from a
true stress-true strain curve using the following algorithm
5 5 4 4 3 3 2 2 1 1
5 5 4 4 3 3 2 2 1 1
(ln ln ) (ln ln ) (ln ln ) (ln ln )(ln ln )
(ln ln ) (ln ln ) (ln ln ) (ln ln ) (ln ln )
i i i i i i i i i ii
i i i i i i i i i i
n
(2.2)
Note that there are other approaches to computing ( )pn n in the literature [ref].
Among these, the method proposed by Krauss and Matlock [49] is of interest:
( )p
p
dn
d
(2.3)
where d
d
is the true strain hardening rate. Equation 2.3 is derived from the Holloman
equation:
n
pK (2.4)
The predicted n from both approaches are essentially identical.
Equation 2.2 was used to calculate the strain hardening rate for each material at
all strain rates, and the results are plotted in Figure 2.10. The strain hardening exponent
was calculated over the 5-10% true strain range and plotted. A decrease in strain
hardening index is observed with increased strain rate in Medium Mn. (10 wt. %) steel,
201LN, and CR5 steels. This reduction is especially pronounced in the Medium Mn (10
wt. %) steel for strain rates > 0.005s-1, and this is again attributed to the effect of strain
rate on the austenite in the microstructure. The DP steels, in contrast, display minimal
51
sensitivity to the strain rate and this insensitivity is likely attributable to the large
population of mobile dislocations in ferrite before the onset of deformation during testing
and to the significant presence of martensite in the microstructure. As for QP980, the
strain hardening exponent remains almost constant up until 5s-1, and then the strain
hardening exponent increases, this behavior will be discussed in section 2.4.
Energy absorbed until the fracture is critical to the performance of the material
during a vehicle crash incident, and this property can be represented by toughness.
Toughness can be calculated by calculating the area under the stress-strain curve. This
means that the increase in strength with strain rates does not necessarily mean that the
material can absorb more energy during a crash, both strength and ductility must be taken
into consideration. To take a closer quantitative look at the effect of strain rate on the
energy absorbing capabilities of the six materials, the percentage change in four
mechanical properties between 0.005 s-1 and 500 s-1 was calculated and plotted in Figure
2.11. Part (a) shows how strain rate affects the yield strength in the materials. As
expected, the largest increase in the yield strength was observed in CR5, which showed
an increase of ~105%, followed by SS 201, which showed an increase of ~79%. As for
the multi-phased materials, the increase is not that significant, starting with DP590 which
showed an increase of 28% while its counterpart DP980 showed only a 2% increase.
While Medium Mn. (10 wt. %) and QP980 showed 14% and 9% respectively. The same
trend was almost observed when comparing the change of tensile strength in Figure
2.11b, but the values were different. The two materials that don’t follow the same trend
52
(a)
(b)
Figure 2.9: (a) Yield strength (MPa) vs log strain rate (b) Tensile strength (MPa) vs log
strain rate
53
Figure 2.10: n-value vs log strain rate
was SS201 and Medium Mn. (10 wt. %). Both showed a significant drop in the
percentage change of the tensile strength when comparing them to the percentage change
in the yield strength. Medium Mn. (10 wt. %) even showed a drop in its tensile strength.
This apparent change in the trend is due to the austenite phase in the microstructure that
is hugely influenced by the strain rate that affects the strain hardening behavior and
subsequently the tensile strength. This will be discussed further for SS201 in the next
section and Medium Mn. (10 wt. %) in the next chapter. However; for the other
materials, the strain hardening is not significantly changed at different strain rates, and
thus, the percentage change in the tensile strength is similar to that of yield strength. As a
result, the percentage increase in ultimate tensile strength for CR5 was ~57%, and that of
54
SS 201 was ~21%. For DP980 and QP980, strain hardenability was not significantly
affected at high rates, confirmed by the similar percentage increase of both the ultimate
tensile strength and the yield strength for the two materials (~7%).
Total elongation is an indication of material ductility and has a strong influence
on the energy absorption capabilities of the materials. As in the case of tensile strength,
the phases in the microstructure affect the response of the material to changes in strain
rate. Figure 2.11c shows the percentage change in total elongation for the six materials.
The total elongation of both CR5 and SS 201 decreased at high rates, agreeing with the
classic theory of strain rate sensitivity, which dictates that ductility decrease with the
increase in strain rate (for materials with a positive strain rate sensitivity). DP590 and
DP980 both showed an increase in the total elongation, more significantly with DP980
(due to its higher martensite content, as discussed in the next section). Medium Mn. (10
wt. %) steel and QP980 also showed an increase but to a very different magnitude.
Medium Mn. (10 wt. %) showed an increase of only ~19%, while QP980 showed the
most significant percentage increase of all materials at ~56%. The results obtained from
the multi-phase materials do not conform to the definition of positive or negative strain
rate sensitivity, as those materials showed a disconnect between the effect of strain on the
strength and ductility, and some of them even showed a disconnect between the response
of yield and ultimate strength to different strain rates.
Finally, the combined effects of all these properties will dictate the toughness of
the materials at different strain rate and therefore, their energy absorbing capabilities at
different strain rates. Figure 2.11d compares the effect of strain rate on the toughness of
55
the four materials. As noted, the percentage increase of toughness for DP590 was the
highest (~69%), which can be easily predicted due to the increase in the strengths (both
yield and tensile) and ductility observed in the material. Followed by both DP980 and
QP980 with ~28% and ~30% respectively, this would not be very easy to predict if you
only look at the stress-strain curves as it shows that the material is barely affected by the
change in strain rates. Then the austenite-based materials follow with almost identical
percentages increase of ~20%, and finally, CR5 showed a negligible increase of only
~5%. In summary, it is clear here that the six materials (driven by their different
microstructures) responded differently to increasing strain rate, leading to a net change in
toughness observed in Figure 2.11d. Total ductility reduction in CR5 at high rates
counter-acted the increase in the strength, leading to a small increase in toughness.
Toughness increase in SS 201 was strength-driven, since total ductility remained
unchanged, while toughness increases in DP980, Medium Mn. (10 wt. %) and QP980
was primarily ductility-driven. Moreover, of course, DP590’s increase in toughness was
both strength and ductility driven.
56
(a) (b)
(c) (d)
Figure 2.11: Percent Change between 0.005s-1 and 500s-1 of (a) Yield Stress (b) Tensile
Strength (c) Total Elongation and (d) Toughness.
57
2.4 Discussion
All the results above suggest that the phases present in the microstructure of steel
greatly affects its response to increased loading rates. Based on Orowan’s principle [50],
the strain rate is directly proportional to dislocation velocity and other dislocation
parameters. When materials are loaded at high rates, the density and the speed of
dislocations are higher. The relationship between dislocation velocity and the critical
resolved shear stress have been studied extensively in the literature, starting with the
work of Gilman and Johnston [51, 52] and followed by many others; such as the work of
Parameswaran et al. on aluminum [53]. Different relations were derived, but all assume
the speed of dislocation movement is directly proportional to the critical resolved shear
stress. Therefore, when materials are loaded at higher strain rates, higher stress is needed
to overcome the critical resolved shear stress to move the dislocations at the speed
required for the strain rate. This, in turn, implies higher yield and flow stresses during
deformation at high rates, and hence reduced ductility.
The above coincides with the behavior of CR5 at high rates, where the ultimate
tensile strength increased significantly; this is mainly due to the ferrite phase that
constitutes 100% of the microstructure of CR5. However, when looking at the results of
the other materials, it is noticeable that the strain rate sensitivity showed different trends.
To explain this, it is important to look closely at what happens during high strain rate
testing. During any kind of a mechanical test, where work is used to produce deformation
on the materials, heat will be generated. Under quasi-static conditions, test duration is
long giving enough time for the heat generated during the test to be dissipated. Thus a
58
small increase in temperature is observed; it is therefore well established in the literature
that isothermal conditions take place during quasi-static testing. This is not the case
during high rate testing where the test occurs in a very short period; the 500s-1 tests
carried out in this study were completed within 0.4ms. In this short period, the generated
heat does not have enough time to be dissipated, and thus the specimen’s temperature
increases significantly, changing the conditions from isothermal to adiabatic. For CR5
mild steel, which consists of soft and ductile ferrite only, the effect of temperature rise
softening is not as significant as the work hardening due to the higher loading rate. Thus
the net effect is a significant increase in strength at higher strain rates. Moving to the
dual-phase steels, which contain both ferrite and martensite in their microstructure which
has different properties and responses to different strain rates. Martensite is a hard and a
brittle phase, and high temperature (at high strain rates) causes significant softening in the
behavior of the material. This leads to two opposing effects taking place when testing
DP590 and DP980 at high strain rates: work hardening which occurs primarily in the
ferrite; and softening due to temperature rise, which occurs primarily in the martensite.
The latter was reported in the work of Hwang et al. [29], which showed a slight decrease
in strength and an appreciable increase in ductility in a Martensitic steel as the result of
an increase in deformation rate. These opposing effects cause the reduction in the
strength increase observed with the dual-phase steels as opposed to CR5. The strength
increase reduced from 58% (CR5) to 30% (DP590) to 6% (DP980). This also confirms
the effect of martensite as the percentage of martensite increase from 0% to 40% to 70%
(see Figure 2.12). Regarding ductility, a similar but opposite trend can be observed, the
59
higher the martensite content, the higher is the increase of ductility with strain rate. This
is because the drop in ductility of ferrite is outbalanced by the increase in ductility in the
martensite.
Another point to add is that the post-necking elongation portion increases (this is
the case with all six steels regardless of the microstructure). Note the post-necking
elongation of ~12-20% strains in all six materials at 500 s-1 is significantly higher than
the quasi-static post-necking strains. This is believed to be primarily driven by the
localized temperature increase, shifting the necking from localized to diffused, and thus
prolonging the deformation.
Figure 2.12: Effect of martensite on the percentage increase of tensile strength
Moving on to the third main phase in steels and the most complex of all; austenite.
Austenite is also a ductile phase like CR5, but it has a unique property that causes its
60
behavior and modeling to be unique and challenging. That unique property is that it
undergoes a transformation to the martensite phase when deformed. The stability of the
austenite phase dictates the rate of formation and the amount of martensite formed during
deformation. The stability mainly depends on the steel’s composition, the rate of
deformation and temperature, as well as the amount of plastic strain, stress state during
deformation, grain size and its morphology [54]. Of the prime three factors, the effects of
composition and temperature on the rate of transformation have been studied extensively
in the literature [55, 56]. It is well-documented, based on many studies on 300 series
austenitic stainless steels (SS300), that temperature rise due to deformation heating
reduces the rate of transformation, thus giving the effect of stabilizing the austenite during
deformation, yet the degree of stabilizing depends on the composition of the material
[57]. The third factor and the factor most relevant to automotive applications and this
study is the strain rate. The changes in the transformation kinetics at high strain rates can
be related to temperature. As mentioned earlier, high deformation rates are considered to
occur under adiabatic conditions, as opposed to isothermal conditions during quasi-static
deformation; the significant increase in temperature via adiabatic heating affects the
kinetics of phase transformation [58]. The latter was studied by some researchers such as
Talonen et al. [57], Lichtenfeld et al. [54] and Isakov et al. [43] who showed that
austenite-to-martensite transformation in SS3xx was generally inhibited at high strain
rates, leading to lower ductility.
Studying austenite-to-martensite transformation kinetics requires accurate
measurements of phase volume fractions and their changes in response to plastic strain.
61
Of the different methods used for such measurements, the most common is magnetic
induction, which takes advantage of the ferromagnetic properties of α’ martensite to
determine its volume content within a specimen [59]. X-ray diffraction (XRD), which
takes advantage of the fact that α’ martensite and austenite have different crystal
structures, is another common method for measuring the austenite/martensite volume
fraction. The intensity of the measured diffraction peaks is proportional to the volume
fraction present. Other methods of phase fraction measurements include neutron
diffraction and electron backscatter diffraction (EBSD); while the former is probably the
least used due to availability or accessibility limitations and the fact that it is relatively
slow, the latter is two dimensional (area fraction measurement technique) and lacks the
statistical representation of the material being analyzed. Comparisons between the
different measurement methods can be found in this reference [60].
Though the majority of efforts in the literature have focused on 300 series
stainless steels and reached similar conclusions regarding the adiabatic heating effects on
inhibiting transformation at high deformation rates, in almost all cases, different alloys
were affected by different degrees, and thus their high rate tensile behaviors and
transformation kinetics were different. To our knowledge, there have been no reports in
the literature on the high rate deformation and transformation kinetics in 200 series
stainless steels (SS2xx).
Beginning with the single-phased austenitic stainless steel 201LN, positive strain-
rate sensitivity is clearly inferred when looking at the yield strength; however, the quasi-
static tensile strength is higher than that observed at the intermediate rate. This is a direct
62
result of the significant reduction in strain-hardening rate (strain hardening exponent n)
with increasing strain rate. Total elongation, on the other hand, is noted to decrease as the
strain rate increases, while there is hardly any change in the total elongation with strain
rate. While the Orowan’s principle mentioned earlier explains the general systematic
increase in yield strength with loading rate, it does not explain the drop in hardening
behavior, and as a result, the unique trend of changes observed in the tensile strength. For
that, a closer look at the phase fraction changes with plastic strain is needed.
The two methods for measuring the volume fraction of martensite produced
similar trends, as shown in the two parts of Figure 2.13, but different values, especially
for the quasi-static condition. It is out of the scope of this paper to discuss the differences
and relative accuracies of the two methods; previous researchers [61, 62] concluded that
there is significant variability in the measurements when using different methods, with no
robust way of identifying which method is the best. The results in Figure 2.13 indicate a
strong strain rate effect on the transformation of austenite to martensite in SS201. The
volume fraction of martensite formed at quasi-static rates at around 50 % strain was
measured to be 13 % using magnetic induction and 24 % using the XRD. These values
dropped significantly for the higher strain rates with no significant difference between the
three strain rates (1, 10 and 500 s-1). This behavior is different from what has been noted
in the literature for SS3xx alloys, which have been studied extensively at different rates.
Though increasing the strain rate has been reported to slow down the austenite-to-
martensite transformation in SS3xx alloys, the transformation is not entirely inhibited; in
fact, over 50% transformation is reported at ~200 s-1 [54, 57]. In the SS201 studied here,
63
the transformation is almost entirely inhibited at low-intermediate rates (as low as 1 s-1).
This can be related to alloy content; the SS201 has a lower nickel content and higher
manganese content compared to SS301, and this was shown to affect the stability of the
austenite phase during quasi-static deformation in the work of Tavares et al. 2009 [63].
The results shown here when compared to the results shown in the literature for SS3xx,
lead to the same conclusion for high strain rates; that is the greater stability of austenite in
SS2xx compared to that in SS3xx. For instance, the plots of martensite formation in [57]
and [54] show a gradual drop at different strain rates, and not as abrupt as the ones sown
here in Figure 2.13.
Moreover, while more than 80% of the austenite transforms in SS3xx, a smaller
fraction of the austenite transforms into martensite in SS201 (up to ~25% at quasi-static
conditions). These disparities in the transformation behavior are responsible for the
significant difference in strain rate sensitivity of the two grades; rate effects on the
stress/strain curves of SS201 (Figure 2.7b) are strong compared to those reported in the
literature for SS3xx [54, 57]. The less stable austenite in the SS3xx results in the
formation of larger fractions of martensite; the latter is known to have low/negative rate
sensitivity, which results in the lower impact on the stress/strain behavior [64].
For QP980, the small increase in strength at higher strain rates is similar to that
observed in DP980, which can be attributed to the similarities in the microstructure
regarding the presence of martensite and ferrite. The slightly lower martensite and
slightly higher ferrite volume fractions in QP980 (compared to DP980) results in the
64
(a) (b)
Figure 2.13: Volume fraction of martensite measured at different plastic using: (a)
magnetic induction (b) XRD.
slightly higher rate sensitivity (Figure 2.11a and b). However, regarding ductility (Figure
2.11c), QP980 showed an increase of ~55% in total elongation at higher strain rates,
which is higher than that exhibited by DP980 and by all of the other materials. Since this
cannot be driven by the ferrite/martensite portion of the microstructure, it is believed to
be associated with the austenite present in QP980. However, the effects and exact
mechanisms are not apparent. While austenite is known to exhibit deformation-induced
transformation into martensite, causing additional resistance to necking and thus
enhanced tensile ductility and strain hardening, it was shown earlier that in SS201 the
austenite-to-martensite transformation is inhibited at high rates since temperature
increase reduces the chemistry driving force and increases the stacking fault energy,
which has the combined effect of inhibiting the transformation. So, this does not agree
nor explain why QP980 exhibits improved total elongation, on the level of ~55%, at high
65
strain rates. The effects of high loading rates on the austenite-to-martensite
transformation in TRIP-assisted steels (including QP and Medium Mn. steels) is not fully
understood. While some researchers claim that transformation should be inhibited at high
rates (based on the results in fully austenitic stainless steels), others indicate the opposite.
For instance, Liu et al. [65] showed that at strain rates higher than 100s-1, QP980 shows
increased ductility (which is similar to the results shown here in Figure 2.11c). Liu et al.
also measured the austenite content in the material after deformation and showed that the
transformation occurs at higher rates when the material is tested at high strain rates.
Current understanding based on data presented in this and a previous study [66] suggests
that the increased ductility with strain rates associated with QP980 and Medium Mn. (10
wt. %) are a result of combined and potentially counteracting effects. Thermal
stabilization of austenite to transformation via adiabatic heating may delay transformation
to greater strains, thereby extending the onset of necking. Additionally, modified
dislocation slip (increasingly planar with increased strain rate and increasingly non-planar
with increased temperature) may compete concerning the creation of dislocation
interactions necessary to induce austenite transformation to martensite.
2.5 Summary and Conclusions
The effects of the phase content on the strain rate sensitivity of six materials with
different microstructures were investigated. The materials were: CR5 (100% ferrite),
SS201 (100% austenite), DP590 (60% ferrite, 40% martensite), DP980 (30% ferrite, 70%
martensite), Medium Mn. (10 wt. %) Steel (34% ferrite, 66% austenite) and QP980 (40%
66
ferrite, 50% martensite, 10% austenite). The materials were deformed at 0.005, 0.5, 5, 50
and 500 s-1, and their mechanical response was examined. The ferrite phase was shown to
have a strong “classical” positive strain rate sensitivity, exhibited by an increase of
strength and a decrease of ductility with strain rate. When introducing martensite in the
mixture, the behavior is changed. This is because martensite in its nature is a hard and a
brittle material and has good sensitivity to the high-temperature conditions of high strain
rate testing. The softening effect of temperature on martensite counteract the effect of
ferrite and causes the increase in strength experienced by DP590 to be lower than that of
CR5 and caused the ductility of DP590 to increase with an increase in strain rate. When
increasing the martensite volume fraction in the microstructure as is the with DP980, it
magnifies the effects that were observed in DP590, and the material showed almost no
rate sensitivity regarding strength and a significant increase of ductility at high strain
rates. The third phase that was studied was the austenite, which was present in three
materials: 1) single-phased SS201, 2) double-phased Medium Mn. (10 wt. %), and 3)
triple-phase QP980. Austenite is known to exhibit a deformation-induced transformation
to martensite (a harder phase), causing significant strain hardening in the plastic phase of
deformation. The effect of strain rate on this phenomenon is what mainly dictates the
behavior of austenite-containing materials. The effect of strain rate on this phenomenon
was measured by two techniques. The results show that the volume fraction of formed
martensite indicate that austenite-to-martensite transformation is only evident at quasi-
static rates (to a maximum of ~25%), and it is almost entirely inhibited at rates above
~0.5s-1. The overall impact of this on the tensile behavior is a completely diminished
67
strain hardening behavior at high strain rates, leading to a reduction in the increase of
tensile strength of the material as compared to CR5. Also, this contributed to the
reduction of the total elongation exhibited by this material at high strain rates. QP980
had a similar martensitic content as DP980 which explains the negligible rate sensitivity
regarding strength shown by both materials; however, QP980 showed notably enhanced
ductility, which cannot be explained based on the known responses of the individual
phases. Even austenite, the highly ductile phase, is known to exhibit a drop in uniform
ductility at high rates due to the inhibition of austenite-to-martensite transformation,
which was also confirmed by the results obtained here for the SS 201. Further
investigation of austenite-to-martensite transformation kinetics and its rate dependence
from the literature and a separate study helped us to understand that is most likely the
result of two counteracting phenomena: the decrease in the transformation rate of the rate
due to adiabatic heating at high strain rates delays the transformation to higher strains
causing an extension to the necking, on the other hand, dislocation slips are affected by
high strain rates and can affect the creation of dislocation interactions that are necessary
to induce austenite transformation to martensite. The combined effect of all these
properties leads to a variation in the property that is most important to the automotive
industry and plays a crucial role in deciding whether this material is deemed viable. The
change of the toughness of these materials with strain rate highlights the importance of
looking at the combined effect of these properties and just the individual effect.
Toughness in both DP980 and QP980 was shown to improve significantly at rates close
to crash conditions, which compared favorably to the conventional CR5 mild steel that
68
showed no increase in toughness. When in fact CR5 showed an increase of 105% in yield
strength as opposed to a negligible increase in the yield strength of DP980 and QP980.
As was mentioned in the introduction, the automotive sector is looking for new
alloys of steel that have the advantage in performing over the current materials that are
being used and thus can lead to a reduction in the weight of the vehicles. The state-of-the-
art research in the steel alloys is AHSSs; which utilize a complex microstructure that can
produce superior properties compared to the conventional steels. In this chapter, four
AHSSs were shown and tested with a range of combinations of phases in the
microstructure, all have shown superior properties when comparing them to the
conventional steel (CR5) which is used heavily in the automotive industry. Then the
question becomes which of these materials must be used, what is the criteria for deciding
the choice of the material, and ultimately is there a way to reverse engineer the
microstructure that can provide the optimum properties for a material to achieve
lightweighting. As for the criteria, there are multiple aspects to keep in mind, such as
cost, formability, corrosion resistance, energy absorption… In this dissertation, our
primary focus was on the performance of these materials when loaded at high strain rates
(i.e., during a crash event) and the criteria we took into account was the energy
absorption capability. We have shown that these new materials do not follow most
classical theorems and it is of the utmost importance to be able to obtain high quality,
accurate experimental data at multiple strain rates to be able to judge the materials, as
some of these materials’ performance enhance at high strain rates while others did not.
We have also shown that the microstructure has a significant influence on the behavior of
69
these materials and their strain rate sensitivity and it is a troublesome task to accurately
quantitively predict how the addition of a new phase or the removal of an already existent
phase would change the properties. Martensite, in particular, showed very promising and
improved energy absorption capabilities despite its negative strain rate sensitivity
regarding strength. Austenite, on the other hand, is a very complex material to predict and
its influence on energy absorption at high rates was not as promising as martensite. Yet, it
is still a desirable phase due to its unique transformation induced plasticity (TRIP) effect
that can substantially improve its ductility and thus its performance. Yet, high strain rates
seem to be working oppositely regarding the TRIP effect and thus more work is required
to fully understand and predict the effects strain rate has on the TRIP effect in austenite.
In summary, the goal of this chapter was to provide a comprehensive overview of
the effect of the three main phases in AHSSs (ferrite, martensite and austenite) on the
performance of the material at high strain rates and provide a basis and a starting point to
allow for designing the microstructure of the steel to have specific properties.
2.6 References
[1] S. Keeler and M. Kimchi, Advanced High-Strength Steels Application Guidelines
V5: WorldAutoSteel, 2015.
[2] S. Curtze, V. T. Kuokkala, M. Hokka, and P. Peura, "Deformation behavior of
TRIP and DP steels in tension at different temperatures over a wide range of
strain rates," Materials Science and Engineering: A, vol. 507, pp. 124-131, 2009.
70
[3] X. Yang, L. G. Hector, and J. Wang, "A Combined Theoretical/Experimental
Approach for Reducing Ringing Artifacts in Low Dynamic Testing with Servo-
hydraulic Load Frames," Experimental Mechanics, vol. 54, pp. 775-789, 2014.
[4] I. Choi, D. Kim, S. Kim, D. Bruce, D. Matlock, and J. Speer, "The effect of
retained austenite stability on high speed deformation behavior of TRIP steels,"
Metals and Materials International, vol. 12, pp. 13-19, 2006.
[5] H. Huh, S.-B. Kim, J.-H. Song, and J.-H. Lim, "Dynamic tensile characteristics of
TRIP-type and DP-type steel sheets for an auto-body," International Journal of
Mechanical Sciences, vol. 50, pp. 918-931, 2008.
[6] S. Oliver, T. B. Jones, and G. Fourlaris, "Dual phase versus TRIP strip steels:
Microstructural changes as a consequence of quasi-static and dynamic tensile
testing," Materials Characterization, vol. 58, pp. 390-400, 2007.
[7] B. A. Gama, S. L. Lopatnikov, and J. W. Gillespie Jr, "Hopkinson bar
experimental technique: a critical review," Appl Mech Rev, vol. 57, pp. 223-250,
2004.
[8] J. Van Slycken, P. Verleysen, J. Degrieck, L. Samek, and B. De Cooman, "High-
strain-rate behavior of low-alloy multiphase aluminum-and silicon-based
transformation-induced plasticity steels," Metallurgical and Materials
Transactions A, vol. 37, pp. 1527-1539, 2006.
[9] G. T. Gray III, "Classic split-Hopkinson pressure bar testing," ASM Handbook,
Mechanical testing and evaluation, vol. 8, pp. 462-476, 2000.
71
[10] H. Huh, W. Kang, and S. Han, "A tension split Hopkinson bar for investigating
the dynamic behavior of sheet metals," Experimental Mechanics, vol. 42, pp. 8-
17, 2002.
[11] W. W. Chen and B. Song, Split Hopkinson (Kolsky) bar: design, testing and
applications: Springer Science & Business Media, 2010.
[12] S. Mates and F. Abu-Farha, "Dynamic Tensile Behavior of a Quenched and
Partitioned High Strength Steel Using a Kolsky Bar," pp. 155-161, 2016.
[13] S. Lee and G. Swallowe, "Direct measurement of high rate stress–strain curves
using instrumented falling weight and high-speed photography," The Imaging
Science Journal, vol. 52, pp. 193-201, 2004.
[14] A. Thompson, "High strain rate characterization of advanced high strength
steels," University of Waterloo, 2006.
[15] B. Yan, Y. Kuriyama, A. Uenishi, D. Cornette, M. Borsutzki, and C. Wong,
"Recommended practice for dynamic testing for sheet steels-development and
round robin tests," SAE Technical Paper 0148-7191, 2006.
[16] D. Zhu, S. Rajan, B. Mobasher, A. Peled, and M. Mignolet, "Modal analysis of a
servo-hydraulic high speed machine and its application to dynamic tensile testing
at an intermediate strain rate," Experimental Mechanics, vol. 51, pp. 1347-1363,
2011.
[17] V. Tarigopula, O. S. Hopperstad, M. Langseth, A. H. Clausen, and F. Hild, "A
study of localisation in dual-phase high-strength steels under dynamic loading
72
using digital image correlation and FE analysis," International Journal of Solids
and Structures, vol. 45, pp. 601-619, 2008.
[18] L. Hector Jr, V. Savic, A. Sachdev, and J. Coryell, "Digital Image Correlation
Studies of Advanced High-Strength Steel Tensile Deformation," in International
Symposium on New Developments in Advanced High-Strength Sheet Steels, 2013,
pp. 209-218.
[19] ISO, "ISO_26203-Metallic materials — Tensile testing at high strain rates," ed,
2011.
[20] S.-E.-P. S. d. S. VDEh, "SEP1230. The determination of the mechanical
properties of sheet metal at high strain rates in high-speed tensile tests," ed, 2006.
[21] H. Huh, J. Lim, and S. Park, "High speed tensile test of steel sheets for the stress-
strain curve at the intermediate strain rate," International Journal of Automotive
Technology, vol. 10, pp. 195-204, 2009.
[22] H. Huh, H. Lee, and J. Song, "Dynamic hardening equation of the auto-body steel
sheet with the variation of temperature," International Journal of Automotive
Technology, vol. 13, pp. 43-60, 2012.
[23] S. Xu, D. Ruan, J. H. Beynon, and Y. Rong, "Dynamic tensile behaviour of TWIP
steel under intermediate strain rate loading," Materials Science and Engineering:
A, vol. 573, pp. 132-140, 2013.
[24] J.-H. Kim, D. Kim, H. N. Han, F. Barlat, and M.-G. Lee, "Strain rate dependent
tensile behavior of advanced high strength steels: Experiment and constitutive
modeling," Materials Science and Engineering: A, vol. 559, pp. 222-231, 2013.
73
[25] J. Qin, R. Chen, X. Wen, Y. Lin, M. Liang, and F. Lu, "Mechanical behaviour of
dual-phase high-strength steel under high strain rate tensile loading," Materials
Science and Engineering: A, vol. 586, pp. 62-70, 2013.
[26] J. Huh, H. Huh, and C. S. Lee, "Effect of strain rate on plastic anisotropy of
advanced high strength steel sheets," International Journal of Plasticity, vol. 44,
pp. 23-46, 2013.
[27] D. Dong, Y. Liu, L. Wang, and L. Su, "Effect of strain rate on dynamic
deformation behavior of DP780 steel," Acta Metall Sin, vol. 49, pp. 159-166,
2013.
[28] W. Li, J. Zhu, Y. Xia, and Q. Zhou, "Testing and modeling the effect of strain-
rate on plastic anisotropy for a traditional High Strength Steel," in ASME 2015
International Mechanical Engineering Congress and Exposition, 2015, pp.
V009T12A084-V009T12A084.
[29] B. C. Hwang, T. Y. Cao, S. Y. Shin, S. H. Kim, S. H. Lee, and S. J. Kim, "Effects
of ferrite grain size and martensite volume fraction on dynamic deformation
behaviour of 0·15C–2·0Mn–0·2Si dual phase steels," Materials Science and
Technology, vol. 21, pp. 967-975, 2013.
[30] S. Li, D. Zou, C. Xia, and J. He, "Effect of Strain Rate on Deformation-Induced
Martensitic Transformation of Quenching and Partitioning Steels," steel research
international, pp. n/a-n/a, 2015.
74
[31] R. Alturk, W. E. Luecke, S. Mates, A. Araujo, K. Raghavan, and F. Abu-Farha,
"Rate effects on transformation kinetics in a metastable austenitic stainless steel,"
Procedia Engineering, vol. 207, pp. 2006-2011, 2017.
[32] M. Rashid, "Dual phase steels," Annual Review of Materials Science, vol. 11, pp.
245-266, 1981.
[33] A. K. Sachdev, "Effect of retained austenite on the yielding and deformation
behavior of a dual phase steel," Acta Metallurgica, vol. 31, pp. 2037-2042, 1983.
[34] L. F. Ramos, D. K. Matlock, and G. Krauss, "On the deformation behavior of
dual-phase steels," Metallurgical Transactions A, vol. 10, pp. 259-261, 1979.
[35] X. H. F. Abu-Farha, X. Sun, Y. Ren, L.G. Hector, Jr, G. Thomas, T. Brown, "In-
situ Local Measurement of Austenite Mechanical Stability and Transformation
Behavior in Third Generation Advanced High Strength Steels," Mettalurgical and
Materials Transactions A, vol. in press., 2018.
[36] J. Speer, D. Matlock, B. De Cooman, and J. Schroth, "Carbon partitioning into
austenite after martensite transformation," Acta materialia, vol. 51, pp. 2611-
2622, 2003.
[37] X. Xiong, B. Chen, M. Huang, J. Wang, and L. Wang, "The effect of morphology
on the stability of retained austenite in a quenched and partitioned steel," Scripta
Materialia, vol. 68, pp. 321-324, 2013.
[38] M. Borsutzki, D. Cornette, Y. Kuriyama, A. Uenishi, B. Yan, and E. Opbroek,
"Recommendations for dynamic tensile testing of sheet steels," International iron
and steel institute, vol. 30, 2005.
75
[39] N. Beynon, T. Jones, and G. Fourlaris, "Effect of high strain rate deformation on
microstructure of strip steels tested under dynamic tensile conditions," Materials
Science and technology, vol. 21, pp. 103-112, 2005.
[40] P. Verleysen, J. Peirs, J. Van Slycken, K. Faes, and L. Duchene, "Effect of strain
rate on the forming behaviour of sheet metals," Journal of Materials Processing
Technology, vol. 211, pp. 1457-1464, 2011.
[41] J. V. Paul and J. Kimberley, "A Desktop Tensile Kolsky Bar for the Dynamic
Testing of Metallic Foils," Journal of Dynamic Behavior of Materials, vol. 1, pp.
439-446, 2015.
[42] J. Kwon, J. Huh, H. Huh, and J. Kim, "Micro-Tensile Tests of Aluminum Thin
Foil with the Variation of Strain Rate," in Dynamic Behavior of Materials,
Volume 1, ed: Springer, 2013, pp. 89-100.
[43] M. Isakov, S. Hiermaier, and V.-T. Kuokkala, "Effect of strain rate on the
martensitic transformation during plastic deformation of an austenitic stainless
steel," Metallurgical and Materials Transactions A, vol. 46, pp. 2352-2355, 2015.
[44] G. T. Gray, "High-Strain-Rate Deformation: Mechanical Behavior and
Deformation Substructures Induced," Annual Review of Materials Research, vol.
42, pp. 285-303, 2012.
[45] J. Butler, "Lüders front propagation in low carbon steels," Journal of the
Mechanics and Physics of Solids, vol. 10, pp. 313-318, 1962.
[46] P. Zavattieri, V. Savic, L. Hector Jr, J. Fekete, W. Tong, and Y. Xuan, "Spatio-
temporal characteristics of the Portevin–Le Châtelier effect in austenitic steel with
76
twinning induced plasticity," International Journal of Plasticity, vol. 25, pp.
2298-2330, 2009.
[47] B. C. De Cooman, P. Gibbs, S. Lee, and D. K. Matlock, "Transmission electron
microscopy analysis of yielding in ultrafine-grained medium Mn transformation-
induced plasticity steel," Metallurgical and Materials Transactions A, vol. 44, pp.
2563-2572, 2013.
[48] H. Paruz and D. Edmonds, "The strain hardening behaviour of dual-phase steel,"
Materials Science and Engineering: A, vol. 117, pp. 67-74, 1989.
[49] G. Krauss and D. Matlock, "Effects of strain hardening and fine structure on
strength and toughness of tempered martensite in carbon steels," Le Journal de
Physique IV, vol. 5, pp. C8-51-C8-60, 1995.
[50] K. Qin, L. Yang, and S.-s. Hu, "Interpretation of Strain Rate Effect of Metals," in
Dynamic Behavior of Materials, Volume 1, ed: Springer, 2014, pp. 21-27.
[51] W. Johnston and J. J. Gilman, "Dislocation velocities, dislocation densities, and
plastic flow in lithium fluoride crystals," Journal of Applied Physics, vol. 30, pp.
129-144, 1959.
[52] J. Gilman and W. Johnston, "Dislocations and mechanical properties of crystals,"
Fisher, J., Johnston, W., Thomson, R., Vreeland, T.(eds.), vol. 116, 1957.
[53] V. Parameswaran, N. Urabe, and J. Weertman, "Dislocation mobility in
aluminum," Journal of Applied Physics, vol. 43, pp. 2982-2986, 1972.
77
[54] J. A. Lichtenfeld, C. J. Van Tyne, and M. C. Mataya, "Effect of strain rate on
stress-strain behavior of alloy 309 and 304L austenitic stainless steel,"
Metallurgical and Materials Transactions A, vol. 37, pp. 147-161, 2006.
[55] K. Nohara, Y. Ono, and N. Ohashi, "Strain-induced martensitic transformation in
metastable austenitic stainless steels in multi-stage tensile deformation at various
temperatures," New Aspects of Martensitic Transformations. Japan Institute of
Metals, Tokyo. 1976, 315-320, 1976.
[56] S. F. Peterson, M. Mataya, and D. Matlock, "The formability of austenitic
stainless steels," JOM Journal of the Minerals, Metals and Materials Society, vol.
49, pp. 54-58, 1997.
[57] J. Talonen, H. Hänninen, P. Nenonen, and G. Pape, "Effect of strain rate on the
strain-induced γ→ α′-martensite transformation and mechanical properties of
austenitic stainless steels," Metallurgical and materials transactions A, vol. 36,
pp. 421-432, 2005.
[58] P. Larour, "Strain rate sensitivity of automotive sheet steels: influence of plastic
strain, strain rate, temperature, microstructure, bake hardening and pre-strain,"
Lehrstuhl und Institut für Eisenhüttenkunde2010.
[59] S. Hecker, M. Stout, K. Staudhammer, and J. Smith, "Effects of strain state and
strain rate on deformation-induced transformation in 304 stainless steel: Part I.
Magnetic measurements and mechanical behavior," Metallurgical Transactions A,
vol. 13, pp. 619-626, 1982.
78
[60] G. T. K.S Raghavan, J. Arnold, A. Gill, and T. Gnaupel-Herold, "A Comparison
of Different Measurement Methods to Evaluate Strain-Induced Martensite
Formation in a Metastable Austenitic Stainless Steel," presented at the Materials
Science and Technology (MS&T), Pittsburgh, Pennsylvania, USA, 2014.
[61] P. Jacques, S. Allain, O. Bouaziz, A. De, A.-F. Gourgues, B. Hance, et al., "On
measurement of retained austenite in multiphase TRIP steels—results of blind
round robin test involving six different techniques," Materials Science and
Technology, vol. 25, pp. 567-574, 2009.
[62] J. Talonen, P. Aspegren, and H. Hänninen, "Comparison of different methods for
measuring strain induced α-martensite content in austenitic steels," Materials
Science and Technology, vol. 20, pp. 1506-1512, 2004.
[63] S. Tavares, J. Pardal, M. G. Da Silva, H. Abreu, and c. R. da Silva, "Deformation
induced martensitic transformation in a 201 modified austenitic stainless steel,"
Materials characterization, vol. 60, pp. 907-911, 2009.
[64] W. Wang, M. Li, C. He, X. Wei, D. Wang, and H. Du, "Experimental study on
high strain rate behavior of high strength 600–1000MPa dual phase steels and
1200MPa fully martensitic steels," Materials & Design, vol. 47, pp. 510-521,
2013.
[65] C. Liu, L. Wang, and Y. Liu, "Effects of Strain Rate on Tensile Deformation
Behavior of Quenching and Partitioning Steel," in Materials Science Forum,
2013, pp. 401-406.
79
[66] K. O. F. W.A. Poling, "High Strain Rate Deformation Response of DP980 and
QP980 Automotive Sheet Steels (Presentation)," presented at the Materials
Science and Technology MS&T, 2014.
80
CHAPTER THREE
3. STRAIN RATE EFFECT ON TENSILE FLOW BEHAVIOR AND
ANISOTROPY OF A MEDIUM MANGANESE TRIP STEEL
ABSTRACT
Plastic anisotropy dependence on nominal strain rate of a medium manganese (10
wt. % Mn) TRIP steel consisting of an initial 66% austenite volume fraction (balance
ferrite) is investigated. This material exhibits yield point elongation, propagative
instabilities during hardening, and austenite transformation to α'-martensite either directly
or through ε-martensite. Uniaxial strain rates within the 0.005-500 s-1 range along the 0º,
45º, and 90 º orientations were selected based upon their relevance to automotive
applications. The plastic anisotropy (r) and normal anisotropy (rn) indices corresponding
to each direction and strain rate were determined using the strain fields from stereo digital
image correlation (DIC) systems that enabled both quasi-static and dynamic
measurements. The results provide evidence of significant, orientation-dependent strain
rate effects on both the flow stress and the evolution of r and rn with strain. This has
implications not only for material performance during forming but also for the
development of future strain rate dependent anisotropic yield criteria. Since tensile data
alone for the subject medium manganese TRIP steel does not satisfactorily connect
microstructural mechanisms responsible for the observed macroscopic-scale behavior
from tensile testing, a discussion of additional tests that must supplement the mechanical
test results presented herein is provided.
81
`3.1 Introduction
Usage of advanced high strength steel (AHSS) continues to increase in
automotive structural applications due to their favorable combinations of strength and
ductility [1]. However, fundamental challenges remain to implementation in vehicle body
structures. Examples are flow behavior dependence on strain rate and temperature,
martensitic transformation dependence upon deformation mode and strain path,
springback, and liquid metal embrittlement in spot welds [2].
Many AHSS grades are characterized, in part, by their multi-constituent
microstructures relative to the primarily single-constituent steels of common use. The
role of manganese (Mn) is especially important in the 1st, 2nd and 3rd generations of
AHSSs and is considered in Figure 3.1 regarding three main deformation mechanisms,
viz. dislocation-mediated plasticity, transformation induced plasticity or the TRIP effect,
and twinning induced plasticity (TWIP) observed in so-called TWIP steels. Manganese
alloy content is denoted by the large open arrow wherein 0% is associated with stable
ferrite in traditional steels and many 1st Gen AHSSs, such as dual-phase steels [3]. As Mn
content increases, increased austenite stabilization is made possible. Application of strain
to austenite bearing steels may result in the TRIP effect which involves the diffusionless
shear transformation of austenite to martensite (denoted by the blue region in Figure 3.1)
[4]. The TRIP mechanism is especially crucial in 3rd generation AHSSs [5]. These steels
can be produced with heat treatments such as quenching and partitioning (Q&P) [6], or
inter-critical annealing wherein ferrite rejects both C and Mn which partition to austenite
[7]. Enhanced work hardening results along with a delay in necking under tension, and
82
greater elongations to fracture for a given steel strength [6, 8]. As Mn content is further
increased, austenite to martensite transformation may occur, in part, through the
formation of ε-martensite (denoted by the red region), the hexagonal close-packed form
of iron [9, 10]. This was recently reported for a so-called medium Mn TRIP steel with 10
wt.% Mn, which is a 3rd generation AHSS [11]. The ε-martensite phase is nucleated at
intersecting shear bands [12]. Steels that exhibit both the TRIP and TWIP mechanisms
have been associated with Mn content as low as 10 wt.% in bulk [13]. Fully austenitic
steels that exhibit the TWIP effect (2nd Gen AHSSs) have the highest Mn content and
lowest stacking fault energies [14].
Figure 3.1: Deformation mechanisms in multiphase AHSSs with Mn alloying content
Phases are as follows: ferrite (, body-centered cubic), austenite (, face-centered cubic),
epsilon martensite (, hexagonal close-packed), martensite (', body-centered tetragonal).
Note that denotes shear stress. The solid red arrows indicate that yield stress decreases,
in general, with increasing Mn content.
83
Automotive body structure materials are subject to different deformation modes,
strain paths and strain rates both during manufacturing and when integrated into a vehicle
body structure [15]. It is well-established that many steels exhibit strain rate dependent
flow properties, such as tensile yield and ultimate strengths as well as ductility [16-25].
Strain rates in forming of automotive components are typically in the range of 0.1-10 s-1,
and the various stress modes and strain paths require extensive materials characterization
to support sufficient formability prediction. Strain rates during vehicle impact, however,
can reach as high as ~1000 s-1. Increased temperatures during stamping, both due to the
effects of friction and adiabatic heating at elevated strain rates, may reduce the rate of
retained austenite transformation, thereby lowering the work hardening rate and tensile
strengths of TRIP steels, in particular [26-28].
Although AHSSs offer enhanced combinations of strength and ductility, practical
limits exist on their formability, and various tests have been devised to characterize and
predict in-vehicle performance [29]. Advanced fracture models, either microstructure
based [30] or based upon a phenomenological model, such as GISSMO (Generalized
Incremental Stress-State Dependent Damage model) [31], require additional tests ranging
from compression of small to ligament-type specimens that probe specific stress
triaxialities. Perhaps the most common test instrument for material data generation is the
screw-driven tensile load frame, which is limited to quasi-static strain rates of
approximately 0.1 s-1 or less. For higher strain rates, a servo-hydraulic load frame is
commonly employed to test automotive sheet steels [17, 23, 32, 33]. Here, a test
specimen is accelerated to a target speed with an actuator that engages a slack adaptor
84
assembly to rapidly deform a test specimen in tension. While the upper limit on nominal
strain rate for servo-hydraulic test machines can achieve 1000 s-1, the 500 s-1 rate is
commonly employed as a useful upper limit. One of several persistent challenges to high
strain rate testing is ringing which appears as a spurious oscillation superimposed onto
the “true” material response (e.g., load vs. displacement) [23]. Ringing is most severe
near the resonant frequency of the load train, and it obscures the desired material
response leading to inaccurate flow data. The split-Hopkinson bar [34-37], also known as
the Kolsky bar [38, 39], is commonly applied in dynamic material tests at strain rates at
or above 1000 s-1, but there are notable exceptions utilizing this technique in which tests
have been conducted at or near 500 s-1. Here, the specimen is plastically deformed by an
elastic wave pulse initiated by a striker through the load train consisting of incident and
transmission bars and the test specimen. Unlike a servo-hydraulic load frame, the test
specimen is often not deformed to fracture in a split-Hopkinson bar test thereby
precluding determination of fracture strains.
The majority of mechanical property studies of AHSS do not consider strain rate
effects on plastic anisotropy, notably the plastic anisotropy index, alternatively termed r
or r-value. Recently, there have been some reports in which plastic anisotropy has been
investigated at different strain rates. For example, Huh et al. [40] showed that the plastic
anisotropy of DP780 and TRIP590 tends to diminish at high strain rates. They also
constructed yield loci at different strain rates and noted changes in the general shape of
yield loci. Rahmaan [41] studied strain rate effects on the anisotropy of DP600 and
TRIP780; however, strain rate had a negligible effect on anisotropy. Li et al. [42] showed
85
that anisotropy changes with strain rate in high strength low alloy (HSLA) steels. These
effects were then characterized by calibrating different anisotropic yield functions and
observing the strain rate sensitivity of the coefficients in those yield functions.
Strain rate effects on the plastic anisotropy of AHSS, 3rd generation AHSS in particular,
require additional investigation. In this study, strain rate effects on the plastic anisotropy
of a developmental Medium Mn (10 wt.%), duplex (ferrite+ austenite) TRIP sheet steel is
investigated. Of particular interest is the effect of strain rate on the tensile flow stresses
along different in-plane loading directions and on the r-value and its evolution with
plastic strain. For this purpose, a servo-hydraulic load frame was used to apply nominal
strain rates in the 0.5 – 500 s-1 range. The extent to which the r-value changes at selected
nominal strain rates are quantified. The value of generated results for calibration and
validation of material constitutive models for CAE analyses is discussed regarding the
need for a methodology that incorporates strain rate effects on sheet anisotropy effects in
plasticity models. Additional experimentation required to connect important
microstructural mechanisms to anisotropy dependence on strain rate under alternative
strain paths is suggested.
3.2 Medium Manganese TRIP Steel
The chemical composition in weight percentage (wt.%) for the uncoated medium
manganese (Fe-0.16C-10.0Mn-1.56Al-0.2Si, wt.%) TRIP steel, from now on Med. Mn
TRIP is listed in Table 3.1 [43]. Processing involved inter-critical batch annealing at
600-700 °C for 96 hours. According to Gibbs et al. [7], this facilitates the enrichment of
86
retained austenite with both C and Mn. The material was then hot rolled in billet form,
and cold rolled into 1500 mm long strips each with a nominal 1.1 mm thickness and 200
mm width. The as-manufactured material contains 66% austenite volume fraction with
the balance of ferrite. Figure 3.2 is a field emission SEM image of the as-manufactured
microstructure showing the morphology of ferrite and austenite phases. The combined
average grain size of ferrite and austenite phases, similar in size and morphology, is ~2
μm. The presence of annealing twins in the austenite grains is noted and is suggestive of a
low austenite stacking fault energy as a result of alloy enrichment in the austenite phase.
Martensitic transformation in the Med. Mn TRIP under quasi-static tension as well as
deformation and strain path dependence in a T-shaped stamping have been detailed in
Abu Farha et al. [11] and Wu et al. [15], respectively. Texture evolution was noted in
synchrotron X-ray data for in-situ quasi-static tests of specimens with the 0º orientation,
but no similar tests were performed on specimens with the 45º and 90º orientations [11].
3.3 Test Specimens
Test specimen geometries are shown in Figure 3.3. All quasi-static tests were
conducted with the ASTM E8/ISO 6892 Type 1 [44] test specimen geometry. A search of
the literature revealed a wide variety of sheet specimen tensile geometries for high strain
rate testing with servo-hydraulic load frames. For example, Yan et al. [45] show four
examples used in their round-robin study on dynamic testing while Borsutzki et al. [46]
show ten different tensile geometries. The high rate test specimen geometry must take
into account the gripping mechanism, facilitate strain gauge application, and not
87
contribute itself to load train ringing and non-uniform strain evolution. After a careful
literature review, the dog-bone test specimen geometry in Figure 3.3 was selected. This
geometry is based upon that in Yang et al. [47]. One gripper end was designed to be
longer than the other to accommodate linear strain gauges used as load cells at strain rates
Table 3.1: Chemical composition (wt.%) of Med. Mn steel.
Elements Mn C Al Si P S Fe
wt.% 10.0 0.16 1.37 0.19 0.002 0.0018 Balance
Figure 3.2: Field emission SEM image of the as-received, medium Mn TRIP steel (two
phases, ferrite (F) and retained austenite (A)) with ~2 μm grain size. Annealing twins
(AT) appear in austenite grains which are indicative of a low stacking fault energy.
88
in excess of 5 s-1. It guarantees strain homogeneity in the gauge section length as
suggested by ISO 29203-1,2 [48] and SEP1230 [49]. The second version with a 10mm
gauge length was tested, but non-uniform strain accumulation along the gauge length
resulted. All test specimens were cut with wire EDM and cleaned before application of
the DIC contrast pattern.
Figure 3.3: Tensile specimen geometry for high rate testing. All dimensions are in mm
3.4 Instrumentation
Quasi-static tension tests were conducted on an Instron 5985 universal
electromechanical load frame. The load frame has a 250 kN static load capacity with an
actuator that can achieve a maximum speed of 17 mm/s. The upper grip is attached to a
precision force transducer (with 250 kN rating) that is connected to the moving
crosshead. The tests were performed at a constant crosshead speed of ~0.3 mm/s with
quasi-static strain rate 0.005 s-1.
Nominal strain rates from 0.5 s-1 to 500 s-1 were achieved with an INSTRON
VHS 65/80-20 servo-hydraulic test unit shown in Figure 3.4. This is a two-column load
frame with an 80 kN dynamic load rating. Deformation of a test specimen occurs via
89
linear motion of an actuator, which can achieve a ~20m/s maximum speed, connected to
the lower grip of a slack adaptor unit. The upper grip is connected to a Kistler 9071A
piezoelectric load washer which is between the upper (stationary) grip and a manually
adjustable yoke. A significant challenge with servo-hydraulic units during high rate
testing is ringing from the sudden engagement of the lower grip, discussed in detail in
Yang et al. [47]. To substantially minimize ringing and to make sure that any minor
specimen bending was eliminated, a linear strain gauge connected to a ¼ bridge circuit
adhered to both sides of the longer grip end on each high rate test specimen. The strain
gauges functioned as the load cell for strains > 5 s-1.
3.5 Strain Measurement
For accurate strain measurements, stereo digital image correlation (DIC) systems
were used to monitor material deformation during testing. A black and white speckle
contrast pattern was applied to the gauge section of each test specimen. This facilitated
strain measurement in DIC post-processing of images captured during a test. For the
quasi-static tests at the nominal 0.005 s-1 rate, the GOM 3D DIC system was used, which
consisted of two black and white 5M cameras with 50mm lenses, two high-intensity
lights, and the ARAMIS software for post-processing. A frame rate of 5 f/s was selected
for the quasi-static tests.
For the high rate tests in the 0.5 – 500 s-1 range, digital images of a specimen
gauge section were captured at a fixed 20,000 f/s rate with a high-speed stereo DIC
system from Correlated Solutions. This system, which is shown in part in Figure 3.4,
90
consisted of the following components: two Photron SA1.1 black and white cameras,
high intensity movie lights, a National Instruments multi-channel data acquisition (DAQ)
box for image capture and data logging (e.g. image number tagged with corresponding
strain gage voltages and load washer loads), and the Vic-Snap image capture software
from Correlated Solutions. After testing, all DIC images were processed consistently to
obtain the full-field history of strains using algorithms detailed in Sutton et al. [50].
Figure 3.4: INSTRON VHS 65/80-20 (VHS=very high speed) servo-hydraulic test unit
91
3.6 Results and Discussion
3.6.1 Tensile Flow Behavior
Figures 3.5(a)-(c) show engineering stress-strain curves for the Med. Mn TRIP
steel along the rolling (RD), diagonal (DD), and transverse (TD) directions, respectively,
at 0.005, 0.5, 5, 50 and 500 s-1. All specimens were tested to fracture. Each curve is
characterized by a protracted yield point elongation (YPE), denoted by the plateaus in the
stress/strain curves which span approximately 7% engineering strain and followed by
hardening before fracture. The YPE is not obvious in the 500 s-1 curves due to an artifact
in the load signal that the strain gauges were unable to eliminate. However, confirmation
that the yield point elongation persists to 500 s-1 was acquired through examination of
DIC strain rate contours in Figures 3.6(a)-(c). Relevant mechanical properties as a
function of strain rate and orientation are listed in Tables 3.2-3.4. An initial negative
strain rate sensitivity regarding the flow stress and ultimate strength occurs for all three
directions. This is observed by the high strain hardening at 0.005 s-1 followed by the
generally significant drop in strength at 0.5 s-1. Beyond 0.5 s-1, flow stress once again
increases indicating a positive strain rate sensitivity, but strength levels observed at 0.005
s-1 are not achieved in the 0.5 s-1 to 500 s-1 range. The reduction in strain hardening is a
result of suppressed martensite transformation from austenite due to adiabatic heating and
austenite stabilization during deformation at high strain rates [24]. Multiple studies have
shown that the increase in temperature inhibits the deformation-induced transformation of
austenite and thus leads to lower strain hardening at higher rates in fully austenitic,
duplex ferrite-austenite, and Q&P steels [51-54]. It is also worth noting that the strain
92
Table 3.2: Tensile Properties Associated with RD Flow Curves in Fig. 4.5a. Engineering
stress and strain are reported.
Nominal Strain
Rate (s-1)
0.2% Yield
Strength (MPa)
Ultimate
Tensile
Strength (MPa)
Strain at
Uniform
Elongation
Strain at
Fracture
0.005 690 1115 0.33 0.41
0.5 743 821 0.30 0.40
5.0 765 826 0.31 0.39
50.0 782 867 0.29 0.41
500.0 815 967 0.24 0.44
Table 3.3: Tensile Properties Associated with DD Flow Curves in Fig. 4.5b. Engineering
stress and strain are reported.
Nominal Strain
Rate (s-1)
0.2% Yield
Strength (MPa)
Ultimate
Tensile
Strength (MPa)
Strain at
Uniform
Elongation
Strain at
Fracture
0.005 702 1030 0.39 0.45
0.5 712 805 0.36 0.51
5.0 736 819 0.35 0.49
50.0 770 840 0.34 0.52
500.0 795 865 0.32 0.53
93
Table 3.4: Tensile Properties Associated with TD Flow Curves in Fig. 4.5c. Engineering
stress and strain are reported.
Nominal Strain
Rate (s-1)
0.2% Yield
Strength (MPa)
Ultimate
Tensile
Strength (MPa)
Strain at
Uniform
Elongation
Strain at
Fracture
0.005 700 1160 0.33 0.37
0.5 770 828 0.28 0.40
5.0 805 852 0.28 0.43
50.0 845 865 0.28 0.42
500.0 890 902 0.25 0.40
(a) (b)
94
(c)
Figure 3.5: Engineering stress-strain curves at the five testing strain rates for: (a) 0º (RD
– rolling direction), (b) 45º (DD – diagonal direction), and (c) 90º (TD – transverse
direction).
hardening varies with tensile loading orientation relative to the sheet rolling direction,
which is due to an orientation dependence of both the texture evolution and martensite
transformation. Previous studies indicate that martensite formation from austenite is
suppressed when the tensile axis is oriented along <001> in both austenitic stainless
steels and high Mn steels that exhibit two stage TRIP behaviors (γ→ε→α’) [55-57]. The
elevated work hardening rates in both the RD and TD orientations relative to the DD
orientation suggest that a considerable <001> austenite component exists along the DD
orientation. Serrations beyond yield point elongation, which are clearly evident in
hardening regions of the 0.005 s-1 curves, are indicative of dynamic strain aging. This is
95
(a) (b) (c)
Figure 3.6: (a) Tensile (2D DIC) strain contours at the ninth image showing the onset of
Lüdering (red contour patch) in the 500 s-1 curve (see Fig. 3.5). (b) Tensile (2D DIC)
strain contours showing Lüders strain at image twelve in the 500 s-1 curve (c) Tensile (2D
DIC) strain contours showing Lüders strain at image twelve in 500 s-1 curve.
manifested by propagative instabilities resembling the highly spatially correlated Type A
Portevin-Le Châtelier bands reported in tensile tests of other AHSS [58, 59]. Band
propagation beyond yield point elongation for all orientations ceases at 500 s-1 as noted
via examination of strain rate contours computed in DIC post-processing. Total
elongation tends to increase for strain rates higher than 0.005 s-1 along the DD and TD.
The data in Tables 3.2-3.4 show that flow properties along the DD experience the
lowest magnitude of strain rate sensitivity of yield strength, wherein the yield stress
increases by 13% between 0.005 s-1 and 500 s-1. By comparison, tests along the RD and
TD exhibit yield strength increases of 18% and 27%, respectively. The same trend by
96
orientation is observed in the change in uniform elongation with strain rate from 0.005 s-1
to 500 s-1. Tests along the DD show a change of 17% in ductility, while the change in
uniform elongation observed along the RD and TD is 27 % and 25%, respectively.
The anisotropy in the flow behavior may be characterized by normalized true flow stress,
σn, calculated as a function of true strain via [40, 42]:
0
o
n
(3.1)
where is the flow stress in either the DD or the TD relative to the RD flow stress, σ0. The
material is isotropic when 1n . Figure 3.7 shows n vs. ε in both the DD and the TD
for all five strain rates. The normalized true flow stress was calculated in the true strain
range of 0.05-0.25 to avoid both YPE and the region of non-uniform deformation (post
necking). Although n at all strain rates varies as a function of total strain in both the DD
and TD, its reduction is especially prominent with the DD at 0.005 s-1. Again, this is
attributed to the reduction in work hardening as a result of martensite transformation
suppression, presumably along <001> fibers in austenite, as a result of as-produced
texture. As strain is increased, the much smaller changes in n suggest that the material
becomes increasingly isotropic. The same trend toward isotropy with increasing strain
rate is also observed, as the plots at higher strain rates tend to move closer towards 1n
in Figure 3.7.
97
(a) (b)
Figure 3.7: Normalized flow stresses from Eq. (3.1) along the (a) diagonal direction (DD)
(b) transverse direction (TD).
3.6.2 Anisotropy
A well-known and important parameter that both denotes the level of anisotropy
and quantifies the resistance to thinning under tension in sheet metals is the Lankford
coefficient or r-value [60], also termed the plastic anisotropy ratio. The r-value, which is
the ratio of the in-plane (transverse) to thickness strains, is used to gauge anisotropy in
stamping as well as calibrate anisotropic yield criteria for finite element simulations [61].
Strain fields parsed in DIC post-processing of image data recorded during the tests
provided a useful alternative to extensometers for computing r-values [62]. Figure 3.8
shows the r-value evolution with true strain at different strain rates for each of the tested
orientations. The measured r-value variation with orientation changes with both strain
rate and total strain. The r-value at 0.005 s-1 in the three loading directions has a high
98
dependency on the strain, which is a strong indicator of the texture evolution with strain
[11, 63]. The curves corresponding to 0.005 s-1 tend to be outliers with those for the
remaining strain rates grouping together. The change in all three directions during quasi-
static loading is towards perfect isotropy (r=1); along the RD the r-value changes from
0.54 to 0.63, along the DD the r-value changes from 1.15 to 1.1 and along the TD the r-
value changes from 0.72 to 0.8. Figure 3.8 shows, in summary, that the plastic anisotropy
index tends toward a common value with strain for all tested orientations due to tensile
texture evolution, although the common value at a given orientation is not equal to that of
an alternate orientation. Additionally, the plastic anisotropy index tends toward isotropy
(r=1) with increased strain rate, and this indicates that material texture effects on plastic
anisotropy are mitigated at increased strain rate. Of particular note in Figure 3.8 are the
relative differences of quasi-static r-values and elevated strain rate r-values in each
orientation. A clear decrease in r-value is observed above quasi-static test rates in the DD
orientation, but both the RD and TD orientations reveal an increased r-value with
increased strain rate. As previously noted, this is presumably an effect of both
transformation kinetics related to as-produced austenite texture and tensile texture
evolution in the ferrite and austenite phases during deformation. Additional studies are
required to characterize the structural evolution during deformation fully and definitively
account for the observed behavior. The change of r-value with strain is, in part, due to
the non- homogenous yielding observed with the Med. Mn. TRIP, i.e., yield point
elongation [40, 63, 64]. This behavior was recently confirmed by coupled stereo digital
99
(a) (b)
(c)
Figure 3.8: r-value evolution with strain at different strain rates for specimens tested
along the (a) rolling direction (RD) (b) diagonal direction (DD), and (c) transverse
direction (TD).
100
image correlation and synchrotron X-ray diffraction where plastic deformation and
austenite transformation were found to be highly inhomogeneous and occur in
propagating bands [11].
Figure 3.9 shows the average r-value as a function of true strain for thee RD, DD,
and TD, calculated across a true strain range of 0.05-0.25. Also shown in Figure 3.9 is the
coefficient of normal anisotropy (nr )
0 45 902
4n
r r rr
(3.2)
which is a useful indicator of formability. The highest r-value is observed along the DD,
which corresponds with the ductility results shown in tables 3.2-3.4 (the highest ductility
is observed when testing along the DD). Similar to Figure 3.8, the anisotropy exhibited in
the three directions is most significant (furthest away from an isotropic value of 1.0) at
the quasi-static rate (r0avg=0.6, r45avg=1.1 and r90avg=0.78). The r-values are closest to 1
(thus approaching isotropic behavior) at the highest rate (500 s-1). These results are in
qualitative agreement with Huh et al. [40] who identified significant strain rate effects on
the r-value of TRIP590 and DP780. The reported r-values of Huh et al. also tended
toward unity as the strain rate was increased. A similar conclusion was reached by Li et
al. [42] who were able to quantify the strain rate effect on the anisotropy for an HSLA
steel tested at different rates. For the Med. Mn. TRIP of interest here, the normal
anisotropy increases with strain rate indicates better resistance to thinning at higher strain
rates; and thus, improved formability.
101
The aforementioned results suggest a significant influence of strain rate on the
texture evolution and austenite transformation of Med. Mn TRIP, leading to a change in
the normalized flow stress and the r-value observed along different loading directions.
This effect, in turn, raises the question of the accuracy of using quasi-static flow data to
model plastic anisotropy. Modeling of plastic anisotropy is often accomplished via an
anisotropic yield function, a critical part of the constitutive relations needed to accurately
describe the behavior of metal sheet plasticity. The strain rate effects shown here on the
flow stress and the r-values suggest that the calibration of the yield function using quasi-
static data is only accurate at strain rates close to quasi-static. As with the different
magnitudes of flow stresses and r-values at different strain and strain rates, comes yield
loci with different shapes and sizes. Therefore, there is a need for the development of a
methodology that will allow for modeling plastic anisotropy during high rate deformation
for steels such as the Med. Mn TRIP considered herein.
3.7 Conclusions
The effect of strain rate on the plastic anisotropy of a development Med. Mn.
TRIP steel with an initial 66% volume fraction of austenite (balance ferrite), was
investigated. During tensile straining, austenite transforms to martensite either directly or
through -martensite, the transient hexagonal close-packed form of martensite. Tests
were conducted in three different orientations, viz, 0º or RD, 45 º or DD, 90 º or TD, in
uniaxial tension at selected nominal strain rates within the 0.005 s-1 to 500 s-1 range. The
change in anisotropy with strain rate and the total strain was characterized using three
102
Figure 3.9: Average r-value and normal anisotropy coefficient (rn) variation with the
logarithm of nominal strain rate for the RD, DD, and TD.
parameters: normalized flow stress, the r-value, and the normal anisotropy ( nr ). Strain
rate significantly affected all parameters; the normalized flow stress was observed to
approach unity as the strain rate increases, and a similar trend was observed with the r-
value. Moreover, the r-value was also shown to be dependent on total true strain. This
behavior has been previously noted in many other steels; however, we believe that is
further accentuated in this material due to the non-homogeneous strain accumulation
associated with the yield-point-elongation (YPE) and dynamic strain aging as well as
austenite transformation localized to propagative instabilities reported in Abu-Farha et al.
[11].
While results from the tensile tests are instructive for identifying strain rate
dependence on anisotropy, they do not establish the role of underlying microstructural
103
mechanisms responsible for the observed macro-scale behavior. Indeed, there are
observations from the extant literature that are suggestive of some of these mechanisms
which likely are at work in the Med. Mn. TRIP steel of interest in this paper. Some of
these mechanisms, such as the effect of strain rate on texture evolution and austenite
transformation to martensite were alluded to in the course of discussing the observed
anisotropy dependence on strain rate. Additional data from electron backscatter
diffraction (EBSD) and X-ray (or neutron) diffraction will be required to address this
issue. For example, variations of austenite volume fraction along with texture evolution
in the RD, DD, and TD over the range of strain rates relevant for automotive applications
would provide a more substantive basis for interpretation of the tensile results presented
herein. A clearer picture of how strain rate affects -martensite formation and subsequent
'-martensite formation would also be forthcoming. A repeat of these measurements for
different deformation modes (e.g., biaxial, plane strain, shear) would provide insights into
how anisotropy evolves in forming processes wherein uniaxial tension is not the
predominant deformation mode. The addition of a high rate infrared camera coupled to
the DIC system would give needed temperature information that can be coupled to
anisotropy and the martensitic transformation. Further investigations should focus on
quantifying the effect of strain rates on the shape and size of the yield locus requiring
calibration of different anisotropic yield functions using uniaxial and biaxial experimental
data. This should be the initial step in developing a rate-dependent anisotropic yield
function.
104
Major conclusions from this study are as follows:
[1] The flow response of the Med. Mn (10 wt.%) TRIP over the 0.005 s-1 to 500 s-1 range
of nominal strain rates is characterized by a protracted yield point elongation followed by
a hardening stage that precedes fracture. Propagative instabilities resembling Type A
Portevin-Le Châtelier bands are suggestive of dynamic strain aging and appear at all
tested strain rates except 500 s-1.
[2] During uniaxial tensile testing of the Med. Mn. TRIP, strain hardening decreased with
increasing strain rate due to the suppression of the austenite to martensite transformation
which was likely a result of adiabatic heating. This resulted in an orientation dependent
drop ranging ~16% to 30% of ultimate tensile strength between 0.005 s-1 the elevated
strain rates.
[3] At high strain rates (0.5 s-1 and greater) the anisotropy of the Med. Mn TRIP steel was
diminished, as demonstrated by the change in the normalized flow stresses and r-values
with increased strain rate, i.e., both approached unity indicative of perfect isotropy.
Additionally, the r-values at strain rates other than the quasi-static rates tend to group
together; these two effects are a result of the mitigation of the material texture effect at
high strain rates (> 0.5 s-1).
[4] The increase in the coefficient of normal anisotropy ( nr ) of the Med. Mn. (10 wt.%)
TRIP sheet at elevated strain rates indicates that the resistance to thinning of the material
is improved at elevated strain rates, which suggest that this aspect of formability is
improved relative to quasi-static strain rates.
105
[5] Anisotropy observed in this study has the following implications for the calibration of
anisotropic yield criteria: the calibration of yield criteria using quasi-static data is valid
only at quasi-static deformation; the changes in the normalized flow stress and the r-value
directly affect the shape and size of the yield locus; to accurately model the anisotropy
during forming, rate-dependent anisotropic yield criteria are necessary.
3.8 References
[1] S. Keeler and M. Kimchi, WorldAutoSteel, V5, 2015.
[2] C. M. E. Curt D. Hovarth, Jatinder P. Singh, Jason J. Coryell, in International
Symposium on New Developments in Advanced High-Strength Sheet Steels, 1
(2017).
[3] M. Rashid, Annu. Rev. Mater. Sci., 11, 245 (1981).
[4] V. F. Zackay, E. R. Parker, D. Fahr, and R. Busch, ASM Trans Quart, 60, 252
(1967).
[5] E. De Moor, P. Gibbs, J. Speer, D. Matlock, and J. Schroth, Iron Steel Technol.,
7, 132 (2010).
[6] A. Clarke, J. Speer, M. Miller, R. Hackenberg, D. Edmonds, D. Matlock, et al.,
Acta Mater., 56, 16 (2008).
[7] P. Gibbs, E. De Moor, M. Merwin, B. Clausen, J. Speer, and D. Matlock, Metall.
Mater. Trans. A, 42, 3691 (2011).
[8] X. Hu, X. Sun, L. G. Hector, and Y. Ren, Acta Mater., 132, 230 (2017).
[9] S. Takaki, H. Nakatsu, and Y. Tokunaga, Mater. Trans., JIM, 34, 489 (1993).
106
[10] B. C. De Cooman, P. Gibbs, S. Lee, and D. K. Matlock, Metall. Mater. Trans. A,
44, 2563 (2013).
[11] X. H. F. Abu-Farha, X. Sun, Y. Ren, L.G. Hector, Jr, G. Thomas, T. Brown,
Metall. Mater. Trans. A, in press., (2018).
[12] G. Olson and M. Cohen, Metall. Mater. Trans. A, 6, 791 (1975).
[13] J. Benzing, W. Poling, D. Pierce, J. Bentley, K. Findley, D. Raabe, et al., Mater.
Sci. Eng., 711, 78 (2018).
[14] P. Zavattieri, V. Savic, L. Hector Jr, J. Fekete, W. Tong, and Y. Xuan, Int. J.
Plast., 25, 2298 (2009).
[15] W. Wu, Y.-W. Wang, P. Makrygiannis, F. Zhu, G. A. Thomas, L. G. Hector, et
al., Mater. Sci. Eng., 711, 611 (2018).
[16] V. Tarigopula, O. S. Hopperstad, M. Langseth, A. H. Clausen, and F. Hild, Int. J.
Solids Struct., 45, 601 (2008).
[17] H. Huh, S.-B. Kim, J.-H. Song, and J.-H. Lim, Int. J. Mech. Sci., 50, 918 (2008).
[18] H. Huh, H. Lee, and J. Song, Int. J. Automot. Technol, 13, 43 (2012).
[19] R. Alturk, S. Mates, Z. Xu, and F. Abu-Farha, TMS 2017 146th Annual Meeting
& Exhibition Supplemental Proceedings, 243 (2017).
[20] S. Xu, D. Ruan, J. H. Beynon, and Y. Rong, Mater. Sci. Eng., 573, 132 (2013).
[21] J.-H. Kim, D. Kim, H. N. Han, F. Barlat, and M.-G. Lee, Mater. Sci. Eng., 559,
222 (2013).
[22] J. Qin, R. Chen, X. Wen, Y. Lin, M. Liang, and F. Lu, Mater. Sci. Eng., 586, 62
(2013).
107
[23] X. Yang, L. G. Hector, and J. Wang, Exp. Mech., 54, 775 (2014).
[24] S. Li, D. Zou, C. Xia, and J. He, Steel Res. Int., 87, 1302 (2016).
[25] B. C. Hwang, T. Y. Cao, S. Y. Shin, S. H. Kim, S. H. Lee, and S. J. Kim, Mater.
Sci. Technol., 21, 967 (2013).
[26] S. F. Peterson, M. Mataya, and D. Matlock, JOM, 49, 54 (1997).
[27] M. P. Pereira and B. F. Rolfe, J. Mater. Process. Technol., 214, 1749 (2014).
[28] I. Choi, D. Son, S.-j. Kim, D. K. Matlock, and J. G. Speer, SAE Technical Paper
0148-7191 (2006).
[29] H. Hayashi and T. Nakagawa, J. Mater. Process. Technol., 46, 455 (1994).
[30] D. Gerbig, A. Srivastava, S. Osovski, L. G. Hector, and A. Bower, Int. J. Fract.,
209, 1 (2017).
[31] F. Andrade, M. Feucht, A. Haufe, and F. Neukamm, Int. J. Fract., 200, 127
(2016).
[32] I. Choi, D. Kim, S. Kim, D. Bruce, D. Matlock, and J. Speer, Met. Mater. Int., 12,
13 (2006).
[33] S. Oliver, T. B. Jones, and G. Fourlaris, Mater. Charact., 58, 390 (2007).
[34] B. A. Gama, S. L. Lopatnikov, and J. W. Gillespie Jr, Appl Mech Rev, 57, 223
(2004).
[35] J. Van Slycken, P. Verleysen, J. Degrieck, L. Samek, and B. De Cooman, Metall.
Mater. Trans. A, 37, 1527 (2006).
[36] G. T. Gray III, ASM Handbook, Mechanical testing and evaluation, 8, 462
(2000).
108
[37] H. Huh, W. Kang, and S. Han, Exp. Mech., 42, 8 (2002).
[38] W. W. Chen and B. Song, Split Hopkinson (Kolsky) bar: design, testing and
applications: (New York, Springer Science & Business Media, 2010).
[39] S. Mates and F. Abu-Farha, Dynamic Behavior of Materials, 1, 155 (2016).
[40] J. Huh, H. Huh, and C. S. Lee, Int. J. Plast., 44, 23 (2013).
[41] M. T. Rahmaan, "Low to high strain rate characterization of DP600, TRIP780,
AA5182-O," University of Waterloo, 2015.
[42] W. Li, J. Zhu, Y. Xia, and Q. Zhou, ASME 2015 International Mechanical
Engineering Congress and Exposition, 2015.
[43] V. Savic, L. Hector, U. Basu, A. Basudhar, I. Gandikota, N. Stander, et al., SAE
Technical Paper 0148-7191, 2017.
[44] E. ISO, "6892-1. Metallic materials-Tensile testing-Part 1: Method of test at room
temperature," International Organization for Standardization, 2009.
[45] B. Yan, Y. Kuriyama, A. Uenishi, D. Cornette, M. Borsutzki, and C. Wong, SAE
Technical Paper 0148-7191, 2006.
[46] M. Borsutzki, D. Cornette, Y. Kuriyama, A. Uenishi, B. Yan, and E. Opbroek,
Rep. Proc., Annu. Conf. [Int. Iron Steel Inst.], 30, (2005).
[47] Y. Wang, H. Xu, D. L. Erdman, M. J. Starbuck, and S. Simunovic, Adv. Eng.
Mater., 13, 943 (2011).
[48] ISO, "ISO_26203-Metallic materials — Tensile testing at high strain rates," ed,
2011.
109
[49] S.-E.-P. S. d. S. VDEh, "SEP1230. The determination of the mechanical
properties of sheet metal at high strain rates in high-speed tensile tests," ed, 2006.
[50] H. Schreier, J.-J. Orteu, and M. A. Sutton, Image correlation for shape, motion
and deformation measurements, (New York, Springer US, 2009).
[51] R. Alturk, W. E. Luecke, S. Mates, A. Araujo, K. Raghavan, and F. Abu-Farha,
Procedia Eng., 207, 2006 (2017).
[52] J. Talonen, H. Hänninen, P. Nenonen, and G. Pape, Metall. Mater. Trans. A, 36,
421 (2005).
[53] J. A. Lichtenfeld, C. J. Van Tyne, and M. C. Mataya, Metall. Mater. Trans. A, 37,
147 (2006).
[54] M. Isakov, S. Hiermaier, and V.-T. Kuokkala, Metall. Mater. Trans. A, 46, 2352
(2015).
[55] T.-H. Lee, H.-Y. Ha, J.-Y. Kang, J. Moon, C.-H. Lee, and S.-J. Park, Acta Mater.,
61, 7399 (2013).
[56] M. Eskandari, A. Zarei-Hanzaki, M. Mohtadi-Bonab, Y. Onuki, R. Basu, A.
Asghari, et al., Mater. Sci. Eng., 674, 514 (2016).
[57] D. Goodchild, W. Roberts, and D. Wilson, Acta Metall., 18, 1137 (1970).
[58] P. Zavattieri, V. Savic, L. Hector, J. Fekete, W. Tong, and Y. Xuan, Int. J. Plast.,
25, 2298 (2009).
[59] J. Min, L. G. Hector, L. Zhang, J. Lin, J. E. Carsley, and L. Sun, Mater. Sci. Eng,
673, 423 (2016).
[60] W. Lankford, Trans. ASM, 42, 1197 (1950).
110
[61] W. F. Hosford, Int. J. Mech. Sci., 27, 423 (1985).
[62] G. Huang, B. Yan, and Z. Xia, SAE Int. J. Mater., Mech. Manuf., 4, 385 (2011).
[63] D. Daniel, J. J. Jonas, and J. Bussiére, Texture, Stress, and Microstructure, 19,
175 (1992).
[64] R. Arthey and W. Hutchinson, Metall. Mater. Trans. A, 12, 1817 (1981).
111
CHAPTER FOUR
4. A RATE DEPENDENT ANISOTROPIC YIELD FUNCTION
ABSTRACT
The results of the previous chapter confirmed the strain rate dependence of the
anisotropic properties of Medium Mn. (10 wt. %) steel. To quantify this dependence, the
calibration of an anisotropic yield function is needed. First, a comprehensive review of
the different yield functions in the literature is provided, showing the strength and
weaknesses associated with each of them. The result of that is a systematic choice of an
anisotropic yield function that can accurately model the strength and strain anisotropic
nature of Medium Mn. (10 wt. %) steel. The experimental results obtained from the
previous chapter are used to calibrate at each strain rate, allowing for plotting the yield
locus at each strain rate. The shape and size of the yield loci were observed to change
with the varying strain rate; thus showing a strong dependence on the strain rate. This
provided the motivation to develop a methodology to model the strain rate dependence of
the yield locus. As mentioned in Chapter 3, there are currently no anisotropic functions
in the literature that can predict the effect of strain rate on anisotropy. Therefore, a
methodology to model the rate dependent anisotropy of materials is proposed at the end
of this chapter. Followed by the detailed numerical procedure of a material card that can
capture the strain rate in the hardening and anisotropy of Medium Mn. (10 wt. %) steel.
112
`4.1 Introduction
Accurate modeling of the behavior of the materials in FEA simulations is very
crucial, especially when the materials in question are AHSSs. This is mainly due to: 1)
the desire to optimize the amount of material needed (and therefore the cost) by
increasing the accuracy of the simulations, and 2) the complex nature of these materials
and the difficulty in predicting their responses to different loading conditions.
Models to predict the response of materials (constitutive models) can be divided
into two types: physical models and phenomenological models. While physical models
are more accurate, yet they have a very complex nature and are difficult and costly to
implement in a finite element code. Therefore, phenomenological models are more
widely used in the automotive industry as they require calibration of a fewer number of
parameters and are relatively more straightforward to implement in a finite element code.
A phenomenological model consists of three components [1]: 1) a yield function, 2) a
flow rule and 3) a hardening law. The yield function is an expression to define the
relationship between the different stress component at the onset of yielding, and through
plastic deformation. The flow rule is an expression which defines the relationship
between the components of strain rate and stress. Finally, the hardening law defines how
the initial yield stress evolves during plastic deformation [2].
In chapter 2, it was shown that the range of different microstructure in the
different AHSSs affects their strain rate sensitivity at speeds close to that encountered
during a crash event (hardening). In chapter 3, it was also shown that the anisotropy of
the material changes at different strain rates (yield function). In this chapter, a modeling
113
approach will be devised to incorporate the strain rate sensitivity in both the hardening
and anisotropy of Medium Mn. (10 wt. %) steels during crash simulations.
4.2 Overview of yield functions
Yield functions define the stress state at which the elastic behavior of the material
ends, and plastic flow begins. During a uniaxial tensile test, this can be easily found by
looking at the stress-strain curve and finding the stress at which plastic flow begins.
However, during a multi-axial loading condition, such as that encountered during a crash
scenario, it is not as straightforward as it is in the case of a uniaxial test. Therefore, an
implicit relationship between the principal stresses is needed to define the onset of plastic
flow, this is known as the yield function. The function will represent a three-dimensional
surface of the principal stresses. The elastic state of the material includes all the points
inside the surface and the points on the surface itself present the plastic state. Points are
not allowed to be outside the surface. One of the oldest and most used yield functions is
the “maximum shear stress criterion” proposed by Tresca [3] in 1864 and the “maximum
distortion energy criterion” by Von Mises [4] in 1913 [5]. The former defines the onset of
plastic flow as when the maximum shear stress reaches a critical value, equation 4.1
1 2 2 3 3 1max , , y (4.1)
where 1 2 3, , are the stresses in the three principal directions and y is the yield stress.
Von Mises function, on the other hand, is a quadratic relation (instead of linear with
Tresca) of the second deviatoric stress invariant, equation 4.2. The basis of the function is
114
that the plastic flow of the material begins when the elastic energy of distortion reaches a
critical value.
2 2 2 2
1 2 3 2 1 3 2 y (4.2)
These functions were followed by other formulations such as Drucker [6] in 1949
which tried to find a formulation to model the material with results lying between Tresca
and von Mises. Followed by Hershey [7] in 1954 which suggested a generalized
formulation of the von Mises equation. Nonetheless, all these equations were proposed on
the assumption of isotropy in the material properties. However, experimental results
showed that some materials exhibited a strong anisotropy in their behavior and therefore
there was a need for an anisotropic yield function. In 1948, Hill [8] made an
improvement on the generalized von Mises equation by adding parameters that depend on
anisotropic properties that can be derived using experimental methods (the yield stresses
in rolling, transverse and diagonal directions, or the r-values in the three directions) as
shown in equation 4.3
2 2 2 2 2 2
22 33 11 22 33 11 23 31 122 ( ) 2 2 2 1ijf F H G L M N
(4.3)
where F, H, G, L, M, N are the new parameters that introduce anisotropy to the equation.
These parameters can be calculated by using either the three r-values or the yield stresses
in the three directions relative to the rolling directions. That means that this equation
assumes an inherent relation between the r-values and the yield stress. Although this is
true for some materials, it is not a universal rule. This is a significant drawback, as some
materials show anisotropy in the yield stresses and the r-values in varying magnitudes
115
(such as Medium Mn. (10 wt. %) shown in chapter 3), and Hill 1948 definitely can’t
capture that behavior. Other drawbacks were highlighted by another researcher when they
used Hill 1948 for the calibration of their materials, such as Pearce [9], Woodthorpe and
Pearce [10] and Banabic et al. [11]. Due to these reasons and others that are detailed in
the references, Hill made improvements to his earlier equation and developed the Hill
1979 yield criterion [12]. The new Hill yield criterion followed a non-quadratic
formulation, and it improved on the original yield criterion while maintaining some of the
simplicity. Nonetheless, there were still drawbacks about this equation, mainly its
assumption that 0 90r r and
0 90 . Due to this, Hill proposed two other yield criteria in
1990 [13] and 1993 [14]. Some of the problems mentioned were resolved by one of these
models, but it came at the cost of increased complications in the formulation (and
therefore higher computational cost) as in the case of Hill 1990 or with the introduction
of more assumptions as in the case of Hill 1993 [2].
The Hill family of yield criteria used the von Mises yield criterion as a basis for
their formulations. There exist another family of yield criteria that used Hershey’s model
[7] as a basis for their formulations. The first person to do this was Hosford [15] in 1972,
who proposed a function that was a generalization of Hershey’s criteria (equation 4.4)
22 33 33 11 11 22
a a a aF G H (4.4)
Looking at this equation and Hill’s equation, one can notice some similarities
between them, yet the main difference is the identification of the parameter “a” in
Hosford’s and “m” in Hill 1979 and beyond. “a” is determined based on the
crystallographic structure of the material [16-18]. Leading to a connection between the
116
microstructure and the yield function, which was considered a significant advantage of
this yield criterion and all Barlat yield criteria as they used equation 4.4 as a basis for
their formulations.
The first criterion developed by Batlat was published in 1987 with Richmond
[19], which used Hershey’s model as a basis with a modification in using generalized
coordinates rather than material coordinates. This criterion was simple and suffered
multiple drawbacks, most importantly is its limitations in predicting the onset of yield for
other than uniaxial conditions [2]. In 1991, Barlat [20] proposed a new yield criterion
(Yld91) formulated in terms of the deviatoric principal stresses (equation 4.5)
1 2 2 3 1 3 2m m m m
yS S S S S S (4.5)
This model was simple, general and easy to manipulate to fit varying conditions,
yet it could not predict the yielding of different kinds of materials. Therefore, Barlat tried
to improve on the 1991 model and proposed a model in 1997 (Yld94) [21] by adding
weighting parameters, in front of the deviatoric principal stresses, that are a function of
the anisotropic properties. These parameters introduced a right amount of flexibility to
the model to allow it to predict the anisotropic yielding of a wide range of materials. Yet,
the model suffered from complexities in application to numerical codes that limited its
use as compared to Yld91 [22]. Other models that were proposed in the literature include
BBC [23], Karafillis-Boyce [24], Budiansky [25], Gotoh [26, 27] and CPB family [28-
30].
117
4.3 Barlat Yld2000-2d
Even with the good applicability of Yld91 and its common use in finite element
simulations of anisotropic materials, there were still some drawbacks associated with it.
The most important of those are the lack of proof of convexity and the inaccuracy in
describing yielding for complex loading conditions [31]. Therefore in 2003, Barlat et al.
[31] developed a new yield function (Yld2000-2d) that can address these drawbacks.
Barlat et al. wanted the yield function to satisfy two primary conditions: 1) the function
must have at least seven parameters to take into account the three uniaxial yield stresses
in the three directions, the r-values in the three directions and the biaxial yield stress. 2)
the function must satisfy the convexity condition.
To ensure convexity, Barlat et al. used an isotropic yield function that was
proposed by Hershey in 1954 [7] and Hosford 1972 [32] which was made up of the sum
of two functions (equation 4.6)
2 a (4.6)
where
1 2 2 1 1 2........ 2 2a a a
s s s s s s (4.7)
This function was made to account for anisotropy by using a linear transformation on the
deviatoric stress tensors, which can be written in terms of the Cauchy stress tensors. This
would lead into two linearly transformed stress tensors ,X X , where
. . . .
. . . .
X C s C T L
X C s C T L
(4.8)
118
where T is the transformation matrix for the Cauchy stress tensor to the deviatoric stress
tensor. Consequentely, this will lead to the replacement of 1s ,
2s with X and
X respectively. With certain conditions imposed on the coefficients of the ,C C , the
,L L matrices reduce to:
11
12 1
21 2
22 7
66
2 / 3 0 0
1/ 3 0 0
0 1/ 3 0
0 2 / 3 0
0 0 1
L
L
L
L
L
(4.9)
311
12 4
21 5
22 6
66 8
2 2 8 2 0
1 4 4 4 01
4 4 4 1 09
2 8 2 2 0
0 0 0 0 9
L
L
L
L
L
(4.10)
The coefficients 1 8 can be found using the seven experimental values
mentioned above, and the extra coefficient can be determined using the biaxial r-value.
The yield function now reduces to the same form as equation (4.6) with the replacement
of ,X X in place of 1 2,s s in equation (4.7). This final form of the yield equation
satisfies the two conditions stated above and on top of that provides an equation that is
relatively easier to derive than Yld91 [31].
Yld2000-2d have proved to be one of the most accurate models that can predict
anisotropy in different materials showing varying degrees of anisotropy in different
properties while keeping the application in a numerical code to a practical level for
industrial applications. Nonetheless, the numerical complexity of implementing this
119
function is still higher than other equations. In regards to the anisotropy observed in
Medium Mn. (10 wt. %), it was observed that it showed anisotropy with both the flow
stress (strength) and the r-value (strain). For that reason, Yld2000-2d was chosen to be
the basis for the modification that will allow for the development of a rate-dependent
anisotropic yield function.
4.4 Calibration of Yld2000-2d
In order to fully calibrate the Yld2000-2d, eight experimental values are needed.
Namely, the yield stresses in the three rolling directions, the r-values in three rolling
directions, the biaxial yield stress and the biaxial r-value. The first six parameters were
obtained at the different strain rates in the previous chapter, while the biaxial yield stress
and r-value at 0.005s-1 were obtained using a hydraulic-bulge test. There are no
experimental capabilities available to allow biaxial testing at high strain rates. Therefore
the biaxial yield stress at strain rates higher than 0.005s-1 was estimated based on how
45 and 90 changed with strain rates with respect to 0 . This necessarily induces some
inaccuracies in the results [33, 34], yet that is a limitation that future research will be
focused on. As for the r-value at the high strain rates, it was calculated analytically using
Yld96 [35]. The calculation was done for the 0.005s-1, and it was very close to the
experimental value. The values that are used in the calibration of the Yld2000-2d are
shown in table 4.1.
120
Table 4.1: Experimental results used to calibrate Yld2000-2d
Strain Rate (s-1) 0
0
90
0
45
0
0r 90r
45r 0
b
br
0.005 1 1.014 1.017 0.47 0.55 0.78 1.240 0.85
0.5 1 1.036 0.958 0.58 0.69 0.70 1.211 0.88
5 1 1.052 0.962 0.58 1.03 1.05 1.210 0.92
50 1 1.080 0.985 0.52 0.85 1.25 1.229 0.95
500 1 1.092 0.975 0.64 0.90 0.93 1.115 0.98
Using a numerical procedure that was developed based on the work done in the
paper published by Barlat et al. [31, 36], the eight parameters (
1 8
) were found for
each strain rate, and the values are shown in table 4.2. It can be deduced that the strain
rate has a significant effect on the value of these eight parameters in varying magnitudes,
with the maximum effect being observed with 3 (~46% increase). This suggests a
powerful influence of strain rate on the anisotropy and the shape of the yield locus, and
therefore ultimately on the accuracy of using the quasi-static parameters to model the
behavior at high strain rates. To further quantify this, the parameters obtained at each
strain rate were used to plot the yield locus on the normalized stress axes as shown in
Figure 4.1. The dependency of the yield locus on the strain rate is significantly apparent
in Figure 4.1. The yield locus moves closer towards isotropy with the increase in strain
rate as the results of the previous chapter led to conclude. This further asserts the
inaccuracies of calibrating yield functions using quasi-static experimental results to
model material behavior at high strain rates as in the case with crash simulations.
121
Table 4.2: The calibrated
1 8
for each strain rate
Strain Rate (s-1) 1 2
3 4
5 6
7 8
0.005 0.7412 0.9234 0.5559 0.9264 0.9442 0.5415 0.9156 1.2494
0.5 0.8586 0.9206 0.6428 0.9173 0.9446 0.5866 0.9379 1.3832
5 0.7774 1.0321 0.7159 0.8940 0.9426 0.5659 1.0001 1.3142
50 1.0080 0.9559 0.8098 0.9214 0.9447 0.7527 1.0341 1.1488
500 0.9241 0.9841 0.8414 0.9645 0.9642 0.7367 1.0144 1.1686
4.5 Rate Dependent Yld2000-2d
The experimental testing results shown in chapter 3 and the calibration results of
the Yld2000-2d both exemplifies the motivation behind developing the methodology to
numerically model the anisotropic nature of materials at multiple strain rates. The
introduction of a strain rate dependent parameter in the formulation of an anisotropic
yield function is a complex task, especially when the function in question is Yld2000-2d.
Yet the observation made by Li et al. [34, 37] regarding the rate dependency of
the parameters of Hill48 when used to model an HSLA steel, inspired the investigation of
this phenomenon for state-of-the-art developmental steel with a complex microstructure
and behavior (Medium Mn. steel). The results shown in table 4.2 and visualized below in
Figure 4.2 showed that indeed the parameters of Yld2000-2d are strain rate dependent.
The next question that comes to mind is whether that dependency can be represented by a
mathermatical relation. After different mathematical forms of equations were tried, the
optimum form that was selected to best model the trend of the parameters is the cubic
form shown below
3 2
0
(log ) (log ) (log )a b c d
(4.11)
122
where 0 is the parameter calibrated with quasi-static results.
The equation developed for the parameters of Yld2000-2d were used to calculate
the
1 8
at the different strain rates and used these values to plot the yield locus at
different strain rates. The results are shown in Figures 4.3-4.6, wherein each figure three
yield loci are shown. A yield locus plotted using the 0.005s-1 results, a yield locus plotted
Figure 4.1: Yield locus calibrated using data from experiments at different strain rates
123
using the results from the strain rate shown, and a yield locus plotted using equation
(4.11) for the strain rate shown. Figure 4.3 shows the comparison of the three yield loci
between the 0.005s-1 and the 0.5s-1. It can be observed that the yield locus plotted using
the parameters obtained by the equation suggested is in a good agreement with the yield
locus plotted using the experimental results performed at 0.5s-1. While the yield locus
plotted using the experimental results obtained at 0.005s-1 varies in shape and size from
the yiled locus plotted using the experimental results performed at 0.5s-1. The variation is
significant especially where the arrows are drawn.
Figure 4.2: Variation of the Yld2000-2d constants (
1 8
) with the strain rate
124
As the strain rate increases to 5, 50 and 500s-1, the difference between the loci
plotted using the experimental results at the specified strain rates and the locus plotted
using the experimental results obtained at 0.005s-1 starts to increase, while the yield loci
Figure 4.3: Comparison of yield locus plotted using the experimental data at 0.005s-1
(pink), at 0.5s-1 (blue) and using the equation (green).
125
plotted with the parameters obtained using equation (4.11) maintains an excellent
agreement at all strain rates.
Figure 4.4: Comparison of yield loci plotted using the experimental data at 0.005s-1
(pink), at 5s-1 (blue) and using the equation (green).
126
Figure 4.5: Comparison of yield loci plotted using the experimental data at 0.005s-1
(pink), at 50s-1 (blue) and using the equation (green).
4.6 Numerical Procedure
4.6.1 Flow Rule
As the materials yield and begin plastic deformation, hardening begins, and a
relation is required to describe how the stress respond to plastic strain increments. This
relation, or a flow rule, usually takes the form shown in equation (4.12) [38]
127
( )ij
ij
dgd d
d
(4.12)
where g is called the plastic potential and d is a scalar function. This general form of
the flow rule describes what is known as a non-associated flow rule. In the particular case
where the plastic potential is the yield function, this becomes an associated flow rule. The
associated flow rule is based on the normality principle which was proposed by Drucker
in 1951 [39]. The principle states that the increment in plastic strain is outward normal to
Figure 4.6: Comparison of yield loci plotted using the experimental data at 0.005s-1
(pink), at 500s-1 (blue) and using the equation (green).
128
the yield locus surface at any given point. This allows for the replacement of the plastic
potential with the yield function and thus simplifying the mathematical formulation. The
use of associated flow rule with different materials have been researched and verified for
metals but not for materials like rock and soil [40].
Another relation is required to adequately describe the behavior of the material,
which is the hardening law. There are a significant amount of work in the literature that
studies the effectiveness of different hardening laws to capture the strain rate sensitivity
exhibited by different materials, such as Johnson Cook (JC) [41-43], Reyes [44, 45],
Zerilli and Armstrong [46], Combined Holloman and Voce equation [47] and many
others. The selection of the equation depends mainly on the material chosen as it was
shown in Chapter 2 that the microstructure, significantly influences the material’s
behavior at multi-strain rates. There is no widely acceptable equation for AHSSs at multi-
strain rates conditions. That is mainly because of the inhomogeneity of the response of
the material’s hardening properties with strain, as was strongly exemplified with the
chosen material. With the results observed for Medium Mn. (10 % wt.) steel, it was
decided to code the material using the hardening curves at different strain rates as the
input to obtain the values otherwise obtained using a hardening equation. The decision
was made based on the high accuracy of the experimental results obtained, and the range
of strain rates at which these experiments were performed. The LS-DYNA user manual
describes the procedure to code a material card using hardening curves at multi-strain
rates instead of a rate dependent hardening law.
129
4.6.2 Stress Integration
The strain rate dependent Yld2000-2d developed and the associated flow rule will
be used to predict the material properties and develop a numerical code that can be used
to code a user-defined material card in LS-DYNA (UMAT) to model the behavior of
Medium Mn. steel during crash simulations. The code must have an algorithm defined to
integrate the constitutive relation at each integration point in the elements in the finite
element model [48, 49]. Therefore, a stress integration algorithm must be selected and
used to develop the code [50, 51]. The two most well-known algorithms for stress
integration are: 1) the closest point projection iterative method, and 2) the cutting-plane
algorithm [52, 53]. The closest point projection iterative method is usually applied to
relatively simple constitutive relations as otherwise, it will lead to computationally
expensive codes and complex mathematical relations to develop [54]. While the cutting-
plane algorithm (shown in Figure 4.7) was proposed to avoid the complex codes and
mathematical operations by abandoning the need to calculate the gradient of the flow rule
and the yield function in the formulation [55]. Therefore, the cutting-plane algorithm
based on the incremental deformation theory of plasticity was selected to write the code
as described by Abedrabbo et al. [53].
The incremental deformation theory of plasticity was selected as it provides
significant advantages over the total deformation theory by avoiding some unnecessary
steps that might affect the accuracy of the formulation as observed by Abedrabbo et al.
[53] following the work of Ortiz and Simo [56, 57]. The incremental theory of plasticity
was applied by researchers starting with Chung and Richmond [58] that used the
130
minimum plastic work path to integrate the constitutive relations, and Yoon et al. [36, 59]
which used the same theory to integrate Yld96 and its associated flow rule. In the
incremental deformation theory, the material’s strain increments are equal to the true
strain increments, and the rotation of the material is also found at each step through polar
decomposition to find the current incremental angle [53].
In this work, the cutting plane algorithm was used to numerically integrate the
constitutive equation with the associated flow rule. In LS-Dyna, the strain increment n ,
the previous stress state value 1n are provided at the beginning of each time step. The
strain is assumed to be elastic, and therefore the stress is calculated through the elastic
relations. That trial stress and the strain rate are used to calculate the stress state using the
newly developed strain rate dependent Yld2000-2d, and the size of the yield function is
calculated using the hardening flow rule. A check is performed to test whether the
calculated trial stress lies inside the yield surface, if the condition is met, then the trial
stress is the actual true stress, however, if it did not then the material has yielded, and the
stress state is plastic. The Newton-Raphson method is then used to iteratively return the
trial stress to the yield surface by calculating the normality parameter (λ), which tries to
“cut” the stress back to the yield surface, and thus the name ‘the cutting plane algorithm’.
Figure 4.8 is a flowchart explaining the flow of calculations used in the stress
integration process. The first step is to read the history variable which stores the values
for the current strain increment n and the previous stress
1n . It is assumed that the
current strain increment is elastic and therefore, the current stress can be calculated based
on the elasticity equation (4.13)
131
1 :n n nC (4.13)
where C is the elasticity matrix. The strain rate is calculated using the incremental strain,
and the strain rate is used to calculate the coefficients (
1 8
) using the equations
developed in this chapter. Then the stress calculated, which will be called “initial stress”,
and the strain rate dependent (
1 8
) will be used to calculate the value of the yield
function using equations (4.6-4.10) and the derivative of the yield function using the
steps below:
1
1 2 1 2 1 2
1( )( 2 2 )2
aa a aX X X X X X
(4.14)
where is the effective stress. The next step is calculating the prime values for ,X X
matrices.
Figure 4.7: A diagram explaining the process of the cutting-plane algorithm
132
2
2
1
222
( )
2 2[ ]
( )
2 2
xx yy xx yy
xy
xx yy xx yy
xy
X X X XX
X
XX X X X
X
(4.15)
2
2
1
22
2
( )
2 2[ ]
( )
2 2
xx yy xx yy
xy
xx yy xx yy
xy
X X X XX
X
X X X X XX
(4.16)
Equation (4.14) can then be differentiated using the chain rule
1 2 1 2 1 2
1
( 2 2 )1
2
a a a
a
X X X X X X
a
(4.17)
2 3
11 1
1( )
2
nc d c d
ac d i k c d k c d k
X X
a X X
(4.18)
where d is xx, yy, and xy.
Once these two values are calculated, the code is instructed to find the value of
the flow stress and the derivative using on the input hardening stress-strain curve based
on the current strain and strain rate. The first check is now performed to determine if the
effective stress calculated by the yield function is elastic or not (i.e. if the stress state falls
within or on the yield locus) by comparing it with the stress obtained from the hardening
curves:
( , , ) ( , ) 0p p
n n n n n nH (4.19)
133
where ( , )p
n nH is the hardening stress obtained from the input curves. If the condition
does not hold, then the material is in the plastic region now, the “initial stress” calculated
is wrong and has to be adjusted and the strain increment is not totally elastic. In order to
achieve that, the normality principle and the associated flow rule are used. As a reminder
the flow rule is:
p F
(4.20)
where F is the yield function or in other words the effective stress and is what is
called the normality parameter [54]. It is required to find the value of to be able to
calculate the increment of the plastic strain. It should be noted that according to the
incremental deformation theory of plasticity, the normality parameter is equal to the
equivalent plastic strain [53]
:: p
p
(4.21)
Once is found, the new stress value can be calculated as follows:
: ( )p
n n nC (4.22)
moreover, it is known that:
1
p p p
n n (4.23)
substituting that into equation (4.22) leads to
1: ( ) :p p
n n nC C (4.24)
substituting the flow rule in place of p , and using the initial stress:
1 :i
i i nn n C
(4.25)
134
where the “i” donates the trial number, so i
n represents the trial stress calculated based
on elasticity, and the “i+1” is the new trial stress that has been “cut back” using equation
(4.25). The next step would be to again to repeat the procedure shown between equation
(4.14) to (4.19) and see if the new stress state is on the yield locus. If not, then the
process must be repeated again, starting with updating the normality parameter, which is
explained next.
Before the steps mentioned in the previous paragraph can be performed, the
normality parameter must be calculated. The method to calculate the normality parameter
will take advantage of linearization using Taylor’s expansion [53, 56, 57] of equation
(4.19)
( ) 1 ( 1) ( )0 ( , ) ( ) ( )i i
i i p i i i p i p i
n n n n n n
(4.26)
from equation (4.25):
1 :i
i i nn n C
(4.27)
and
( 1) ( )p i p i
n n (4.28)
substituting both back into equation (4.26)
( )0 ( , ) ( : )ii i
i i p i nn n C
(4.29)
Therefore, the normality parameter can be defined as:
( )( , )
: :
i i p i
n n
i i i
n n
p
HC
(4.30)
135
The linearized Taylor’s expansion on the yield function represents the tangent
lines shown in Figure 4.7. As mentioned before, this value of the normality parameter is
fed back into equation (4.25) to calculate the new stress value, which is again verified
using equation (4.19), however this time replacing the zero with a small number like (10-
9) which is the convergence criteria. Once that is met, the stress and the equivalent plastic
strain are returned to the FEM code to continue the simulation.
4.7 FEA Validation
The developed UMAT card is compiled and implemented in LS-DYNA. The first
step is to validate the accuracy of this model by running simulations of the experiments
performed at different strain rates and comparing the simulation results with the
experimental results. Figure 4.9 shows the finite element model of the tensile test
specimen used to run the simulations. The model consisted of fully integrated shell
elements with five integration point across the thickness. The elements had a size of
0.5mm, giving enough elements across the width (10) to ensure the good accuracy of the
results.
The first simulation was performed at a strain rate of 0.005s-1 using the Mat133 as
the material card, which is an already developed material card of Yld2000-2d in LS-
DYNA. Another simulation was performed at the same strain rate using the newly
developed material card, and load-displacement curves of both simulations were
compared with the experimental load-displacement curves shown in Figure 4.10.
136
Figure 4.8: Flowchart showing the steps followed in the cutting plane algorithm
The results shown confirm that the formulation and coding of the material card
are accurate and matches the coding of the LS-DYNA material card at quasi-static rates.
137
Figure 4.9: Finite element model used to verify the developed equation
The next simulation aimed at testing the validity of the strain-rate dependent Yld2000-2d
model developed in this chapter. Therefore, tensile test simulations were performed at the
strain rates at which the experimental data were performed, and again the load-
displacement of the simulations and the experiments were compared in Figure 4.11 and
Figure 4.12. In addition to that, load-displacement curves from the simulation run using
Mat133 (rate-independent Yld2000-2d) are also shown in the figure.
138
Figure 4.10: Comparison of the experimental results and simulations at 0.005s-1
Figure 4.11: Load vs. displacement comparison between the experiments and simulations
at 50s-1 using the Yld2000-2d calibration at 0.005s-1 (red curve) and using the developed
equation (green curve)
139
Figure 4.12: Load vs. displacement comparison between the experiments and simulations
at 500s-1 using the Yld2000-2d calibration at 0.005s-1 (red curve) and using the developed
equation (green curve)
The results shown in the figures validate that the material card was coded and
formulated correctly and that the equation developed provide a significant improvement
in the accuracy of predicting the behavior of the material at multi-strain rates
4.8 Conclusions and Summary
In this chapter, a new rate-dependent yield criterion was proposed to model the
anisotropy and hardening of Medium Mn. (10 % wt.) at multi-strain rates. In the
beginning, a comprehensive review of constitutive relations was performed, explaining
their different types and the different components. Then different yield criteria were
140
discussed, and the Yld2000-2d was chosen due to its accuracy and flexibility in capturing
different modes of anisotropy. Yld2000-2d was calibrated using experimental data from
multi-strain rates, and the results showed that the yield locus’ shape and size is strain rate
dependent and thus it is necessary to capture this dependency to improve the accuracy of
constitutive relations that are used to model similar anisotropic material at multi-strain
rates.
With that motivation, it was concluded that the parameters of the Yld2000-2d are
strain rate dependent and can be mathematically modeled in relation to the strain rate.
These relations were developed and used to develop a User Defined Material (UMAT)
card to be used in a commercial FEM software (LS-DYNA). The numerical procedure to
develop the code was described in detail with most of the mathematical manipulation
involved, and then the material card was compiled and implemented in LS-DYNA.
Several simulations were performed using the newly developed material card, the
results of which were compared with the experimental results showing an excellent
agreement between the two. The results highilighted the inaccuracies of using a yield
function calibrated using the quasi-static results to model the behavior of anisotropic
materials, in this case a third-generation AHSS, during loading at multi-strain rates.
4.9 References
[1] J. Hu, "Characterization and Modeling of Deformation, Springback, and Failure
in Advanced High Strength Steels (AHSSs)," 2016.
141
[2] D. Banabic, Sheet Metal Forming Processes: Constitutive Modelling and
Numerical Simulation, 1 ed. Berlin, Germany: Springer-Verlag Berlin Heidelberg,
2010.
[3] H. Tresca, "On the yield of solids at high pressures," Comptes Rendus Academie
des Sciences, vol. 59, p. 754, 1864.
[4] R. v. Mises, "Mechanik der festen Körper im plastisch-deformablen Zustand,"
Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen,
Mathematisch-Physikalische Klasse, vol. 1913, pp. 582-592, 1913.
[5] M. J. Michno Jr and W. N. Findley, "An historical perspective of yield surface
investigations for metals," International Journal of Non-Linear Mechanics, vol.
11, pp. 59-82, 1976.
[6] D. C. Drucker, "Relation of experiments to mathematical theories of plasticity,"
Journal of Applied Mechanics-Transactions of the Asme, vol. 16, pp. 349-357,
1949.
[7] A. Hershey, "The plasticity of an isotropic aggregate of anisotropic face-centered
cubic crystals," Journal of Applied Mechanics-Transactions of the Asme, vol. 21,
pp. 241-249, 1954.
[8] R. Hill, "A theory of the yielding and plastic flow of anisotropic metals," Proc. R.
Soc. Lond. A, vol. 193, pp. 281-297, 1948.
[9] R. Pearce, "Some aspects of anisotropic plasticity in sheet metals," College of
Aeronautics Cranfield1968.
142
[10] J. Woodthorpe and R. Pearce, "The anomalous behaviour of aluminium sheet
under balanced biaxial tension," International Journal of Mechanical Sciences,
vol. 12, pp. 341-347, 1970.
[11] D. Banabic, "Determination of yield loci cross tensile tests assuming various
kinds of yield criteria," in Proc. 20th Biennial IDDRG Congress, 2000, pp. 343-
349.
[12] R. Hill, "Theoretical plasticity of textured aggregates," in Mathematical
Proceedings of the Cambridge Philosophical Society, 1979, pp. 179-191.
[13] R. Hill, "Constitutive modelling of orthotropic plasticity in sheet metals," Journal
of the Mechanics and Physics of Solids, vol. 38, pp. 405-417, 1990.
[14] R. Hill, "A user-friendly theory of orthotropic plasticity in sheet metals,"
International Journal of Mechanical Sciences, vol. 35, pp. 19-25, 1993.
[15] W. F. Hosford, "On yield loci of anisotropic cubic metals," in Proceedings of the
Seventh North American Metal working Conference SME, 1979, pp. 191-197.
[16] R. W. Logan and W. F. Hosford, "Upper-bound anisotropic yield locus
calculations assuming-pencil glide," 1980.
[17] W. F. Hosford, "Comments on anisotropic yield criteria," International journal of
mechanical sciences, vol. 27, pp. 423-427, 1985.
[18] W. F. Hosford, "The mechanics of crystals and textured polycrystals," Oxford
University Press(USA), 1993, p. 248, 1993.
[19] F. Barlat and O. Richmond, "Prediction of tricomponent plane stress yield
surfaces and associated flow and failure behavior of strongly textured fcc
143
polycrystalline sheets," Materials Science and Engineering, vol. 95, pp. 15-29,
1987.
[20] F. Barlat, D. J. Lege, and J. C. Brem, "A six-component yield function for
anisotropic materials," International journal of plasticity, vol. 7, pp. 693-712,
1991.
[21] F. Barlat, R. Becker, Y. Hayashida, Y. Maeda, M. Yanagawa, K. Chung, et al.,
"Yielding description for solution strengthened aluminum alloys," International
Journal of Plasticity, vol. 13, pp. 385-401, 1997.
[22] D. Banabic, "Plastic behaviour of sheet metal," in Sheet Metal Forming
Processes, ed: Springer, 2010, pp. 27-140.
[23] D. Banabic, T. Kuwabara, T. Balan, D. Comsa, and D. Julean, "Non-quadratic
yield criterion for orthotropic sheet metals under plane-stress conditions,"
International Journal of Mechanical Sciences, vol. 45, pp. 797-811, 2003.
[24] A. Karafillis and M. Boyce, "A general anisotropic yield criterion using bounds
and a transformation weighting tensor," Journal of the Mechanics and Physics of
Solids, vol. 41, pp. 1859-1886, 1993.
[25] B. BUDIANSKY, "Anisotropic plasticity of plane-isotropic sheets," in Studies in
Applied Mechanics. vol. 6, ed: Elsevier, 1984, pp. 15-29.
[26] M. Gotoh, "A theory of plastic anisotropy based on a yield function of fourth
order (plane stress state)-I," 1977.
144
[27] M. Gotoh, "A theory of plastic anisotropy based on yield function of fourth order
(plane stress state)—II," International Journal of Mechanical Sciences, vol. 19,
pp. 513-520, 1977.
[28] O. Cazacu and F. Barlat, "A criterion for description of anisotropy and yield
differential effects in pressure-insensitive metals," International Journal of
Plasticity, vol. 20, pp. 2027-2045, 2004.
[29] O. Cazacu, B. Plunkett, and F. Barlat, "Orthotropic yield criterion for hexagonal
closed packed metals," International Journal of Plasticity, vol. 22, pp. 1171-1194,
2006.
[30] B. Plunkett, O. Cazacu, and F. Barlat, "Orthotropic yield criteria for description of
the anisotropy in tension and compression of sheet metals," International Journal
of Plasticity, vol. 24, pp. 847-866, 2008.
[31] F. Barlat, J. Brem, J. W. Yoon, K. Chung, R. Dick, D. Lege, et al., "Plane stress
yield function for aluminum alloy sheets—part 1: theory," International Journal
of Plasticity, vol. 19, pp. 1297-1319, 2003.
[32] W. Hosford, "A generalized isotropic yield criterion," Journal of Applied
Mechanics, vol. 39, pp. 607-609, 1972.
[33] J. Huh, H. Huh, and C. S. Lee, "Effect of strain rate on plastic anisotropy of
advanced high strength steel sheets," International Journal of Plasticity, vol. 44,
pp. 23-46, 2013.
[34] W. Li, J. Zhu, Y. Xia, and Q. Zhou, "Testing and modeling the effect of strain-
rate on plastic anisotropy for a traditional High Strength Steel," in ASME 2015
145
International Mechanical Engineering Congress and Exposition, 2015, pp.
V009T12A084-V009T12A084.
[35] F. Barlat, Y. Maeda, K. Chung, M. Yanagawa, J. Brem, Y. Hayashida, et al.,
"Yield function development for aluminum alloy sheets," Journal of the
Mechanics and Physics of Solids, vol. 45, pp. 1727-1763, 1997.
[36] J.-W. Yoon, F. Barlat, R. E. Dick, K. Chung, and T. J. Kang, "Plane stress yield
function for aluminum alloy sheets—part II: FE formulation and its
implementation," International Journal of Plasticity, vol. 20, pp. 495-522, 2004.
[37] J. Zhu, Y. Xia, G. Gu, and Q. Zhou, "Extension of non-associated Hill48 Model
for characterizing dynamic mechanical behavior of a typical high-strength steel
sheet," in ASME 2014 International Mechanical Engineering Congress and
Exposition, 2014, pp. V009T12A097-V009T12A097.
[38] W. F. Hosford and R. M. Caddell, Metal forming: mechanics and metallurgy:
Cambridge University Press, 2011.
[39] D. C. Drucker, "A more fundamental approach to plastic stress-strain relations,"
in Proc. of 1st US National Congress of Applied Mechanics, 1951, 1951, pp. 487-
491.
[40] A. Anandarajah, Computational methods in elasticity and plasticity: solids and
porous media: Springer Science & Business Media, 2011.
[41] D. Zhu, B. Mobasher, S. Rajan, and P. Peralta, "Characterization of dynamic
tensile testing using aluminum alloy 6061-T6 at intermediate strain rates,"
Journal of Engineering Mechanics, vol. 137, pp. 669-679, 2011.
146
[42] W. Kang, S. Cho, H. Huh, and D. Chung, "Modified Johnson-Cook model for
vehicle body crashworthiness simulation," International journal of vehicle design,
vol. 21, pp. 424-435, 1999.
[43] G. R. Johnson, "A constitutive model and data for materials subjected to large
strains, high strain rates, and high temperatures," Proc. 7th Inf. Sympo. Ballistics,
pp. 541-547, 1983.
[44] V. Tarigopula, O. Hopperstad, M. Langseth, A. Clausen, and F. Hild, "A study of
localisation in dual-phase high-strength steels under dynamic loading using digital
image correlation and FE analysis," International Journal of Solids and
Structures, vol. 45, pp. 601-619, 2008.
[45] A. Reyes, O. S. Hopperstad, O.-G. Lademo, and M. Langseth, "Modeling of
textured aluminum alloys used in a bumper system: Material tests and
characterization," Computational Materials Science, vol. 37, pp. 246-268, 2006.
[46] F. J. Zerilli and R. W. Armstrong, "Dislocation‐mechanics‐based constitutive
relations for material dynamics calculations," Journal of Applied Physics, vol. 61,
pp. 1816-1825, 1987.
[47] J.-H. Kim, D. Kim, H. N. Han, F. Barlat, and M.-G. Lee, "Strain rate dependent
tensile behavior of advanced high strength steels: Experiment and constitutive
modeling," Materials Science and Engineering: A, vol. 559, pp. 222-231, 2013.
[48] F. Armero and A. Pérez‐Foguet, "On the formulation of closest‐point projection
algorithms in elastoplasticity—part I: The variational structure," International
Journal for Numerical Methods in Engineering, vol. 53, pp. 297-329, 2002.
147
[49] A. Pérez‐Foguet and F. Armero, "On the formulation of closest‐point projection
algorithms in elastoplasticity—part II: Globally convergent schemes,"
International Journal for numerical Methods in Engineering, vol. 53, pp. 331-
374, 2002.
[50] F. Armero and J. Simo, "A priori stability estimates and unconditionally stable
product formula algorithms for nonlinear coupled thermoplasticity," International
journal of plasticity, vol. 9, pp. 749-782, 1993.
[51] K. Hashiguchi, "Generalized plastic flow rule," International Journal of
Plasticity, vol. 21, pp. 321-351, 2005.
[52] P. Tuğcu and K. Neale, "On the implementation of anisotropic yield functions
into finite strain problems of sheet metal forming," International Journal of
Plasticity, vol. 15, pp. 1021-1040, 1999.
[53] N. Abedrabbo, F. Pourboghrat, and J. Carsley, "Forming of aluminum alloys at
elevated temperatures–Part 2: Numerical modeling and experimental
verification," International Journal of Plasticity, vol. 22, pp. 342-373, 2006.
[54] N. Abedrabbo, F. Pourboghrat, and J. Carsley, "Forming of aluminum alloys at
elevated temperatures–Part 1: Material characterization," International Journal of
plasticity, vol. 22, pp. 314-341, 2006.
[55] J. C. Simo and T. J. Hughes, Computational inelasticity vol. 7: Springer Science
& Business Media, 2006.
148
[56] J. Simo and M. Ortiz, "A unified approach to finite deformation elastoplastic
analysis based on the use of hyperelastic constitutive equations," Computer
methods in applied mechanics and engineering, vol. 49, pp. 221-245, 1985.
[57] M. Ortiz and J. Simo, "An analysis of a new class of integration algorithms for
elastoplastic constitutive relations," International Journal for Numerical Methods
in Engineering, vol. 23, pp. 353-366, 1986.
[58] K. Chung and O. Richmond, "A deformation theory of plasticity based on
minimum work paths," International Journal of Plasticity, vol. 9, pp. 907-920,
1993.
[59] J. W. Yoon, D.-Y. Yang, K. Chung, and F. Barlat, "A general elasto-plastic finite
element formulation based on incremental deformation theory for planar
anisotropy and its application to sheet metal forming," International Journal of
Plasticity, vol. 15, pp. 35-67, 1999.
149
CHAPTER FIVE
5. CRASH SIMULATIONS OF A RAILROAD TANK CAR
ABSTRACT
Following the development and the validation of the user-defined material card
(UMAT), that includes the constitutive relations describing the behavior (hardening and
anisotropy) of Medium Mn. (10 % wt.) at multi-strain rates, the next step is to use the
newly developed material card to perform crash simulations and analyze the significance
and impact on the modeling accuracy. The material card was to perform finite element
crash simulations on a railroad tank car. The simulations were performed in accordance
with the regulations of the FRA regarding crash tests of tank cars. First, the finite element
model was developed based on the information provided by a commercial tank car
manufacturer. The accuracy of the model was verified by running simulations of an
actual crash test performed by a tank manufacturer, ensuring to correctly define the initial
and boundary conditions, and the material properties of TC128-B, which is the material
used in railroad tank cars. The results of the simulation were compared with the results of
the actual crash test allowing for the validation of the finite element model developed.
Following that, a different set of simulations were performed using the Medium Mn. (10
% wt.) steel as the material for the tank. The simulations were run using three types of
material cards: 1) A material card based on the quasi-static results, 2) a material card
predicting the strain rate sensitivity of the material (hardening only) and 3) the material
card coded in chapter 4 which can predict the strain sensitivity of the material (hardening
150
and anisotropy). The results of these simulations were compared and analyzed based on
the force-displacement history and energy absorbed until fracture. The impact and
significance of the work done are highlighted by comparing the results of these
simulations.
5.1 Introduction
Due to the great difficulty in obtaining automotive crash models and crash test
data from an OEM, and the availability of both from a tank car manufacturer, it was
decided to use the material card developed to perform crash simulations on railroad tank
cars. The simulations will follow the standards and regulations instated by the
Association of American Railroads (AAR) and the Federal Railroad Administration
(FRA) regarding crash tests and simulations [1].
The interest in studying the crashworthiness of tank cars is not recent, as there are
studies that date back to the 1970s that were performed with the association of the AAR
[2]. However, there has been an increased interest in recent years to improve the
modeling capabilities due to the increase in derailment accidents that led to the spilling of
hazardous materials into the environment [3-5]. The accidents have shown that there is a
need for more research to be done to improve the structural integrity of the tank cars.
Consequently, steps were taken by the government and the industry, manifested by the
Next Generation Rail Tank Car (NGRTC) project [6], which aimed to research and study
the different crash scenarios that can lead to structural failure, and consequent spilling, to
develop an accurate test methodology to replicate the scenarios provide solutions. The
151
studies focused on the prevention of the release of hazardous materials from the tank
during a crash.
Studying the recent crash incidents, two accident scenarios were highlighted and
were considered most critical. The first scenario involves a derailment accident, which
causes the tank car to derail from the track, causing different track cars to collide together
(Figure 5.1). The other scenario involves a train on a train collision. These two scenarios
have led to the development of two cross-sectional profiles for the impactor to be used in
tests and simulations [7, 8]. The cross sections correspond to two different parts of the
tank car: 1) the coupler shank, and 2) the draft sill. The structure of the tank must be able
to maintain its integrity when collided by either of these profiles. The structural integrity
of the tank is characterized mainly using two parameters: 1) the impact energy absorbed
by the structure before puncture, and 2) the minimum velocity of the impactor that causes
tank puncture (impact velocity) [8].
Figure 5.1: A schematic showing the collisions associated with a derailment accident [3].
152
The finite element model development has also been a significant point of interest
for governmental and industrial agencies due to the complex nature of the model. The
complexity associated with the simulations mainly comes from the necessity of
accurately modeling the response of the tank material at multi-strain rates and the fact
that tank cars are actually carrying fluids inside of them. The first complexity has been
the goal of this research and the previous chapter has led to the development of a material
card that is capable of accurately modeling the multi-rate behavior. However, to tackle
the second complexity, it is crucial to accurately define the fluid and the fluid-structure
interaction (FSI) using appropriate material models and equations of state. This is
complex because, in finite element analysis, the methodology used to model solids is
different from the one used with fluids. Numerical discretization in finite element
simulation can be performed using different methods [9], such as Lagrangian, arbitrary
Lagrangian Eulerian (ALE), smooth particle hydrodynamic (SPH), element-free Galerkin
(EFG), and discrete element. Lagrangian Methods are the most commonly used in
forming and crash simulations as they can model the behavior of solid materials under
small deformations accurately at a low computational cost. Lagrangian methods are
characterized by a deformable mesh, with every material point coinciding with every
node (i.e., the mesh moves and deforms) [10]. While Eulerian methods, on the other
hand, consist of a fixed non-deformable geometrical mesh within which the material can
move freely. A combination of both methods is known as an Arbitrary Lagrangian
Eulerian formulation which allows the mesh translational and rotational motions while
still maintaining the non-deformability and the separation between the material and the
153
mesh [11, 12]. Fluids flow when being loaded, and the flowing would cause a
considerable distortion in the element which would cause the simulation to fail (negative
Jacobian); therefore, fluids can only be modeled using Eulerian or ALE elements.
Solving a problem using ALE formulation requires an additional time step (advection
time step) and requires 3D-Solid elements. A model containing both element
formulations need special care to correctly define the interaction between the ALE mesh
and the Lagrangian mesh, and this was studied in the literature regarding railroad tank
cars [8, 13, 14].
This chapter will focus on the development of the finite element model of the
railroad tank car. Beginning with some details about the parameters in the model and the
procedure followed to verify the accuracy of the base model developed. Moreover, the
model will then be used to run crash simulations using Medium Mn. (10% wt.) steel,
using the constitutive relation developed in the previous chapter. The simulations will
aim to quantify the impact and importance of using the constitutive relation on the results
of an impact.
5.2 Finite Element Model
Figure 5.2 shows a schematic of a typical railroad tank car, describing its different
components. The particular tank car modeled in LS-DYNA is based on the
DOT117J100W design as per the definition and standards of the AAR [1]. The auxiliary
parts on the tank such as (manways, pressure relief devices, bolsters, etc…) are not
included in the model to reduce computation times, and their inclusion was proven to not
154
affect the results of the simulations [15-17]. The tank heads were designed based on the
FRA regulations as an ellipsoid of revolution with the major axis equaling the diameter of
the tank and the minor axis to be half of the major axis. The tank was designed with an
outer diameter of 120.5” and a thickness of 9/16”, while the jacket had a thickness of
11GA (~3mm). The developed model is shown in Figure 5.3a. The model was meshed
using shell elements of a length of 200mm, while the center area of the tank where the
impact will occur, had a refined mesh of size of 20mm. The final model with mesh is
shown in Figure 5.4. The impactor chosen was based on the standards of the AAR and
FRA for impact testing. It had face dimensions of 12”x12” with an edge radius of 0.5”
(Figure 5.3b). The impactor was also meshed wih similar size to the impact area mesh
despite the fact that no computation is required on those elements. However, this is
crucial to ensure a smooth accurate contact algorithm.
Figure 5.2: A schematic of a typical North American railroad tank car [18].
155
(a)
(b)
Figure 5.3: The geometry of the (a) tank (b) impactor
Figure 5.4: The meshed finite element model of the tank
156
The boundary and initial conditions in the crash simulations were defined with the
goal to replicate the real scenarios encountered during an actual crash test [19]. The
actual test includes impacting the tank on one side with a ram car, while a wall rests on
the other side to prevent the tank from rolling. A similar setup was used for our model,
with the wall on the left side of the tank car (Figure 5.5). Another rigid wall is added on
the lower side of the tank to act as the floor and prevent any vertical motion.
Figure 5.5: (a) A tank car used for an actual crash test [8] (b) the model developed to
replicate the boundary conditions encountered during a crash test.
157
Removing the wheels from the model mandates adding boundary conditions that
can correctly represent their effect. Failing to do so will give extra degrees of freedom for
the elements of the tank leading to an inaccurate deformation behavior in the impact area
and the lack of the actual structural integrity. The impactor was assigned an initial
velocity of 18 mph toward the longitudinal and vertical center of the tank.
The material used in the simulation is TC128-B, which is a typical carbon steel
material used for plate materials. The material has a yield strength of about 400MPa and
ultimate tensile strength of about 700MPa. The TC128-B used in tank materials
undergoes a heat treatment process known as normalizing, which is described in detail by
Hayashi et al. [20]. With the material and boundary conditions of the simulations well
defined, the simulations are ready to be performed. A summary of the parameters of the
simulations are summarized in table 5.1
Table 5.1: the parameters of the simulations
Lading Chlorine
Gross Weight 286000lbs
Tank Inside Diameter 119 3/8”
Thickness 9/16”
Internal Pressure 100psi
Outage 5%
Weight of Impactor 286000lbs
158
As mentioned earlier, the tank cars modeled transport liquified goods, the fluid, in
this case, was chlorine. Under the normal conditions, the material being transported exist
in two phases: liquid and vapor. In order to correctly define the pressure in the fluids and
the interaction between the fluid, the vapor, and the tank, it is necessary to model the
coexisting of the two phases inside the tank in addition to the atmospheric air surrounding
the tank. Therefore, the fluid was modeled using three regions: the liquid chlorine in the
tank, the vapor chlorine in the tank, and the air surrounding the tank from outside. The
three fluids were modeled using solid elements. The interaction between the fluids and
the tank was defined using the FSI capabilities of LS-DYNA. The Mie-Gruneisen
equation of state (EOS) [21] already defined in LS-DYNA was used to model the state of
the chlorine, while the ideal gas EOS was used for the air.
Figure 5.6: The model showing the initialization of the fluids inside the tank. Green
(liquid), blue (vapor).
159
5.3 Crash simulations using TC128-B
The initial crash simulations were performed without modeling fluids using ALE
formulation, but rather as just an additional mass on the tank. The objective of the initial
run was to ensure the structural part of the model was correctly defined and leads to
accurate results. To verify that, the results of the initial run of simulations were compared
to results published in the literature specifically the results published by Kirkpatrick in
2010 whose performed crash tests and simulations on a tank car with similar design and
parameters. First, the acceleration profile of the impactor was compared to that obtained
in the literature by Kirkpatrick in 2010. The results are shown in Figure 5.7. Second, the
force-displacement curves extracted from the impactor were compared with the results
obtained from Kirkpatrick (Figure 5.8). The results of the simulations performed in this
study agree to a certain degree of accuracy (in terms of the magnitude and overall trend)
with the results from Kirkpatrick publication. The difference can be mainly attributed to
the lading modeling.
The results obtained from the first set of simulations ensured the validity and
accuracy of the structural part of the model. Therefore, the next set of simulations
involved modeling lading using the ALE formulation as described earlier. Other than the
lading modeling, the parameters used in the initial run were kept the same, and the
simulations were run at the same initial impactor speed. The load-displacement curve was
extracted from these simulations and compared with the same results by Kirkpatrick in
Figure 5.9. The circle is drawn on the maximum load value in the second set of
simulation to be compared with the actual crash test results presented by Kirkpatrick
160
rather than with his simulation results. To better visualize the accuracy and improvement
over the first run, the load-displacement curve from the crash test published by
Kirkpatrick was digitized and plotted with the results of the second set of simulation
(Figure 5.9c). Our simulation predicted a maximum force of 0.98 Mlbs, while during the
actual crash test, the force recorded was 0.96 Mlbs, which is accurate to about 2%. This
means that the finite element model developed is accurate and can capture the actual
deformation of the tank during a crash event.
5.4 Crash simulation using medium Mn (10% wt.) steel
With the results of the previous section validating the accuracy of the finite
element model developed, the material of the tank was now replaced with Medium Mn.
(10% wt.) steel and the developed material card described in chapter 4 was used. Three
different simulations were performed to quantify the impact and importance of the newly
(a)
161
(b)
Figure 5.7: Comparison of the acceleration profiles of the impactor between (a) the
simulation without lading and (b) Kirkpatrick results [6].
(a)
162
(b)
Figure 5.8: Comparison of the force-displacement curves of the impactor between (a) the
simulation without lading and (b) Kirkpatrick results [6].
(a)
163
(b)
(c)
Figure 5.9: Comparison of the force-displacement curves of the impactor between (a) the
simulation with lading, (b) Kirkpatrick results [6], and (c) direct comparison between the
simulation results and the actual test results published by Kirkpatrick.
164
developed constitutive relation. The first simulation was performed using the
experimental results obtained using the quasi-static experimental results, i.e., that only
one hardening curve was used in the simulation (quasi-static curve) and the anisotropic
Yld2000-2d was used as the yield function calibrated using the quasi-static data. The
second simulation was performed using the experimental results obtained using the multi-
rate experimental results for hardening and the quasi-static experimental results for the
yield function, i.e., that the hardening curves at the different strain rate were used in the
simulation and the anisotropic Yld2000-2d was used as the yield function calibrated
using the quasi-static data. Finally, the last simulation was performed using the
experimental results obtained using the multi-rate experimental results for hardening and
multi-rate experimental results for the yield function, i.e. that the hardening curves at the
different strain rate were used in the simulation, and the newly developed rate-dependent
anisotropic Yld2000-2d (chapter 4) was used as the yield function calibrated using the
multi-rate data (the material card coded in chapter 4 was used). Figure 5.10 shows the
force-displacement curves obtained from the three simulations. The figure clearly shows
that the three simulations have distinct load-displacement profiles from the three different
input cards and the dependency of the simulation results on capturing the anisotropy
(notice the difference between the red and the green curves). The results show that
simplifying the input card by modeling only the quasi-static response of the material lead
to overestimation of the puncture resistance of the tank, whereas when improving the
accuracy of the input card by modeling the multi-rate response in terms of hardening and
anisotropy the load-displacement curve tend to significantly drop in terms of the area.
165
Figure 5.10 represents a qualitative comparison between the three scenarios,
however, to further prove the necessity and impact of accurate modeling, the energy
absorbed to fracture was calculated in the simulation for the three different scenarios.
Figure 5.11 shows a bar graph representation of the energy. Scenario 1 corresponds to the
Figure 5.10: Comparison of the three load-displacement profiles of simulations ran using:
1) quasi-static hardening and yielding (green), 2) multi-rate hardening and quasi-static
yielding (blue), and 3) the developed constitutive relation, i.e., multi-rate hardening and
yielding (red)
166
first simulation where only quasi-static experimental results were used. Scenario 2
corresponds to capturing multi-rate hardening while only modeling quasi-static yielding,
and finally, scenario 3 corresponds to using the rate-dependent anisotropic Yld2000-2d
developed in the previous chapter. The results show that capturing the multi-rate
hardening behavior of Medium Mn. (10 % wt.) steel reduces the energy absorbed by the
tank before puncture by 13%. Furthermore, accurately modeling the full spectrum of the
multi-rate response of the material (hardening and anisotropy) manifested by the material
card coded in chapter 4 reduces the energy absorbed by 27% as compared to the
conventional modeling technique (quasi-static only). The significant drop in energy
shown here emphasize the necessity and impact of the work performed, as the errors
introduced by following the conventional approach for AHSSs can prove to be fatal.
Figure 5.11: Energy absorbed to fracture during a crash simulation of the three different
scenarios described.
167
5.5 Summary
A finite element model of a railroad tank car was built to perform crash
simulations. The model was verified by running crash simulations using TC128-B (the
material used in the industry) and the results were compared to the results of simulations
and crash tests in the literature. Following that, crash simulations were performed on the
tank with Medium Mn. (10% wt.) steel as the material of the tank. Several variations of
the constitutive relations were used, and three scenarios of simulations were run. The first
scenario involved running the simulations while modeling only the quasi-static behavior
of the material; the second scenario involved modeling the multi-rate hardening behavior
and quasi-static yielding and finally, the third scenario involved using the developed rate-
dependent anisotropic yield function with multi-rate hardening (coded in a material card).
When comparing the energy absorbed before tank puncture in the three simulations, it
was apparent that the difference between the three scenarios is significant, emphasizing
the necessity and impact of accurately modeling the different properties of the materials
at multi-strain rates and not only the hardening.
5.6 References
[1] A. o. A. Railroads, "MANUAL OF STANDARDS AND RECOMMENDED
PRACTICES: SECTION C—PART II," in DESIGN, FABRICATION, AND
CONSTRUCTION OF FREIGHT CARS, ed. Washington, D.C. 20024: The Association
of American Railroads, 2015, p. 497.
168
[2] E. Phillips and L. Olsen, "Final Phase 05 Report on Tank Car Head Study," RPI-
AAR Tank Car Safety Research and Test Project, RA-05-17, 1972.
[3] S. W. Kirkpatrick, "Detailed Impact Analyses for Development of the Next
Generation Rail Tank Car: Part 1—Model Development and Assessment of Existing
Tank Car Designs," in ASME 2009 Rail Transportation Division Fall Technical
Conference, 2009, pp. 43-51.
[4] NTSB, "Collision of Norfolk Southern Freight Train 192 With Standing Norfolk
Southern Local Train P22 With Subsequent Hazardous Materials Release at Graniteville,
South Carolina," Railroad Accident Report NTSB/RAR-05/04 PB2005-916304. Notation
7710A, 2005.
[5] D. Jeong, Y. Tang, H. Yu, and A. B. Perlman, "Engineering analyses for railroad
tank car head puncture resistance," in ASME 2006 International Mechanical Engineering
Congress and Exposition, 2006, pp. 1-7.
[6] S. Kirkpatrick, "Detailed Puncture Analyses of Various Tank Car Designs,"
Applied Research Associates, Mountain View, CA, 2009.
[7] D. Tyrell, D. Jeong, K. Jacobsen, and E. Martinez, "Improved tank car safety
research," in ASME 2007 Rail Transportation Division Fall Technical Conference, 2007,
pp. 43-52.
[8] H. Yu and D. Y. Jeong, "Impact dynamics and puncture failure of pressurized
tank cars with fluid–structure interaction: A multiphase modeling approach,"
International Journal of Impact Engineering, vol. 90, pp. 12-25, 2016.
169
[9] L.-D. K. U. s. Manual and I. Volume, "Version 971," Livermore Software
Technology Corporation, vol. 7374, p. 354, 2007.
[10] T. J. Hughes, W. K. Liu, and T. K. Zimmermann, "Lagrangian-Eulerian finite
element formulation for incompressible viscous flows," Computer methods in applied
mechanics and engineering, vol. 29, pp. 329-349, 1981.
[11] J. Donea, S. Giuliani, and J.-P. Halleux, "An arbitrary Lagrangian-Eulerian finite
element method for transient dynamic fluid-structure interactions," Computer methods in
applied mechanics and engineering, vol. 33, pp. 689-723, 1982.
[12] N. Takashi and T. J. Hughes, "An arbitrary Lagrangian-Eulerian finite element
method for interaction of fluid and a rigid body," Computer methods in applied
mechanics and engineering, vol. 95, pp. 115-138, 1992.
[13] H. Yu, Y. H. Tang, J. E. Gordon, and D. Y. Jeong, "Modeling the effect of fluid-
structure interaction on the impact dynamics of pressurized tank cars," in ASME 2009
International Mechanical Engineering Congress and Exposition, 2009, pp. 59-69.
[14] H. Yu, D. Jeong, J. Gordon, and Y. Tang, "Analysis of impact energy to fracture
unnotched charpy specimens made from railroad tank car steel," in ASME 2007 Rail
Transportation Division Fall Technical Conference, 2007, pp. 123-130.
[15] D. Jeong, J. Gordon, Y. Tang, A. Perlman, and H. Yu, "Analysis of railroad tank
car shell impacts using finite element method," in IEEE/ASME/ASCE 2008 Joint Rail
Conference, 2008, pp. 61-70.
170
[16] Y. Tang, H. Yu, J. Gordon, and D. Jeong, "Finite element analyses of railroad
tank car head impacts," in ASME 2008 Rail Transportation Division Fall Technical
Conference, 2008, pp. 117-125.
[17] Y. Tang, H. Yu, J. Gordon, M. Priante, D. Jeong, D. Tyrell, et al., "Analyses of
full-scale tank car shell impact tests," in ASME 2007 rail transportation division fall
technical conference, 2007, pp. 29-37.
[18] C. P. Barkan, S. V. Ukkusuri, and S. T. Waller, "Optimizing the design of railway
tank cars to minimize accident-caused releases," Computers & operations research, vol.
34, pp. 1266-1286, 2007.
[19] D. Tyrell, K. Jacobsen, B. Talamini, and M. Carolan, "Developing strategies for
maintaining tank car integrity during train accidents," in ASME 2007 Rail Transportation
Division Fall Technical Conference, 2007, pp. 59-68.
[20] K. Hayashi, A. Nagao, and Y. Matsuda, "550 and 610 MPa class high-strength
steel plates with excellent toughness for tanks and penstocks produced using carbide
morphology controlling technology," JFE Technical Report, vol. 11, p. 20, 2008.
[21] A. I. Burshtein, Introduction to thermodynamics and kinetic theory of matter:
John Wiley & Sons, 2008.
171
CHAPTER SIX
6. SUMMARY AND CONCLUSIONS
In summary, this research focused on developing a methodology to characterize
the behavior of new grades of AHSSs at strain rates equivalent to that encountered during
a crash event. First, the effects of microstructure on the strain sensitivity of different
AHSSs was studied by performing experiments at multiple strain rates on a variety of
AHSSs with a wide array of phases compositions in their microstructure. The
microstructures of these different materials contained a variety of different combination
of the three main phases of steel. The results showed the significant impact the
microstructure has on the behavior of the steels at multi-strain rates. Following that, the
focus of the work shifted to a specific third-generation AHSS, which was Medium Mn.
(10% wt.) steel. This steel is considered one of the most promising third-generation
AHSSs, due to its unique properties and microstructure.
Further experiments were performed on this material to study its multi-rate
behavior deeply. The results showed that the material exhibit anisotropy at quasi-static
strain rates and at high strain rates, however, the anisotropy varied at different strain
rates. This strain rate sensitive anisotropy was significant and therefore had to be
accurately modeled in the crash simulations performed on this material. A survey of the
literature showed that to the best knowledge of the author, there is no strain rate
dependent anisotropic yield function capable of capturing this kind of behavior.
Therefore, after a thorough literature review of the different anisotropic yield functions,
172
Yld2000-2d was selected to be the basis to develop a new rate-dependent anisotropic
yield function. The developed yield function was numerically coded in a material card
to be used in the commercial FEA software, LS-DYNA. The material card was first
validated by running simulations of tensile tests and comparing it with the experimental
results. Once this was accomplished, the material card was used to perform crash
simulations on a railroad tank car. The results of the simulations showed that using the
material card developed had a significant impact on the results, which was quantified by
calculating the energy absorbed by the tank before puncture occurred. The energy
absorbed dropped by 27% when using the newly developed model as compared to the
conventional method of modeling (quasi-static only). The significant difference in the
energies highlights the impact of capturing the full spectrum of the multi-rate behavior of
AHSSs (hardening and anisotropy) when performing crash simulations. The significance
is manifested in the fatal implications of having errors in that magnitude when running
crash simulations, most importantly the safety of the humans and the environment.
6.1 Contributions
The main contributions of this work are:
1) Developed the experimental capabilities and approach to enable robust material
testing at a wide range of strain rates (up to very high strain rates).
2) Performed a comprehensive and detailed experimental study on the effects of the
different steel phases (ferrite, austenite and martensite) on the macro-mechanical
173
response of AHSSs at different rates up to the highest strain rates encountered
during a crash event.
3) Investigated and captured hardening and anisotropy in one of the latest
generations of AHSSs (Med. Mn Steel Grade 10%wt); a promising material with
very complex behavior.
4) Developing a new rate-dependent anisotropic yield function based onYLD2000-
2d that can capture the change in anisotropy observed in medium Mn. (10% wt.)
steel at different strain rates. This presents a significant modification to the
methodology of modeling and characterizing the behavior of material especially
AHSSs at multi-rates.
5) A UMAT card, for the modified constitutive relations for Med. Mn Steel, was
developed for use in LS-DYNA; the card was validated through a series of
simulations of tension testing that show the ability to accurately capture hardening
and anisotropy in the material at different strain rates.
6) Full impact simulations of rail tank cars were performed as a final overall
validation of the developed material card; the simulations provide a quantitative
measure of the importance of accurately modeling the hardening and anisotropy at
multi-strain rates in complex materials.
6.2 Future Work
The future work will focus on developing a relation that can predict the behavior
of the materials at high strain rates based on the initial phase composition of the
microstructure. This is particularly important in minimizing the costs associated with the
174
development of the new generations of AHSSs. Also, further work needs to be done
regarding the new methodology manifested by the new anisotropic yield function.
Several experiments need to be performed to improve the accuracy of the model at
different plastic work. Finally, a fracture model must be added to the constitutive relation
to increase the accuracy of crash simulations. The research will focus on developing a
rate-dependent fracture model.
It is also worth noting that the current rate dependent anisotropic yield function
includes a parameter that depends on the crystallographic configuration of the
microstructure. When austenite, which has an FCC crystallographic structure, transforms
into martensite, which has a BCT crystallographic structure, the yield function needs to
be updated in accordance with the new structure. In the current work, it was shown that
the transformation is inhibited at high strain rates, and therefore even with the
transformation taking place, the microstructure will be majorly an FCC structure and thus
justifies the choice for the parameter. However, it is worth investigating the effect of the
change in the structure on the yield function in the future work.