as_bohemen_20040224.
An acoustic emission study of
martensitic and bainitic
transformations in carbon steel
The research described in this thesis was performed in the department of Materials
Science and Technology, Delft University of Technology, Rotterdamseweg 137, 2628
AL Delft, The Netherlands.
The research described in this thesis was carried out in the framework of the Strategic Research Programme of the Netherlands Institute for Metals Research in the
Netherlands (www.nimr.nl).
An acoustic emission study of
martensitic and bainitic
transformations in carbon steel
PROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof.dr.ir. J.T. Fokkema,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op dinsdag 24 februari 2004 om 13.00 uur
door
Stefanus Matheus Cornelis VAN BOHEMEN
doctorandus in de natuurkunde
geboren te Wassenaar
Dit proefschrift is goedgekeurd door de promotor:
Prof.dr. I.M. Richardson
Toegevoegd promotor: Dr.ir. J. Sietsma
Samenstelling promotiecommissie:
Rector Magnificus, voorzitter
Prof.dr. I.M. Richardson, Technische Universiteit Delft, promotor
Dr.ir. J. Sietsma, Technische Universiteit Delft, toegevoegd promotor
Prof.dr. G. den Ouden, Technische Universiteit Delft
Prof.dr.ir. M. Wevers, Katholieke Universiteit Leuven, Leuven, België
Prof.dr. R. Boom, Technische Universiteit Delft
Dr. P.J. Jacques, Université catholique de Louvain, Louvain-la-Neuve, Belgium
Dr.ir. M.J.M. Hermans, Technische Universiteit Delft
Prof.dr.ir. S. van der Zwaag, Technische Universiteit Delft, reservelid
Dr.ir. M.J.M. Hermans heeft als begeleider in belangrijke mate aan de
totstandkoming van het proefschrift bijgedragen.
Published and distributed by: DUP Science
DUP Science is an imprint of
Delft University Press
P.O. Box 98
2600 MG Delft
The Netherlands
Telephone: +31 15 2785678
E-mail: [email protected]
ISBN 90-407- 2477 -6
Keywords: Phase transformations, acoustic emission, steel, martensite, bainite
c 2004 by S.M.C. van Bohemen
Copyright °
All rights reserved. No part of the material protected by this copyright notice may
be reproduced or utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and retrieval system, without written permission from the publisher: Delft University Press.
Printed in The Netherlands
Contents
1 Introduction
1
2 Acoustic emission and phase transformations
2.1 Historical review of the AE technique . . . . .
2.2 Basic theory of acoustic emission . . . . . . . .
2.2.1 Material and transducer response . . . .
2.2.2 Sensors and pre-amplifiers . . . . . . . .
2.2.3 Attenuation and noise . . . . . . . . . .
2.3 Phase transformations in steel . . . . . . . . . .
2.3.1 Martensitic transformation . . . . . . .
2.3.2 Bainitic transformation mechanism . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
7
9
10
12
14
15
15
18
3 Experimental
3.1 Acoustic emission system . . . . . . . . . . . .
3.1.1 Sensor mounting and noise precautions .
3.1.2 Attenuation due to waveguides . . . . .
3.1.3 Source location . . . . . . . . . . . . . .
3.2 Gas tungsten arc welding . . . . . . . . . . . .
3.3 Thermo-mechanical simulator . . . . . . . . . .
3.4 Dilatometer . . . . . . . . . . . . . . . . . . . .
3.5 Furnace . . . . . . . . . . . . . . . . . . . . . .
3.6 Materials . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
23
23
27
28
31
32
38
41
43
43
4 Acoustic emission monitoring of phase transformations in steel
4.1 Study of steel C45 . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Thermo-mechanical simulator experiments . . . . . . . . . .
4.1.2 Welding experiments . . . . . . . . . . . . . . . . . . . . . .
4.2 Study of steel 42CrMo4 . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Welding experiments . . . . . . . . . . . . . . . . . . . . . .
4.2.2 Dilatometer experiments . . . . . . . . . . . . . . . . . . . .
4.2.3 Furnace experiments . . . . . . . . . . . . . . . . . . . . . .
4.3 Study of low carbon steels . . . . . . . . . . . . . . . . . . . . . . .
4.4 Study of a high-alloyed steel . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
47
48
48
56
58
58
60
61
64
65
i
ii
4.5
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 A study of acoustic emission energy generated
martensite formation
5.1 Theoretical background . . . . . . . . . . . . .
5.2 Martensite formation . . . . . . . . . . . . . . .
5.2.1 Travelling arc welding of steel 42CrMo4
5.2.2 Spot welding of steel 42MnV7 . . . . . .
5.3 Bainite and martensite formation . . . . . . . .
5.3.1 Spot welding of steel C45 . . . . . . . .
5.3.2 Travelling arc welding of steel C45 . . .
5.4 Conclusions . . . . . . . . . . . . . . . . . . . .
67
during bainite and
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6 Kinetics of the martensitic transformation studied by means
acoustic emission
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Theoretical background . . . . . . . . . . . . . . . . . . . . . . . .
6.3 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . .
6.4 Study of steels C50, C60, C70 and C80 . . . . . . . . . . . . . . . .
6.4.1 Calculation of the martensite volume fraction . . . . . . . .
6.4.2 Proportionality factors k and dislocation densities ρ . . . .
6.4.3 Koistinen-Marburger kinetics . . . . . . . . . . . . . . . . .
6.4.4 A different analysis of the results for steel C80 . . . . . . .
6.4.5 Microstructural analysis . . . . . . . . . . . . . . . . . . . .
6.4.6 Martensite-start temperature Ms . . . . . . . . . . . . . . .
6.4.7 Rate constant C1 . . . . . . . . . . . . . . . . . . . . . . . .
6.5 Analysis of the results for steel 42CrMo4 . . . . . . . . . . . . . . .
6.6 Study of a shape memory alloy . . . . . . . . . . . . . . . . . . . .
6.6.1 Acoustic emission experiments . . . . . . . . . . . . . . . .
6.6.2 Optical Confocal Laser Scanning Microscopy observations .
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
74
75
76
78
82
83
84
87
.
.
.
.
.
.
.
.
of
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
91
92
93
94
96
96
97
99
105
106
107
109
112
113
113
114
116
7 Analysis of acoustic emission signals originating from bainite and
martensite formation
121
7.1 Acoustic emission during plastic deformation . . . . . . . . . . . . . 122
7.2 Dislocation dynamics during displacive transformations . . . . . . . 125
7.2.1 Nucleation and growth of martensite . . . . . . . . . . . . . . 125
7.2.2 Nucleation and growth of bainite . . . . . . . . . . . . . . . . 127
7.3 Analysis of continuous acoustic emission . . . . . . . . . . . . . . . . 127
7.4 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.6 Discussion of proportionality factors k . . . . . . . . . . . . . . . . . 135
7.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Summary
141
iii
Samenvatting
147
List of publications
153
Curriculum Vitae
155
Nawoord
157
iv
Chapter 1
Introduction
Acoustic emission (AE) is the name given to the phenomenon of elastic waves
being generated by the rapid release of strain energy from localized sources within
a material [1]. As an AE event occurs at a source, elastic waves are generated and
propagate in all directions and ultimately reach the surface of the material.
Phenomena that are classified today as acoustic emission have been observed
since the beginning of technology. For example during pottery making the early
potters learned to associate the sound of pottery cracking as it cooled with the
formation of cracks in their creations. Another familiar example of audible acoustic emissions is the so-called ‘tin cry’, heard by tin smiths during the deformation
of tin, which is due to mechanical twinning [2]. These observations date back to
approximately 3000 BC. The first documented observation of acoustic emission during forging of steel (iron) was made in the eighth century by an Arabian alchemist.
These audible emissions were most likely produced by the formation of martensite during cooling. Around the start of the twentieth century, the martensitic
microstructure was observed for the first time by the German metallurgist Adolf
Martens (1850-1914). In 1936 Forster and Scheil reported that the martensitic
transformation in steel is accompanied by ”clicks” [3]. This may be considered as
the first study of acoustic emission during martensite formation.
A unified (unambiguous) explanation of the source of acoustic emission does not
yet exist. Usually the source is a process which involves a mechanism of deformation
or fracture. Sources that have been identified in metals include dislocation motion
(plastic deformation) [4], crack growth [5], oxidation [6], magnetic domain motion
(the acoustic Barkhausen effect/emission) [7], twinning and displacive phase transformations [8]. In this thesis the acoustic emission during phase transformations in
steel is discussed, mainly focussing on martensitic and bainitic transformations.
A martensitic transformation is a diffusionless first-order phase transition during
which the lattice distortion is mainly described by a combination of shears [9, 10].
It involves a cooperative and almost simultaneous shear movement of atoms from
parent to product phase, often indicated as a displacive process. The strain energy
1
2
Chapter 1: Introduction
produced during growth of the new lattice is reduced by plastic deformation [11].
In this process of martensitic transformation acoustic emission is generated. This
causes transient surface displacements that can be detected with a transducer. The
voltage signal from the transducer is then usually amplified in a pre-amplifier and
analyzed with a computer to study the underlying processes in real-time.
In the development of new high strength steels the martensitic transformation,
in combination with thermal and/or mechanical treatments, plays an important
role. Due to the change in lattice structure and the fact that the transformation is
displacive, several physical properties can be used to investigate the characteristics
of the transformation, such as the transformation-start temperature and the kinetics of the transformation. The most common methods used to study the martensitic
transformation in-situ are electrical resistivity, dilatometry and calorimetry. The
acoustic emission technique used in this work is a rather new and promising technique and has not often been used to study the martensitic transformation in realtime. Moreover, the acoustic emission technique is considered to be a good method
to investigate the displacive character of a phase transformation [12]. Since there is
still no general agreement about the mechanism of bainite formation [13, 14, 15, 16],
acoustic emission measurements during bainite formation will give valuable information concerning its mechanism of growth.
Outline
The acoustic emission experiments described in this thesis have been performed
mainly on medium carbon steels with the aim to study the martensitic and bainitic
transformation in these materials under continuous cooling conditions. Continuous
cooling of steel is achieved during/after welding, and in a thermo-mechanical simulator (welding simulator). One of the major merits of measurements under natural
continuous cooling conditions is the absence of possible external noise from a heating source, which is required for isothermal transformation conditions. Moreover,
the transformation rate is usually faster during continuous cooling, which results
in a better signal to noise ratio.
In chapter 2 the basic concepts of acoustic emission are presented, including an
overview of the development and applications of the AE technique. The effect of
the material and sensor response to the original waveform at the source is described
followed by a discussion of sensitivity, attenuation and noise sources. Finally, the
theory of the phase transformations from austenite to bainite and martensite is
discussed.
The acoustic emission instrumentation, the signal processing technique and the
experimental equipment is described in chapter 3. In most experiments, a welding
apparatus or a thermo-mechanical simulator was used to apply a thermal cycle to
the steel studied. The thermal treatment of a spot weld is usually very similar to the
thermal treatment of a specimen used in the thermo-mechanical simulator. In order
to investigate relatively large samples, some experiments were performed using a
furnace and a salt bath. For a proper comparison between the AE technique and
3
dilatometry, simultaneous measurements of acoustic emission and dilatation were
performed by connecting the AE system to a conventional dilatometer apparatus.
The principles of these techniques are discussed and the procedure for measuring
acoustic emission is given. This chapter concludes with an overview of the materials
used in this study.
In chapter 4 the results of AE experiments on various carbon steels are presented, which show that not only the martensitic transformation, but also the
bainitic transformation is accompanied by acoustic emission. The implication of
this observation for the transformation mechanism of bainite is discussed. For comparison with the AE results during the martensitic and bainitic transformation, AE
experiments were performed on a low carbon steel, which transforms from austenite
to ferrite via a diffusion controlled mechanism. Furthermore, the AE signal measured during martensite formation is compared with the change in dilatation of the
sample during transformation.
In chapter 5 the relationship between the AE energy released during the martensitic transformation and the volume of martensite formed is studied by means of
welding experiments. The AE energy due to the released strain energy accompanying martensitic transformations is theoretically and experimentally studied. It
is shown that both for martensite and bainite formation a specific relation exists
between the AE energy (rate) and the volume (rate) of the transformation.
Besides the fundamental interest in the nature of the martensitic and bainitic
transformation in steels, the results of the study presented in chapter 5 may be used
to develop an AE monitoring system to detect martensite and bainite formation
during welding. Since these hard regions in the weld and heat-affected-zone (HAZ)
are susceptible to cold cracking, real time monitoring of the welding process is of
considerable practical importance.
The kinetics of the martensitic transformation in four carbon steels (C50, C60,
C70 and C80) are discussed in chapter 6. By using the relation derived in chapter
5, the volume fraction of martensite f as a function of time t during cooling can
be calculated from the measured AE signal. The results are compared with the
kinetics predicted by the Koistinen and Marburger (KM) equation. At the end
of chapter 6 the kinetics of the martensitic transformation in a CuAl-based shape
memory alloy are studied by means of both acoustic emission and optical confocal
laser scanning microscopy.
In chapter 7 the frequency spectra of acoustic waves generated during bainite
and martensite formation are studied. The change in the mean frequency corresponding with the transition from bainite to martensite formation during cooling of
a spot weld is attributed to differences in the interface motion of the two transformations. This chapter concludes with an overview and discussion of the proportionality factors between the AE energy and the volume transformed, which were
determined for a number of steels studied in this thesis.
4
Chapter 1: Introduction
References
[1] H.N.G. Wadley, C.B. Scruby and J. Speake, Int. Metals Rev. 3, 41 (1980).
[2] R.B. Liptai, D.O. Harris and C.A. Tatro, Acoustic Emission (ASTM STP 505),
3 (1972).
[3] F. Forster and E. Scheil, Zeitschrift Für Metallkunde 9, 245 (1936).
[4] N. Kiesewetter, P. Schiller, Phys. Stat. Sol. A 38, 569 (1976).
[5] S.H. Carpenter and M.R. Gorman, J. Acoustic Emission 13, s1 (1995).
[6] F. Ferrer, H. Idrissi, H. Mazille, P. Fleischmann and P. Labeeuw, NDT & E Int.
33, 363 (2000).
[7] C.C.H. Lo and C.B. Scruby, J. Appl. Phys. 85, 5193 (1999).
[8] G.R. Speich and A.J. Schwoeble, Acoustic Emission (ASTM STP 571), 40
(1975).
[9] Z. Nishiyama, Martensitic Transformation, Academic Press, London (1978).
[10] D.A. Porter and K.E. Easterling, Phase transformations in metals and alloys,
Chapman & Hall, London (1992).
[11] R.W.K. Honeycombe and H.K.D.H. Bhadeshia, Steels, Edward Arnold, London (1995).
[12] P.C. Clapp, J. Phys. IV 5, 11 (1995).
[13] R.F. Hehemann, K.R. Kinsman and H.I. Aaronson, Metall. Trans. 3, 1077
(1972).
[14] H.I. Aaronson and H.J. Lee, Scripta Metallurgica 21, 1011 (1987).
[15] W.T. Reynolds, Jr., H.I. Aaronson and G. Spanos: Mat. Trans. JIM 32, 737
(1991).
[16] Y. Ohmori and T. Maki, Mat. Trans. JIM 32, 631 (1991).
5
6
Chapter 1: Introduction
Chapter 2
Acoustic emission and phase
transformations
The acoustic emission measurements during phase transformations in steels discussed in this thesis did not only result in a better understanding of the martensitic
and bainitic transformation, but also of the AE technique itself. In this chapter
some basic aspects of acoustic emission are explained, in as far as they are relevant for the subject of this thesis. First the background of the AE technique is
briefly discussed in section 2.1. Subsequently, the effect of the material and sensor
response to the original waveform at the source is described in section 2.2. Furthermore, a discussion of sensitivity, attenuation and noise sources is given. The theory
of the phase transformations from austenite to bainite and martensite is addressed
in section 2.3.
2.1
Historical review of the AE technique
It is generally considered that acoustic emission as a technology started in the
early 1950s with the work of Joseph Kaiser [1] who monitored the emissions of
(audible) sound from materials subjected to external loads. He and his coworkers
were the first to use electronic instrumentation to detect acoustic waves produced
by metals during deformation. They reported that many metals such as zinc, steel,
aluminium, lead and copper produce elastic waves under applied stress and that
acoustic emission activity was irreversible: acoustic emissions were not generated
during reloading until the previous stress level was exceeded. This phenomenon
has become known as the Kaiser effect.
Investigators in the early 1960s realized extensive improvements in the instrumentation of the acoustic emission technique. They found that many problems
concerning background noise could be eliminated, or at least minimized, by working with instrumentation whose frequency range was well above the audible range.
7
8
Chapter 2: Acoustic emission and phase transformations
As a result, many engineers and scientists became interested in acoustic emission
and utilized this technique in studies relating to materials research, structural evaluation and non-destructive testing.
Most papers on acoustic emission were published in the 1970s and 1980s. In
those years, an increasing effort was devoted to the understanding of the fundamental aspects of acoustic emission, such as the nature of the source, and the
way in which the elastic waves are propagated and detected. A long-term goal of
these studies was to learn how to calculate a description of the source event from
the voltage signal of the sensor. To solve the problem, scientists adopted analysis
techniques from earthquake engineering in an attempt to model acoustic emission
sources. Regarding the generation of waves, an earthquake is physically very similar to an acoustic emission event; it is actually only different in scale. Both the
seismic and microseismic activity are initiated by a sudden release of strain energy
at a ‘source’ and in both cases the vibrations propagate through the structure.
Whereas the problem in the case of a semi-infinite material could be solved to
some extent using this approach, in the case of a metal plate the waveform is quite
complicated because of reflections, interference etc. Therefore, most applications
of the AE technique were limited to a qualitative level over these years. Unsuccessful attempts in those years to utilize AE for investigation of the source properties
and elementary mechanisms of martensitic transformations and plastic deformation
discouraged researchers from these topics.
Although the application of AE for fundamental research declined in the late
1980s, the AE technology became more and more popular as a tool for nondestructive testing (NDT) in industry. In these areas use can be made of AE
for in-situ detection of crack evolution and safety monitoring because of the intrinsic nature of AE signals. Important reasons for its increasing acceptance and use
were the improvements in microelectronics and in computer-based recording and
analysis techniques to handle the high signal rates. Certain AE waveform parameters, such as amplitude and duration became the standard quantities to describe
the AE signals. The display of these parameters allowed a better analysis of results
in comparison with the number of counts on X-Y recorders, which was used in the
early days of AE technology.
The acoustic emission technique is considered quite unique among the nondestructive testing methods. In contrast to other NDT methods the detected energy
is released from within the object rather than being supplied externally by the NDT
method, as in e.g. ultrasonics, eddy current or radiography. This has the advantage
that with the AE technique the entire structure can be monitored using only one
sensor at a certain position. Moreover, with two or more sensors the location of
the source can be determined. An inherent drawback is that the AE technique can
only detect active sources like crack growth or plastic deformation; cracks or flaws
which have formed previously do not radiate elastic waves. Another limitation of
the AE technique is that it can be considerably influenced by ambient noise. Great
care must be taken to distinguish AE signals from the ambient noise, especially
during practical application of the technique in an industrial environment.
Basic theory of acoustic emission
signal
9
pre-amplifier
AE system
sensor
source
AE waves
Figure 2.1: The basic principles of the acoustic emission technique. After Pollock
[2].
Recent progress in electronics and computer technology has created new possibilities to use the AE technology for laboratory studies through essential improvements in AE measuring systems and analyzing tools. For instance, the fullwaveform recording and processing of individual signals became possible with the
aid of powerful computers. Nevertheless, the interpretation of detected signals
should always be performed with considerable care because they are never simply
related to the mechanism of the source. This and other fundamental aspects of
acoustic emission are addressed in section 2.2.
2.2
Basic theory of acoustic emission
Acoustic emission is a highly sensitive technique for detecting active microscopic
events in a material. This section gives a brief overview of the most important aspects of the AE technique; more details about the equipment and signal-processing
are addressed in section 3.1. In Fig. 2.1 the process of generation and detection is
illustrated. The AE technique involves a source, which is active in a material, and
AE instrumentation for the detection of the waves: a sensor, a pre-amplifier and
signal processing equipment. The event at the source causes a release of energy
which propagates in the form of a transient stress wave. This wave propagates
through the material, until it reaches the sensor. The sensor converts the small
surface displacements into an electrical signal, which is transmitted to a nearby
pre-amplifier and subsequently to the signal-processing equipment.
The acoustic emission waveform at the source is generally thought of as a simple
pulse [3]. This is related to the nature of the generating source, and therefore the
emissions contain a broad spectrum of frequencies. Depending on the source, the
10
Chapter 2: Acoustic emission and phase transformations
frequency of the waves extends from tens of kHz up to tens of MHz [2, 4]. In
general, the detected signal has a complex waveform, which depends on both the
characteristics of the AE event and the wave propagation effects (wave modes, wave
velocity, attenuation, reflections, and interference) between source and sensor. In
addition to the wave propagation behaviour, the waveform is further changed by
the sensor response. When a sensor is excited by a broadband transient pulse, it
’rings like a bell’ at its own natural frequencies of oscillation. These two effects, the
material response and transducer response, can make the actual signals observed
very different from the original pulses emitted by the source. This is discussed in
more detail in section 2.2.1.
2.2.1
Material and transducer response
The simplest model for an acoustic emission event with non-zero rise time is the
force dipole F (t), whose time variation is pulse-like giving a step-like displacement
S(t) at the source [3] as shown in Fig. 2.2. The width and height of this pulse
depend on the dynamics of the source process. During the event at a source elastic
waves are generated. In order to evaluate the surface displacements due to an
event at the source, it is important to understand the wave propagation behaviour
in materials.
There are different wave modes for acoustic waves in materials: longitudinal
(compression), transverse (shear), surface and plate waves. These waves travel
at different velocities, which are a function of the density of the material, the
Young’s modulus and the Poisson’s ratio. In the case of steel, the velocity of
longitudinal waves is approximately 5000 m/s and transverse waves travel at a speed
of approximately 3000 m/s. The wave velocity of the fastest mode in a material
can be measured by using two sensors at different locations and is particularly
important for source location, because it is used in the computation of the location
of the source (see section 3.1.3).
For a semi-infinite material, it is in principle possible to relate the time variation
of the displacement waveform at the sensor to the event life time at the source [3].
In a plate as shown in Fig. 2.2, however, the surface displacement waveform at
the sensor has usually little resemblance to the original waveform at the source.
Especially the later part of the waveform may have undergone significant changes
due to multiple reflections, interference and mode conversions.
There are usually many wave paths connecting source and sensor, as illustrated
in Fig. 2.2. Waves are reflected at the boundaries of the material. The amount
of energy reflected depends on the geometric (angle of incidence) and material
mismatch at the reflecting boundary. In the case that the damping of waves at reflecting boundaries is low, the detected waveform is made up of many components
reaching the sensor by different paths. This also implies that the signal amplitude
does not necessarily result from the first component, but may result from the constructive interference of several components arriving later at the sensor. The AE
wave bounces around the specimen, thereby exciting the sensor each time it passes,
Basic theory of acoustic emission
11
T
F
U (t )
X (t )
S (t )
M
F
t
T
U (t )
X (t )
S (t )
M
t
t
Figure 2.2: The original waveform at the source S(t) is significantly changed after
propagation through a plate (M ) and subsequent conversion in the transducer (T )
to an electrical signal U (t).
until it finally decays; the decay time depends on the dimensions of the specimen
and the damping of the material. When the damping at the reflecting boundary is
high, only the first component (the direct wave path) is measured. Typically, the
duration of the detected waveform is much longer than the event life time at the
source. Since the measured waveform displays the response of the specimen to the
initial waves at the source, the frequency information in the waveform may therefore be more related to the specimen geometry than to the event characteristics of
the source.
The above described effects of the material (material properties and geometry
of the specimen) on the source function S(t) can be written as [3]
X(t) = M ∗ S(t)
(2.1)
with X(t) the displacement waveform at the surface and M the material response
function. Such a mathematical description, indicated with the asterisk in Eq. (2.1),
is known as a convolution of the signal. The opposite operation, i.e. the calculation of the source description from the displacement is called deconvolution of the
signal. It can be easily verified by taking the Fourier transform of Eq. (2.1) that
an equivalent relation is valid for the source spectrum S(f ), with f the frequency
of the waves. However, in Fourier space convolution is just multiplication, and
therefore deconvolution is analytically possible in Fourier space.
12
Chapter 2: Acoustic emission and phase transformations
The conversion of the displacement X(t) to an electrical signal U (t) is made
by a highly sensitive transducer. AE transducers are typically based on a ceramic
wafer of piezoelectric material (see Fig. 2.3). This material converts a mechanical
deformation into an electrical voltage. In addition to the propagation through the
material, the original signal is further changed during conversion in the transducer.
The displacement waveform X(t) is convolved with the transducer response function
T , as shown in Fig. 2.2, according to
U (t) = T ∗ X(t) = T ∗ M ∗ S(t)
(2.2)
where U (t) is the voltage output of the sensor. It should be noted that this waveform
is radically different from the signal at the source. The original signal is significantly changed during propagation through the material and after conversion by
the transducer. Although the transducer response function can be measured with
reasonable accuracy, it should be realized that the attachment of the transducer
to the material changes the mechanical boundary conditions at the previously free
surface; the surface displacement is altered by the presence of the sensor. Furthermore, the material response function is in practice difficult to determine, because
an accurate simulation of AE sources is complicated, especially inside a material.
All these complications mean that deconvolution of the measured voltage signal
to evaluate the source function is extremely difficult, and in general has therefore
not been pursued in the literature. Recently, some simulation studies on acoustic
emission have been carried out [5, 6].
2.2.2
Sensors and pre-amplifiers
Sensitivity and bandwidth are the most important factors when choosing a sensor
for AE monitoring. The sensor most often used nowadays for AE monitoring is
the piezoelectric transducer [4]. The active element in a piezoelectric transducer is
usually a special ceramic such as lead-zirconate-titanate (PZT). The piezoelectric
crystal converts the displacement at its surface into an electrical voltage. It exhibits
the piezoelectric effect: when the crystal is deformed, the electric voltage across the
crystal is changed. The design of a typical AE sensor with piezoelectric element is
shown in Fig. 2.3.
The sensor housing and electronics are designed to minimize electromagnetic
interference (EMI). Regarding the electronics, the sensors can be divided into two
types: single-ended or differential. A single-ended sensor contains one crystal and is
susceptible to EM noise signals. In contrast, a differential sensor has a design such
that common noise signals due to EMI are rejected. It contains two crystal elements
of opposite polarity, and the signal outputs of these elements are transmitted to the
two inputs of a differential pre-amplifier, where the difference of the two signals is
amplified. A detected AE signal produces two voltage signals of opposite polarity
and thus the difference is two times the signal output from one element. EMI signals
picked up by the two electronic circuits produce signals of the same polarity, which
cancel out in the pre-amplifier.
Basic theory of acoustic emission
13
Figure 2.3: Schematic illustration of a typical acoustic emission sensor with piezoelectric element. After Miller [4].
Pre-amplifiers are used to provide a higher voltage, which is more usable for
further processing. It is preferable to place the pre-amplifier close to the sensor
to minimize pick-up of electromagnetic interference; sometimes the pre-amplifier is
integrated in the sensor housing. Pre-amplifiers contain a frequency filter to reject
unwanted noise signals, and have a wide dynamic range. They inevitably generate
electronic noise (thermal noise), and it is this background noise (and that of the
sensor) that determines the smallest microscopic movement detectable with AE.
The sensitivity of an AE transducer (detection threshold) can be defined as the
minimum level of the signal amplitude that can be detected above the background
noise. Whereas in other types of experiments such white noise can be reduced by
signal averaging, this does not hold for AE experiments, because the relevant AE
signals are also changing with time, and are in fact noise-type signals themselves.
It is important that the pre-amplifier (and transducer) generate the minimum electronic background noise. Typically for modern equipment, the smallest signal that
can be well distinguished from the electronic noise is approximately 4 µV at the
output of a typical transducer, corresponding to a surface displacement of about
10−14 m. This illustrates that piezoelectric transducers are extremely sensitive.
For comparison, atomic radii are in the order of 10−10 m, thus displacements of
1/10000 of an atomic radius can produce well-distinguishable AE signals. The dynamic range of a transducer is normally 105 , from 10−14 m to 10−9 m. Usually,
the amplitudes of AE signals are expressed on a logarithmic (decibel) scale, with 1
µV corresponding to 0 dB and 100 mV corresponding to 100 dB (each 20 dB is a
factor 10).
Sensitivities of sensors are typically shown as frequency response diagrams (output voltage versus frequency). In order to fully characterize a source in terms of
its time scale and/or frequency content, the transducer bandwidth should match
or even overlap the bandwidth of the surface displacements. This is a difficult condition to fulfil because the frequency range of the surface displacements typically
extends from 500 Hz to 500 MHz [8, 9] whereas the bandwidth of the transducer
is usually in the range of 100 kHz to 1 MHz [3]. Therefore, the amount of source
14
Chapter 2: Acoustic emission and phase transformations
information that can be retrieved from the detected signal is limited.
Two other types of sensors, alternative to piezoelectric crystals, that have been
considered in the past are the laser interferometer and the capacitive transducer.
However, their characteristics regarding sensitivity or bandwidth were not found
to be optimal for acoustic emission monitoring. Laser interferometers have a too
small bandwidth and therefore insufficient sensitivity for the typical bandwidth of
acoustic emission. Capacitive transducers can be constructed to be sensitive over
a wide frequency range with a flat frequency response. However, they are less
sensitive; the typical minimum displacement that can be measured is in the order
of 10−10 m, which is normally insufficient for acoustic emission monitoring.
Sensor coupling and reproducibility of response are important factors. Calibration checks should be performed after mounting the transducer on the specimen
to ensure that the sensor is operating properly at the correct sensitivity. This is
discussed in more detail in section 3.1.
2.2.3
Attenuation and noise
Whether a signal can be detected is in the first place determined by the sensitivity
of the AE instrumentation and the amplitude of the elastic waves emitted by the
source. Furthermore, the detectability of the generated AE signals depends on the
attenuation and the noise over the frequency range of the detecting instrumentation.
Attenuation refers to the reduction of the wave amplitudes during propagation. The
major mechanisms governing attenuation are geometric spreading of the wavefront,
loss of AE energy into adjacent media and damping in the propagating material
[4]. The attenuation due to geometric spreading of the wavefront is dominant
close to the source, because due to geometric spreading the amplitude falls off
inversely with distance. Due to absorption, the amplitude falls off exponentially
with distance; thus this attenuation mechanism becomes predominant far from the
source. Also grain boundary scattering and scattering against welds may contribute
to attenuation; their effects cannot be predicted quantitatively. The attenuation
due to welds in the waveguides relevant for the work described in this thesis is
discussed in section 3.1.2.
In laboratory studies the attenuation due to damping and geometric spreading
does not normally limit the detectability because the specimens employed are usually small. On the other hand, the use of a waveguide between the specimen and
the sensor may influence the detectability very strongly in the following way: In
case the waveguide is a rod with a cross-section that is relatively small compared
to the size of the specimen, the waveguide acts as an acoustic resistor; not all the
available AE energy in the specimen can be transmitted to the sensor. A systematic
study of the attenuation of some different waveguides is given in section 3.1.2.
Sources of noise fall into two main categories, electrical and mechanical [4].
Noise sources should be examined to discriminate between noise and relevant acoustic emission signals. Both relevant AE signals and noise signals can be classified
as burst signals or continuous signals. The distinction is based on the rate of oc-
Phase transformations in steel
15
currence. For burst AE signals, start and end points are clearly visible whereas for
continuous AE signals amplitude and frequency variations can be observed but the
duration of the signal is relatively long.
Continuous noise signals may be caused by leaking air lines in the vicinity of the
set-up or electromagnetic interference (EMI). The EMI noise signals are coupled
to the acoustic emission equipment by radiation or electrical conduction. Some
examples of sources of EMI are transformers, powerful lamps and electric motors.
Many mechanical noise sources give rise to burst-type noise signals. In principle,
any movement of mechanical parts in contact with the test object forms a potential
source of noise. Fortunately, most mechanical noise diminishes in amplitude at
frequencies above 100 kHz, i.e. in the operating range of the sensor.
Effort must be made to reduce ambient acoustic noise and EMI. In the absence
of external noise sources, which can be obtained for example under laboratory
conditions, the sensitivity is still limited by the noise produced by the pre-amplifier,
i.e. the background noise.
2.3
Phase transformations in steel
During cooling of steel, phase transformations from austenite to ferrite, pearlite,
bainite and martensite can occur in order of increasing undercooling below the A r3
temperature [7]. The microstructure formed during cooling mainly depends on the
chemical composition of the steel together with the cooling rate and the prior thermal history. The mechanical properties of steel, such as strength and toughness,
are strongly correlated to the microstructure that is formed during cooling. Understanding the various phase transformations is therefore of primary importance in
order to optimize the microstructure and the mechanical properties.
Any phase transformation in the solid state involves nucleation and growth,
and based on the mechanism by which the new phase is formed, two types of phase
transformations can be distinguished: diffusional and diffusionless transformations.
Excellent overviews of the characteristics of the diffusion-controlled transformations
from austenite to ferrite and pearlite are given in references [7, 10, 11]. For the steels
studied in this thesis, the final microstructure is mainly controlled by the austenite
to bainite and martensite phase transformations. In this section the most important
characteristics of these phase transformations are discussed.
2.3.1
Martensitic transformation
The martensitic transformation is a diffusionless first-order phase transformation
during which the lattice distortion can be described by a combination of shears
[12]. It involves a cooperative and almost simultaneous movement of atoms from
parent to product phase. Sometimes this type of phase transformation is also called
a displacive or shear transformation.
Martensitic transformations can occur in many metals provided the conditions
are such that diffusion-controlled transformations are prevented [7]. The transi-
16
Chapter 2: Acoustic emission and phase transformations
Figure 2.4: The Bain correspondence: (a) two fcc austenite cells to show that a
tetragonal
cell can be outlined in austenite (b) Bain strain of this cell with axial
√
ratio 2 into bct martensite with c/a ratio dependent on the carbon content (After
Christian [10]).
tion from austenite to martensite in steels is the best-known and most important
martensitic transformation because of the technological importance of hardened
steel. About a century ago, the martensitic microstructure in steel was first observed with a microscope by the German metallurgist Adolf Martens. Nowadays,
martensite is the term commonly used to describe the transformation product in a
system where the phase transformation occurs in a displacive manner, and correspondingly, the martensitic transformation is the generic name for these transitions.
Owing to the diffusionless (displacive) character of the transformation, the
martensite has exactly the same composition as its parent austenite. In most practical cases the amount of carbon exceeds the solubility in ferrite, and consequently
the martensitic phase in steel can be simply described as a super-saturated solution
of carbon in the ferritic phase, in which the carbon content leads to a tetragonal
distortion of the lattice. The correspondence in lattice structure between austenite and martensite was first pointed out by Bain. He showed that a body-centered
tetragonal (bct) unit cell could be constructed between two face-centered cubic (fcc)
unit cells as illustrated in Fig. 2.4a. The strain necessary to transform this bct unit
cell into a martensite cell is known as the ’Bain strain’. There is a contraction along
the z axis, and a uniform expansion along the x and y axes (see Fig. 2.4b).
Martensitic transformations usually occur under conditions of rapid cooling;
then there is little time at high enough temperature for the carbon atoms to diffuse
and consequently the carbon atoms are trapped in the octahedral sites of the body-
Phase transformations in steel
17
centered cubic (bcc) lattice structure. The equilibrium solubility of carbon in the
bcc lattice is exceeded, and as a consequence of the transformation mechanism this
results in the bct structure, the distorted form of the bcc structure. Accordingly, the
tetragonality of the bct structure increases with increasing carbon concentration of
martensite. Owing to the high carbon content, the martensitic crystal structure is
actually a meta-stable phase. In case the temperature is increased (the martensite
is heated) the carbon atoms become mobile and will diffuse from the martensite
lattice to form carbides. During this so-called tempering, martensite decomposes
into a mixture of ferrite and cementite, with concentrations according the Fe-C
phase diagram.
In steels, the transformation of an austenitic microstructure to a martensitic
microstructure usually takes place due to a decreasing temperature rather than as
a function of time, which is referred to as an athermal transformation. A necessary
condition for the transformation to start is that the free energy G of martensite
(α0 ) is lower than that of austenite (γ). Since additional energy, such as surface energy and strain energy, is required for the transformation to take place, martensitic
0
transformations do not begin at T0 , where ∆Gγ→α = 0, but start at a lower tem0
perature, the martensite-start temperature Ms . The free energy change ∆Gγ→α ,
which corresponds to the temperature difference between T0 and Ms , constitutes
the driving force for the transformation [12].
Besides the thermodynamics, which determine the available driving force for
transformation, the occurrence of a phase change is governed by the kinetics. The
kinetics of a martensitic transformation depend solely on nucleation, because the
growth of a martensitic crystal usually occurs rapidly. It is well known that the
mechanism of growth is displacive, i.e. the growth takes place by the cooperative
movement of atoms. How the phase nucleates, however, is even today not completely understood. This is mainly due to the great speed of formation, which
makes the martensitic transformation a difficult process to study experimentally.
The kinetics of the transformation are the main subject of chapter 6.
Below the martensite-start temperature, the nucleation of martensite during
cooling is believed to take place at structural imperfections in the parent phase
and these pre-existing embryos (defects) are stimulated to grow into martensite
crystals at different degrees of undercooling below Ms ; they have different energy
barriers to activation [11]. Since growth is very fast, each nucleation event almost
instantaneously leads to the formation of a certain volume of the new phase. Because of the different energy barriers to nucleation the volume fraction of martensite
varies only with the degree of undercooling expressing the athermal character of
the transformation. Although the exact nature of the nucleation sites is not completely understood, a nucleus is usually visualized as an embryo of the new phase
(martensite) which has a semi-coherent interface with the parent phase (austenite).
This glissile interface consists of arrays of parallel dislocations, which glide on appropriate slip planes as the interface moves [11]. A more detailed description of the
interface motion including the generation of acoustic emission during this process
in addressed in chapter 7.
18
2.3.2
Chapter 2: Acoustic emission and phase transformations
Bainitic transformation mechanism
Bainite is the transformation product that forms below the pearlite formation temperature and above the martensite-start temperature. Due to additional changes
during and/or after the phase transformation a strong diversity exists in the microstructural appearance of bainite. Although the bainitic reaction has been studied
extensively since the discovery of bainite in 1930 by Bain and Davenport, there is
still no general agreement about the mechanism of bainite formation [13, 14]. Two
alternative models have been proposed to describe the transformation kinetics: the
diffusional model and the displacive model [15, 16, 17].
The diffusional model assumes that the transformation mechanism involves reconstructive diffusion of substitutional atoms, i.e. ferrite and cementite are precipitated from austenite by diffusive mechanisms [16]. This mechanism thus is similar
to the formation of pearlite, although the typical lamellar structure of pearlite does
not occur.
In the displacive model, the atomic rearrangements during bainite formation
are believed to occur in a diffusionless fashion as far as the substitutional atoms
are concerned [15, 17]. In fact, it is assumed that a plate of bainite forms according
to a martensite-like mechanism without diffusion, followed by a rejection of excess
carbon into the remaining austenite which subsequently forms carbides. It should
be emphasized that in this model the growth occurs without diffusion, whereas the
nucleation at the austenite grain boundaries might still require some partitioning
of carbon [17, 18].
For martensite, crystallographic analysis can be used to verify that the transformation takes place without diffusion since the local compositions before and after
the transformation are equal. Bainite, however, forms at somewhat higher temperatures, at which the carbon can still escape from the bainitic ferrite. This implies
that by crystallographic means it is difficult to determine the nature of the bainitic
reaction mechanism.
Regarding the morphology of bainite, two main structures can be identified
which are called upper and lower bainite. Upper bainite forms at a relatively
high temperature, usually in the range of 450 – 600 ◦ C, and lower bainite between
300 and 450 ◦ C. The change in morphology with transformation temperature is
a direct consequence of the change in diffusivity of carbon. In the case of upper
bainite the diffusivity of carbon is relatively high and therefore carbide precipitates
from the carbon-enriched austenite between the ferrite plates. For lower bainite
also carbide precipitation within the bainitic ferrite occurs owing to the decrease in
diffusivity of carbon in austenite at lower temperatures; the diffusivity in bainitic
ferrite is much higher. Therefore, two kinds of cementite can be recognized in lower
bainite: cementite particles that precipitated from the carbon-enriched austenite
and cementite particles that precipitated from supersaturated ferrite. The latter
precipitation shows a strong resemblance with the tempering of martensite. The
layers of carbide in lower bainite are usually extremely fine compared with those in
upper bainite. Consequently, a steel with a lower bainitic microstructure is tougher
Phase transformations in steel
19
than a steel with an upper bainitic microstructure. Moreover, lower bainite is
stronger since the precipitates are finer.
The amount of cementite particles in bainite does not only depend on the carbon
concentration but also on the alloying elements. For example, by increasing the
silicon concentration the cementite precipitation can be greatly retarded because
silicon has a negligible solubility in cementite. This can improve the toughness
of bainitic steels, and is also of importance in the production of TRansformation
Induced Plasticity (TRIP) steels.
Acoustic emission and the displacive character of transformations
In the past many debates have taken place dealing with the exact nature of displacive (martensitic) transformations and how to define such a transition. Although
it is still not completely understood how growth of the new (martensitic) phase
occurs, there is general agreement between researchers that a displacive transformation involves the cooperative movement of atoms that causes a shape change.
Despite this theoretical agreement, it is sometimes very difficult to prove that a
transformation is diffusionless, especially in the case of bainite discussed above.
Usually the high transformation kinetics at the relatively low transformation temperature is given as an argument that the transformation must be diffusionless.
However, this argument is no real proof since what speed is high and what temperature is low is open to question. The best known and well recognized proof for
the displacive character of a transformation is probably the observation of surface
upheavals on a polished surface.
Recently, it was argued by Clapp [19] that acoustic emission is the best test to
prove the displacive character of a phase transformation. This is based on the fact
that the emission of acoustic energy is strongly related to the coordinated movement
of atoms. In contrast to the above mentioned metallographic test, acoustic emission
offers a relatively simple in-situ test to investigate whether or not the transformation
is displacive. In this respect it is quite surprising that acoustic emission has only
rarely been used to monitor phase transformations, especially since it has been
known for a long time that the formation of martensite in steel is accompanied
by acoustic emission. In view of above arguments, acoustic emission monitoring
during bainite formation will give valuable information about its transformation
mechanism.
A general study of acoustic emission during phase transformations in carbon
steels is presented in chapter 4, with the emphasis on the bainitic transformation.
In chapter 5 the acoustic emission energy is studied as a function of the transformed
volume. How the acoustic emission technique can be used to follow the progress of
martensitic transformations is discussed in chapter 6. Finally, in chapter 7 the characteristics of the acoustic waves generated during the bainitic and the martensitic
transformation are studied.
20
Chapter 2: Acoustic emission and phase transformations
References
[1] J. Kaiser, Untersuchungen uber das auftreten Gerauschen beim Zugversuch,
Ph.D. thesis, Technische Hochschule, Munich (1950).
[2] A.A. Pollock, Practical guide to acoustic emission testing, PAC, Princeton
(1988).
[3] H.N.G. Wadley, C.B. Scruby and J. Speake, Int. Metals Rev. 3, 41 (1980).
[4] R.K. Miller, P. McIntire, Acoustic Emission Testing, Vol 5, 2nd ed., Nondestructive Testing Handbook, (American Society for Nondestructive Testing,
1987).
[5] J. Cerv, M. Landa and A. Machova, Scripta Mater. 43, 423 (2000).
[6] W.M. Mullins, R.D. Irwin, J.C. Malas III and S. Venugopal, Scripta Mater.
36, 967 (1997).
[7] D.A. Porter and K.E. Easterling, Phase transformations in metals and alloys,
Chapman & Hall, London (1992).
[8] C. Scruby, H. Wadley and J.J. Hill, J. Phys. D: Appl. Phys. 16, 1069 (1983).
[9] W.J.P. Vink, Niet-destructief onderzoek, 1st ed., Delftse Uitgevers Maatschappij, Delft (1995).
[10] J.W. Christian, Theory of Transformations in Metals and Alloys, 3rd ed., Elsevier Science, Oxford (2002).
[11] R.W.K. Honeycombe and H.K.D.H. Bhadeshia, Steels, Edward Arnold, London (1995).
[12] Z. Nishiyama, Martensitic Transformation, Academic Press, London (1978).
[13] R.F. Hehemann, K.R. Kinsman and H.I. Aaronson, Metall. Trans. 3, 1077
(1972).
[14] H.I. Aaronson and H.J. Lee, Scripta Metallurgica 21, 1011 (1987).
21
22
Chapter 2: Acoustic emission and phase transformations
[15] H.K.D.H. Bhadeshia, Bainite in steels, The Institute of Materials, London
(2001).
[16] W.T. Reynolds, Jr., H.I. Aaronson and G. Spanos: Mat. Trans. JIM 32, 737
(1991).
[17] Y. Ohmori and T. Maki, Mat. Trans. JIM 32, 631 (1991).
[18] G.B. Olson, H.K.D.H. Bhadeshia and M. Cohen, Acta Metall. 37, 381 (1989).
[19] P.C. Clapp, J. Phys. IV 5, 11 (1995).
Chapter 3
Experimental
In this chapter the experimental equipment is described that was used to measure
the acoustic emission signals generated during phase transformations in steel. A
description of the AE system used is addressed in section 3.1, including a discussion of noise suppression precautions, attenuation due to waveguides and source
location. In most experiments an arc welding device or a thermo-mechanical (welding) simulator was used to apply a thermal cycle to the studied specimen, which
upon cooling was monitored by means of acoustic emission. These two heat cycling
methods are characterized by a relatively high cooling rate due to the metallic heat
conduction. The set-ups for AE measurements during welding and during thermal
cycling in the thermo-mechanical simulator are described in section 3.2 and section
3.3 respectively. A few experiments were performed using a conventional dilatometer in order to facilitate comparison between acoustic emission and dilatometry. In
section 3.4 the set-up and measurement procedure for experiments with the AE
system connected to the dilatometer is given. Furthermore, some experiments were
performed using a furnace in order to study large-sized samples, which give a better
signal to noise ratio. During quenching in a salt bath the acoustic emission was
measured as described in section 3.5. At the end of the chapter an overview is given
of the steels studied in this thesis.
3.1
Acoustic emission system
In order to measure the AE signals a PAC (Physical Acoustics Corporation) AE
system was used [1]. This is a fully digital two-channel acoustic emission system
that performs AE waveform and AE signal parameter measurements and stores and
displays the resulting data. The main components are the AEDSP board, which is
integrated in a standard computer, and the Mistras software.
In addition to the two AE channels, the system has two input connections for
external signals, which are known as parametric 1 and parametric 2. The external
signal can be derived from a load cell or a thermocouple if the applied stimulus to
23
24
Chapter 3: Experimental
(a)
(b)
Figure 3.1: (a) Typical burst AE waveform. (b) Part of continuous AE waveform.
produce AE is respectively a force or a temperature change. This parametric input,
which is recorded along with the AE data, can be used for the interpretation of the
acquired data, in order to relate the AE signal to the stimulus (force, temperature).
Usually, the measured AE parameters are plotted against the parametric input.
Acoustic emission is normally described in terms of parameters associated with
the magnitude and rate of occurrence of acoustic emission events [2, 3]. Depending
on the rate of occurrence of AE signals at the sensor, two types of signals can be
distinguished: burst and continuous emission. A burst-type signal is thought of
as a signal from a single, discrete event. When the rate of occurrence is high, the
individual burst signals overlap and combine to form continuous emission. It should
be realized that sometimes the distinction can be rather arbitrary, for instance
when the successive individual signals are visible in the overall continuous wave. In
Fig. 3.1 an example of both types of AE signals are shown.
Related to the two types of signals, two types of data are recognized: time
driven data and hit driven data for continuous and burst emission respectively [1].
In Fig. 3.2 a block-diagram is given which illustrates how the data acquisition is
performed, what kind of data can be measured and which are the control settings.
Continuous acoustic emission is characterized by the root mean square (rms)
voltage Urms of the recorded waves. The mean square voltage U 2 corrected for the
background noise is defined by [4]
U 2 (t) =
1
τ
t+τ
Z
t
Up2 (t0 )dt0 − Un2
(3.1)
where Up (t) is the voltage output at the pre-amplifier and τ a time constant usually
chosen as τ = 0.1 s. The amplification of signals is standard 40 dB (100 ×) and
throughout this thesis the amplified values are displayed; they are not converted
back to the voltage output at the transducer. For measurements of the rms voltage
with an amplification of 60 dB, the results are divided by 10, i.e. converted back to
Acoustic emission system
25
>
=
=
τ#
+τ
τ
"
!
"
Figure 3.2: Diagram showing the two types of acquired AE data with the most
important settings and parameters: time driven data for continuous emission and
hit driven data for burst emission.
the 40 dB scale. The rms voltage is measured with a resolution of 0.02 mV relative
to a background noise level (Un2 )1/2 of 0.24 to 0.28 mV. It should be mentioned
that early measurements were performed with a resolution of 0.2 mV.
For characterizing burst-type AE signals a threshold level is set somewhat above
the background noise level; the chosen threshold value depends on the amplitude of
the AE signals and the desired amount of data acquired. If the AE signal exceeds
the threshold in either positive or negative direction a so-called hit is recorded. A
typical example of a corresponding waveform is shown in Fig. 3.3. Such a waveform
is produced by joining many single points called samples. They correspond to single
measurements at constant time intervals. The system can sample with 1, 2, 4 or
8 MHz; normally a sample rate of 4 MHz is sufficient for measuring signals with
frequencies up to 1 MHz.
In general, some hundreds or thousands of bursts are recorded for evaluation.
To evaluate all the waveforms corresponding to the bursts requires a huge amount
of memory, and interpretation of the waveforms themselves is difficult. Therefore
the most important features of each waveform are determined, which are called the
AE parameters. These allow an easier comparison with other results. The main
signal parameters describing the waveform are the signal amplitude, the signal rise
time and the signal duration. They are illustrated in Fig. 3.3. The time of the
first threshold crossing is called arrival time and is needed for the calculation of the
location of the AE event. The parameter ’counts’ gives the number of times the
signal crosses the threshold. The amplitude is the peak voltage of the AE waveform
and can be a useful measure of the signal size. The time from the first threshold
26
Chapter 3: Experimental
Figure 3.3: The system timing parameters for capturing of a burst-type AE signal,
and the commonly used AE parameters to describe the waveform.
crossing (count) to the peak voltage is the rise time; it can for instance be used to
filter out noise signals, since these have usually very short rise times. The duration
is the time from the first to the last threshold crossing.
The above-mentioned signal parameters cannot simply be related to the characteristics of the source because they strongly depend on the threshold setting and the
system timing parameters. For so-called hit driven data the measurement process
begins when the voltage signal from the pre-amplifier first crosses the threshold.
This threshold crossing triggers certain timers which determine when the hit has
passed and the system is ready for the next hit, hence, this is not trivial. The timers
that capture a hit are the Peak Definition Time (PDT), the Hit Definition Time
(HDT) and the Hit Lockout Time (HLT) [1]. They determine to a large extent the
measured AE parameters described above, such as rise time, duration and peak
amplitude.
The function of the PDT is to enable the determination of the true peak amplitude and rise time of the AE waveform. The PDT circuitry is triggered by the
first maximum after the threshold crossing and retriggered if a new maximum is
measured within the set PDT. The function of the HDT is to enable the system
to determine the end and thus the duration of the waveform. The HDT circuitry
is (re)triggered by the threshold crossing(s). When no threshold crossing occurs
within the set HDT, the end of the hit is defined by the last threshold crossing.
The HDT should be set as short as possible to ensure that two (or more) separate
hits will not be treated as a single hit; but also not too short to avoid fragmenta-
Acoustic emission system
27
Figure 3.4: The typical frequency response of a wide-band sensor (PAC model WD).
tion of the burst signal. The function of the HLT is to exclude the measurement of
reflections and late-arrival parts of the AE signal. The HLT circuitry is triggered
by the time out of the HDT.
Common AE plots are based on the measured parameters of the AE signal,
together with the external parametric input variables, such as load, temperature,
dilatation etc. The plots can be classified into different types such as history plots,
distribution plots and location plots. It should always be remembered that measured hit driven data cannot simply be compared with the results obtained by
other researchers because they depend strongly on the system settings. In this
thesis mainly time driven data (Urms ) is used, whilst in chapter 7 hit driven data
is used for the frequency analysis of waves.
3.1.1
Sensor mounting and noise precautions
Surface displacements were measured with a wide-band (100 – 1000 kHz) differential
AE sensor (PAC model WD). The frequency response of this sensor is shown in
Fig. 3.4. In order to determine this sensitivity of the sensor, a calibration was
carried out according to the face-to-face technique, which is based on the voltage
output per unit of pressure input.
The signal from the sensor is amplified by 40 or 60 dB with a low-noise broadband (100 – 1200 kHz) pre-amplifier (PAC 1220A). The measured rms voltage due
to the electronic noise of the sensor and the pre-amplifier is 0.24 to 0.28 mV. Although the sensor has a differential design, the shielding to guard against external
EMI noise is not sufficient under all circumstances. For example, it was found that
in the case the pre-amplifier was positioned closer than approximately 20 cm to the
computer monitor, the background noise level increased significantly. Positioning
of the sensor close to a computer monitor also leads to an increase in background
noise, typically to a value of 0.5 to 1 mV.
Usually, a noise survey will be performed before the main experiment is carried
28
Chapter 3: Experimental
out. Many noise problems will become apparent during setting up of the experiment
and can be dealt with before data acquisition starts. Good observation and a careful
and systematic approach are very important for diagnosis of noise problems. For
example, if the noise has a continuous character, the cause is probably an electrical
problem (e.g. bad shielding of the BNC cable). On the other hand, if the noise
pattern is rather irregular, the noise source is for instance the descaling of an oxide
layer during cooling of the specimen. At best, the background noise is just the
electronic noise of the pre-amplifier and the sensor.
An essential requirement in mounting a sensor is sufficient acoustic coupling
between the wear-plate of the sensor and the surface of the specimen. Before
mounting, the wear-plate and the surface need to be cleaned. In case the surface
of the specimen is not smooth, it has to be polished/ground with silicon-carbide
paper. Then the couplant, e.g. vacuum grease, can be smeared on the wearplate. Subsequently, the sensor can be pressed on the surface of the specimen. The
couplant layer should be thin and fill all the gaps caused by the surface roughness
to ensure a good acoustic transmission. The sensor should also be attached firmly
to the mounting surface at all times during operation. This can be achieved by a
holding device such as magnetic hold-down or just tape. Electrical contact between
the sensor case and the structure needs to be avoided.
In the case the specimen becomes very hot or very cold, a waveguide is required
for two reasons. The temperature range in which the sensor can operate is typically
−40 – 180 ◦ C . Secondly, commonly used couplants may become unstable at very
high or low temperatures. Waveguides are also necessary when the size of the
specimen is smaller than the diameter of the sensor or when access to the specimen is
difficult. These reasons for using a waveguide are especially relevant for laboratory
studies. A waveguide is typically a metal rod which conducts the acoustic signal
from the specimen to the sensor. One end is designed for acoustic coupling with the
specimen; the other end is usually conical to accommodate the mounting of an AE
sensor. To minimize attenuation, the diameter of the waveguide should be as large
as possible, and the waveguide should have an acoustic impedance similar to that
of the specimen. Furthermore, it is preferred that the joints are made by welding
to obtain a good acoustic conductance.
3.1.2
Attenuation due to waveguides
Preliminary measurements showed that the use of waveguides reduces the measured
AE energy in comparison with the case where the sensor is mounted directly onto
the workpiece. The results indicated that the attenuation is primarily governed by
the diameter of the waveguide and the quality of the welded joints.
To investigate the attenuation due to the presence of a waveguide in a quantitative manner, three waveguides with different diameters (d = 1, 2 and 4 mm) made
of plain steel were welded onto the workpiece as shown in Fig. 3.5. The length of
each waveguide was 100 mm; the disc-shaped mounting plates for the sensor were
identical (diameter = 24 mm, thickness = 10 mm).
Acoustic emission system
29
Figure 3.5: The experimental set-up used to measure the attenuation due to waveguides: (1) welding torch; (2) spot weld; (3) workpiece; (4) waveguide; (5) sensor-1;
(6) sensor-2; (7) pre-amplifiers (60 dB); (8) AE analyzing system.
14
plate
d = 4 mm
d = 2 mm
d = 1 mm
12
Urms [mV]
10
8
6
4
2
0
0
1
2
3
4
5
t [s]
Figure 3.6: The rms voltage as a function of time for a sensor on three different
waveguides and another sensor mounted on the plate.
30
Chapter 3: Experimental
1.0
normalized detected energy
0.9
d = 1 mm
d = 2 mm
d = 4 mm
d = 2 mm, l = 200 mm
d = 4 mm, RVS
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
4
8
12
16
π d2/4 [mm2]
Figure 3.7: The detected energy for each waveguide normalized to the detected
energy on the sensor that was mounted directly onto the plate.
As a highly reproducible source of continuous acoustic emission the martensitic
transformation, which occurs during cooling of a spot weld, was employed [5, 6].
After the production of a spot weld, the acoustic emission due to the formation of
martensite in the weld was measured with two identical sensors: sensor-1 mounted
on a waveguide and sensor-2 mounted on the plate. With both sensors the rms voltage of the generated acoustic emission signals was measured as a function of time.
Repeated measurements with a certain waveguide did not reveal any significant
systematic differences.
The results obtained for the three waveguides are plotted in Fig. 3.6. As expected, it can be seen that the intensity of the signal decreases with decreasing
waveguide diameter. For each waveguide the AE energy detected on sensor-1 was
determined from the area under the peak in a plot of U 2 against t, i.e. the integrated value ∫ U 2 dt [5, 6]. The calculated values, normalized to the AE energy
detected on sensor-2, are plotted against 41 πd2 in Fig. 3.7. It can be seen that the
acoustic conductance of a waveguide is proportional to the cross-section. Although
theoretical predictions for comparison do not exist, the relationship observed seems
reasonable in view of the fact that a similar relationship exists for electric and heat
conduction.
In order to investigate the influence of the welded joint between the waveguide
and the plate on the measured AE energy, the measurements were repeated with
identical waveguides spot welded on the plate. The results from this second series
of experiments indicated that the quality of the welds have a quite strong effect on
the measured signal, which is expressed by the error bars in Fig. 3.7.
Acoustic emission system
31
1.0
plate
d = 4 mm
d = 2 mm
d = 1 mm
U 2 / U 2maximum
0.8
0.6
0.4
0.2
0.0
0
1
2
3
4
t [s]
Figure 3.8: The mean square voltage U 2 (normalized to the maximum value) as
a function of time for the sensor mounted on the plate and the sensor mounted
on three different waveguides. This shows that the attenuated signals and the
non-attenuated signal from Fig. 3.6 can be mapped onto each other by a single
multiplication factor.
To examine the influence of the waveguide material, a measurement with a
stainless steel rod (d = 4 mm, RVS) was performed. The result obtained was not
significantly different from the result obtained for the steel rods discussed above
(see Fig. 3.6). Furthermore, the measured signal did not change significantly when
the length of the waveguide rod was increased to 200 mm (d = 2 mm, l = 200
mm), which indicates that the waveguide diameter is much more important than
its length. Using the same set-up with the two sensors, it was also found that a
thick layer of grease between the sensor and the plate has a strong attenuation
effect on the measured signal.
With respect to the power (U 2 ) of the signals shown in Fig. 3.6, it is interesting
to mention that for all measurements, the signals obtained can be mapped onto each
other by a single multiplication factor as shown in Fig. 3.8. This indicates that the
attenuated signals contain the same information of the martensitic transformation
as the non-attenuated signal.
3.1.3
Source location
By using multiple sensors the position of an AE source can be determined. Computation of the source location is possible by using the wave velocity of the acoustic
32
Chapter 3: Experimental
waves and the arrival times at different sensors. An AE wave propagates in concentric circles from its source and arrives at different sensors with certain delays.
The delay is proportional to the distance between the sensor and the source. In
AE terminology a detected AE wave with one sensor is called a hit; a located hit
is called an event.
The simplest form of source location is linear (one dimensional) location, which
can be used on rods and long thin specimens. Linear location requires only two
sensors to locate a source, whereas planar location requires three sensors. In the
case of linear location, ∆t is defined as the time interval between detections of the
same waveform at the two sensors. The distance difference between a source and
the two different sensors is equal to the product of ∆t and the wave velocity c L ,
according to
| x1 − xS | − | x2 − xS |= cL ∆t
(3.2)
where x1 , x2 and xS are the positions of sensor-1, sensor-2 and the source respectively. The wave velocity of the longitudinal waves in steel was measured, yielding
cL = 5.2 ± 0.3 × 103 m/s.
In order to investigate the practical use of source location, many pencil lead
breaks (used to cause an artificial AE event) were made on the middle of a 800
mm long steel rod (d = 25 mm) with sensors on both ends (see Fig. 3.9). For
such an artificial source midway (± 2 mm) between the sensors, ∆t is theoretically
zero; in the ideal case both sensors are hit at the same time. However, the result
in Fig. 3.9 shows a considerably broad distribution of located events. It can be
seen that most events (more than 50 %) are calculated to be located within a
distance of 40 mm from the real location, i.e. an inaccuracy of approximately 10
%. Such an accuracy is usually sufficient for industrial applications but not for
laboratory studies and therefore the source location is not pursued further in this
thesis. The accuracy of source location is mainly governed by small differences in
sensitivity of the transducers employed. It can be visualized that for a sensor with
a better sensitivity the voltage signal U is relatively high, which implies that the
threshold is crossed a fraction earlier by the same displacement input; this leads
to an inaccurate source location. Also uncertainties in the measurement of arrival
times can contribute to inaccuracies in the source location.
3.2
Gas tungsten arc welding
Gas tungsten arc (GTA) welding is an arc welding process in which heat is produced
by an electric arc, which operates at an arc length of a few millimeters between a
non-consumable tungsten electrode and a metal workpiece. Under normal conditions the arc voltage is in the range of 10 – 20 V and the arc current ranges from
20 A to 300 A.
The experimental set-up used to study the AE signals generated during GTA
welding is schematically illustrated in Fig. 3.10. An ESAB DTA 300 welding power
Gas tungsten arc welding
33
!"#
$
Figure 3.9: One-dimensional location plot showing the number of events detected
as a function of position on a 800 mm steel rod.
(7)
(1)
(8)
(2)
(4)
(6)
(3)
(5)
Figure 3.10: Schematic drawing of the experimental set-up used for AE measurements during GTA welding: (1) automated welding system; (2) welding torch; (3)
bead-on-plate weld; (4) spot weld; (5) workpiece; (6) transducer; (7) pre-amplifier
(60 dB); (8) AE analyzing system.
34
Chapter 3: Experimental
unit provided the welding current. The torch can be adjusted in the vertical direction allowing different arc lengths between the tungsten electrode and the plate.
Argon was used as the shielding gas with a flow rate of approximately 6 l/min to
protect the electrode, the arc and the weld pool from the surrounding atmosphere.
During welding, arc voltage and arc current were continuously measured as a function of time. The travel unit, power supply and measuring system were controlled
by a computer using Labview (National Instruments), a graphical programming
language. The commonly used welding conditions are given in Table 3.1.
Table 3.1: Commonly used welding conditions.
electrode material
tungsten, 2% thoria
polarity
electrode negative
electrode diameter
2.4 mm
electrode tip angle
60◦
arc voltage
10 – 12 V
arc current
30 – 100 A
travel speed
0 or 2 mm/s
arc length
2.0 – 3.0 mm
shielding gas
argon
shielding gas flow rate
6 l/min
shield cup inner diameter
16 mm
electrode tip extension
5 mm
The welding heat transferred to the steel workpiece is primarily used for melting,
which leads to the formation of a weld pool; the volume of the weld pool that has
solidified after welding is referred to as the weld metal. Due to the heat flow in
the workpiece, the zone next to the weld pool is also exposed to the welding heat.
At ambient temperature after welding this heat affected zone (HAZ) and the weld
metal have a different microstructure compared to the parent material.
The high cooling rates of welds determine that the phase transformations do
not usually occur under equilibrium conditions. The main variable in the welding
process that determines the cooling rate is the heat input H (J/m), which is a
function of welding power and travel speed v according to
H=η
VI
v
(3.3)
where η is the heat transfer efficiency, V is the arc voltage, and I is the arc current.
For welds made under quasi-stationary conditions (travelling arc), the cooling
rate during welding can be estimated using Rosenthal’s equation for bead-on-plate
welds (the 3-dimensional heat flow case) [7], given by
2πλv(T − T0 )2
dT
=
dt
H
(3.4)
Gas tungsten arc welding
35
Figure 3.11: The CCT diagram of steel 42CrMo4 determined by the Max-PlanckInstitut für Eisenforschung [9]; A = austenite; F = ferrite; P = pearlite; Zw =
bainite; M = martensite. The solid lines are cooling curves T (t).
with λ the thermal conductivity, and T0 the plate temperature prior to welding.
The cooling time ∆t8/5 between 800 ◦ C and 500 ◦ C, which can be determined from
Eq. 3.4, is usually considered to be independent of the position in the weld [8]
and useful for the prediction of phase transformations that may take place during
cooling. It follows directly from Rosenthal’s equation that a higher/lower heat
input leads to a lower/higher cooling rate. For the welding conditions (heat inputs)
used in this work ∆t8/5 varies from 0.5 s to 2 s, i.e. cooling rates between 800 ◦ C
and 500 ◦ C in the range of 150 ◦ C/s to 600 ◦ C/s.
For welds made under static conditions (spot welds), the cooling rate cannot
be predicted by theory; it depends on the heat input, the welding time and the
size (heat capacity) of the workpiece. However, it can be easily visualized that for
equal heat inputs the cooling rate after spot welding is higher than the cooling rate
during travelling arc welding.
To predict the solid state phase transformations that can occur in a weld which
is continuously cooled, it is common practice to use a continuous cooling transformation (CCT) diagram. As an example the CCT diagram of steel 42CrMo4 is
shown in Fig. 3.11. In order of increasing undercooling below the Ac3 temperature,
the following transformation products may be formed: ferrite, pearlite, bainite and
martensite. It should be noted that the positions of the C-curves in a CCT diagram, which indicate the start of the transformation, do not depend only on the
36
Chapter 3: Experimental
chemical composition (alloying elements) of the steel, but also on the austenitizing
temperature (the maximum temperature during welding Tp ) and the austenitizing
time, which both determine the austenite grain size. This means that a number
of possible transformation products may appear in a weld at ambient temperature. Since the peak temperature varies with distance from the weld center, within
the HAZ different sub-zones can be distinguished with characteristic microstructures. The sub-zones that can be identified in a bead-on-plate weld or a spot weld,
schematically shown in Fig. 3.12, are the following:
(1) Weld metal (Tp > Tm ): Previously melted zone.
(2) Coarse-grained zone (1100 ◦ C < Tp < Tm ): In this sub-zone of the HAZ the peak
temperature Tp is between 1100 ◦ C and the melting temperature Tm . This part will
be completely austenitized and grain growth will take place during austenitizing.
(3) Fine-grained zone (Ac3 < Tp < 1100 ◦ C): Also in this sub-zone the base metal
will be completely austenitized. Due to nucleation and growth processes the austenite grain size is small in comparison with the unaffected parent metal. Because of
the relatively low peak temperature, the austenite grain sizes remain small.
(4) Partially austenitized zone (Ac1 < Tp < Ac3 ): In this sub-zone a partial transformation takes place; only the pearlite in the base metal is austenitized. The
austenite in this zone has a relatively high carbon content.
(5) Tempered zone (Tp < Ac1 ): In this sub-zone no transformation takes place.
Nevertheless, some microstructural changes may occur, such as globularization of
carbides and other aging effects.
The steels studied have a relatively high carbon content, and under normal
welding conditions with the cooling rates given above, usually only martensite is
formed in the weld metal and the HAZ; sometimes bainite is formed in the HAZ.
Whether bainite or martensite is formed at a specific location in the weld depends
on the austenitizing temperature, the austenitizing time and the cooling rate at
that position.
Measurement procedure
To measure the acoustic emission during and directly after welding a wideband (100
– 1000 kHz) differential piezoelectric transducer was used, as shown in Fig. 3.10.
The transducer signal was pre-amplified by 60 dB, and subsequently recorded and
analyzed by the AE system. The transducer was mounted on the workpiece with a
magnetic hold-down, and vacuum grease was used to achieve good acoustic coupling
between workpiece and transducer. In normal welding situations a waveguide will
be required between the workpiece and the transducer in order to protect the transducer from heating up above its operating temperature. However, in the present
laboratory set-up the welding times were relatively short, and consequently the temperature increase was relatively small. This allowed working without waveguides,
with the advantage that attenuation of the signal was kept to a minimum.
In general, the welding experiments were performed using medium carbon steels
in the form of plates with dimensions 250 × 200 mm2 and a thickness in the range
Gas tungsten arc welding
(5) (4) (3) (2)
37
(1)
(6)
Figure 3.12: Schematic diagram of the transverse cross-section of a bead-on-plate
(or cross-section of a spot weld) showing (1) the weld metal; (2) the coarse-grained
zone; (3) the fine-grained zone; (4) the partially austenitized zone; (5) the tempered
zone; (6) parent metal.
of 5 to 10 mm. Prior to welding, the workpiece was cleaned with acetone and
attached to a copper base plate to assure good electrical contact. The angle of
the electrode tip was checked regularly during the production of the welds because
it has a significant influence on the heat transfer. Welds were made under static
conditions (spot welding) and under stationary conditions (travelling arc).
During welding AE noise signals are generated, which are caused by the interaction between arc and plate. In this study an investigation of the exact origin of this
arc noise was not attempted because there are many forces involved in the welding
process which can contribute to the observed noise. Welds made on plain carbon
steel revealed that the arc noise increases with welding power. The rms voltage of
the noise signals was in the range of 0.5 mV to 1.2 mV for arc currents of 20 to 120
A; this is significantly higher than the background noise of approximately 0.26 mV.
It was also found that the arc noise was only slightly dependent on the composition
of the welded material. Tests with the sensor detached from the plate but close to
the arc showed that the high electromagnetic fields, which are present around the
arc, do not give rise to EMI noise onto the sensor.
Experiments using different specimen configurations showed that the results
obtained are independent of plate size. Moreover, the results did not change significantly when using a sample in the shape of a rod, in which case the distance
between source and sensor was 800 mm. This indicates that in the laboratory setup, where distances are relatively short, attenuation of the signals due to geometric
spreading or damping does not play a significant role. Furthermore, the reflections
of waves are presumably damped sufficiently before they arrive at the sensor.
38
Chapter 3: Experimental
(9)
(7)
(8)
(1)
(3)
(6)
(4)
(5)
(2)
Figure 3.13: Schematic overview of the experimental set-up for AE measurements
using the Gleeble thermo-mechanical simulator: (1) specimen; (2) grips; (3) waveguide; (4) transducer; (5) pre-amplifier; (6) AE system; (7) dilatometer; (8) thermocouple; (9) Gleeble system.
3.3
Thermo-mechanical simulator
A thermo-mechanical simulator is a combined thermal and mechanical system for
physical simulation of processes that can occur during the production or processing
of steel. In the experiments described in this thesis the thermal system of the
Gleeble 1500 (Dynamic Systems Inc.) was used. The Gleeble heating system can
heat specimens at rates of up to 10000 ◦ C per second, which is fast enough for
welding simulation. The heating system of the Gleeble is based on the electrical
resistance of the specimen, and therefore this heating method is called resistance
heating (or Joule heating). Other heating methods commonly used in the laboratory
are induction heating and furnace heating, which are described in section 3.4 and
section 3.5 respectively.
For resistance heating the current density is approximately uniform throughout
the volume of the specimen. Therefore, when a current passes through the specimen,
the specimen is heated at the same time in the whole volume. The heat loss to the
exposed surfaces is negligible compared to the heat loss to the water cooled copper
grips via (metallic) conduction. Therefore, isothermal planes exists perpendicular
to the specimen axis. The temperature profile as well as the natural cooling rate
can be controlled by the resistance of the grips and the shape of the specimen. The
natural cooling rate is mainly determined by conduction through the water cooled
jaws and to some extent also by convection and radiation through the surface of
the specimen in the vacuum or inert gas atmosphere.
Measurement procedure
The experimental set-up used for AE measurements in the Gleeble is schematically
illustrated in Fig. 3.13. In general, the specimen under study was subjected to
continuous cooling from high temperatures during which the generated acoustic
Thermo-mechanical simulator
39
emission signals were measured. Optionally, the dilatation of the specimen was
measured simultaneously with a CCT dilatometer. The Gleeble offers the possibility
to export data, such as temperature and dilatation, to the parametric inputs of the
AE system.
The specimen under study was mounted in the copper grips of the Gleeble system and a K-type (chromel/alumel) thermocouple was spot-welded onto the middle
of the specimen for monitoring and controlling the temperature during the thermal
cycle. With a thermocouple welder the thermocouple wires, with a diameter of 0.2
mm, were welded about 1 mm apart in the same cross-section perpendicular to the
specimen axis. This is required to avoid a voltage potential along the specimen axis,
which can lead to an error in the temperature measurement. During positioning
of the specimen in the grips uniform tightening is required to avoid induction of a
mechanical stress or bending moment within the specimen. Moreover, the specimen
should be tightened enough to minimize the thermal and electrical resistance at the
interface of the specimen and the grips; this is important when a high cooling rate
is desired. For further information about the mounting of a specimen, see the users
manual [10].
A waveguide was welded to one end of the specimen to transport the acoustic
emission waves to the transducer and to prevent the transducer from overheating.
On the conical end of the waveguide, a wideband (100 – 1000 kHz) differential
piezo-electric transducer was mounted and vacuum grease was used in between to
achieve good acoustic coupling. The transducer was connected to a pre-amplifier
with 40/60 dB gain and 100 – 1200 kHz band-pass filter. From there the AE
signals were transmitted to the analyzing system (Mistras 2001) for recording and
analysis. The dilatometer was used to monitor the radial dimension length change
of the specimen.
The specimen materials employed were medium carbon and low carbon steels.
The specimens for AE and dilatometric measurements were machined from steel
rods with a diameter d2 of 12 mm. In order to achieve the required cooling rate,
different specimen configurations were tested. In general, the experiments were conducted using specimens with an effective free span l1 of approximately 10 mm and
a diameter d1 of approximately 5 mm (see Fig. 3.14). Due to the thermal gradient
along the axis of the specimen, the austenitized volume depends on the austenitizing temperature Ta , which is controlled by the thermocouple in the middle of
the specimen: the higher the austenitizing temperature, the larger the austenitized
volume. The thermal gradient is similar to the thermal gradient that exists in the
HAZ during welding. Thus different zones can be recognized in the specimen just
as in a weld, see section 3.2.
Prior to measuring, the chamber of the Gleeble system was evacuated to about
0.2 mbar, twice filled with argon gas and purged, and again filled with argon gas.
These precautions were taken to minimize decarburization and oxidation of the
specimen at high temperature. In a typical measurement, the specimen was electrically heated in 20 seconds to the austenitizing temperature Ta , austenitized for 20
seconds, and continuously cooled to ambient temperature, during which the acoustic
40
Chapter 3: Experimental
Figure 3.14: Schematic illustration of the thermal gradient in the specimen during
austenitizing and the design parameters that determine the thermal gradient and
the natural cooling rate when the heating stops: (1) part of specimen in between
the grips, indicated by the length l2 ; (2) austenitized volume, indicated by the
striped region.
emission was measured. The heating system of the Gleeble generates (mechanical)
noise, which makes it impossible to study the acoustic emission generated during
isothermal holding. For the same reason controlled cooling, slower than natural
continuous cooling to obtain for example bainite formation, was not feasible.
In order to change the natural continuous cooling rate, the specimen design
and free span can be changed. The free span l2 , the distance between the copper
grips, has a relatively small influence on the cooling rate. The cooling rate is to a
large extent determined by the specimen diameter d1 and the effective free span l1 .
Increasing each of the parameters d1 , l1 or l2 results in a lower cooling rate.
In addition to the mechanical noise of the heating system, it should be noted
that noise resulting from oxidation may take place. Since it is difficult to obtain
a high vacuum, oxidation of the specimen will always occur to some extent. In
particular the descaling of oxidation products results in noise. The scale (iron
oxide) has a different coefficient of thermal expansion than the steel, and being
brittle it tends to fracture during cooling. This results in burst-type noise, which
is usually clearly visible as scatter superimposed on the continuous rms voltage.
This is most prominent after repeating experiments with the same specimen. As
in the case of welding, high electromagnetic fields are present in the vicinity of the
specimen during heating, however, they do not give rise to EMI noise at the sensor.
In comparison with the welding experiments described before, the attenuation
of the generated AE signals is much higher since a waveguide is used. The wave-
Dilatometer
41
(6)
(8)
(7)
(5)
(1)
(2)
(4)
(3)
Figure 3.15: Schematic drawing of the experimental set-up for simultaneous measurements of acoustic emission and dilatation using the Bähr 805A/D dilatometer:
(1) specimen; (2) quartz rods; (3) thermocouple; (4) induction coil; (5) waveguide;
(6) transducer; (7) pre-amplifier; (8) AE system.
guide, with a diameter of 3 mm is MIG welded onto the specimen, which results
in a full fusion weld with good acoustic conductance. Although the attenuation
due to a waveguide can be estimated from the waveguide study (see section 3.1.2),
the results obtained with the Gleeble cannot be readily compared with the results
from welding experiments because the shape of the specimen used for the Gleeble
experiments is significantly different, and may cause attenuation by itself.
3.4
Dilatometer
A dilatometer can measure longitudinal length changes with high accuracy (≈ 0.1
µm), and can be used to monitor the volume change of a sample which is subjected
to a thermal treatment. Measuring the dilatation is a commonly used method to
study phase transformations in steel. To facilitate comparison between acoustic
emission and dilatation measurements, the AE system was connected to the Bähr
805A/D dilatometer in order to measure both signals simultaneously. In this section
the combined experimental set-up is described.
In Fig. 3.15 a schematic drawing of the set-up is given. The samples used in the
dilatometer were 10 mm long with a diameter of 5 mm. The sample was clamped
between quartz rods, and a thermocouple was spot welded onto the middle of the
sample for measuring and controlling the temperature. A waveguide was welded
onto the sample as shown in Fig. 3.15 to transport the AE waves to the sensor. In
view of the attenuation a thick waveguide wire is desired, however, a wire with a
diameter of 1 mm was used because the wire needed to be bent as shown in Fig. 3.15.
Moreover, a large wire, with a large heat capacity, may affect the temperature of
the sample.
42
Chapter 3: Experimental
(8)
(7)
(6)
(2)
(1)
(3)
(4)
(5)
(9)
Figure 3.16: Schematic illustration of the experimental set-up for AE measurements
using the furnace and the salt bath to apply a heat cycle to the specimen: (1)
specimen; (2) thermocouple; (3) waveguide; (4) transducer; (5) pre-amplifier; (6)
AE system; (7) data logger; (8) furnace; (9) salt bath.
In the dilatometer the induction method was used to heat the sample, which
is surrounded by an electrically conductive coil carrying a high frequency current.
The sample is heated by the eddy current generated in the surface layer of the
specimen. Induction heating is usually faster than furnace heating but slower than
resistance heating. For induction heating it takes some time to have a uniform
temperature from the surface to the center of the specimen.
In the experiments, the sample is heated to an austenitizing temperature of 900
◦
C, austenitized for 2 minutes, and subsequently continuously cooled to ambient
temperature in a helium atmosphere. The use of high pressure gas spray to cool
the sample had to be avoided because it results in an additional significant noise
signal.
An important difference between dilatometry and acoustic emission is that the
measured dilatation is proportional to the amount of volume transformed, whereas
the AE signal is proportional to the transformation rate [5, 6]. Moreover, the AE
signal is measured relative to a background noise level, whereas for dilatometry
only the sample size determines the magnitude of the signal. In the case of acoustic
emission, a high volume transformation rate (determined by the volume and the
cooling rate) is important to obtain a good signal to noise ratio. In addition, the
necessity of using a waveguide as explained above has major consequences for the
detectability of the generated signals.
It appeared that especially the small-sized sample and the small waveguide
limit the possibilities of AE measurements using the dilatometer. In an attempt to
circumvent these problems, additional experiments were performed using a conventional convection furnace.
Furnace
3.5
43
Furnace
With a furnace large-sized samples can be austenitized, and a relatively thick waveguide rod can be used. In Fig. 3.16 a schematic overview of the experimental set-up
is given. Typically, a sample with a diameter of approximately 15 mm and a length
of 40 mm was employed. A waveguide was welded onto the sample; this waveguide
consisted of a stainless steel rod with a diameter of 5 mm and a disc-shaped plate
on which the AE sensor was mounted.
To measure the temperature inside the sample with a special chromel/alumel
thermocouple, a seven millimeter deep hole (2 mm diameter) was drilled in the
sample. With a data-logger the temperature was recorded and simultaneously the
acoustic emission (AE) was measured with the AE analyzing system; both with an
acquisition rate of 1 Hz.
In order to austenitize a sample with the waveguide attached, the door of the
furnace was modified such that the sensor was outside the furnace. After austenitizing the sample at 900 ◦ C for 20 minutes, the sample was taken out of the furnace
and quenched in the fluid salt at a temperature in the range of 300 – 500 ◦ C. During
cooling in the salt bath the acoustic emission was measured.
3.6
Materials
In the following chapters the results are presented of the AE experiments on a
number of steels, mainly focusing on the commercially available medium carbon
steels C45 and 42CrMo4. The compositions of all steels studied in this thesis
are given in Table 3.2. The high-alloyed steel 75MnSiCr was received from prof.
H.K.D.H. Bhadeshia (Cambridge University), and steel 16MnSi is a TRIP steel
received from Corus.
Table 3.2: Chemical composition of steels (wt%). (- not detected)
Steel
C45
42CrMo4
Fe360
St52-3
St50K
42MnV7
75MnSiCr
16MnSi
C50
C60
C70
C80
C
0.44
0.42
0.07
0.17
0.32
0.41
0.75
0.16
0.5
0.6
0.7
0.8
Si
0.27
0.24
0.32
0.61
0.28
0.36
1.63
1.64
0.44
0.39
0.37
0.41
Mn
0.70
0.66
0.74
1.33
0.76
1.64
1.95
1.67
0.51
0.50
0.68
0.61
P
0.022
0.024
0.030
0.015
0.013
0.029
0.003
0.080
0.024
0.020
0.027
0.012
S
0.010
0.007
0.031
0.030
0.003
0.009
0.046
0.043
0.042
0.049
Cr
0.20
1.08
0.14
0.03
0.05
0.26
1.48
0.19
0.20
0.23
0.29
0.28
Mo
0.18
0.02
0.05
0.28
0.01
0.01
0.02
0.02
V
0.005
0
0.08
0.10
-
Cu
0.22
0.12
0.27
0.24
0.21
0.22
0.23
Ni
0.03
0.06
0.10
0.07
0.11
0.01
0.10
0.07
0.16
0.15
44
Chapter 3: Experimental
The heat treatments of the steels were carried out using the methods described
in this chapter; the various heat treatments were applied in order to generate a
bainitic and/or martensitic structure as indicated in Table 3.3.
Table 3.3: Schematic diagram showing which steels are studied, which method is
used, and which transformations occur: B = bainite, M = martensite.
Steel
C45
42CrMo4
Fe360
St52-3
St50K
42MnV7
75MnSiCr
16MnSi
C50
C60
C70
C80
Welding
(section 3.2)
M+B
M
Simulator
(section 3.3)
M+B
M
Dilatometer
(section 3.4)
Furnace
(section 3.5)
M
B
M+B
M+B
M
B
M+B
M
M
M
M
References
[1] Users Manual Mistras 2001, PAC, Princeton (1995).
[2] A.A. Pollock, Practical guide to acoustic emission testing, PAC, Princeton
(1988).
[3] R.K. Miller, P. McIntire, Acoustic Emission Testing, Vol 5, 2nd ed., Nondestructive Testing Handbook, American Society for Nondestructive Testing,
(1987).
[4] W. Schaarwachter and H. Ebener, Acta Metall. Mat. 38, 195 (1989).
[5] S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden, J. Phys. D: Appl.
Phys 34, 3312 (2001).
[6] S.M.C. van Bohemen, M.J.M. Hermans, G. den Ouden and I.M. Richardson,
J. Phys. D: Appl. Phys. 35, 1889 (2002).
[7] D. Rosenthal, Welding Journal 20 (5), 220-s (1941).
[8] G. den Ouden, Lastechnologie 3rd ed., Delftse Uitgevers Maatschappij, Delft
(1993).
[9] A. Rose, W. Peter, W. Strassburg and L. Rademacher, Atlas zur
Wärmebehandlung der Stähle Teil II, Verlag Stahleisen G.B.H., Düsseldorf
(1956).
[10] Users Manual Gleeble 1500, Dynamic Systems Inc. (1988).
45
46
Chapter 3: Experimental
Chapter 4
Acoustic emission
monitoring of phase
transformations in steel
Although it has been known for a long time that acoustic emission is generated
during martensite formation, only a few acoustic emission studies of the various
transformations taking place in carbon steels can be found in the literature. About
30 years ago Speich and Fisher measured acoustic emission during martensite formation in an FeNi alloy [1]. They compared their AE results with electrical resistivity data and results from quantitative metallography. Some years later Speich
and Schwoeble presented a general study of acoustic emission during the various
phase transformations in a wide variety of steels [2]. Their results showed that
the intensity of acoustic emission generated during martensite formation decreased
strongly as the carbon content of the steel decreased, becoming nearly undetectable
in maraging steel. Furthermore, they showed that no acoustic emission was detected
during the formation of ferrite, pearlite and bainite, which was attributed to the
diffusion-controlled growth of these transformation products. Especially the latter
observation was important because the mechanism of growth of bainite was a subject of debate in those years; two alternative models had been proposed to describe
the transformation kinetics: the diffusional model and the displacive model (see
section 2.3.2). Despite the ongoing debate concerning the bainitic reaction mechanism no attempts have been made to verify the results of the study of bainite by
Speich and Schwoeble [2].
The monitoring of acoustic emission during bainite formation in various carbon
steels is the main topic of this chapter. In section 4.1 the results of AE experiments
during various phase transformations in steel C45 are discussed. Subsequently,
the results for martensite and bainite formation in steel 42CrMo4 are presented
in section 4.2. Finally, the results of AE measurements during bainite formation
47
48
Chapter 4: Acoustic emission monitoring of phase transformations in steel
in the low carbon steels St 52-3 and St 50K (section 4.3) and the high-alloyed
steel 75MnSiCr (section 4.4) are discussed. All steels were investigated using the
techniques described in chapter 3.
4.1
Study of steel C45
In this section the AE measurements on steel C45 are discussed. The chemical
composition of the steel is given in Table 3.2. Section 4.1.1 contains the results of
AE measurements during the bainitic and the martensitic transformation in steel
C45 specimens carried out using the Gleeble 1500 thermo-mechanical simulator.
The results of AE measurements during welding of steel C45 are discussed in section
4.1.2.
4.1.1
Thermo-mechanical simulator experiments
The experiments were carried out as described in section 3.3. Based on the CCT
diagram of the medium carbon steel C45 and possible cooling rates during continuous cooling in the Gleeble, it is expected that bainite and/or martensite may be
formed. For comparison, measurements under equal conditions were performed on
a steel Fe360 specimen, in which neither bainite nor martensite will be formed.
In order to achieve the required cooling rate, different specimen configurations
were tested. After some preliminary experiments two types of specimen were made
with different design parameters (see Fig. 3.14). In order to obtain martensite
formation specimen type A with an effective free span l1 = 8 mm, a diameter d1 =
4 mm, and distance between the grips l2 = 22 mm was used. Specimens of type B
(with l1 = 10 mm, d1 = 5 mm and l2 = 32 mm) were used to generate both bainite
and martensite formation during continuous cooling.
For all measurements, the specimen was electrically heated in 20 seconds to the
austenitizing temperature Ta , austenitized for 20 seconds, and continuously cooled
to ambient temperature. Measurements were repeated several times to check the
reproducibility of the experiments. It is known that repeated rapid heating of a
martensitic microstructure can result in austenite grain refinement and this was
found to play a role for the first few thermal cycles applied to each specimen [3].
However, after some (≈ 5) repeated thermal cycles this grain refinement effect
saturates and the measurements were found to become reproducible.
After thermal cycling, the specimens were cut in the middle longitudinally and
the microstructure was analyzed using an optical microscope (Olympus). The
micro-hardness of the specimens was measured by means of a micro Vickers hardness tester (Buehler Ltd.) using a load of 100 g.
Martensite formation
For the experiments using specimen type A the cooling rate was relatively high
(∆t8/5 ≈ 2 s) and only martensite formation occurred. A typical result of these
Study of steel C45
49
900
800
T [ºC]
700
600
500
M
400
300
200
100
0
0
5
10
15
20
t [s]
Figure 4.1: The temperature of steel C45 (specimen type A) during cooling in the
Gleeble thermo-mechanical simulator (Ta = 830 ◦ C). M: inflection in cooling curve
due to the release of latent heat during martensite formation.
(a)
∆d [µm]
60
50
40
30
20
10
(b)
Urms [mV]
5
4
3
2
1
0
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.2: Plot of (a) the dilatation and (b) the rms voltage as a function of
temperature for steel C45 (specimen type A) austenitized at Ta = 830 ◦ C, cooling
according Fig. 4.1.
50
Chapter 4: Acoustic emission monitoring of phase transformations in steel
Figure 4.3: Optical micrograph of the microstructure of steel C45 (specimen type
A) after continuous cooling (see Fig. 4.1) from Ta = 830 ◦ C.
experiments is discussed below. The specimen was austenitized at 830 ◦ C, and the
cooling curve is depicted in Fig. 4.1.
For comparison with the AE data, the cross-strain of the specimen was also
measured simultaneously. The dilatation of the steel specimen as a function of
temperature is shown in Fig. 4.2a, and reveals the formation of martensite. The
phase transformation is reflected as a change in the slope due to the difference in
specific volume and thermal expansion between austenite and martensite.
The measured rms voltage during cooling is shown in Fig. 4.2b. It can be seen
that the rms voltage measured during austenitizing is very high due to the electrical
current flow through the specimen. When the heating stops, the rms voltage drops
back to the background noise level. The onset of a peak in the AE data is observed
at the martensite-start temperature: the martensite peak. The signal level increases
to a maximum and then tails off to the background noise level. The microstructure
of the specimen after cooling was found to be completely martensitic, see Fig. 4.3,
with a measured micro-hardness of approximately 700 HV0.1 .
Bainite formation
In order to obtain bainite formation, a lower cooling rate is required, and the
experiments were conducted using specimens of type B austenitized at 820 ◦ C. In
Fig. 4.4 the temperature is plotted as a function of time. In the cooling curve a
clear inflection is visible, which is caused by the release of latent heat during bainite
formation; the inflection corresponding to martensite formation is less pronounced.
During cooling the cross-strain and the AE signals were measured simultaneously. The dilatation of the steel specimen, austenitized at 820 ◦ C, is plotted against
temperature in Fig. 4.5a. The dilatation plot reveals that besides bainite formation
Study of steel C45
51
martensite formation also occurs at lower temperatures.
In Fig. 4.5b the rms voltage is plotted as a function of temperature. When
the heating stops, the rms voltage drops back to the background noise level, and
subsequently the onset of a peak in the AE data is observed at bainite transformation temperatures: the bainite peak. The signal level increases to a maximum
and then tails off to the background noise level. Upon further cooling, a second
peak is observed at temperatures of martensite formation: the martensite peak.
The shapes of the bainite and martensite peaks reflect the evolution of bainite
and martensite formation. In fact, the mean square voltage is proportional to the
volume transformation rate [4, 5]. This is explained in more detail in chapter 5.
The positions of the maximum of the peaks in the AE data, indicated by the
vertical lines, agree rather well with the positions in the dilatation plot where the
slope changes. It should be noted that due to the inhomogeneous temperature of the
specimen an unambiguous comparison of the dilatation signal and the AE signal
is complicated and therefore not attempted for the experimental data obtained
from this set-up. By using a different set-up a more reliable comparison of both
techniques will be given in section 4.2.2.
Recent work [6] has indicated that the martensite-start temperature is best
determined by the temperature corresponding to signal maximum rather than the
temperature corresponding to the onset of the signal. The data points during the
rise time of the signal may be attributed to the thermal gradients in the sample
leading to localized transformations as the sample approaches the martensite-start
temperature. This results in a broadening of the peak. In relation to this, it is
interesting to note that the onset of the bainite peak occurs at an unexpected
high temperature of approximately 600 ◦ C. This can be explained by the thermal
gradients in the sample, which are more prominent at higher temperatures. It
should be realized that the data are plotted against the temperature in the middle
of the sample. The acoustic emission measured, however, may be generated in
regions with a lower temperature.
The formation of both bainite and martensite was confirmed by metallographic
analysis. The microstructure of steel C45 is shown in Fig. 4.6 and contains both
martensite and bainite. The fine needle structure visible in the micrograph, white
areas (M), resembles a martensitic structure with a measured micro-hardness of
approximately 700 HV0.1 , and the dark areas (B) are identified as bainite with a
micro-hardness of approximately 400 HV0.1 .
The observation of the bainite peak, with a magnitude of the same order as the
martensite peak has major implications for the interpretation of the mechanism for
bainite formation. Since only processes involving shear and release of strain energy
generate acoustic emission, the observed acoustic emission during bainite formation strongly indicates that the mechanism of growth of bainite is displacive in the
studied steel. It should be noted that the mechanism of growth of Widmanstätten
ferrite is also considered to be displacive [7], and that this transformation product
will give rise to a peak in the AE data at temperatures just above those of bainite
formation. However, microscopic analysis did not reveal any Widmanstätten ferrite
52
Chapter 4: Acoustic emission monitoring of phase transformations in steel
900
800
T [ºC]
700
B
600
500
M
400
300
200
100
0
0
5
10
15
20
25
t [s]
Figure 4.4: Cooling curve for steel C45 (specimen type B) after austenitizing at T a
= 820 ◦ C in the Gleeble thermo-mechanical simulator.
(a)
∆d [µm]
80
70
60
50
40
30
(b)
Urms [mV]
5
4
3
2
1
0
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.5: The dilatation (a) and the rms voltage (b) against temperature for steel
C45 (specimen type B) austenitized at Ta = 820 ◦ C, cooling according Fig. 4.4.
Study of steel C45
53
Figure 4.6: Optical microscopy image of the microstructure of steel C45 (specimen
type B) after continuous cooling from Ta = 820 ◦ C.
in the specimen after thermal cycling. It should also be mentioned that transformation plasticity caused by the density difference between γ and α might occur
during bainite formation, and the associated stress can in principle lead to acoustic
emission. It is believed that this does not play a significant role in the present experiments because the transformation plasticity occurring during ferrite formation
in mild steel (Fe360), which is expected to be of the same order of magnitude, did
not result in detectable acoustic emission. The AE measurements on steel Fe360
are discussed in more detail below.
The peak in the AE data cannot be explained by diffusional growth of bainite,
since such a structural change does not involve a cooperative displacement of atoms.
In fact, thermal cycling experiments using mild steel (Fe360), which transforms via
diffusive mechanisms to ferrite during continuous cooling, did not reveal any peak
in the AE data (see Fig. 4.7). On this point it should be noted that the results
obtained in this study are in contradiction with the results obtained by Speich and
Schwoeble [2]. The fact that they did not observe acoustic emission during bainite
formation can probably be attributed to the relatively small size of their samples
(10 mm in length × 3 mm in diameter). This in combination with the relatively
slow isothermal transformation to bainite probably results in a very small signal
that does not exceed the background noise level and is thus undetectable. The fact
that bainite is undetectable under the above mentioned conditions is explained in
more detail in section 4.4.
Austenite grain size effect
To evaluate the effect of the austenite grain size on the evolution of bainite and
martensite formation, AE measurements were also performed during cooling of
54
Chapter 4: Acoustic emission monitoring of phase transformations in steel
Urms [mV]
5
4
3
2
1
0
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.7: The rms voltage during cooling of steel Fe360 after austenitizing at T a
= 880 ◦ C indicating that the formation of ferrite is not accompanied by AE. Note
that bainite and martensite formation gives a signal of Urms ≈ 4 mV (see Figs. 4.2
and 4.5). The intensity at Ta = 880 ◦ C is due to noise from the heating system of
the Gleeble.
steel C45 (specimen type B) after austenitizing at Ta = 770 ◦ C and Ta = 870 ◦ C. It
should be noted that changing the austenitizing time ta can also alter the austenite
grain size. However, preliminary tests showed that large differences in austenitizing
time are required to obtain a significant effect on the austenite grain size, implying
that an approach based on changing the austenitizing time is less suitable.
It is well known that the austenite grain size increases for higher Ta , however,
estimating the prior austenite grain sizes from a martensite-bainite microstructure is
rather difficult. Therefore, quantitative metallographic analysis was carried out on
the pearlite-ferrite microstructure formed in steel C45 after slow controlled cooling
from elevated temperatures. Microscopic observations showed that the austenite
grain size is approximately 10 µm for Ta = 770 ◦ C and 30 µm for Ta = 870 ◦ C.
The results of the AE measurements during cooling from the different austenitizing temperatures are plotted in Fig. 4.8, together with the result for Ta = 820 ◦ C,
which was discussed in the previous sub-section. The magnitude of the bainite peak
relative to that of the martensite peak at each Ta gives insight into the evolution
of both phase transformations, i.e. the relative amounts of bainite and martensite
formed. In comparison with the result for Ta = 820 ◦ C, it can be seen that for the
lowest austenitizing temperature, Ta = 770 ◦ C, the martensite peak is small relative
to the bainite peak (see Fig. 4.8a). Furthermore, the martensite-start temperature
Ms , as determined by the maximum of the martensite peak, yields a value of 220
◦
C. This is in good agreement with the fact that in the CCT diagram of steel C45
the noses move to shorter times when the austenitizing temperature decreases and
that the martensite-start temperature of the remaining austenite decreases when
Study of steel C45
55
6
(a)
Urms [mV]
5
(d)
4
3
Ms
2
1
50 µ m
0
Urms [mV]
(e)
(b)
5
Ms
4
3
2
1
50 µm
0
(c)
Urms [mV]
5
Ms Ta = 870 ºC
(f)
4
3
2
1
0
50 µ m
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.8: Plot of the rms voltage against temperature for steel C45 austenitized
at (a) Ta = 770 ◦ C, (b) Ta = 820 ◦ C and (c) Ta = 870 ◦ C; the corresponding
microstructure images are given in (d), (e) and (f) respectively.
56
Chapter 4: Acoustic emission monitoring of phase transformations in steel
bainite is formed [8].
The plot for Ta = 870 ◦ C (Fig. 4.8c) shows that the martensite peak is dominant
and the bainite peak is smaller. The martensite-start temperature in this case is
approximately 290 ◦ C, which is in line with the change of the CCT diagram for
this austenitizing temperature. It should be remembered that the magnitude of the
peaks for different Ta cannot be compared in an absolute sense since the amount of
material involved is dependent on the austenitizing temperature (see section 3.3).
However, the ratio of the bainite and martensite peak is proportional to the relative
amounts of both phases.
The effect of the austenitizing temperature on the bainite and martensite peak
can be well explained in terms of a model describing the relationship between the
austenite grain size and the bainitic transformation rate in this type of steel [9]. This
model presumes that the growth rate of a bainite sheaf into the grain is relatively
small compared with the nucleation rate at the austenite grain boundary. Under
these conditions the overall transformation rate is governed by the number density
of grain boundary nucleation sites, which is determined by the mean austenite grain
size.
At low Ta , the austenite grain size is small and this results in a large number of
nucleation sites and thus a relatively large bainite peak. The remaining austenite
is enriched in carbon due to the partitioning of excess carbon from the bainite
sheaves, and this causes a relatively low martensite-start temperature. This can
be understood by realizing that the carbon enrichment increases the strength or
shear resistance of the remaining austenite, and therefore the required driving force
to initiate the shear for martensite formation is higher. Since the driving force for
martensite formation increases with undercooling, the martensite-start temperature
decreases.
Upon increasing the austenitizing temperature Ta , the austenite grain size increases and this results in less nucleation sites for bainite formation. Consequently,
only a small part of the total volume is transformed into bainite and all the remaining austenite is transformed to martensite resulting in a relatively large martensite
peak. Since only a small amount of bainite has formed, the carbon enrichment of
the remaining austenite is limited and the martensite-start temperature is relatively
high.
4.1.2
Welding experiments
As shown in the previous section, bainite formation can be monitored by means of
AE during continuous cooling in the Gleeble welding simulator. Regarding austenite grain sizes and cooling rates the transformation behaviour of a spot weld is
expected to be very similar (see section 3.3), and the AE measurements on steel
C45 during/after welding are discussed in this section.
The AE measurements during welding were carried out using the set-up as
described in section 3.2. A typical example of the results obtained in the case of
spot welding with moderate arc currents is presented in Fig. 4.9a. In this figure
Study of steel C45
57
10
Urms [mV]
8
6
4
(2)
(1)
2
0
c
(3)
(4)
8
Urms [mV]
(a)
(4)
(b)
(5)
6
(2)
4
(1)
2
0
d
(3)
0
5
10
15
20
25
t [s]
Figure 4.9: a)b) The rms voltage Urms as a function of time during and after spot
welding of steel C45 with (a) I = 55 A and (b) I = 90 A: (1) noise level due
to the arc; (2) extinction of the arc; (3) background noise level; (4) peak due to
martensite formation; (5) peak due to bainite formation. c)d) cross-sections of spot
welds produced with an arc current of (c) 55 A and (d) 90 A.
the rms voltage (Urms ) is plotted as a function of time for an arc current of 55
A. The rms voltage measured during welding (1) is due to the noise of the arc.
Upon extinction of the arc (2), Urms drops back to the background noise level
(3). After a short period in which the spot weld cools down to the martensite-start
temperature Ms , the onset of the martensite peak in the Urms data is observed. The
signal increases to a maximum (4) and then tails off to the background noise level.
In this case the heat input is relatively low, resulting in a high cooling rate of the
spot weld that leads to a complete transformation of the austenite to martensite.
To verify the formation of martensite, the spot weld produced was cut in the
middle and in Fig. 4.9c the cross-section is shown. Microscopic observation shows
that martensite is formed in both the weld metal and the heat-affected zone, the
structure being more pronounced in the weld metal. For both zones hardness values
of 600 – 700 HV0.1 were found, indicative of a fully martensitic structure.
A typical example of the results obtained for spot welding with relatively high
heat input (relatively low cooling rate) is shown in Fig. 4.9b. In this figure two peaks
can be observed: a bainite peak (5) and a martensite peak (4). The austenite in
58
Chapter 4: Acoustic emission monitoring of phase transformations in steel
the zone close to the base metal, which has a relatively small grain size, transforms
to bainite giving rise to the first peak (5). Upon further cooling, the remaining
austenite in the centre of the weld transforms to martensite resulting in the second
peak (4).
The cross-section of the spot weld produced with an arc current of 90 A shows a
dark structure in the zone adjacent to the interface between the HAZ and the base
metal, which is identified as bainite (see Fig. 4.9d). The measured micro-hardness
values in this zone were in the range 300 – 500 HV0.1 . The weld metal has a fully
martensitic structure with hardness values exceeding 600 HV0.1 .
In summary, the results during welding of steel C45 are in good agreement with
the results obtained using the Gleeble thermo-mechanical simulator. Since it is
extremely difficult to measure the temperature of a spot weld, the data shown in
this section could only be plotted against time, which complicates a quantitative
comparison between the two results. Moreover, a waveguide was used during the
Gleeble experiments which has a strong attenuating effect.
4.2
Study of steel 42CrMo4
In addition to the experiments on steel C45 described in the previous section, similar
experiments were performed on steel 42CrMo4 using the techniques described in
chapter 3. The chemical composition of steel 42CrMo4 is given in Table 3.2. A few
experiments on this steel were performed using the thermo-mechanical simulator
under the same conditions as for steel C45 (see section 4.1.1). During cooling of steel
42CrMo4 in the Gleeble thermo-mechanical simulator only martensite is formed.
The results obtained are very similar to the result obtained for steel C45 shown
in Fig. 4.2 and are discussed in detail in Ref. [10]. In section 4.2.1 the results of
welding experiments on steel 42CrMo4 are discussed. Owing to the relatively high
hardenability of this steel in comparison with steel C45, only martensite is formed
in the weld and the HAZ during cooling.
It can be concluded from the experiments described in the previous section that
a quantitative analysis of the measurements using the thermo-mechanical simulator
is complicated because of the thermal gradient in the sample during austenitizing;
the measured AE signals originate from all the regions of austenite, with different
grain sizes and/or carbon content, decomposing into bainite or martensite. On the
other hand, samples austenitized in a dilatometer or a furnace do have a homogeneous microstructure as described in sections 3.4 and 3.5. The acoustic emission
experiments on steel 42CrMo4 using the dilatometer set-up and the furnace set-up
are presented in section 4.2.2 and section 4.2.3 respectively.
4.2.1
Welding experiments
The measurement of acoustic emission during welding was carried out according
the procedure described in section 3.2. In Fig. 4.10 a typical example of the results
obtained in the case of spot welding of steel 42CrMo4 is shown. After extinction
Study of steel 42CrMo4
59
10
Urms [mV]
8
6
4
(2)
2
0
18
(1)
(4)
(3)
20
22
24
26
28
t [s]
Figure 4.10: The rms voltage as a function of time during and after spot welding
of steel 42CrMo4 with I = 75 A: (1) noise level due to the arc; (2) extinction of
the arc; (3) background noise level; (4) peak due to martensite formation.
Figure 4.11: (a) Cross-section of spot weld produced with an arc current of 75 A.
(b) Optical microscopy image of the weld metal showing a martensitic structure.
60
Chapter 4: Acoustic emission monitoring of phase transformations in steel
∆d [µm]
15
(a)
10
5
0
M sd
-5
Urms [mV]
1.0
(b)
0.8
0.6
0.4
MsAE
0.2
0.0
200
300
400
T [ºC]
Figure 4.12: Plot of the dilatation (a) and the rms voltage (b) against temperature
for steel 42CrMo4 austenitized at Ta = 900 ◦ C.
of the arc, the spot weld cools down to the martensite-start temperature and the
formation of martensite is reflected by a peak in the AE data. The signal observed
is in good agreement with the result obtained for steel C45 shown in Fig. 4.9a.
The cross-section of the spot weld is given in Fig. 4.11a. Hardness values of
approximately 700 HV0.1 were found in both the weld metal and the heat-affected
zone, indicative of a fully martensitic structure. In Fig. 4.11b an optical image of
the weld metal is shown, which clearly reveals the martensitic structure.
4.2.2
Dilatometer experiments
To facilitate comparison between the AE technique and conventional dilatometry,
the acoustic emission and dilatation were measured simultaneously using the combined AE – dilatometer (Bähr 805A/D) set-up as described in section 3.4 (see
Study of steel 42CrMo4
61
Fig. 3.15). Steel 42CrMo4 was selected as a suitable material based on the CCT
diagram and estimated (natural) cooling rates in the dilatometer. The sample was
heated to an austenitizing temperature of 900 ◦ C, austenitized for 2 minutes, and
subsequently continuously cooled to ambient temperature in a helium atmosphere.
The generated AE and change in dilatation measured during cooling are shown
in Fig. 4.12. It can be seen that the transformation to martensite is reflected in both
the dilatation and the AE data. The AE signal rapidly increases to a maximum
value, and at approximately the same temperature the slope of the dilatation curve
changes. Msd determined from the dilatation signal yields 306 ± 8 ◦ C. For the
determination of Msd , the criterion is used that when the dilatation signal starts
to deviate significantly from the austenite line during cooling; the corresponding
temperature is the start temperature of the transformation. From the AE signal
MsAE is determined as the temperature corresponding to signal maximum, yielding
302 ± 6 ◦ C. The inaccuracy in both measurements is mainly due to the temperature
differences in the sample; the maximum temperature difference is estimated to be
5 ◦ C based on measurements using two thermocouples. The inaccuracy in the
determination of MsAE and Msd from respectively the AE signal and the dilatation
signal, is only a few degrees.
The agreement between both results is satisfactory. The few data points during the rise time of the AE signal may be attributed to the thermal gradients
in the sample, leading to localized transformations as the sample approaches the
martensite-start temperature: a broadening of the peak.
A major drawback of experiments with the dilatometer described above, is the
low signal level caused by the small sample and the attenuation due to the waveguide. This made it impossible to measure bainite formation. Although the sample
sizes used are of the same order of magnitude, the waveguide in the case of the
Gleeble experiments has a larger diameter, and the transformation rates are higher
owing to the higher cooling rates.
4.2.3
Furnace experiments
As explained in the previous sub-section, the detectability of AE signals is low
using the dilatometer set-up, and therefore experiments were performed using a
conventional furnace and a salt bath as described in section 3.5. Such experiments
have the advantage that large-sized samples and a relatively thick waveguide rod
can be used. The objective of this study was to obtain bainite formation in steel
42CrMo4 and measure it by means of acoustic emission.
A sample of steel 42CrMo4 with a diameter of 16 mm and a length of 40 mm
was employed. The sample was austenitized at a temperature of 900 ◦ C for 20
minutes and subsequently cooled in a salt bath at a temperature of 400 ◦ C. During
cooling, the rms voltage of the generated AE waves and the temperature of the
sample were measured. The results, Urms (T1 = 900 ◦ C) and T as a function of t,
are shown in Fig. 4.13; the measurement was started (t = 0 s) approximately 50
seconds before cooling. It can be seen that during austenitizing (T1 = 900 ◦ C), Urms
62
Chapter 4: Acoustic emission monitoring of phase transformations in steel
16
Urms (T1 = 900 ºC)
T
Urms (T1 = 650 ºC)
14
10
800
700
8
600
6
500
4
400
2
0
T [ºC]
Urms [mV]
12
900
300
50
100
150
200
t [s]
Figure 4.13: The rms voltage and the temperature as a function of time during
cooling of steel 42CrMo4 in the salt bath.
8
Urms (corrected)
T
800
700
4
600
500
2
T [ºC]
Urms [mV]
6
900
400
300
0
50
100
150
200
t [s]
Figure 4.14: The rms voltage against time for steel 42CrMo4 corrected for the noise
due to oxidation.
Study of steel 42CrMo4
63
= 0.3 mV, the background noise value. At approximately t = 50 s the sample was
taken out of the furnace, and quenched directly in the salt bath. During the first
few seconds of quenching very high signals were observed, which are attributed to
oxidation and/or another chemical reaction between the steel and the salt. During
cooling this noise signal decreases strongly and at t = 70 s (T = 550 ◦ C) the onset
of a bainite peak can be seen. This is in good agreement with the bainite-start
temperature of this steel calculated from the empirical equation given in Ref. [7]
yielding Bs = 563 ◦ C. The rms voltage increases to a maximum value and then tails
off to a constant noise level. At t = 170 s (T = 400 ◦ C) the transformation has
probably finished, and the observed constant signal level is predominantly caused
by oxidation.
To investigate the influence of the oxidation process, a dummy experiment was
performed; the same sample was heated up to 650 ◦ C and subsequently quenched
in the salt bath. Since the sample was not austenitized, no bainitic transformation
will occur during cooling. Consequently only the oxidation process is monitored.
The result of this experiment is also shown in Fig. 4.13: Urms (T1 = 650 ◦ C). It can
be seen that the noise peak decreases relatively fast compared to the total signal
Urms (T1 = 900 ◦ C) resulting from both bainite formation and oxidation. In an
attempt to correct for the noise, the mean square voltage of the noise signal was
subtracted from the mean square voltage of the total signal. The result for the
corrected signal, the bainite peak, is shown in Fig. 4.14. Since the noise peak is
somewhat dependent on the initial temperature of the sample, the correction for
the noise (for t < 70 s) is not very accurate. Furthermore, it can be seen that the
noise peak has a rather irregular appearance; it is not really a continuous AE signal,
which complicates the correction. Nevertheless, the noise correction demonstrated
above confirms that the peak signal observed in the temperature range of 400 –
550 ◦ C can be attributed to bainite formation. This is in good agreement with
the acoustic emission measurements during bainite formation in steel C45 (section
4.1.1) and is consistent with the displacive model of the bainitic transformation.
It should be noted that the bainitic transformation in steel 42CrMo4 (see
Fig. 4.14) takes place during slow continuous cooling, and not during the isothermal
part of the treatment as was initially attempted. The lower than expected cooling
rate is predominantly caused by the low heat transfer from the sample to the liquid
salt due to the oxide layer formed on the surface. Also the large sample size (large
heat capacity) might play a role, but this is considered to be of minor importance.
This is supported by the temperature measurement with a second thermocouple,
spot welded on the surface. During cooling, the temperature difference inside the
sample is small and diminishes rapidly; when the temperature of the sample is homogeneous the temperature difference between the sample and the salt is still 100
◦
C.
In addition, AE experiments under the same conditions were performed on steel
50CrV4 and steel 16MnCr5. The observed AE signals during cooling of these steels
showed a similar trend as found for steel 42CrMo4 (Fig. 4.13), and support the
results obtained for this steel.
64
Chapter 4: Acoustic emission monitoring of phase transformations in steel
Urms [mV]
3
2
1
0
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.15: The rms voltage as a function of temperature for steel St52-3 during
continuous cooling in the thermo-mechanical simulator (Ta = 850 ◦ C).
4.3
Study of low carbon steels
In order to examine whether the carbon content has an effect on the bainitic reaction
mechanism, the low carbon steels St52-3 and St50K were studied using the thermomechanical simulator. In Table 3.2 the chemical compositions of the steels are listed.
The experiments were performed as described in section 3.3, and the analysis of
results is analogous to the analysis of the AE measurements on steel C45 described
in section 4.1.1.
In order to generate both bainite and martensite formation during continuous
cooling, the experiments were performed using specimens with an effective free
span l1 = 10 mm, a diameter d1 = 5 mm, and distance between the grips l2 =
30 mm. For both steels, the specimen was electrically heated up in 20 seconds
to the austenitizing temperature Ta = 850 ◦ C, austenitized for 20 seconds, and
continuously cooled to ambient temperature. During cooling the rms voltage of the
generated AE signals was measured. After thermal cycling, the microstructure of
the specimens was analyzed using an optical microscope (Olympus) and a microhardness tester (Buehler Ltd.).
In Fig. 4.15 the rms voltage is plotted against temperature during cooling of
steel St52-3; the result for steel St50K is shown in Fig. 4.16. It can be seen that
for both steels the evolution of bainite and martensite formation is reflected by two
distinct peaks in the AE data at temperatures of 500 – 600 ◦ C and 200 – 300 ◦ C
respectively. This result is in good agreement with the result obtained for steel C45
(see section 4.1.1). Therefore, the observed bainite peaks during cooling of steel
St52-3 and steel St50K prove that for low carbon steels the mechanism of growth
of bainite is also displacive.
Study of a high-alloyed steel
65
Urms [mV]
3
2
1
0
0
100 200 300 400 500 600 700 800 900
T [ºC]
Figure 4.16: Plot of the rms voltage against temperature for steel St50K austenitized at Ta = 850 ◦ C.
4.4
Study of a high-alloyed steel
This section reports the AE measurements on the high-alloyed steel 75MnSiCr during cooling in the salt bath. The experiments were performed using the furnace
set-up as described in section 3.5. The chemical composition of steel 75MnSiCr is
given in Table 3.2. This steel is characterized by a very low bainite-start temperature; the bainite formed at such a low temperature results in a very high strength
of this steel. The sample was block-shaped with dimensions of approximately 100
× 40 × 20 mm3 .
In order to minimize oxidation, an austenitic stainless-steel bag was folded
around the sample (not completely sealed due to the waveguide). After austenitizing at T = 900 ◦ C the sample was placed in the salt bath at T = 260 ◦ C. In
Fig. 4.17 the measured rms voltage and temperature are plotted as a function of
time during cooling in the salt bath. The cooling rate was relatively low due to the
austenitic stainless-steel bag (air layer in between the bag and the sample). From
the cooling curve the cooling time ∆t8/5 was determined yielding approximately
200 seconds. The AE and temperature measurements were started just before the
sample was quenched in the salt bath. In the beginning the background noise level
of 0.26 – 0.30 mV was measured. After 500 seconds, when the temperature of the
sample was approximately T = 350 ◦ C, the onset of a peak due to bainite formation
was observed; the signal increased to a maximum at t = 900 s, yielding U rms = 2.2
mV. At that point the sample is still slowly cooling (T = 300 ◦ C); the temperature
differences in the sample are expected to be very small. After the maximum signal
was reached, the signal tailed off to the background noise level. At approximately
t = 2500 s the sample temperature was equal to the bath temperature and U rms =
1.1 mV. After further isothermal holding the signal level slowly decreased towards
66
Chapter 4: Acoustic emission monitoring of phase transformations in steel
3.0
Urms
T
800
2.0
1.5
600
1.0
T [ºC]
Urms [mV]
2.5
1000
400
0.5
0.0
0
1000
2000
3000
4000
5000
6000
200
t [s]
Figure 4.17: Plot of the rms voltage and the temperature against time after quenching of steel 75MnSiCr in the salt bath.
the background noise level. From the AE and temperature data shown in Fig. 4.17
it can be concluded that a transition of bainite formation under continuous cooling
conditions to bainite formation under isothermal conditions takes place. Noise signals can be identified in the AE data; these are burst signals superimposed on the
continuous AE, which are probably caused by oxidation phenomena. In order to understand that oxidation related signals are still measured when using the austenitic
stainless-steel bag, it should be realized that the austenitic stainless-steel bag is
making contact with the sample or waveguide on some points and thus acoustically
coupled to the sensor.
The result obtained for steel 75MnSiCr shown in Fig. 4.17 is in line with result of
the measurement on steel 42CrMo4 discussed in section 4.2.3. The main difference
is that in this case the transformation takes place over a much longer period of
time. The observed peak in the AE data implies that the bainitic transformation
in the high-alloyed steel 75MnSiCr is displacive, which is in good agreement with
the results of the AE experiments on medium carbon steels C45 and 42CrMo4,
and the low carbon steels St50K and St52-3, which were discussed in the previous
sections of this chapter.
Also experiments were performed without using the austenitic stainless-steel
bag. In this case the cooling rate was much higher, ∆t8/5 was approximately 40
seconds, and no peak in the AE data was observed during cooling to the bath temperature. For steel 75MnSiCr the incubation time for bainite formation is relatively
long [11], and in comparison with the above described experiment the cooling rate
Conclusions
67
was probably too fast for bainite formation to take place under continuous cooling
conditions. Subsequently, isothermal holding at the bath temperature T = 290 ◦ C
for more than 12 hours did not reveal detectable acoustic emission. Experiments
on the same steel by other researchers have revealed that the bainite-start temperature lies above 300 ◦ C [11], the martensite-start temperature at T ≈ 150 ◦ C. It is
therefore expected that bainite forms at T = 290 ◦ C, but at a slow rate. The total
AE energy due to bainite formation that was measured in Fig. 4.17 is given by the
area under the curve, which is 3300 mV2 s. If a similar energy is released during 10
hours [11], the average level for U 2 is about 0.09 mV2 , which is not distinguishable
above the background level.
Maybe with a larger sample of the studied steel or another steel with higher
transformation kinetics it is possible to measure AE under true isothermal conditions. It should be remembered that the background noise level cannot be reduced; the detectability of acoustic emission is completely governed by the volume
transformation rate and the amount of released elastic energy per unit volume of
transformation product. The simple calculation given below shows that it is not
surprising that bainite formation was not detected under isothermal conditions.
4.5
Conclusions
The acoustic emission measurements discussed in this chapter have resulted in a
better understanding of the growth mechanism of the various phase transformations
in steel, in particular the bainitic transformation. Based on the results obtained
using the different techniques for thermal cycling as described in chapter 3, the
following conclusions can be drawn.
The chapter started with the discussion of acoustic emission measurements during continuous cooling of steel C45 using the Gleeble 1500 thermo-mechanical simulator. During cooling two distinct peaks in the AE data were observed at temperatures of 500 – 600 ◦ C and 200 – 300 ◦ C, which are attributed to bainite and
martensite formation respectively. The occurrence of acoustic emission during the
bainitic transformation implies that the bainite reaction mechanism in steel C45 is
diffusionless and is best described in terms of the displacive model. In contrast, no
acoustic emission was detected during the diffusion-controlled transformation from
austenite to ferrite in steel Fe360.
Acoustic emission measurements for different austenitizing temperatures revealed changes in the evolution of bainite and martensite formation in steel C45.
The ratio of the bainite and martensite peak gives information about the amounts
of bainite and martensite formed. It was found that decreasing the austenitizing
temperature enhanced bainite formation, as reflected by an increase of the bainite peak relative to the martensite peak. Moreover, the change in position of the
martensite peak reflected that the martensite-start temperature decreased in case
of enhanced bainite formation due to carbon enrichment of the remaining austenite
after bainite formation.
68
Chapter 4: Acoustic emission monitoring of phase transformations in steel
In addition, AE measurements during/after welding of steel C45 were performed
to study the phase transformation occurring in a weld during cooling. In the case
of spot welding with moderate arc currents a martensite peak was observed during
cooling. At high heat input, the cooling rate decreased and a bainite peak followed
by a martensite peak were observed after extinction of the arc. It can be concluded
that the results obtained for welding of steel C45 are in good agreement with the
results of AE measurements on steel C45 using the Gleeble thermo-mechanical
simulator.
During cooling of steel 42CrMo4 after welding and in the thermo-mechanical
simulator only martensite was formed. The results obtained are in line with the results obtained for martensite formation in steel C45. Subsequently, the martensitic
transformation in steel 42CrMo4 was studied using the combined AE – dilatometer (Bähr 805A/D) set-up. The simultaneous measurement of acoustic emission
and dilatation during the martensite formation in steel 42CrMo4 allowed a proper
comparison between the AE technique and conventional dilatometry. The transformation to martensite was reflected in both the dilatation and the AE data. Analysis
of the data showed that the martensite-start temperature Msd determined from the
dilatation signal was only 4 ◦ C higher than the MsAE determined from the AE signal. This indicates that the sensitivity of both techniques in the set-up used is of
the same order of magnitude. In general, however, it is difficult to draw unambiguous conclusions about the sensitivity of acoustic emission compared to dilatation.
This is due to the fact that the intensity of the AE signal is not only dependent on
the sample volume but also dependent on the transformation rate. Moreover, the
intensity of the signal is strongly affected by the use of a waveguide.
To study the bainitic transformation in steel 42CrMo4, experiments were performed using a conventional furnace and a salt bath. Due to oxidation very high
signals were observed directly after the sample was quenched in the salt bath.
However, this noise signal decreased strongly and the observed AE signal in the
temperature range of 400 – 550 ◦ C could be attributed to bainite formation. In
order to correct for the oxidation process, a dummy experiment was performed: the
same sample was heated up to 650 ◦ C and subsequently quenched in the salt bath.
The result obtained for the dummy experiment showed only the AE signal due to
oxidation. These data were used to extract the AE signal due to bainite formation
from the total signal.
In addition to the study of the medium carbon steels C45 and 42CrMo4, AE
measurements were performed during bainite and martensite formation in two low
carbon steels, St50K and St52-3, and during bainite formation in the high-alloyed
steel 75MnSiCr. The results of the AE measurements on the low carbon steels using
the thermo-mechanical simulator revealed two peaks in the AE data corresponding
bainite and martensite formation, similar to the case of steel C45. The chapter ends
with the discussion of AE measurements on steel 75MnSiCr, which is characterized
by a very low bainite-start temperature and a relatively long incubation time for
bainite formation. During cooling of this high-alloyed steel in the salt bath a peak in
the AE data was observed at bainite transformation temperatures, which reflected
Conclusions
69
that the transformation takes place over approximately two hours.
The overall results presented in this chapter show that acoustic emission is
generated during bainite formation in a wide variety of carbon steels. This gives
strong evidence that the bainitic transformation is displacive.
70
Chapter 4: Acoustic emission monitoring of phase transformations in steel
References
[1] G.R. Speich and R.M. Fisher, Acoustic Emission (ASTM STP 505), 140
(1972).
[2] G.R. Speich and A.J. Schwoeble, Acoustic Emission (ASTM STP 571), 40
(1975).
[3] C.R.F. Azevedo, A.A. Garboggini and A.P. Tschipitschin, Mater. Sci. Technol.
9, 705 (1993).
[4] S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden, J. Phys. D: Appl.
Phys 34, 3312 (2001).
[5] S.M.C. van Bohemen, M.J.M. Hermans, G. den Ouden and I.M. Richardson,
J. Phys. D: Appl. Phys. 35, 1889 (2002).
[6] S.M.C. van Bohemen, J. Sietsma, M.J.M. Hermans and I.M. Richardson, Acta
Mat. 51, 4183 (2003).
[7] H.K.D.H. Bhadeshia, Bainite in steels, The Institute of Materials, London
(2001).
[8] P.J. van der Wolk, PhD thesis, Delft University of Technology, DUP Science,
Delft (2001).
[9] A. Matsuzaki and H.K.D.H. Bhadeshia, Mater. Sci. Technol. 15, 518 (1999).
[10] S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden, Mater. Sci. Technol. 18, 1524 (2002).
[11] F.G. Caballero, H.K.D.H. Bhadeshia, K.J.A Mawella, D.G. Jones and P.
Brown, Mater. Sci. Technol. 18, 279 (2002).
71
72
Chapter 4: Acoustic emission monitoring of phase transformations in steel
Chapter 5
A study of acoustic emission
energy generated during
bainite and martensite
formation
In this chapter a quantitative study is presented of acoustic emission signals generated during bainite and martensite formation. A relationship is derived and validated that relates a measurable AE signal parameter to the volume of martensite
or bainite formed.
As explained in section 2.3.1 a large amount of the Gibbs free energy differ0
ence ∆Gγ→α is dissipated in the process of slip, which occurs within a martensite
crystal in order to relax the stresses accompanying the lattice transformation [1].
During the transformation a part of the released strain energy is radiated as transient elastic waves called acoustic emission. The dynamics of the transformation on
a microscopic scale has a strong dependence on the chemical composition, and consequently the amount of the free energy change ∆G that is converted into acoustic
emission energy is expected to be material dependent.
Since many material properties of the specimen and transducer influence the
AE waves during the period between generation at the source and detection at the
transducer it is impossible to determine the absolute value of the energy of an AE
source, unless a proper calibration can be carried out [2]. The energy value of the
electrical signal measured at the transducer is affected by many factors such as the
acoustic impedance of the transducer and the specimen and the system bandwidth.
However, for experiments using the same specimen configuration these factors may
be treated as constants, and under such conditions the results of experiments using
the same type of sources with different magnitudes can be compared.
In section 5.1 a relation for the AE energy generated during martensite for73
74
Chapter 5: A study of acoustic emission energy ...
mation will be derived from basic principles, and subsequently extended to include
bainite formation. The relation is tested for various steels that transform to martensite and/or bainite during welding. Both spot welding and travelling arc welding
experiments are performed. The validation of the relation for martensite formation
in steel 42CrMo4 and steel 42MnV7 is presented in section 5.2. In section 5.3 the
results of welding experiments on steel C45 are discussed. After spot welding of this
steel only martensite is formed in the weld; under travelling arc welding conditions
bainite is also formed. Based on both sets of experiments the relationship for AE
energy during bainite formation is tested.
Besides the fundamental interest in the generation of acoustic emission during
displacive phase transformations, the results of this study may be used to develop an
AE monitoring system to detect martensite and bainite formation during welding.
Since these hard regions in the weld and heat-affected-zone (HAZ) are susceptible
to cold cracking due to hydrogen embrittlement [3], real time monitoring of the
welding process is of considerable practical importance. Although it has been known
for a long time that acoustic emission can be detected during the formation of
martensite, little effort has been made to study this transformation during welding.
Most recently, Liu and Kannatey-Assibu performed acoustic emission experiments
during welding of high carbon steel [4]. They showed that after welding the AE
signals originating from martensite formation could be distinguished from other
signals. However, to identify martensite formation during welding was found to be
difficult due to the noise of the gas metal arc (GMA) welding process.
5.1
Theoretical background
Consider a volume of austenite transforming to martensite immediately after welding when the temperature T falls below the martensite-start temperature M s . A
0
fraction of the free energy change ∆Gγ→α at the Ms temperature is converted into
AE energy. This fraction is assumed to be constant over the temperature range M s
0
to Mf , where Mf is the martensite-finish temperature. The increase in ∆Gγ→α at
lower temperatures is necessary to start the nucleation at lattice distortions in the
parent phase that are less favourable. When it is assumed that the energy dE AE of
AE waves that are generated during the formation of a volume dVm of martensite
is proportional to this volume, then
dEAE ∝ dVm
(5.1)
The energy emission rate of an AE source can be expressed in terms of the mean
square voltage U 2 of the acoustic waves detected at a transducer [5], or
dEAE
∝ U2
dt
(5.2)
with U 2 defined by Eq. (3.1). By substitution of Eq. (5.2) in Eq. (5.1) and integration it follows that
Martensite formation
75
Z
U 2 dt ∝
Z
dVm
(5.3)
with the integration running over the entire transformation range. In the case of
spot welding, the integral term on the right hand side is equivalent to the volume
Vm of martensite produced in the spot weld and Eq. (5.3) can be written as:
Z
2 dt = k V
Um
(5.4)
m m
with km the proportionality factor between the AE energy and the transformed
volume, which is presumably material dependent.
The integral on the left hand side can be calculated by measuring the area
under the curve in a U 2 versus t plot. Assuming that this theory is valid for every
displacive transformation involving acoustic emission, it follows that when bainite
and martensite formation occur successively, Eq. (5.4) generalizes to
Z
U 2 dt = kb Vb + km Vm
(5.5)
with kb the proportionality factor for bainite formation and Vb the volume of bainite. This equation can be written in differential form as:
U 2 = kb
dVm
dVb
2
+ km
= Ub2 + Um
dt
dt
(5.6)
It should be noted that Eqs. (5.5) and (5.6) may be viewed as manifestations
of the conservation of energy (power). When bainite and martensite are formed
simultaneously, the total measured AE power at each moment is the sum of the AE
powers due to bainite and martensite formation respectively. For example, in the
case of travelling arc welding martensite and bainite may be formed at the same
time. Under steady state welding conditions the amounts of bainite and martensite
formed per unit time are constant, and the validity of Eq. (5.6) can be tested.
It should be noted that for travelling arc welding experiments, the mean square
voltage of the electrical noise and the welding noise need to be subtracted from the
measured mean square voltage.
5.2
Martensite formation
In order to examine the AE energy generated during martensite formation, welds
were made on steel 42CrMo4 and steel 42MnV7. For both spot welding and travelling arc welding a complete transformation to martensite is achieved in these
commercially available steels. In section 5.2.1 the results obtained for travelling arc
welding of steel 42CrMo4 are discussed. The analysis of spot welding experiments
on steel 42MnV7 is presented in section 5.2.2.
76
Chapter 5: A study of acoustic emission energy ...
10
Urms [mV]
8
(4)
6
4
(1)
2
(2)
(5)
(3)
0
0
10
20
30
40
50
60
t [s]
Figure 5.1: The rms voltage plotted versus time during travelling arc welding of
steel 42CrMo4 with I = 75 A: (1) background noise level; (2) arc ignition; (3)
increase signal level due to martensite formation; (4) steady state condition; (5)
extinction of arc.
5.2.1
Travelling arc welding of steel 42CrMo4
Steel 42CrMo4 in the form of plates with dimensions 250 mm × 200 mm × 8 mm
was used for the experiments. The chemical composition of steel 42CrMo4 is given
in Table 3.2, and the AE measurements during welding were performed using the
set-up as described in section 3.2. To vary the amount of martensite formed, a
large number of welding experiments with different heat inputs was carried out
under travelling arc conditions with a travel speed v = 2 mm/s. An example of the
results obtained is shown in Fig. 5.1. In this figure the rms voltage is plotted as a
function of time for an arc current of 75 A. The plot shows that under non-welding
conditions the background noise is measured (1). When the arc is ignited and the
welding cycle starts (2), the signal abruptly rises to a higher level. This level is
considered as the noise signal level produced by the arc during welding. After a
certain time the first part of the weld has cooled down to the Ms temperature and
starts to transform to martensite, which leads to an increase in signal level (3). This
continues until a steady state situation is reached (4). At the end of the welding
cycle (5) a peak is observed, which is due to the fact that after extinction of the
arc the cooling rate increases and, hence, a larger volume of metal is transformed
to martensite per unit time.
To test the validity of Eq. (5.6) for martensite formation the rms voltage was
Martensite formation
77
10
9
35 A
60 A
75 A
100 A
8
Urms [mV]
7
6
5
4
3
2
1
0
0
10
20
30
40
50
60
70
t [s]
Figure 5.2: The rms voltage as a function of time during travelling arc welding of
steel 42CrMo4 for different arc currents.
Figure 5.3: The cross-sections of welds produced on steel 42CrMo4 with an arc
current of (a) 35 A, (b) 60A, (c) 75 A and (d) 100 A.
78
Chapter 5: A study of acoustic emission energy ...
measured during 14 welding runs with arc currents in the range of 35 up to 100 A
with steps of 5 A. The results of four typical welding runs performed with 35 A, 60
A, 75 A and 100 A are plotted in Fig. 5.2. In order to minimize overlapping of data
points, the time basis is shifted for the four welding runs. It is seen that the signal
level increases with arc current and that also the peak after extinction of the arc
becomes higher and broader due to the fact that the amount of material involved
increases.
After each welding run, the volume rate dVm /dt of martensite formation was
determined by measuring the area Am of the martensitic phase from weld crosssections and multiplying this area by the travel speed v. In Fig. 5.3 the crosssections corresponding the four welding runs of Fig. 5.2 are shown. Hardness measurements confirmed that in all cases both the weld metal and the heat-affected
zone (HAZ) are fully martensitic. It should be noted that the martensitic structure
is more pronounced in the weld metal than in the HAZ due to the fact that in the
weld metal the mean grain size is larger than in the HAZ. The results obtained for
the 14 welding runs are listed in Table 5.1.
2 under steady state conditions (corrected
In Fig. 5.4 the average AE power Um
for welding noise) is plotted as a function of dVm /dt. The error bars reflect the
2 ; the value of dV /dt can be determined
inaccuracy in the determination of the Um
m
with high accuracy. The data obtained from the welds produced with arc currents
in the range of 35 – 100 A can be fitted to Eq. (5.6) with reasonable accuracy. The
result for km from the fit yields 0.85 ± 0.1 × 103 V2 s/m3 . Regarding the accuracy of
the fit, it should be noted that for low heat inputs the welding noise has a relatively
large contribution to the total measured AE power (see Table 5.1). Moreover, the
accuracy in the determination of the AE power due to the welding noise is limited.
2 at low heat inputs are relatively inaccurate,
Therefore, the results obtained for Um
as expressed by the error bars in Fig. 5.4.
5.2.2
Spot welding of steel 42MnV7
In this section the results of spot welding experiments on steel 42MnV7, in the
form of plates with dimensions 250 mm × 200 mm × 8 mm, are discussed. The
chemical composition of steel 42MnV7 is given in Table 3.2. Two typical examples
of the results obtained in the case of spot welding are presented in Figs. 5.5a and
5.5b. The martensite peak observed during cooling of the weld is very similar to
the peak observed after spot welding of steel C45 with moderate heat input, see
section 4.1.2.
In order to check the reproducibility of the experiments two series of spot welds
were made; first a series with arc currents in the range of 65 to 110 A, and secondly
a series with arc currents in the range of 50 to 95 A. Comparing the results makes
it possible to estimate the accuracy in determining the volume of martensite in a
spot weld; hence, for an accurate determination of the transformed volume it is
required that the spot weld is cut exactly in the middle.
2 dt was determined for each weld. Microscopic
The generated AE energy ∫ Um
Martensite formation
79
Table 5.1: Results of travelling arc experiments on steel 42CrMo4.
Arc
current
(A)
35
40
45
50
55
60
65
70
75
80
85
90
95
100
Arc
voltage
(V)
10.1
10.8
10.2
10.2
10.1
10.3
10.1
10.3
10.3
10.6
10.5
10.6
10.5
10.5
Arc
length
(mm)
2.1
2.5
2.8
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
3.0
dVm /dt
Urms
total
(mV)
1.4
1.5
1.7
2.1
2.3
2.8
3.1
3.5
4.0
4.3
4.5
4.9
5.1
5.3
(mm3 /s)
3.1
4.4
4.9
6.3
7.1
9.9
12.0
14.4
16.8
20.3
22.1
25.8
27.3
30.0
Urms
noise
(mV)
0.8
0.8
0.8
0.9
0.9
0.9
0.9
0.9
0.9
0.9
1.0
1.0
1.0
1.0
2
Um
martensite
(mV2 )
1.3
1.7
2.3
3.6
4.5
7.0
8.8
11.4
15.4
17.7
19.3
23.0
25.0
27.3
30
Um2 [mV2]
25
20
15
10
5
0
0
5
10
15
dVm/dt
20
25
30
[mm3/s]
2 against dV /dt for welding on steel
Figure 5.4: Plot of the average AE power Um
m
42CrMo4 with arc currents in the range of 35 to 100 A.
80
Chapter 5: A study of acoustic emission energy ...
10
Urms [mV]
8
(a)
(4)
6
4
(2)
2
(1)
0
(c)
(3)
(b)
Urms [mV]
8
(4)
6
(2)
4
(1)
2
0
(d)
(3)
0
5
10
15
20
t [s]
Figure 5.5: a)b) The rms voltage Urms as a function of time during and after spot
welding of steel 42MnV7 with (a) I = 60 A and (b) I = 85 A: (1) noise level due
to the arc; (2) extinction of the arc; (3) background noise level; (4) peak due to
martensite formation. c)d) Transverse cross-sections of spot welds produced with
an arc current of (c) 60 A and (d) 85 A.
analysis of the cross-sections (see Figs. 5.5c and 5.5d) revealed that in all cases
martensite is formed in both the weld metal and the HAZ; for both zones microhardness values of 600 – 700 HV0.1 were found, indicative of a fully martensitic
structure. The cross-sections were also used to measure the radius r and depth d
of the martensitic zone from which the volume Vm of martensite in the spot weld
was calculated. The volume of a spot weld can be calculated from the measured
radius r and depth d of the spot weld using the illustration shown in Fig. 5.6. This
figure shows that the angle θ is a function of r and d as
cos θ =
(r − d)
r
(5.7)
By inserting the values of θ and r in
Vm =
Z
θ
0
³
(cos θ)3
1´
−
π(r sin θ 0 )2 r sin θ 0 dθ0 = πr3 − cos θ + 1 +
3
3
(5.8)
the volume Vm is obtained. The results obtained for the spot welds on steel 42MnV7
are listed in Table 5.2.
Martensite formation
81
θ
r
d
Figure 5.6: A typical spot weld can be approximated to fall inside a sphere with
radius r. Based on this and the depth d of the spot weld, the volume of the spot
weld can be calculated.
2 dt (corrected for the thermal background noise)
In Fig. 5.7 the values of ∫ Um
are plotted as a function of Vm . From the fit of the data to Eq. (5.4) it follows
that km yields a value of 1.15 ± 0.05 × 103 V2 s/m3 . Comparison of the results of
the two series shows that the measurements are fairly reproducible, especially for
relatively high heat inputs. Only for the low heat inputs significant differences are
seen between the two series. This is due to the fact that it is relatively difficult to
cut small spot welds exactly in the middle.
In summary, it has been demonstrated that both spot welding (sub-section 5.2.2)
and travelling arc welding (sub-section 5.2.1) can be used to determine the proportionality factor k. The advantages and disadvantages of each method are discussed
below. Since the thermal background noise is much lower than the welding noise,
2 dt after spot welding can be determined with relatively high accuthe integral ∫ Um
2 during travelling arc welding. Furthermore,
racy compared with the AE power Um
the value of the welding noise cannot be determined with high accuracy. On the
other hand, it is in principle more difficult to determine the volume of martensite
Vm in a spot weld than to determine the volume transformation rate dVm /dt during
travelling arc welding. This is due to the fact that it is experimentally difficult to
cut a spot weld exactly in the middle; this does not play a role for travelling arc
welds. The results obtained for steel 42MnV7, however, show that the volume of
martensite in a spot weld can be determined with sufficient accuracy, especially
for the welds produced with high heat inputs. Overall, it can be concluded that
spot welding is a more accurate method to study the AE energy generated during
82
Chapter 5: A study of acoustic emission energy ...
martensite formation.
In addition, a few spot welds were produced on steel 42CrMo4 in order to verify
whether a similar value of km can be obtained as in the previous sub-section. A
typical example of a result obtained for spot welding is shown in Fig. 4.10 and
discussed in the corresponding section. The values of km were determined by mea2 against t, and the volume V
suring the area under the peak in a plot of Um
m of
martensite formed, and inserting the values obtained in Eq. (5.4). For a few spot
welds produced with different heat inputs, a value of 1.0 ± 0.2 × 103 V2 s/m3 was
found for km . This is in good agreement with the results obtained in the case of
travelling arc welding of steel 42CrMo4 (section 5.2.1).
Table 5.2: Results for two series of spot
R
2 dt
I
r
d
Vm
Um
3
(A)
(mm) (mm) (mm ) (mV2 s)
50
3.2
1.3
15
22
60
3.3
1.6
23
26
65
3.5
1.8
30
34
70
3.7
2.0
39
44
75
3.7
2.3
49
57
80
3.9
2.4
57
67
85
4.0
2.8
74
84
90
4.2
2.9
85
101
95
4.4
3.0
99
113
5.3
welding experiments on steel 42MnV7.
R
2 dt
I
r
d
Vm
Um
3
(A)
(mm) (mm) (mm ) (mV2 s)
65
3.5
1.9
31
27
70
3.6
2.1
40
35
75
3.7
2.4
52
59
80
3.9
2.5
59
71
85
4.0
2.7
73
87
90
4.2
2.9
85
98
95
4.4
3.1
103
115
100
4.5
3.2
112
128
105
4.6
3.4
129
147
110
4.9
3.6
151
178
Bainite and martensite formation
In this section the study of the released AE energy during welding of steel C45
is presented. The chemical composition of steel C45 is given in Table 3.2. The
experiments were performed using plates with dimensions 250 mm × 200 mm ×
5 mm, and the welds were made using the computer controlled GTA welding unit
as described in section 3.2. Based on the results discussed in section 4.1.2 it is
known that bainite and/or martensite are formed in the weld. Whether bainite or
martensite is formed at a specific location in the weld depends on the austenitizing
temperature, the austenitizing time and the cooling rate at that position.
To vary the amounts of bainite and martensite formed, experiments were performed with different arc currents under two conditions. Spot welds were made
to examine the AE energy generated during the martensitic transformation. It
should be noted that after spot welding the cooling rate is relatively high, which
is required for martensite formation. The results of the AE measurements after
spot welding are discussed in section 5.3.1. Subsequently, travelling arc welding
Bainite and martensite formation
2
2
Um dt [mV s]
160
83
first series
second series
linear fit
120
80
40
0
0
20
40
60
80
100
120
140
160
Vm [mm3]
2 dt against V for two series of spot welds produced on steel
Figure 5.7: Plot of ∫ Um
m
42MnV7 with arc currents in the range of 50 – 110 A.
experiments were performed to study the AE energy generated during the bainitic
transformation. The results of this study are presented in section 5.3.2.
5.3.1
Spot welding of steel C45
A series of spot welds was made using different heat inputs. The spot welding time
was limited to 10 seconds to avoid excessive heating of the workpiece. It should
be noted that a high temperature of the workpiece leads to a decrease in cooling
rate after spot welding. After extinction of the arc, the rms voltage Urms of the
continuous acoustic emission was measured.
A typical example of the results obtained in the case of spot welding with
moderate arc currents is presented in Fig. 4.9a and discussed in the corresponding
section. To test the validity of Eq. (5.4), Urms was measured for a number of spot
welds produced with arc currents in the range of 35 to 80 A. Of each weld the
2 against t was calculated in order to
area under the martensite peak in a plot of Um
2
determine the generated AE energy ∫ Um dt. Cross-sections were made of all welds,
and microscopic observation showed that for all welds in both the weld metal and
the HAZ a fully martensitic structure, with micro-hardness values of 600 – 700
HV0.1 , was formed. The cross-sections were also used to measure the radius r and
depth d of the martensitic zone from which the volume Vm of martensite in the spot
weld was calculated. The results obtained are listed in Table 5.3.
2 dt (corrected for the background noise) are plotted
In Fig. 5.8 the values of ∫ Um
as a function of Vm . It can be seen that for welds obtained with arc currents in
84
Chapter 5: A study of acoustic emission energy ...
the range of 35 to 80 A a linear relation exists between AE energy and the volume
of martensite formed, as predicted by Eq. (5.4). The proportionality factor k m
extracted from the linear fit yields a value of 2.38 ± 0.05 × 103 V2 s/m3 .
The results obtained for martensite formation in steel C45 are in good agreement
with the results obtained in the previous section using steels 42CrMo4 and 42MnV7.
The observed differences in the values of km can be attributed to the differences
in chemical composition between the steels. The fraction of the free energy change
0
∆Gγ→α that is released in the form of acoustic waves is probably dependent on
the chemical composition of the material.
I
(A)
35
40
45
50
55
60
65
70
75
80
5.3.2
Table 5.3: Results for spot welding of steel C45.
R
2 dt
r
d
Vm
Um
(mm3 )
(mm)
(mm)
(mV2 s)
2.3
1.1
7
22
2.4
1.3
11
29
2.7
1.4
14
43
3.0
1.7
23
74
3.1
2.0
31
83
3.2
2.3
41
101
3.6
2.5
53
123
3.7
2.8
66
150
3.8
2.9
72
172
3.9
3.0
83
195
Travelling arc welding of steel C45
Welding runs were made under travelling arc conditions with v = 2 mm/s and
different arc currents. Before, during and after welding Urms was measured. A
typical example of the results obtained is presented in Fig. 5.9a, in which U rms is
plotted as a function of welding time for an arc current of 55 A. After a certain
welding time, the first part of the weld starts to cool down and transforms partly to
bainite, and the remaining austenite transforms subsequently to martensite. Both
transformations contribute to the increase in signal level (3). This continues until
a steady state situation is reached (4). At the end of the welding cycle (5) a peak
is observed, which is due to the fact that after extinction of the arc the cooling rate
increases.
In order to test the validity of Eq. (5.6) for the case of bainite and martensite
formation, Urms was measured during a number of welding runs with arc currents
in the range of 40 to 60 A. Under steady state conditions, bainite and martensite
formation occur simultaneously and the measured AE power U 2 is the sum of the
AE powers due to both transformations.
Transverse cross-sections were made of all welds and in Fig. 5.9b the transverse
Bainite and martensite formation
85
200
Um2dt [mV2s]
160
120
80
40
0
0
20
40
60
80
100
3
Vm [mm ]
2 dt against V
Figure 5.8: Plot of ∫ Um
m for spot welds produced on steel C45 with
arc currents in the range of 35 – 80 A.
10
(a)
Urms [mV]
8
(b)
(4)
6
(2)
4
2
(3)
(1)
0
0
10
(5)
20
30
40
50
t [s]
Figure 5.9: (a) The rms voltage Urms plotted against time during travelling arc
welding of steel C45 with I = 55 A: (1) background noise level; (2) arc ignition; (3)
increase signal level due to martensite and bainite formation; (4) steady state condition; (5) extinction of the arc. (b) Transverse cross-section of the weld produced
with an arc current of 55 A.
86
Chapter 5: A study of acoustic emission energy ...
cross-section for I = 55 A is shown. Optical microscopy observation and microhardness measurements showed that the microstructure in the central part of the
weld (white area) is fully martensitic whilst in the zone adjacent to the interface
between the HAZ and the base metal bainite is formed (dark area). In that part
of the weld, the peak temperature during welding (austenitizing) is relatively low,
which leads to a small mean austenite grain size, which in turn enhances bainite formation. The transverse cross-sections were used to measure the areas A m and Ab of
the martensitic phase and the bainitic phase respectively. The volume transformation rates dVm /dt and dVb /dt of martensite and bainite formation were determined
by multiplying the areas by the travel speed v.
For each welding run, the average AE power under steady state conditions was
corrected for the electrical noise and the welding noise to obtain the AE power
2
Utotal
due to martensite and bainite formation. Using the km from the fitting of the
spot welding data, the value of the AE power due to bainite formation (U b2 ) can be
2 ). The results obtained are listed
separated from that of martensite formation (Um
in Table 5.4.
Table 5.4: Results for travelling arc welding of steel C45.
I
(A)
40
45
50
55
60
Am
(mm2 )
2.1
2.3
2.5
3.4
4.2
Ab
(mm2 )
0.4
1.1
1.6
1.7
2.3
2
Utotal
(mV2 )
9.9
13.2
15.4
20.4
25.0
2
Um
(mV2 )
9.6
10.6
11.5
15.7
19.3
Ub2
(mV2 )
0.3
2.6
3.9
4.7
5.7
The values of Ub2 are plotted as a function of dVb /dt in Fig. 5.10. It can be
seen that the AE power of bainite formation is proportional to the volume rate of
bainite formation, as predicted by theory. From the least-squares fit of the data to
Eq. (5.6) it follows that kb = 1.2 ± 0.1 × 103 V2 s/m3 . It should be noted that this
is approximately a factor 2 smaller than the value of km for steel C45. This can be
explained by the fact that the bainitic transformation involves less strain energy
[6], and therefore the amount of plastic deformation during the growth of bainite
is smaller than during the growth of martensite. Assuming that the AE energy
generated during growth of the new phase is proportional to the strain energy
involved, it immediately follows that the amount of AE energy produced per unit
volume of bainite is smaller than that produced per unit volume of martensite, i.e.
kb is smaller than km .
Conclusions
87
7
Ub2 [mV2]
6
5
4
3
2
1
0
0
1
2
3
4
5
6
dVb /dt [mm3/s]
Figure 5.10: Plot of Ub2 as a function of dVb /dt for welding runs on steel C45
produced with arc currents in the range of 40 – 60 A.
5.4
Conclusions
In this chapter the measurements of acoustic emission signals during welding of a
number of steels are presented and discussed. The results obtained are interesting
both from a fundamental and practical point of view.
The results of AE measurements during travelling arc welding show that the
rms voltage due to martensite formation is significantly larger than the rms voltage
due to the noise of the welding process. This indicates that acoustic emission can be
applied as a real-time monitoring technique for the detection of martensite formation during welding of steel. Moreover, after extinction of the arc a characteristic
peak in the rms voltage is observed during cooling, which reflects the evolution of
martensite formation.
In the beginning of this chapter a relationship was derived that describes the
AE energy generated during bainite and martensite formation. In the subsequent
sections this relation was tested using three carbon steels: 42CrMo4, 42MnV7 and
C45.
In the case of travelling arc welding of steel 42CrMo4 only martensite is formed.
The analysis of the measured acoustic emission signals showed that the mean square
2 ) is proportional to the volume rate dV /dt of martensite formation,
voltage (Um
m
as predicted by theory. The relation was also validated for spot welding of steel
42MnV7 and steel C45. For both steels it was found that the integrated AE signal
2 dt, is proportional to the volume
(AE energy) during cooling of a spot weld, ∫ Um
Vm of martensite formed.
88
Chapter 5: A study of acoustic emission energy ...
In the case of travelling arc welding of steel C45, bainite and martensite formation occur simultaneously and both contribute to the measured rms voltage.
A combination of the spot welding and the travelling arc welding results showed
that AE power due to bainite formation (Ub2 ) is proportional to the volume rate of
bainite formation (dVb /dt). This indicates that the relation is generally valid for
displacive transformations involving acoustic emission.
References
[1] R.W.K. Honeycombe and H.K.D.H. Bhadeshia, Steels, Edward Arnold, London (1995).
[2] Y. Berlinsky, M. Rosen, J.A. Simmons and H.N.G. Wadley, Rev. of progress
in quant. nondestr. eval. 5, 1345 (1986).
[3] J.F. Lancaster, Metallurgy of Welding, Allen & Unwin Press, London (1986).
[4] X. Liu and Jr. E. Kannatey-Asibu, Welding Journal 69, 389s (1990).
[5] E. Kannatey-Asibu Jr, and P. Dong, ASME J. of Eng. for Ind. 108, 328 (1986).
[6] H.K.D.H. Bhadeshia, Bainite in steels, The Institute of Materials, London
(2001).
89
90
Chapter 5: A study of acoustic emission energy ...
Chapter 6
Kinetics of the martensitic
transformation studied by
means of acoustic emission
The transition from austenite to martensite is governed by thermodynamics and
kinetics as explained in detail in chapter 2. The underlying thermodynamics determine the available driving force for transformation; the kinetics of a martensitic
transformation depend solely on nucleation, because the growth of a martensitic
crystal usually occurs rapidly. It is well known that the mechanism of growth is
displacive. How the phase nucleates, however, is even today not completely understood. This is mainly due to the great speed of formation, which makes the
martensitic transformation a difficult process to study experimentally. Nevertheless, the results presented in this chapter show that acoustic emission, which has
rarely been used in the past, is a very suitable technique to study the kinetics of
the martensitic transformation.
In section 6.1 the techniques commonly used to study the transformation kinetics are reviewed. The derivation of the Koistinen-Marburger (KM) equation
describing the progress of martensite formation in a sample is given in section 6.2.
In section 6.3 the experimental procedure is described. The martensitic transformation in the four carbon steel specimens (C50, C60, C70 and C80) is monitored
by means of acoustic emission using the Gleeble 1500 thermo-mechanical simulator;
the dilatation is measured simultaneously. In section 6.4 the AE measurements during cooling of the steels are analyzed and discussed. For the analysis of the AE data
the results obtained in chapter 5 are used. In this chapter it was shown that the
measured power of acoustic waves generated during the martensitic transformation
is proportional to the transformation rate. Based on this relation, the volume fraction of martensite as a function of undercooling below the Ms temperature can be
calculated, which is demonstrated in section 6.4.1. This allows a direct comparison
91
92
Chapter 6: Kinetics of the martensitic transformation ...
of experimental data with theoretical predictions of the transformation kinetics,
and leads to a better understanding of the transformation characteristics such as
nucleation rates and average volume of martensite crystals. At the end of the chapter the experiments on a model system, a shape memory alloy, are discussed. Both
acoustic emission and confocal microscopy were employed to study the martensitic
transformation in this material.
6.1
Introduction
During cooling, the martensitic transformation starts at a certain temperature,
which is designated the martensite-start temperature Ms , and proceeds only upon
further cooling below this temperature [1, 2]. In order to gain more insight into the
kinetics of the martensitic transformation, it is necessary to investigate quantitatively how the transformation progresses upon continuous cooling. Characterization
of the volume fraction of martensite can be realized by means of several techniques,
such as electrical resistivity [3], quantitative metallography [4, 5], dilatometry [6]
and acoustic emission [3, 7].
About forty years ago, Koistinen and Marburger presented an empirical equation that describes the volume fraction of residual austenite as a function of temperature below the martensite-start temperature [8]. This equation, named the KM
relation, was obtained by fitting their (and other investigators’) data from X-ray
measurements on plain carbon steels. Some years later, Magee [9] showed that the
KM relation can be derived theoretically from basic principles, taking into account
certain assumptions regarding the nucleation and growth of martensite. The most
important assumption of this theory is that the average volume of the martensite
crystals is constant over the extent of the transformation. On this point the KM
relation is different from a theory formulated by Fisher [10], which assumes that the
average volume decreases strongly as the transformation proceeds. The derivation
of the KM relation and the partitioning model of Fisher are discussed in detail in
section 6.2.
Speich and Fisher [3] showed that electrical resistivity and acoustic emission
measurements during martensite formation in an FeNi alloy are in agreement with
the partitioning model of Fisher [10]. Khan and Bhadeshia studied the development
of martensite formation in partially bainitic steel using dilatometry [6]. For volume
fractions up to 0.6 they found reasonably good agreement between experimental
and calculated values using the KM model. The discrepancy in the above modelling
results can probably be explained by differences in the morphology of the martensite
crystals and the austenite grain size. In the case of FeNi the grain size is large and
plate martensite is formed, whilst in the partially bainitic steel the grain size is
small and lath martensite is formed.
Theoretical background
6.2
93
Theoretical background
Theories of the kinetics of the athermal martensitic transformation attempt to describe the progress of martensite formation in a sample. The overall transformation
rate is governed by nucleation and growth. The nucleation of martensite during
cooling is believed to take place at structural imperfections in the parent phase and
these pre-existing embryos (defects) are stimulated to grow into martensite crystals
at different degrees of undercooling below Ms ; they have different energy barriers
to activation [2]. Since growth can be very fast, each nucleation event directly leads
to the formation of a typical volume of the new phase. Thus the volume fraction
of martensite varies only with the degree of undercooling expressing the athermal
character of the transformation.
It was postulated by Koistinen and Marburger [8] that the evolution of martensite formation in a sample that is initially fully austenitic, can be described by
f = 1 − e−C1 (Ms −T )
(6.1)
where f is the volume fraction of martensite in the sample at a temperature T
below Ms , and C1 is a constant. This fraction is defined as the volume of martensite
divided by the volume of austenite that exists in the sample prior to the formation
of martensite.
Magee [9] showed that this empirical equation (the KM relation) can be derived
from first principles assuming that in a temperature decrease dT , the number dN
of new martensite crystals (plates or laths) that form per unit volume of austenite
0
is proportional to the increase in driving force ∆Gγ→α due to the temperature
decrease
0
0
dN
d(∆Gγ→α )
(6.2)
= −C2
(∆Gγ→α < 0)
dT
dT
with C2 a positive constant expressing the proportionality between the increase in
driving force and the consequent increase in density of activated nucleation sites.
The change in the volume fraction of martensite corresponding to the temperature
decrease dT is then given by
dN
df
= Ω(1 − f )
dT
dT
(6.3)
where (1−f ) is the volume fraction of austenite available for further transformation
and Ω is the average volume per newly formed crystal. Combining the Eqs. (6.2)
and (6.3) yields
0
df
d(∆Gγ→α )
= −Ω(1 − f )C2
dT
dT
0
(6.4)
Assuming that Ω, C2 , and d∆Gγ→α /dT are constant over the extent of the
transformation and integrating from Ms (f = 0) to T gives
94
Chapter 6: Kinetics of the martensitic transformation ...
0
ln(1 − f ) = ΩC2
d(∆Gγ→α )
(Ms − T )
dT
(6.5)
This equation is equivalent to Eq. (6.1) with the positive parameter C1 expressed
by
0
d(∆Gγ→α )
C1 = −ΩC2
dT
(6.6)
Thus ln(1 − f ) is expected to vary linearly with T when the nucleation and
growth of the martensite crystals in a sample obeys the characteristics as proposed
by Magee.
As pointed out earlier, the assumption that Ω is constant is in contradiction
with the Fisher model [10], which assumes that Ω decreases strongly as the transformation progresses. The decrease of the average volume of martensite crystals
can be rationalized by considering that the initial martensite crystals divide the
austenite grain into progressively smaller compartments, which is known as the
geometrical partitioning effect. This model assumes that nucleation occurs homogeneously as a result of random fluctuations. If on the other hand the nucleation
is assumed to be heterogeneous (autocatalysis) it can be shown that the average
volume of martensite crystals is independent of the volume fraction of martensite
[9]. Although the exact nature of the nucleation sites is not completely understood,
calculations of the energy barriers to nucleation indicate that the nucleation must
take place heterogeneously on pre-existing embryos, which is in favour of the Magee
model [9].
6.3
Experimental details
The chemical compositions of the four steels investigated (C50, C60, C70 and C80)
are given in Table 3.2. The AE measurements during the martensitic transformation
in the carbon steel specimens were carried out using the Gleeble 1500 thermomechanical simulator as described in section 3.3. The schematic diagram of the
acoustic emission analyzing system in combination with the Gleeble 1500 thermomechanical simulator is shown in Fig. 3.13. For comparison with the AE data, the
cross-strain of the specimens was also measured simultaneously.
The specimens for acoustic emission and dilatometric measurements were machined from as-received steel rods with a diameter of 12 mm. The experiments
were conducted using specimens with an effective free span (l1 ) of 10 mm and 5
mm diameter (see Fig. 3.14). The distance between the grips (l2 ) is adjustable
and has a minor influence on the cooling rate and thermal gradient in the axial
direction. Due to the large thermal gradient during austenitizing only a part of the
specimen is austenitized; a small part is partially austenitized since its temperature
is in the two-phase region, and a part remains ferrite and pearlite (see Fig. 6.1).
This will have major implications for the interpretation of the measured data in the
Experimental details
T
95
850 ºC
Ac3
Ac1
α+γ
α+γ
α
γ
α
Vm
Figure 6.1: Schematic illustration of the temperature gradient over the sample. The
Ac1 temperature is fixed, whilst the Ac3 temperature depends on carbon content.
present study. Moreover, it should be noted that the austenitized volume (= sample
volume) depends on the austenitizing temperature, which is controlled by the thermocouple in the middle of the specimen: the higher the austenitizing temperature,
the larger the austenitized volume. However, the austenitizing temperature was
the same for all specimens, resulting in approximately equal austenitized volumes
for all specimens. The prior austenite grain size was approximately 20 µm for steel
C50, 50 µm for steel C60, 40 µm for steel C70 and 30 µm for steel C80. The grain
size of the austenite formed during austenitizing is expected to be of the same order
of magnitude for all specimens.
In the Gleeble 1500 thermo-mechanical simulator each specimen was electrically
heated in 20 seconds to the austenitizing temperature Ta = 850 ◦ C, austenitized
for 10 seconds, and continuously cooled to ambient temperature. The cooling rate
was approximately 150 ◦ C/s between 800 ◦ C and 500 ◦ C. During cooling, the root
mean square (rms) voltage of the continuous acoustic emission signals was recorded.
Measurements were repeated several times to check the reproducibility of the experiments; comparison of the measured data revealed no obvious or systematic
differences.
In contrast to the AE parameter used in this work, the rms voltage, Speich and
Fisher [3] and Malygin [7] measured the ringdown counts of the acoustic emission
signals to investigate the progress of martensite formation. This value is the number
of times the transducer output voltage exceeds a pre-set threshold. Since this
parameter depends on instrument settings, it has a non-unique relationship to the
properties of the source and is usually not very suitable for quantitative analysis. On
96
Chapter 6: Kinetics of the martensitic transformation ...
800
C50
C60
C70
C80
T [ºC]
600
400
200
0
0
5
10
15
20
25
t [s]
Figure 6.2: Cooling curves for steel C50, C60, C70 and C80 austenitized at T a =
850 ◦ C.
500
T [ºC]
400
the other hand, the integrated squared voltage of all events (AE waves) is directly
proportional to the source intensity and thus more suitable for quantitative source
analysis [11].
After thermal cycling, each specimen was cut in the axial direction and the
microstructure was analyzed using an optical microscope (Olympus). The microhardness of the different microstructures was measured by means of a Vickers hardness tester (Buehler Ltd.) using a load of 100 g. From the cross-section images the
volume of martensite in the sample after cooling to room temperature (Vm ) was
measured.
C50
C60
C70
C80
300 6.4 Study of steels C50, C60, C70 and C80
200
To investigate the carbon dependence of the kinetics, the steels C50, C60, C70
and C80 were examined under identical conditions. In Fig. 6.2 the cooling curves
of the four steels are plotted, which exhibit inflections at temperatures where the
transformation takes place, due to the release of latent heat.
6.4.1
100
Calculation of the martensite volume fraction
Here the results of steel C60 are used as an example to demonstrate how the volume
fraction of martensite can be directly evaluated from the measured AE power. It
can be understood from Fig. 6.3a, where the mean square voltage U 2 of acoustic
emission is plotted as a function of time, that the martensite formation is reflected
0
5
10
15
t [s]
20
25
Study of steels C50, C60, C70 and C80
97
by a peak in the AE data. When the transformation starts, the AE power rapidly
increases to a maximum value and then tails off to the background noise level (U n2 )
as the transformation proceeds. The transformation is assumed to reach completion
(f → 1) where U 2 becomes equal to Un2 .
It has been shown in chapter 5 that the AE energy generated between the start
and the finish of the transformation is proportional to the volume of martensite
formed with a proportionality factor k. In other words, the AE power U 2 is proportional to the volume transformation rate dV /dt,
dV
(6.7)
dt
with V the volume of martensite in the sample. With the integrated value ∫ U 2 dt,
with the integration running over the entire transformation range, and the value
for the final volume of martensite (Vm = 89 mm3 measured from the cross-section),
the proportionality factor k can be calculated as k = 0.61 × 103 V2 s/m3 . Using
this value of k, the AE power as a function of time can be converted to the volume
transformation rate dV /dt. The change of the martensite volume fraction f is the
normalized volume transformation rate, which in combination with Eq. (6.7) can
be written as
U2 = k
1 dV
U2
df
=
=R
dt
Vm dt
U 2 dt
(6.8)
Thus df /dt ∝ U 2 and the calculated values for the transformation rate are plotted
in Fig. 6.3b. The fraction of martensite f at a certain time can now be evaluated as
the area under the peak up to that specific time, divided by the total area under the
peak (Fig. 6.3c). It should be noted that f (t) is independent of the values found
for k and Vm . The analysis of the data was started from the signal maximum.
The few data points during the rise time of the signal may be attributed to the
thermal gradients in the sample leading to localized transformations as the sample
approaches the martensite-start temperature. This results in a broadening of the
peak; at the signal maximum the thermal gradients are believed to have diminished
due to the release of latent heat in the regions where transformations first began.
Probably, in the ideal case when thermal gradients are negligible the rise time of
the signal is even shorter.
The measured AE data for steel C50, C70 and C80 were analyzed according
to the above procedure and the results for Vm (from the cross-sections) and k are
given in Table 6.1.
6.4.2
Proportionality factors k and dislocation densities ρ
A remarkable correlation was found between the values of k obtained in this work
and values of the dislocation densities in martensite observed in similar carbon
steels by other researchers. This correlation is important since it gives better insight into the origin of acoustic emission during martensitic transformations. The
98
Chapter 6: Kinetics of the martensitic transformation ...
(a)
2
U2 [mV ]
40
30
20
10
0
(b)
1.0
df/dt [s-1]
0.8
0.6
0.4
0.2
0.0
1.0
(c)
0.8
f
0.6
0.4
0.2
0.0
0
2
4
6
8
10
12
14
16
18
t [s]
Figure 6.3: The values of (a) the AE power, (b) the transformation rate and (c) the
volume fraction of martensite plotted against time during the martensitic transformation in steel C60.
Study of steels C50, C60, C70 and C80
99
Table 6.1: Values for the volume of martensite Vm and the proportionality factor k
(Eq. 6.7) for the studied steels.
Steel
C50
C60
C70
C80
Vm
(mm3 )
76
89
90
92
k
(103 V2 s/m3 )
0.28
0.61
0.48
0.41
dislocation densities ρ were measured using TEM micrographs for steels with 0.38,
0.61 and 0.78 wt% C [12]. Both the values of k and the dislocation densities are
plotted in Fig. 6.4 as a function of carbon content. It can be seen that both parameters show the same tendency including an apparent maximum around 0.6 % C.
The increase of the dislocation density as the carbon content increases upto 0.6%
C can be attributed to an increase in slipping; for higher carbon steel more dislocations are involved in the slipping process to accommodate deformation strain
due to the shape change. The decrease of the dislocation density beyond 0.6% C
may be due to twinning partly accounting for the accommodation of transformation
strain; twinning involves relatively less dislocations as compared to slipping. The
correlation between k and ρ indicates that the AE energy generated per unit volume of martensite is proportional to the dislocation density in the martensite. The
dislocations arise due to the shear transformation mechanism, which results in slipping to accommodate the shape deformation [13]. Thus the shear mechanism and
the movement of dislocations are strongly related and the limited number of data
available does not permit an unambiguous distinction between both types of AE
sources. In many previous AE studies on twinned martensites the shear mechanism
itself was regarded as the major source of AE during martensitic transformations
[14, 15].
The exact origin of AE generated during martensite formation is important for
the analysis of the acoustic waves (events). For twinned martensites one AE event
was usually assumed to be related to the formation of one martensite crystal [14, 15].
The validity of this assumption is strongly doubted when the dislocation movements
accompanying the martensitic transformation are the actual source of acoustic emission. Because the dislocation movements keep up with the discontinuous (jerky)
interface motion, the formation of a lath or plate of martensite presumably results
in many AE events. A model describing the generation of acoustic waves during
interface motion based on dislocation dynamics is presented in chapter 7.
6.4.3
Koistinen-Marburger kinetics
Analogous to the analysis of the AE data of steel C60 described in section 6.4.1,
the volume fraction of martensite in steels C50, C70 and C80 is derived from the
100
Chapter 6: Kinetics of the martensitic transformation ...
0.7
-3
2
0.2
1.5
0.1
1.0
k [10 V sm ]
2.0
0.4
0.5
0.6
0.7
-2
0.3
2.5
15
0.4
3
3.0
0.5
Disl. density ρ [10 m ]
3.5
k
ρ
0.6
0.8
Carbon content (wt%)
Figure 6.4: The values of k and the dislocation density ρ [12] as a function of carbon
content. The corresponding tendency of both entities suggests a close relation
between the dislocations and the actual source of acoustic emission during the
martensitic transformation.
measured AE power. For all four steels the values of df /dt, f and −ln(1 − f ) are
plotted against temperature in Fig. 6.5. It can be seen that for steel C80 ln(1 − f )
varies linearly with temperature over the entire extent of the transformation (up
to f = 0.995 at least). In comparison, the plots for steel C50, C60 and C70 are
initially (20 – 30 K below Ms ) also linear. Below a certain temperature, however,
a crossover to another regime is observed. Finally, as the transformation reaches
completion (f = 0.99), the dependence is linear again. The crossover to the nonlinear (intermediate) regime indicates that the transformation kinetics decrease at
relatively high volume fractions (see right-hand axis in Fig. 6.5c). It is argued
that this is due to the increase of carbon content of the transforming austenite. It
is easily verified by the KM relation that the progress of the transformation at a
certain temperature is lower for higher carbon contents, since an increase in carbon
content implies a decrease of Ms . In relation with this it should be remembered that
the sample volumes of C50, C60 and C70 where transformations can take place (i.e.
where the local temperature Ta > Ac1 ), consist of a completely austenitized region
A (Ta > Ac3 ) and a partially austenitized region B (Ac1 < Ta < Ac3 ) (Fig. 6.1).
Region B consists of a ferrite-austenite dual-phase microstructure, in which the
austenite is, according to the phase diagram, enriched in carbon. The martensitic
transformation in that region starts at a lower temperature and probably becomes
predominant beyond the crossover observed in Fig. 6.5c. In the intermediate regime
the measured (total) volume transformation rate dV /dt is a superposition of the
volume transformation rates from different austenite parts in region B, each with
Study of steels C50, C60, C70 and C80
-1
df/dt [s ]
1.0
101
(a)
C50
C60
C70
C80
0.8
0.6
0.4
0.2
0.0
(b)
1.0
f
0.8
0.6
0.4
0.2
0.0
(c)
0.99
4
0.98
3
0.95
2
0.87
1
0.65
0
100
150
200
250
f
-ln(1-f)
5
300
T [ºC]
Figure 6.5: The progress of the martensitic transformation in four carbon steels
as a function of undercooling: (a) the change in volume fraction; (b) the volume
fraction and (c) the volume fraction data plotted on a logarithmic scale. The solid
lines through the data are least-squares fits to the KM equation. The dashed line
indicates the second linear regime for steel C50, C60 and C70.
102
Chapter 6: Kinetics of the martensitic transformation ...
a specific carbon content. This transformation behaviour is consistent with the
differences in carbon content of the formed austenite in region B based on the Fe-C
phase diagram.
After cooling the martensite volume fractions in region A and B are denoted
as fA and fB = (1 − fA ) respectively. To describe the progress of the martensitic
transformation in region A with the KM equation a pre-factor fA is introduced.
Based on the underlying theory, the KM equation can only describe the change in
fraction in a sample with a homogeneous composition and thus a single value for
Ms and for C1 ; it does, in its original form, not account for a sample with a range
of carbon contents. However, region B is relatively small and the KM equation for
the entire sample can be approximated as
f = fA (1 − e−C1 (Ms −T ) ) +
n
X
i=1
´
³
fBi 1 − e−CBi (MBi −T )
(6.9)
for which region B is assumed to consist of n volume elements and fBi represents
the volume fraction of martensite formed in element i of region B. CBi and MBi
represent the rate constant of the KM kinetics and the martensite-start temperature
of volume element i. The sum of all pre-factors fA and fBi equals unity.
Table 6.2: Extracted fitting parameters for the KM equation for the fully austeniγ→α0
from MTDATA.
tized region A (f < fA ), and calculated values for d∆GdT
0
Steel
42CrMo4
C50
C60
C70
C80
fA
(-)
1.00
0.83
0.90
0.95
1.00
Ms
(◦ C)
302
317
282
248
211
C1
(K−1 )
0.051
0.054
0.067
0.055
0.046
d∆Gγ→α
dT
◦
(J/mol C)
7
7.1
7.2
7.0
6.9
ΩC 2
(mol/kJ)
7
7.6
9.3
7.9
6.7
In Fig. 6.5c, the solid lines through the data represent least-squares fits to the
first term of Eq. (6.9), i.e. up to the crossover (0 < f < fA ). The extracted
fit parameters are listed in Table 6.2. The fitting results for steel 42CrMo4 are
also given in this table, and are discussed later (section 6.5). Because steel C80 is
‘pearlitic’, the sample is completely austenitized and fA = 1. Since the first term in
Eq. (6.9) cannot describe all measured data the fitting of the data was executed in
an iterative way by extending the fit up to higher fractions until the next data point
starts to deviate relatively strongly from the least-squares fit through the previous
data points. From the fits it follows that the volume fraction of martensite in region
A is approximately fA = 0.83, fA = 0.90 and fA = 0.95 for steel C50, C60 and C70,
respectively. The increase in fA can be attributed to the decrease of the partially
austenitized region for higher overall carbon contents, since the Ac3 temperature
decreases with increasing carbon content. The fitting curves in Fig. 6.5c show that
Study of steels C50, C60, C70 and C80
Table 6.3: Fitting results for the KM equation for the
MB2 CB2
fB2
MB1 CB1
Steel
fB1
(◦ C) (K−1 )
(◦ C) (K−1 ) (-)
(-)
C50
0.095 247
0.052 0.072 190
0.050
C60
0.060 222
0.060 0.035 170
0.055
C70
0.025 180
0.053 0.010 150
0.050
103
two-phase region B.
MB3 CB3
fB3
(◦ C) (K−1 )
(-)
0.005 110
0.046
0.005 110
0.046
0.005 115
0.046
the fits to the first term of Eq. (6.9) also describe a part of the deviation from
the initial linear regime. It can be easily verified that when the KM equation is
modified with a pre-factor fA the data in a ln(1 − f ) vs. T plot start to deviate
from linearity before fA is reached.
For f > fA the transformation in region B becomes predominant and the measured data result from the simultaneous transformation of a number of austenite
parts in this region as explained above. Consequently, comparison of experimental
data with the KM theory is more complex in this regime. However, at approximately ln(1−f ) = 5 (f = 0.993) a second linear regime can be observed in Fig. 6.5c,
which is small and covers only one per cent of the transformation. Moreover, the
fraction data of steel C60 and C70 coincide and approach the data of steel C80.
Thus the transformation kinetics are approximately the same for all samples as
f → 1, which indicates that when the transformation in the C50, C60 and C70
sample reaches completion, austenite with almost 0.80% C is transformed.
To fit the data of steel C50, C60 and C70 for fA < f < 1 three volume elements
in region B were considered in the second term of Eq. (6.9) (n = 3). The fitting
results for C50, C60 and C70 are shown in Fig. 6.6 and the extracted fit parameters
are given in Table 6.3. A physical interpretation of the MBi is not meaningful
because of the correlation between the fitting parameters and the limited number
of elements (n = 3). This is supported by the fact that the fit near each M Bi
deviates strongly from the data. It is important to notice that beyond f = 0.99 all
fitting curves show a good correspondence with the measured data.
The good agreement between the experimental data and the KM relation up
0
to f = fA confirms that C1 (= ΩC2 d∆Gγ→α /dT ) (Eq. (6.6)) is constant over the
extent of the transformation for the steels studied. This does not necessarily imply
0
that Ω, C2 , and d∆Gγ→α /dT are all constant, because in the derivation of the KM
equation the product of these parameters is assumed constant to solve differential
Eq. (6.4); mathematically no explicit assumptions about each individual parameter
0
are made. Using a thermodynamical database, the value of ∆Gγ→α was calculated
as a function of temperature in the range of 0 – 400 ◦ C for the studied steels. Over
0
the whole temperature range the slope d∆Gγ→α /dT is approximately constant for
each carbon content and the values obtained are given in Table 6.2.
On combining the above results, it follows that the product ΩC2 is constant
over the extent of the transformation (Table 6.2). It should be noted that these
two parameters are not separable; if it would be possible for example to determine
104
Chapter 6: Kinetics of the martensitic transformation ...
(a)
0.99
4
0.98
3
0.95
2
0.87
1
0.65
f
-ln(1-f)
5
0
(b)
0.99
4
0.98
3
0.95
2
0.87
1
0.65
f
-ln(1-f)
5
0
(c)
0.99
4
0.98
3
0.95
2
0.87
1
0.65
0
100
150
200
250
300
f
-ln(1-f)
5
350
T [ºC]
Figure 6.6: The progress of the martensitic transformation in steel C50 (a), C60
(b) and C70 (c) beyond the crossover modelled by 3 volume elements in region B.
Study of steels C50, C60, C70 and C80
0.08
0.09
0.06
C1 [K-1]
0.08
C1 [K-1]
0.10
105
0.04
0.02
0.07
100
140
0.06
180
220
T [ºC]
0.05
0.04
0.03
0.0
0.2
0.4
0.6
0.8
f
Figure 6.7: Plot of C1 against f for steel C80 resulting from the direct evaluation
of Eq. (6.4). The inset shows the values of C1 as a function of temperature. The
horizontal line represents the average value of C1 .
the average volume of martensite crystals by other experiments, information about
the distribution of nucleation sites can be obtained from the value of C2 .
6.4.4
A different analysis of the results for steel C80
Especially the high accuracy of the fit of the C80 data to the KM equation shown
earlier implies that C1 is constant, which strongly indicates that Magee’s assumptions [9] are justified. Here it is shown that the data of C80 can be analyzed
without making Magee’s assumptions of a constant C1 in order to test the differential Eq. (6.4). This equation can be validated directly over the whole extent of
the transformation since the values of df /dT can be calculated as
df
df
=
dT
dt
Ã
dT
dt
!−1
(6.10)
with dT /dt the cooling rate values deduced from the cooling curve (Fig. 6.2) by
numerical differentiation. With the values of (1 − f ), the value of C1 can be evaluated as a function of fraction or temperature (Eq. (6.4) and Eq. (6.6)) as shown in
Fig. 6.7. Despite the considerable scatter in the data it can be seen that C1 does
not change significantly during the transformation. The scatter in the data may
be partly due to the inaccuracies in the calculated values of the cooling rate. This
becomes more pronounced for high volume fractions where df /dt → 0. A constant
value fitted to the data yields C1 = 0.047 K−1 , which is in good agreement with
106
Chapter 6: Kinetics of the martensitic transformation ...
0.2
B
2.2 mm
A
5 mm
Figure 6.8: The cross-section of the steel C60 sample. The volume ratio of region A
and region B is in good agreement with the observed crossover at f = 0.90 shown
in Fig. 6.5c.
0
the previously obtained value C1 = 0.046 K−1 . With d∆Gγ→α /dT being constant
it follows that the product ΩC2 is constant.
If the assumption of a constant Ω and C2 is called into question, from a mathematical point of view Ω ∝ 1/C2 is a valid solution. Assuming now that Ω varies
according to Fisher [10] as Ω = Ω0 (1 − f )10 , C2 would increase progressively with
undercooling which is strongly doubted from a physical point of view. It should be
remembered that C2 describes the distribution of effectiveness of nucleation sites as
a function of undercooling. A constant value of C2 (Magee [9]) means that the number density of activated nucleation sites varies linearly with activation energy. On
the other hand, an increasing C2 with decreasing T implies that the number density of activated nucleation sites increases progressively for temperatures far below
Ms . The constant distribution of effectiveness of nucleation sites is considered to
be more realistic. Thus the argument by Fisher that Ω decreases progressively with
increasing volume fraction [10] seems invalid for the steels studied in the present
work.
6.4.5
Microstructural analysis
After completing the experiments each specimen was cut in the axial direction for
microscopic analysis. The cross-section of steel C60 depicted in Fig. 6.8 clearly
shows a region A, which is fully martensitic, and a region B, which contains a
Study of steels C50, C60, C70 and C80
107
Figure 6.9: Optical micrographs of the microstructure of steel C50 (a), C60 (b),
C70 (c) and C80 (d) showing the parent phase (left), the multi-phase microstructure
formed in the partially austenitized region B (middle part) and the fully martensitic
microstructure formed in the completely austenitized region A (right).
mixed microstructure. It can be seen in this figure that the ratio of the two regions
is approximately 9:1, which is in good agreement with fA = 0.90. In Fig. 6.9
the magnifications of the microstructures of the partially austenitized regions for
all four steels are shown. The partially austenitized regions of steel C50, C60
and C70 show a mixed microstructure of ferrite and martensite. On the left-hand
side the pearlite-ferrite base material of steel C50, C60 and C70 can be seen. For
Ac1 < Ta < Ac3 the ferrite remains untransformed during austenitizing, whereas the
pearlite transforms to austenite with a carbon concentration depending on T a . The
microstructure on the right-hand side corresponds to martensite with a measured
micro-hardness of approximately 600 HV0.1 . The micro-hardness of the ferrite and
pearlite is approximately 250 HV0.1 . The magnification of the partially austenitized
regions show less untransformed ferrite for the higher carbon steels; for C80 there
is no untransformed ferrite.
6.4.6
Martensite-start temperature Ms
As expected, the martensite-start temperature Ms resulting from the fit of the data
to the KM equation decreases with increasing carbon content. The results are
plotted in Fig. 6.10. From the least-squares fit through the Ms data it is deduced
that Ms varies with carbon content xC as
108
Chapter 6: Kinetics of the martensitic transformation ...
320
Ms [ºC]
300
280
260
240
220
200
0.50
0.55
0.60
0.65
0.70
0.75
0.80
Carbon content (wt%)
Figure 6.10: The values of Ms extracted from the fit for steel C50, C60, C70 and
C80. The solid line represents the least-squares fit of the data to Eq. 6.11
Ms = 495 − 355xC
◦
(6.11)
with Ms in C and xC in wt% carbon. As mentioned earlier, the transformation
kinetics of the four carbon steels are approximately the same as the transformation
reaches completion. The last per cent of the transformation in C50, C60 and C70
can be described by the KM equation with Ms = 220 ◦ C and C1 = 0.046 K−1 . This
value was obtained by drawing a line in Fig. 6.5c parallel to the data of C80 and
through the data of C50, C60 and C70 at ln(1 − f ) = 5 (dashed line in Fig. 6.5c).
This line intersects the T -axis at T = 220 ◦ C. Inserting this value for Ms in the
above equation yields a carbon content for the last part of the transformation of
approximately 0.77, which is in good agreement with the expected value from the
Fe-Fe3 C phase diagram.
For comparison with the AE data, the dilatation of the samples was measured
simultaneously. The results in Fig. 6.11 show that the changes in slope are in
reasonable agreement with the martensite-start temperatures derived from the fits
through the AE data. It should be noted that the transformation kinetics can also
be derived from the dilatation curves; in principle this would allow comparison
with the transformation kinetics from the AE measurements. However, with the
cross-strain dilatometer the radial length change in the middle of the specimen is
measured, which may not be representative for the total volume change taking place
during the martensitic transformation. Therefore, an unambiguous comparison of
the transformation kinetics derived from the dilatation signal and the AE signal is
complicated and not pursued in this work.
Study of steels C50, C60, C70 and C80
100
C50
C60
C70
C80
80
∆d [µm]
109
60
40
20
0
-20
0
100
200
300
400
500
600
700
800
T [ºC]
Figure 6.11: Plot of the dilatation against temperature for steel C50, C60, C70
and C80. The arrows indicate the temperature at which the slope changes: the M s
temperature.
6.4.7
Rate constant C1
According the KM equation the progress of the transformation from austenite to
martensite in carbon steels is quantitatively described by the Ms temperature and
the value of C1 . The effect of material and process parameters such as chemical
composition and austenitizing temperature on the transformation can be expressed
in terms of Ms and C1 . The martensite-start temperature Ms is determined by
thermodynamics and the rate constant C1 describes the kinetics, i.e. the progress
of the transformation for a certain undercooling.
The values of C1 obtained from the fits are plotted as a function of carbon
content in Fig. 6.12. It can be seen that C1 has a maximum value for steel C60
and decreases approximately linearly by 30% as the carbon content increases from
0.6 to 0.8% C. This means that the transformation rate for a given degree of undercooling is the highest for steel C60. Now the question can be raised how the
carbon dependence of C1 can be explained in terms of the underlying physical pa0
0
rameters Ω, C2 and d∆Gγ→α /dT . The carbon dependence of d∆Gγ→α /dT is
relatively small compared to the decrease in C1 (see Table 6.2). The values for
0
ΩC2 = C1 /(d∆Gγ→α /dT ) therefore show a trend similar to C1 (see Fig. 6.12).
The martensite formed in the investigated steels is considered to have a similar
morphology. Therefore, it is justified to assume that Ω is approximately the same
for the steels studied. Thus the carbon dependence of C1 is expected to mainly
result from a change of C2 with carbon content.
The dependence of C2 on carbon content can in its turn be explained by the
110
Chapter 6: Kinetics of the martensitic transformation ...
0.07
C1
ΩC2
-1
C1 [K ]
9
0.05
8
0.04
ΩC2 [mol/kJ]
0.06
10
7
0.5
0.6
0.7
0.8
Carbon content (wt%)
Figure 6.12: The carbon dependence of C1 and the product ΩC2 . With Ω assumed
constant, the rate constant C2 appears to correlate with the dislocation density
shown in Fig. 6.4.
change in dislocation density with carbon content shown earlier; both have a maximum at approximately 0.6 % C. When a martensite crystal is formed, dislocations
are created in the neighboring austenite due to the fact that the shear stress accompanying the shape change exceeds the yield strength of the austenite [13]. It
should be noted that the dislocations in the austenite (strain energy) do not increase the driving force for transformation because the dislocations are inherited
in the martensite upon transformation. However, these dislocations have a strong
effect on C2 because the dislocation debris leads to extra nucleation sites apart
from the embryos initially present in the austenite. This is consistent with the
heterogeneous nucleation model, which Magee used to validate that Ω does not
change during transformation [9]. Martensitic crystals tend to occur in clusters;
the probability of nucleation is greater in the vicinity of a previously formed crystal (dislocation debris) than it is for random nucleation in untransformed regions.
This effect is known as strain-induced autocatalysis [13]. Thus for steels with high
dislocation densities (C60) the autocatalytic effect is stronger, which is expressed
by a high value of C2 . A higher C2 means that the number density N of activated
nucleation sites for a given driving force increases and thereby the progress of the
transformation is enhanced.
Study of steels C50, C60, C70 and C80
111
0.06
-1
df/dt [s ]
0.05
(a)
0.04
0.03
0.02
0.01
0.00
1.0
f
0.8
(b)
0.6
0.4
0.2
0.0
5
(c)
0.98
3
0.95
2
0.87
1
0.65
0
200
220
240
260
280
300
f
-ln(1-f)
4
0.99
320
T [ºC]
Figure 6.13: The progress of the martensitic transformation in steel 42CrMo4 as
a function of undercooling: (a) the change in volume fraction; (b) the volume
fraction and (c) the volume fraction data plotted on a logarithmic scale. The solid
line through the data is least-squares fit to the KM equation.
112
Chapter 6: Kinetics of the martensitic transformation ...
(6)
(5)
(1)
(2)
(4)
(3)
Figure 6.14: Schematic drawing of the experimental set-up used for AE measurements during cooling of the CuAl-based shape memory alloy: (1) specimen; (2)
waveguide; (3) transducer; (4) pre-amplifier (60 dB); (5) AE analyzing system; (6)
thermocouple.
6.5
Analysis of the results for steel 42CrMo4
In this section the result of the AE measurement on the steel 42CrMo4 specimen
during continuous cooling in the dilatometer, presented in section 4.2.2, is analyzed
in order to compare the transformation kinetics in this steel with the results for the
carbon steels discussed earlier. It should be remembered that for the dilatometer
set-up as described in section 3.4 (see Fig. 3.15), the thermal gradients in the sample
during austenitizing are negligible compared to the thermal gradients when using
the thermo-mechanical simulator set-up. In fact, no partially austenitized region
occurs in comparison with the case of steels C50, C60 and C70; all the decomposing
austenite has the same carbon content. This is the main advantage of the dilatometer set-up compared to the thermo-mechanical simulator set-up. On the other hand,
the disadvantage of the dilatometer set-up is the relatively low signal level, which is
predominantly caused by the strong attenuation due to the small-sized waveguide.
Moreover, the natural cooling rate of a sample in the dilatometer is low compared
to the typical cooling rates obtained using the Gleeble thermo-mechanical simulator. Consequently, only for steel 42CrMo4 a complete transformation to martensite
is achieved in the dilatometer; under the same conditions steels C50 – C80 do not
transform to martensite completely.
In order to facilitate comparison with the KM kinetics, the measured rms voltage
during martensite formation in steel 42CrMo4, shown in Fig. 4.12b, was analyzed
according the procedure described for steel C60 (see section 6.4.1). The results are
shown in Fig. 6.13. From the least-squares fit of the data to the KM equation it
follows that Ms = 302 ◦ C and C1 = 0.051 K−1 . It is important to notice that the fit
is very accurate over the whole extent of the transformation, like the result for steel
C80. The overall results for steel 42CrMo4 are in line with the results obtained for
steels C50 – C80 discussed earlier, and support the conclusions drawn in that part
of the study.
Study of a shape memory alloy
113
(a)
U2 [mV2]
800
600
400
200
0
T [ºC]
60
(b)
40
20
0
-20
0
10
20
30
40
50
60
t [s]
Figure 6.15: (a) The mean square voltage measured during the martensitic transformation in the CuAl-based shape memory alloy, and (b) the cooling curve of the
sample.
6.6
Study of a shape memory alloy
Since in-situ determination of the size of martensite crystals in steel is not feasible,
additional experiments on a model system were performed in order to obtain further
insight into the evolution of martensite crystals as the transformation progresses.
Optical Confocal Laser Scanning Microscopy (CLSM) observation of the surface
of a shape memory alloy (SMA) was carried out. This CuAl-based SMA has a
martensite-start and -finish temperature of approximately 25 ◦ C and 0 ◦ C respectively. In addition to the optical in-situ observations, acoustic emission experiments
on the SMA were performed to study the kinetics of the transformation.
6.6.1
Acoustic emission experiments
The AE experiments were conducted using a CuAl-based SMA sample with a diameter of 15 mm and a thickness of 1 mm (see Fig. 6.14). A waveguide was spot welded
onto the sample to transport the acoustic waves to the sensor and to protect the
114
Chapter 6: Kinetics of the martensitic transformation ...
sensor from excessive heating and/or cooling. The sample was austenitized (heated
to a temperature of 90 ◦ C) using an air heater. Subsequently the sample was cooled
to −20 ◦ C using a Dewar filled with liquid nitrogen. During cooling, the root mean
square (rms) voltage of the continuous acoustic emission signals was measured. The
AE power and the temperature of the sample are plotted in Fig. 6.15.
Analogous to the analysis of the carbon steels, the values of df /dt, f and
−ln(1−f ) are plotted against temperature in Fig. 6.16. It can be seen that ln(1−f )
varies linearly with temperature over the entire extent of the transformation. In
Fig. 6.16c, the solid line through the data represents a least-squares fit to the KM
equation (Eq. (6.1)). The extracted fit parameters are: Ms = 27 ◦ C and C1 =
0.21 K−1 . In comparison with the results obtained for the carbon steels, the rate
constant C1 is approximately four times larger expressing that the transformation
kinetics are much faster in the CuAl-based SMA. Again, the high accuracy of the
fit implies that C1 is constant over the extent of the transformation, which strongly
indicates that Magee’s assumptions are justified, i.e. the average size of martensite
crystals Ω is constant over the extent of the transformation. On this point the
results obtained for the shape memory alloy are in good agreement with the results
obtained for the carbon steels. Furthermore, optical CLSM observation (and video
recording) of the surface upheavals on the shape memory alloy was carried out
in order to gain a better insight into the evolution of martensite crystals as the
transformation progresses.
6.6.2
Optical Confocal Laser Scanning Microscopy observations
The SMA sample for the CLSM-observations, with a diameter of 15 mm and a
thickness of 10 mm, was placed in a petri-dish and by filling the petri-dish with icewater the martensitic transformation was induced. Prior to the measurements, the
sample surface was ground and polished at a high temperature by using hot water.
Although the grinding scratches could not be removed completely, the polishing
was found adequate to reveal the surface upheavals due to the transformation.
During cooling the confocal microscope acquired 1 image per second, which
revealed interesting information about the growth of martensite plates; six characteristic micrographs are shown in Fig. 6.17. The transformation starts with the
formation of thin plates of martensite with distinct orientations. A thickening of
these plates is observed as the transformation proceeds, and for some plates this
thickening clearly occurs in a discontinuous manner. This thickening effect makes
the interpretation in terms of nucleation and growth complex; it is not clear whether
a thickening plate may be treated to form in a single nucleation and growth step.
Nevertheless, the observations show that small and large plates are formed both at
the beginning and at the end of the transformation. A partitioning effect and a
progressive decrease of Ω as proposed by Fisher [10] is clearly not observed. This
is in good agreement with the results of the AE measurements on the SMA discussed in the previous sub-section. It should be noted that the characteristics of
Study of a shape memory alloy
-1
df/dt [s ]
0.15
115
(a)
0.10
0.05
0.00
1.0
f
0.8
(b)
0.6
0.4
0.2
0.0
0.99
(c)
4
0.98
3
0.95
2
0.87
1
0.65
0
-20
-10
0
10
20
30
f
-ln(1-f)
5
40
T [ºC]
Figure 6.16: The progress of the martensitic transformation in the shape memory
alloy as a function of undercooling: (a) the change in volume fraction; (b) the
volume fraction and (c) the volume fraction data plotted on a logarithmic scale.
The solid line through the data represents the least-squares fit to the KM equation.
116
Chapter 6: Kinetics of the martensitic transformation ...
Figure 6.17: Observation of the change in microstructure during cooling of the SMA
sample over a period of 30 seconds.
the martensitic transformation in this model system may be different from those in
steel to some extent. However, the observed evolution of martensite plates in the
model system is in line with the Magee model, and supports the results of the AE
experiments on the carbon steels.
6.7
Conclusions
It has been demonstrated that acoustic emission is a very suitable technique to
study the kinetics of the martensitic transformation. This is based on the fact that
the volume fraction of martensite f as a function of time t during cooling can be
derived directly from the measured AE power U 2 , since U 2 ∝ df /dt. In general, it
can be concluded that the acoustic emission measurements during the martensitic
transformation in the studied carbon steels (C50, C60, C70, C80 and 42CrMo4) and
the CuAl-based shape memory alloy have resulted in a better understanding of the
nucleation and average volume of martensite crystals. A more detailed summary
of the results obtained is described below.
The analysis of results obtained for the four carbon steels shows that for steel
C80 the fraction data as a function of temperature T can be fitted to the KM
equation with high accuracy up to f = 0.99, whilst for steel C50, C60 and C70
deviations occur beyond f = 0.83, f = 0.90 and f = 0.95 respectively, due to the
partially austenitized region in the samples. The decrease of the transformation
Conclusions
117
kinetics at these relatively high volume fractions is caused by the relatively high
carbon content of the transforming austenite in the partially austenitized region.
The extent of this region decreases with carbon content, which was confirmed by
microstructural analysis. It was shown that a modified KM equation could take
account for the presence of the partially austenitized region.
The rate constants C1 extracted from the fits together with the values for
0
d∆Gγ→α /dT calculated to be constant, imply that ΩC2 is constant over the extent of the transformation. These parameters are not separable; however, from a
physical point of view it is most realistic that both Ω and C2 are constant during
martensite formation. Furthermore, the values of Ms obtained from the fits were
found to be in reasonable agreement with the values extracted from the dilatation
curves, and with literature values.
The observed carbon dependence of the rate constant C1 can be attributed to
the dependence of C2 on the carbon content, which is consistent with the change
in dislocation density with carbon content. The dislocation debris in the residual
austenite leads to extra nucleation sites (autocatalysis) which is expressed by a
higher value of C2 . Also, the acoustic emission energy generated per unit volume
of martensite (k factor) has been found to scale with the change in dislocation
density in the formed martensite as the carbon content is varied. This is considered
to be important for the interpretation of AE measurements during martensitic
transformations.
In contrast to the thermo-mechanical simulator set-up used for the study of
steels C50 – C80, no significant thermal gradients are present during austenitizing
of steel 42CrMo4 in the dilatometer set-up. Consequently, the sample is completely
austenitized (no partially austenitized regions) and therefore the experimental data
could be fitted to the KM equation with high accuracy up to f = 0.99, like steel
C80.
Optical CLSM observation of the surface of a CuAl-based shape memory alloy
gave valuable information about the sizes of martensite crystals formed during
transformation. It was shown that both small and large plates are formed both at
the beginning and at the end of the transformation, which is in good agreement with
the Magee model. Furthermore, AE experiments on the shape memory alloy were
performed and the fraction data obtained could be described by the KM equation
with high accuracy over the whole extent of the transformation, like for the carbon
steels.
The overall results presented in this chapter validate the KM model that describes the kinetics of the martensitic transformation. According to Magee this
means that Ω is constant over the extent of the transformation in the studied materials. It is important to notice that this transformation behaviour is different
from the evolution of martensite crystals in the Fisher model, which predicts that
Ω strongly decreases as the transformation proceeds.
118
Chapter 6: Kinetics of the martensitic transformation ...
References
[1] D.A. Porter and K.E. Easterling, Phase transformations in metals and alloys,
Chapman & Hall, London (1992).
[2] R.W.K. Honeycombe and H.K.D.H. Bhadeshia, Steels, Edward Arnold, London (1995).
[3] G.R. Speich and R.M. Fisher, Acoustic Emission (ASTM STP 505), 140
(1972).
[4] D.G. McMurtrie and C.L. Magee, Metall. Trans. 1, 3185 (1970).
[5] H.C. Shin, S.H. Lee, J.H. Jun and C.D. Choi, Mater. Sci. Technol. 18, 429
(2002).
[6] S.A. Khan and H.K.D.H. Bhadeshia, Mater. Sci. Eng. A 129, 257 (1990).
[7] G.A. Malygin, Phys. Solid State 35, 1470 (1994).
[8] D.P. Koistinen and R.E. Marburger, Acta Metall. 7, 59 (1959).
[9] C.L. Magee, The nucleation of martensite, In Phase Transformations, American Society of Metals, 115 (1970).
[10] J.C. Fisher, J.H. Hollomon and D. Turnbull, AIME Trans 185, 691 (1949).
[11] Y. Berlinsky, M. Rosen, J.A. Simmons and H.N.G. Wadley, Rev. of progress
in quant. nondestr. eval. 5, 1345 (1986).
[12] T. Furuhara, S. Morito and T. Maki, J. de Phys. IV 112, 255 (2003).
[13] H.K.D.H. Bhadeshia, Bainite in steels, The Institute of Materials, London
(2001).
[14] Ll. Manosa, A. Planes, D. Rouby and J.L. Macqueron, Acta Metall. and Mat.
38, 1635 (1990).
[15] Z.Z. Yu and P.C. Clapp, Metall. Trans. A 20, 1601 (1989).
119
120
Chapter 6: Kinetics of the martensitic transformation ...
Chapter 7
Analysis of acoustic emission
signals originating from
bainite and martensite
formation
In this chapter the characteristics of acoustic waves generated during bainite and
martensite formation are studied. The results are discussed in a semi-quantitative
manner, a thorough quantitative analysis of signals is not feasible because of the
limited frequency bandwidth of the system and the unknown sample response. The
frequency spectra of acoustic emission signals are interpreted using a dislocation
source model adopted from acoustic emission studies of plastic deformation [1, 2, 3].
It is assumed that the predominant source of acoustic emission during displacive
transformations is the movement of dislocations, that is the slip taking place during growth in order to relieve internal stresses. The results show that the mean
frequency of waves generated during bainite formation is significantly larger than
that of martensitic waves. This difference in the spectral density of the waves can
be explained by the difference in interface motion of the two transformations and
the consequently different behaviour of the dislocations involved.
Before discussing the generation of acoustic waves during growth of martensite and bainite in section 7.2, the theoretical model concerning acoustic emission
during plastic deformation is reviewed (section 7.1). In section 7.3 the analysis of
continuous acoustic emission is discussed in general terms, and subsequently the
experimental procedure is described in section 7.4. In section 7.5 the measured
frequency spectra during the bainitic and the martensitic transformation in steel
C45 and steel 16MnSi are discussed. This chapter ends with a discussion of the
proportionality factors between the acoustic emission energy and the transformed
volume (k factors) obtained for the steels studied in chapters 4, 5 and 6.
121
122
7.1
Chapter 7: Analysis of acoustic emission signals originating from ...
Acoustic emission during plastic deformation
The observed correlation between the proportionality factor k and the dislocation
density ρ (section 6.4.2) suggests that the dislocation motion (slip) accompanying
displacive transformations is the source of acoustic emission, like in the case of
conventional plastic deformation. During displacive transformations a large amount
of the free energy release is dissipated in the process of slip [4, 5], which occurs when
the interface moves into the austenite matrix. Previous studies of acoustic emission
during plastic deformation have resulted in a model for the generation of acoustic
waves during dislocation motion, which is discussed in this section.
The movement of dislocations is the dominant mechanism of plastic deformation
in most crystalline solids, especially metals. A large amount of the total deformation
energy is released as thermal energy in the specimen, which means that most of
the energy generated by the moving dislocations is dissipated as heat in the crystal
lattice. This thermal energy is the cumulative effect of the generation of phonons by
dislocations as they pass through the crystal lattice. The phonons with relatively
low frequencies can be considered as acoustic emissions, since they can be detected
as such.
The phonon generation process is based on the fact that there are specific positions in the lattice which provide a minimum energy configuration for a dislocation.
As a dislocation moves from one position to the next it must overcome the Peierls
barrier before moving. As the dislocation moves away from a minimum-energy position the elastic lattice strain is increased until the ’over-barrier’ position is reached.
Then the increment of elastic strain is suddenly released to produce a vibrational
wave in the lattice. This process is considered as the basic mechanism by which
dislocations can cause acoustic emissions.
To explain the phonon generation, Kiesewetter and Schiller presented the analogy with electrons, which emit photons (Bremsstrahlung) when they start to move
or stop [1]. They argued that dislocations behave similarly and can be considered as
a source of AE (acoustic Bremsstrahlung) when they are accelerated or decelerated.
In the past several attempts were undertaken to derive quantitative information
about the acoustic waves generated by moving dislocations. Scruby et al. used the
Green’s function to obtain the surface displacement resulting from the expansion of
dislocation loops [2]. In another approach Kiesewetter and Schiller proposed that
the acoustic waves are produced at active Frank-Read sources where dislocations
are generated and move outwards [1]. In this work [1] the moving dislocations are
considered to radiate acoustic waves in those instances where they are accelerated.
The main difference between the above models is that the former approach takes
into account the directionality of the source, i.e. the orientation of the moving dislocation with respect to the surface upon which the transducer is mounted. However,
for quantifying the wave energy from dislocation motion this is not required. Actually, the two approaches lead to essentially the same expression of the AE energy
radiated per second (Ė), which is demonstrated below.
Scruby et al. [2] showed that the maximum displacement amplitude at the
Acoustic emission during plastic deformation
123
surface u caused by the growth of a dislocation loop at a constant velocity v is
proportional to the final radius a of the loop, or
u=
c22
bav
Dc31
(7.1)
with b the Burgers vector, D the distance from the loop to the surface at which the
measurement takes place, and c1 and c2 the longitudinal and the transverse wave
velocities respectively. The power Ė radiated by one dislocation can be assumed to
be proportional to u2 [6]. Then for N dislocations moving at a velocity v, Ė can
be written as
c4
Ė = αN 22 6 b2 a2 v 2
(7.2)
D c1
with α the proportionality constant between Ė and u2 .
Kiesewetter and Schiller [1] showed that, according to Eshelby’s theory, the
power generated by N screw dislocations of length l, which oscillate harmonically
with frequency ω0 , is
πGω0 2 2 2 2
Ė = N
b l v
(7.3)
5c31
with G the shear modulus and c1 the velocity of the longitudinal waves. Since the
length l can be viewed as proportional to the radius a of a dislocation loop, both
formalisms give the same dependence on the dislocation characteristics N , b, l and
v. It should be noted that the measured AE power is proportional to the AE energy
radiated per second: U 2 ∝ Ė.
For plastic deformation the strain rate ²̇ can be written, according to Orowan’s
equation, as
²̇ = ρm bv
(7.4)
where ρm is the mobile dislocation density. A change in ²̇ can in principle be due
to a change in ρm or a change in v [7]. Many experimental results [1, 2, 8] of AE
during tensile testing showed that the mean square voltage (the AE power) of the
continuous acoustic emission is proportional to the strain rate, which is expressed
by
U 2 ∝ ²̇
(7.5)
A typical plot of U 2 against ² during tensile testing is shown in Fig. 7.1. It can be
seen that the peak signal increases with strain rate. Fig. 7.2 shows that AE power
U 2 at a given strain is proportional to the strain rate. This linear dependence
indicates that the change in strain rate is primarily due to a change in the mobile
dislocation density, because Eqs. (7.2) and (7.3) show a linear dependence on N .
Therefore, v can be assumed constant. If a change in ²̇ would be due to a change
in v, a quadratic dependence would have been found, U 2 ∝ ²̇2 , see Eqs. (7.2) to
(7.4).
On this point it should be noticed that above relation (Eq. (7.5)) is very similar
to the relation between the AE power and the volume transformation rate during
displacive transformations, given by
124
Chapter 7: Analysis of acoustic emission signals originating from ...
Figure 7.1: The solid lines represent the acoustic emission power as a function of
strain for strain rates of 1, 5 and 20 mm/min; the gauge length LG of the specimens
was 30 mm (From Ref. [8]). It can be seen that the intensity of the signal increases
with increasing strain rate. The dashed line is the stress-strain curve.
Figure 7.2: The acoustic emission power measured at a given strain during tensile
testing of polycrystalline aluminium with different strain rates (From Ref. [2]. The
measured values are normalized to the AE power at a strain rate of 0.5 mm/min.
The straight line through the data indicates that the AE power is proportional to
the strain rate.
Dislocation dynamics during displacive transformations
U2 = k
dV
dt
125
(7.6)
which was derived from basic principles and validated for different steels in chapter
5. If it is assumed that the strain rate that occurs during displacive transformations
is proportional to the volume transformation rate, Eqs. (7.5) and (7.6) are in fact
identical.
The similarities between plastic deformation and displacive transformations concerning acoustic emission discussed above, support the argument made earlier that
it is actually the movement of dislocations during displacive transformations that
constitutes the generation of acoustic waves. How the motion of dislocations takes
place during displacive transformations is explained in the next section.
7.2
7.2.1
Dislocation dynamics during displacive transformations
Nucleation and growth of martensite
As discussed in chapter 6, the nucleation of martensite takes place heterogeneously
on defects in the parent phase, which nuclei are stimulated to grow at different
degrees of undercooling [4]. Since the growth is relatively fast, the overall rate of
transformation is governed by nucleation. Nevertheless it is the process of growth
that is relevant for AE waves being radiated, and therefore martensite growth is
discussed in more detail in this section.
A nucleus is usually visualized as an embryo of the new phase (martensite),
which has a semi-coherent interface with the parent phase (austenite). The most
likely sites for such nuclei are the interfaces of inclusion particles and grain boundaries [4]. The growth of the new phase is constituted by the movement of the glissile
interface, which consists of arrays of parallel dislocations [4, 5]. As the interface
moves into the austenite matrix the dislocations glide on appropriate slip planes.
After the nucleation barrier has been overcome, the interface motion is usually
assumed to be discontinuous until the growth of a plate or lath is completed. In order for the interface to move, the existing interfacial dislocations have to overcome
possible obstacles (crystal defects) and new dislocation loops have to be generated,
which requires an activation energy [9]. Both types of barriers, which are smaller
than the nucleation barrier, are the cause of the jerky motion of the interface [9]
leading to the thickening of plates (or laths) [5].
Assume that at a certain position, as illustrated in Fig. 7.3, a nucleus is formed
and has grown to a certain size at time t0 . The overall rate of the discontinuous
interface motion is governed by the time it takes the interface to overcome a barrier.
After the interface has overcome the barrier, the interface moves until another
barrier is encountered where the interface comes to a halt again. This is illustrated
in Fig. 7.3 for two jerky movements (events), with τ the mean life time of such
126
Chapter 7: Analysis of acoustic emission signals originating from ...
(a)
(b)
(c)
(d)
l
t0
t0 + τ
Figure 7.3: Schematic drawing of the discontinuous interface motion during the
growth of a nucleus (a) to a final martensite crystal (d). The distance l in (b)
represents the displacement of the interface during a jerky movement with mean
life time τ .
an event and l the mean displacement of the interface in this time interval. This
displacement corresponds to a mean free path of dislocations lying in the interface,
and can be written as
l = τv
(7.7)
with v the mean interface velocity during movement between two barriers. During
such an event (displacement), dislocation motion takes place similar to the case of
conventional plastic deformation as described in section 7.1, and this can be regarded as the origin of acoustic wave generation during martensitic transformation.
Since the martensitic transformation takes place by cooperative atomic movement,
the growth of a martensite crystal across grain boundaries cannot occur. Consequently, after several events the interface comes to a rest and a plate or lath of
martensite is formed (Fig. 7.3d).
From the frequency analysis of acoustic waves information can be obtained about
the time scales involved in the underlying process. The frequency fs of an acoustic
wave is inversely proportional to the time τ that the source operates [10], which is
expressed by
1
(7.8)
2τ
By inserting Eq. (7.7) into Eq. (7.8) the mean frequency of acoustic waves at the
source f s becomes
fs =
v
(7.9)
2l
On this point it should be remembered that the mean frequency of acoustic emission
measured at the transducer f may be significantly different from f s due to the
fs =
Analysis of continuous acoustic emission
127
sample response and the limited bandwidth of the transducer (see section 2.2.1).
The dislocation source model employed in this work for the interpretation of
acoustic emissions during displacive transformations is different from the source
models found in the literature on this subject. In many previous studies the volumetric and shape change [11, 12], and the shear mechanism of the transition
[13, 14, 15] were considered as the most important source candidates; in these models one AE event was usually related to the formation of one martensite crystal.
Although the amount of dislocation motion is presumably strongly related to the
volume change and the shear mechanism, the time and length scales at the source
corresponding to an AE event are significantly different in view of the dislocation
source model. Since the dislocation motion is related to the jerky interface movement, the formation of a single lath or plate of martensite can involve many (small)
AE events. Therefore, the dislocation source model is believed to provide a better framework for a suitable analysis of the acoustic waves, which carry dynamical
information of the source mechanism.
7.2.2
Nucleation and growth of bainite
In contrast to martensite, bainite grows at relatively small driving forces, and therefore nucleation events are almost always confined to the austenite grain boundaries, which contain the most potent nucleation sites (smallest activation energy).
Whereas the martensite nucleation is diffusionless, the nucleation stage of bainite
probably involves partitioning of carbon, which leads to a greater reduction in free
energy [16]. The growth of bainite is diffusionless and the crystallographic features
(slip) and surface relief effects are identical to those associated with martensite formation. After transformation most of the excess carbon in the bainitic ferrite will
partition into the residual austenite [17].
Compared with martensite, the overall interface motion (growth rate) is relatively slow, presumably because of the plastic work that is done as bainite grows
[17]. Concerning the discontinuous interface motion, the interface mobility and the
mean displacement of the interface, no clear differences between martensite and
bainite have been reported in the literature.
7.3
Analysis of continuous acoustic emission
Plastic deformation and displacive transformations are usually classified as sources
of continuous AE. For continuous emission the recorded wave is a superposition
of a number of individual waves, with frequencies depending on the characteristic
times of the individual sources. Due to the superimposed character of the wave,
the extracted rise time and amplitude of the combined wave, which are in principle
meaningful AE parameters, cannot unambiguously be related to source characteristics. Moreover, rise time and amplitude depend on the threshold setting and
the wave-recording settings, such as peak definition time and hit definition time
128
Chapter 7: Analysis of acoustic emission signals originating from ...
(see section 3.1). Therefore, results obtained in different set-ups cannot easily be
compared, which makes these AE parameters less useful.
Since the actual wave can be written as the summation of the original waves due
to the individual events, the frequency information of each individual AE source
is contained in the observed signal, regardless of the strong overlapping. It is important to realize that the frequency bandwidth of the system is relatively small
(see section 2.2.1) and usually does not correspond with that of the AE source.
This together with the unknown sample response implies that no quantitative information about time scales of the source processes can be obtained. However, by
using the same set-up a qualitative (comparative) study of different processes can
be undertaken, as is done in this work.
In the literature, no research has been reported about frequency analysis of
acoustic emission during displacive phase transformations. On the other hand,
elaborate studies of the acoustic waves generated during plastic deformation have
been carried out. In the early 1980s, Rouby et al. presented a thorough analysis of
the acoustic waves generated by dislocations moving between two steady positions
[3]. They showed that by using the elastic Green’s function an expression for the
elastic waves could be obtained, in the time and frequency domain. Some years
later, Schaarwachter and Ebener performed AE measurements on deforming copper
using two resonant transducers [18]. They showed that the AE power spectral
density shifts to higher frequencies with increasing strain, which could be well
explained by changes in the event lifetime in the course of work hardening. Recently,
Vinogradov et al. [19] investigated the acoustic emission spectral density during
cyclic deformation of copper single crystals. They observed an increase in the mean
frequency of the AE power spectral density and attributed this to the decrease in
the dislocation mean free path.
7.4
Experimental details
The experiments were performed using steel C45 (plate thickness of 5 mm) and
steel 16MnSi (plate thickness of 3 mm). The chemical compositions of the steels are
given in Table 3.2. Spot welds were made as described in section 3.2. Monitoring
the transformations in a spot weld has the advantage that no waveguides need
to be used, and therefore attenuation and mode-conversion effects are minimized.
Reflections of waves may play a role but they are presumably damped sufficiently
before they arrive at the transducer (see section 3.2). Since the geometry of the
plate remains unchanged during each experiment, the effects of reflections and
interference are constant, and therefore the variations in the power spectral density
should be attributed to changes related to the forming microstructure. In other
words, since the effects of sample response are identical for bainite and martensite
formation, it is justified to compare the wave characteristics in terms of differences
between the two microstructural processes.
In chapter 4 it was shown that in the case where spot welds are produced on
Results and discussion
129
steel C45 with high heat input, bainite and martensite are formed during cooling,
see Fig. 4.9. Preliminary experiments showed that spot welding of steel 16MnSi
also resulted in bainite and martensite formation upon cooling.
After extinction of the arc, several AE parameters were measured simultaneously
during cooling of the sample and recorded for further analysis. First, the rms voltage
Urms of the continuous acoustic emission was measured. Second, the waves (AE
events) constituting the total AE power were recorded. As discussed earlier, the
continuous AE signal with average amplitude Urms can be regarded as a resultant
of individual events. Since the frequency of the waves is analyzed, it is not required
to isolate each AE event, which would be practically impossible. For statistical
averaging on the other hand it is important to record many AE events. To measure
the events with high amplitudes a floating threshold was set with a minimum value
of 32 dB; the floating threshold automatically adjusts the detection threshold in
order to obtain an acquisition rate of approximately 300 events/second. By using
such a threshold only the highest amplitudes are measured, which are generated
during bainite and martensite formation.
For each wave U (t) its power spectral density P (f ) (with f the frequency obtained by Fast Fourier Transformation, FFT) was calculated. The distribution
P (f ) can be most simply represented by the mean frequency f , which is related to
P (f ) by
R
P (f )f df
f= R
P (f )df
(7.10)
with the integration running over the bandwidth of the system. The mean frequency
data were averaged in each time interval of 1 second, which means averaging over
approximately 300 measurements.
7.5
Results and discussion
According to the procedure described above, Urms and f were measured as a function of time during cooling of steel C45 and steel 16MnSi. The result for steel C45
is shown in Fig. 7.4. It can be seen that the mean frequency of waves generated
during bainite formation f b is approximately 460 kHz, whilst the mean frequency
of waves generated during martensite formation f m is approximately 380 kHz. Although the absolute difference is not very large, in view of the bandwidth (100 –
1000 kHz) and the accuracy, this difference in f is significant.
According to Eq. (7.8) the difference in f implies that the mean life time of
events during bainite formation, τb , is shorter. This in its turn can be attributed
to a smaller interface mean free path l b under the assumption that the interface
mobility (or velocity v) is not significantly different for bainite and martensite
formation. The limited number of data available does not permit an unambiguous
explanation for the relatively small mean free path of interfacial dislocations in
130
Chapter 7: Analysis of acoustic emission signals originating from ...
f
500
10
Urms
8
B
300
6
M
4
200
W
100
0
0
Urms [mV]
f [kHz]
400
2
2
4
6
8
10 12 14 16 18 20 22 24
0
t [s]
Figure 7.4: The mean frequency f and the average voltage Urms of waves generated
during bainite and martensite formation in steel C45 after spot welding (W). The
rms voltage shows two peaks: B = bainite formation, M = martensite formation.
The standard deviation in the value of f is expressed by the error bars.
6
600
f
Urms
500
4
300
M
3
2
200
W
B
100
0
2
4
6
Urms [mV]
f [kHz]
400
0
5
1
0
8 10 12 14 16 18 20 22 24 26 28 30
t [s]
Figure 7.5: The mean frequency f and the average voltage Urms of waves generated during bainite and martensite formation in steel 16MnSi after spot welding
(W). The rms voltage shows two peaks: B = bainite formation, M = martensite
formation. The standard deviation in the value of f is expressed by the error bars.
Results and discussion
131
case of bainite, but it could be related to the self-induced strain during bainite
formation.
In Fig. 7.5 the AE parameters measured during bainite and martensite formation in steel 16MnSi are plotted. It can be seen that the results obtained are similar
to the results obtained for steel C45. During the bainitic and martensitic transformation in this steel the values obtained for f b and f m are approximately 480
kHz and 400 kHz respectively. Compared with the result for steel C45, the values
of both f b and f m for steel 16MnSi are approximately 20 kHz higher. This small
difference may be caused by the difference in chemical composition of the steels, or
the difference in plate thickness (5 mm for C45 versus 3 mm for 16MnSi).
Sample response and transducer bandwidth
As mentioned earlier, no quantitative information about time and length scales of
the source can be obtained due to the unknown sample response and the limited
transducer bandwidth. This is explained in more detail below. Concerning the
source, the interface motion during the transformation, it is believed that the mean
interface velocity and the mean interface displacement have a value in the order
of magnitude of v = 100 m/s and l = 100 nm respectively. Inserting these values
into Eq. (7.9) yields a mean frequency of waves at the source of f s = 500 MHz. It
should be noted that this is approximately three orders of magnitude larger than
the measured values of f at the transducer output. This simple calculation confirms
that no quantitative information about the source mechanism can be derived from
the measured f . The observation that surface displacements with much lower
frequencies are measured at the position of the transducer can be explained by
the fact that the source function S(t) is convoluted with the material response
function M (see section 2.2.1).
Consider a frequency spectrum of the source S(f ) with a mean frequency of
500 MHz as shown in Fig. 7.6. In order to evaluate the material response M on
the source function S, the theoretical work of Scruby et al. is adopted [20]. They
calculated the frequency spectra of surface displacements generated by force dipoles
with life times in the range of 30 ns to 1 µs (f s = 0.5 – 15 MHz). The result of their
calculations is shown in Fig. 7.7. It can be seen that the width of the spectrum
M S(f ) is inversely proportional to the life time. For example, in case of a life
time of 30 ns the spectrum extends up to approximately 30 MHz. Furthermore, it
can be seen that the low frequency components in the spectrum have a relatively
high intensity. Based on this result, the material response on the source function
considered in this work was evaluated. The response on a source with a mean life
time of 1 ns, M S(f ), is schematically shown in Fig. 7.6. In the inset of Fig. 7.6
the low frequency spectrum of the surface displacements M S(f ) is shown together
with the transducer bandwidth T (100 – 1000 kHz) taken from Fig 3.4.
In summary, the measured values of f cannot be interpreted quantitatively,
however, the result f b > f m implies that the mean life time of events at the
bainitic source is smaller than the mean life time of events at the martensitic source.
132
Chapter 7: Analysis of acoustic emission signals originating from ...
power
power
MS( f )
T
0.0
0.2
0.4
0.6
0.8
frequency [MHz]
1.0
S( f )
MS( f )
0
200
400
600
800
1000
frequency [MHz]
Figure 7.6: Schematic drawing of the source spectrum S(f ), the spectrum of the
surface displacements M S(f ) resulting from the material response on the source
function, and the frequency response of the transducer T .
Since the material response M is constant, a source spectrum with a higher mean
frequency results in a frequency spectrum of the surface displacements with stronger
high frequency components, and thus a higher measured mean frequency.
Noise analysis
Regarding the electronic noise contributing to Urms (0.28 mV) it should be mentioned that the noise waves have highest amplitudes of approximately 20 dB (= 1
mV) and a mean frequency f n of approximately 500 kHz. They may give a small
contribution to each measured wave, and thus the calculated mean frequency may
be affected by that of the noise to some extent; the noise signals are continuously
present on the background. However, by setting the minimum threshold at 32 dB it
is assumed that the strength of the bainitic and martensitic signals is predominant.
When using a floating threshold with a minimum of 32 dB the measured waves
from which f is calculated have amplitudes that are at least four times larger than
the amplitudes of the noise signals.
In Fig. 7.8 the amplitudes of the waves measured during the bainitic and martensitic transformation in steel C45 are plotted. Although the rms voltage is relatively
low during bainite formation, the maximum amplitudes are of the same order of
magnitude as in the case of martensite formation. In order to understand this,
it should be realized that the rms voltage, being a measure of continuous emission, is more strongly dependent on the rate of occurrence of waves than on the
Results and discussion
133
Figure 7.7: Fourier transforms of the surface displacements M S(f ) generated by
force dipoles with a life time of (a) 30 ns, (b) 100 ns, (c) 300 ns and (d) 1000 ns
(from Ref. [20]). The observed change in the frequency spectrum from (a) to (d)
means that the mean frequency in the detection range f (100 – 1000 kHz) decreases
for (a) to (d).
134
Chapter 7: Analysis of acoustic emission signals originating from ...
60
ampl.
Urms
B
20
6
M
4
0
Urms [mV]
amplitude [dB]
40
8
2
W
-20
0
2
4
6
8
10 12 14 16 18 20 22 24
0
t [s]
Figure 7.8: Plot of the amplitudes of waves generated during bainite and martensite
formation in steel C45 after spot welding. W = welding, B = bainite formation, M
= martensite formation.
520
100 % noise
Bainite
Martensite
f [kHz]
500
N
480
B
460
440
420
M
400
380
0
20
40
ηn
60
80
100
Figure 7.9: Mean frequency data of steel C45 from Fig. 7.4 plotted against the percentage of noise contributing to the signals measured during (B) bainite formation
and (M) martensite formation; (N) represents the mean frequency for the background noise. The solid line indicates that for martensite formation f m is linearly
dependent on ηn .
Discussion of proportionality factors k
135
amplitudes of the waves. Furthermore, it should be remembered that with the
floating threshold setting only the waves with the highest amplitudes are measured
as shown in Fig. 7.8; the waves with amplitudes lower than the threshold, however,
also contribute to the measured Urms . During martensite formation the overlapping
of waves is presumably very strong. The amplitudes of the signals measured during
the bainitic and martensitic transformation in steel 16MnSi were in the same range
as for steel C45 shown in Fig. 7.8.
Since the amplitudes of bainitic and martensitic waves are of the same order of
magnitude, the possible influence of noise would have a similar contribution to both
signals. It is therefore unlikely that the noise can cause the significant difference
between f b and f m . Moreover, increasing and decreasing of the threshold did not
result in significant different values for f b and f m .
In both the results of steel C45 and steel 16MnSi a slight increase in f m is
observed when the martensite signal tails off to the background noise. In order
to investigate whether this is due to noise or related to a change in the growth
dynamics as the martensitic transformation reaches completion, a systematic noise
analysis was carried out using the results obtained for steel C45.
In order to quantify the noise contribution to each measured wave, it is assumed
that the percentage of noise in each wave ηn is equal to the typical amplitude of
the noise waves (20 dB) divided by the amplitude of the total wave. Based on this
assumption ηn was calculated for the bainitic and the martensitic waves using the
amplitude values from Fig. 7.8. For example, ηn corresponding to the maximum of
the martensite peak (46 dB) is 5 %, whilst ηn for the tail of the martensite peak (34
dB) is 20 %. The corresponding values of f m (see Fig. 7.4) are approximately 380
kHz and 400 kHz. To verify if this increase in f m can be attributed to an increase
in ηn , f is plotted against ηn in Fig. 7.9, together with the result for background
noise: ηn = 100 %, f n = 500 kHz. The straight line that can be drawn through the
data points for martensite formation and the data point for 100 % noise indicates
that the increase in f m at the tail of the martensite peak is due to a larger noise
contribution.
The result for bainite formation, f b as a function of ηn , is also plotted in Fig. 7.9.
It can be seen that these data points deviate strongly from the straight line through
the martensite data points discussed above. This supports the argument made
earlier that the difference between f m and f b cannot be attributed to the influence
of noise.
7.6
Discussion of proportionality factors k
In section 6.4.2 it was reported that a clear correlation exists between the AE
energy per volume unit of forming martensite (km ) and values of the dislocation
densities (ρ) in martensite observed by other researchers [21]. This indicates that
an interpretation of k factors based on dislocation densities is most suitable. For
completeness, the plot of k and ρ as a function of carbon content from section
136
Chapter 7: Analysis of acoustic emission signals originating from ...
0.8
-3
3
1.0
0.4
6
0.2
0.0
0
0.8
1
3
2
-3
km [10 V sm ]
7
2
1.2
8
0.6
0.6
2
3
15
-2
ρ [10 m ]
km
ρ
4
3
-2
2
0.2
0.0
5
15
0.4
4
Disl. density ρ [10 m ]
km [10 V sm ]
1.4
1
0.4
0.6
0.8
0
Carbon content (wt%)
Figure 7.10: The values of the proportionality factor km and the dislocation density
ρ [21] as a function of carbon content. In the inset km is plotted against ρ.
6.4.2 is shown again in Fig. 7.10, including the values of km for steel C45 and steel
42CrMo4.
All values of km and kb , obtained for martensite and bainite respectively, are
listed in Table 7.1. The km factors for steel C45 and steel 42CrMo4 are evaluated
for both welding and thermo-mechanical simulator experiments. It can be seen
in Table 7.1 that the km factors obtained in the case of the thermo-mechanical
simulator experiments are one order of magnitude smaller, which is primarily due
to the attenuation caused by the waveguide. This suggests that the k m factors of
steels C50, C60, C70 and C80 (thermo-mechanical simulator experiments) should
be multiplied by approximately a factor 10 for a proper comparison with the k m
factors of steels C45, 42CrMo4 and 42MnV7 (welding experiments).
For steel C45 it was found that km > kb , and this is in line with experimental
data from electron microscopy showing that the dislocation density in martensite is
higher than in bainite of similar composition [22]. It can be seen in Table 7.1 that
km of steel C45 is approximately a factor two larger than km of both steel 42CrMo4
and steel 42MnV7.
In the inset of Fig. 7.10 the values of k are plotted against the values of ρ.
It should be remembered that both entities are not obtained for the same carbon
Conclusions
137
steels; the carbon contents of the steels studied in this work are slightly different
from the carbon contents of the steels studied in Ref. [21]. Although only a few
values of ρ are found in the literature to compare with, it appears from the inset that
k is proportional to ρ, indicated by the solid straight line. Based on this and the
above discussed results of the k factors for different steels and different experimental
techniques, it is proposed that the proportionality factor k of a certain steel can be
written as
k = ρW
(7.11)
with W the total transfer function, which is to a large extent affected by the waveguide (if used).
Table 7.1: Values of the proportionality factors km and kb for the studied steels.
kb
km
Steel
km
(103 V2 s/m3 )
(103 V2 s/m3 )
(103 V2 s/m3 )
(Welding)
(Gleeble)
(Welding)
C45
2.38
0.23
1.2
42CrMo4
0.85
0.11
42MnV7
1.15
C50
0.28
C60
0.61
C70
0.48
C80
0.41
7.7
Conclusions
In this chapter the measured frequency spectra of acoustic waves generated during
bainite and martensite formation are presented and discussed. It is explained that
a quantitative analysis of signals is not feasible mainly because of the unknown
sample response. However, the specimen configuration is the same during bainite
and martensite formation, and therefore the results can be interpreted in a semiquantitative manner. The frequency spectra are interpreted using a dislocation
source model adopted from acoustic emission studies of plastic deformation. This
model is utilized because the movement of dislocations is considered to be the
actual source of acoustic emission during displacive transformations; the dislocation
motion is related to the jerky interface movement and therefore the formation of a
single unit of martensite or bainite can involve many small AE events.
The results for both steel C45 and steel 16MnSi show that f b is significantly
larger than f m . This difference is attributed to the characteristics of (jerky) interface movement of the two transformations: the mean life time of events τ , the mean
velocity v and the mean free path l. The difference in f implies that τb < τm . By
assuming that v is not significantly different for bainite and martensite formation
this difference in τ means that l b < lm .
138
Chapter 7: Analysis of acoustic emission signals originating from ...
Through a systematic analysis of noise signals it is shown that noise signals
have only a small influence on the measured frequency spectra when the floating
threshold is at least 10 dB above the highest amplitude values of noise signals. The
background noise signals give a small contribution to measured signals and have a
relatively high mean frequency. They can explain the slight increase in the mean
frequency observed at the end of the martensitic transformation, however, the large
difference between the mean frequency of bainitic and martensitic waves cannot be
attributed to the noise signals.
Finally, all the proportionality factors between the acoustic emission energy and
the transformed volume k obtained for the steels studied in this thesis are compared
and interpreted based on dislocation densities ρ. The analysis of values of k and
ρ indicates that k is proportional to ρ. This means that the measured AE power
during displacive transformations is linear dependent on the dislocation density,
which is identical to the dependence of U 2 on ρ found in cases of acoustic emission
measurements during tensile testing.
References
[1] N. Kiesewetter, P. Schiller, Phys. Stat. Sol. A 38, 569 (1976).
[2] C. Scruby, H. Wadley and J.E. Sinclair, Phil. Mag A 44, 249 (1981).
[3] D. Rouby, P. Fleischmann and C. Duvergier, Phil. Mag. A 47, 671 (1983).
[4] R.W.K. Honeycombe and H.K.D.H. Bhadeshia, Steels, Edward Arnold, London (1995).
[5] D.A. Porter and K.E. Easterling, Phase transformations in metals and alloys,
Chapman & Hall, London (1992).
[6] S. Hao, S. Ramalingam and B.E. Klamecki, J. of Mat. Process. Techn. 101,
124 (2000).
[7] J. Chung and E. Kannatey-Asibu, Jr, J. Appl. Phys. 72, 1812 (1992).
[8] H. Hatano, J. Appl. Phys. 47, 3873 (1975).
[9] M. Grujicic and G.B. Olson, Int. Sc. 6, 155 (1998).
[10] B. Raj, B.B. Jha and P. Rodrguez, Acta Metall. 37, 2211 (1989).
[11] J.A. Simmons and H.N.G. Wadley, J. of Research of the NBS 89, 55 (1984).
[12] H.N.G. Wadley and R. Mehrabian, Mat. Sc. Eng. 65, 245 (1984).
[13] Ll. Manosa, A. Planes, D. Rouby and J.L. Macqueron, Acta Metall. and Mat.
38, 1635 (1990).
[14] Z. Yu and P.C. Clapp, J. Appl. Phys. 62, 2212 (1987).
[15] Z.Z. Yu and P.C. Clapp, Metall. Trans. A 20, 1601 (1989).
[16] G.B. Olson, H.K.D.H. Bhadeshia and M. Cohen, Acta Metall. 37, 381 (1989).
[17] H.K.D.H. Bhadeshia, Bainite in steels, The Institute of Materials, London
(2001).
139
140
Chapter 7: Analysis of acoustic emission signals originating from ...
[18] W. Schaarwachter and H. Ebener, Acta Metall. Mat. 38, 195 (1989).
[19] A. Vinogradov, V. Patlan and S. Hashimoto, Phil. Mag. A 81, 1427 (2001).
[20] C. Scruby, H. Wadley and J.J. Hill, J. Phys. D: Appl. Phys. 16, 1069 (1983).
[21] T. Furuhara, S. Morito and T. Maki, J. de Phys. IV 112, 255 (2003).
[22] R.W.K. Honeycombe, Steels, Edward Arnold, London (1991).
Summary
Steel is one of the most commonly used materials today, especially in industrial
sectors such as ship building and the automotive industry. In order to meet the
requirements for steel applications, new multi-phase steels are being developed. The
microstructure of these steels consists of a variety of different phases, which leads to
superior material properties - a combination of high strength with good formability.
For the development of such steels research is required to gain more insight into
the underlying microstructure and the mechanisms by which it is formed.
This thesis describes unique acoustic emission experiments during martensitic
and bainitic transformations in steel. The main objective of this work is to obtain
a better understanding of the growth mechanism and kinetics of these solid-state
phase transformations that can occur in carbon steel. In view of fact that acoustic
emission is an unexplored technique in this kind of steel research, this study also
aims to give a good overview of the possibilities and limitations of acoustic emission
as a real-time monitoring technique for the evolution of bainite and martensite
formation.
A concise introduction to acoustic emission (AE) and phase transformations in
steel is provided in chapter 1. Subsequently, in chapter 2 a elaborate description
of the basic principles of the acoustic emission technique is given, including an
overview of the developments and applications of the technique. Special attention
is paid to the influence of noise sources on the measured signal, the sensitivity
of the technique and the material and sensor response to the original AE signal.
Furthermore, the theory of martensitic and bainitic phase transformations in steel
is discussed.
The experimental equipment for measuring acoustic emission generated during phase transformations in steel is described in chapter 3. First, some technical
details concerning the used AE equipment are provided. Secondly, the methods
utilized for applying a thermal cycle to the studied specimens are described. In
most experiments an arc welding device or a thermo-mechanical (welding) simulator is used; in some cases a dilatometer or a furnace in combination with a salt
bath is employed. For each set-up the procedure is described for the measurement
of acoustic emission during continuous cooling of the studied specimens from elevated temperatures. Furthermore, it is discussed that the detection of AE may be
141
142
Summary
complicated due to the electrical (background) noise of the pre-amplifier. This can
become apparent for example when measurements are performed on small samples,
which transform slowly due to a low cooling rate. Also the use of a waveguide can
strongly attenuate the signal. A waveguide is typically a metal rod which conducts
the acoustic signal from the specimen to the sensor. One end is designed for acoustic coupling with the specimen; the other end is usually conical to accommodate
the mounting of an AE sensor. The use of a waveguide is required when the size
of the specimen is smaller than the diameter of the sensor or when the specimen
becomes very hot during the experiment.
The results of AE measurements during various phase transformations in a
number of steels are discussed in chapter 4. During cooling of a medium carbon
steel specimen (C45) in the Gleeble simulator two distinct peaks in the AE data are
observed, which can be attributed to bainite and martensite formation. Consistent
with this, after spot welding of steel C45 two peaks in the AE data are also observed.
The occurrence of acoustic emission during the bainitic transformation implies that
the bainite reaction mechanism in steel C45 is diffusionless and is best described in
terms of the displacive transformation model. In contrast, no acoustic emission is
detected during the diffusion-controlled transformation from austenite to ferrite in
steel Fe360.
The effect of the austenite grain size on the evolution of the bainite and martensite formation in steel C45 is studied by varying the austenitizing temperature. It
is found that upon lowering the austenitizing temperature, i.e. with decreasing
austenite grain size, the bainite peak increases while the martensite peak decreases.
The magnitude of the bainite peak relative to that of the martensite peak gives insight into the evolution of both phase transformations, i.e. the relative amounts of
bainite and martensite formed. Furthermore, it is seen that the martensite-start
temperature decreases with decreasing austenitizing temperature. Both effects can
be well explained with the existing theory for bainitic and martensitic transformations.
To facilitate comparison between the AE technique and conventional dilatometry, the acoustic emission and the dilatation are measured simultaneously during
the martensitic transformation in steel 42CrMo4. From both signals a similar value
of the martensite-start temperature is derived, which indicates that the sensitivity
of both techniques are of the same order of magnitude in the combined AE –
dilatometer set-up employed. It should be noted, however, that the signal to noise
ratio of acoustic emission is relatively low in this set-up due to the small size of
samples studied and the necessity of using a waveguide. At the end of chapter 4
the results of AE measurements during cooling of specimens in the salt bath are
discussed. These experiments make it possible to study large-sized samples, and the
typical cooling rate is such that bainite is formed in the sample. After correcting
the raw data for the noise due to oxidation, a clear peak in the AE data is obtained
that can be attributed to bainite formation.
In chapter 5 the study of the acoustic emission energy generated during bainite
and martensite formation is presented. A theoretical model is derived from basic
143
principles that predicts that the measured AE energy at the sensor is proportional
to the volume of martensite or bainite formed. Both travelling arc welding and
spot welding are employed in order to test the theoretical prediction. Travelling
arc welding has the advantage that the volume transformation rate can be easily
and accurately determined from the transverse cross-sections of the welds. On
the other hand, determining the AE power originating from martensite and/or
bainite formation during travelling arc welding is rather complicated due to the
noise from the arc. Fortunately, for common welding conditions the contribution of
the welding noise to the total measured AE power is very small. Moreover, the noise
contribution can be determined and thus corrected for. During the first seconds of
welding no transformations take place, and consequently only the welding noise is
measured. After a certain time the first part of the weld cools down and a significant
increase in signal level is observed. Under steady state conditions, the measured
AE power is the sum of the powers due to transformations and welding noise. At
the end of the welding cycle a peak is observed, reflecting that after extinction of
the arc the volume transformation rate increases due to a higher cooling rate. Both
observations, the increase during welding and the peak after welding, indicate that
acoustic emission can be used as a real-time monitoring technique for martensite
formation during welding. This can be of practical importance since martensite in
welds can lead to cold cracking.
In order to validate the above mentioned theoretical prediction, AE experiments
are performed during welding with different heat inputs. The analysis of the results
shows that the AE power is linearly dependent on the volume transformation rate,
as predicted. Furthermore, spot welds are made to test the validity of the relation
between the AE energy and the volume transformed. During cooling of a typical
spot weld a peak signal is observed similar to the peak signal after travelling arc
welding. By integrating the area under a peak in a plot of AE power versus time, the
AE energy is obtained. Also the results of spot welding experiments with different
arc currents confirm that the AE energy is proportional to the volume transformed.
This approach has the advantage that no welding noise is measured. On the other
hand, it is quite difficult to determine the volume of martensite in a spot weld
because this requires that the spot weld is cut exactly in the middle. In accordance
with the procedure described above, the proportionality factor between the AE
energy and the transformed volume k is determined for a number of carbon steels.
The small differences in the values obtained can be well explained with differences
in the chemical composition of the steels.
In chapter 6 the kinetics of the martensitic transformation in a number of carbon steels are studied using the acoustic emission technique. By studying the
kinetics of the martensitic transformation one can obtain fundamental knowledge
about the transformation such as nucleation rates and average volume of martensite crystals. The experimental results are compared to a theoretical model for the
transformation kinetics, the Koistinen and Marburger (KM) relation. According to
the KM equation, the progress of the transformation from austenite to martensite
is quantitatively described by the martensite-start temperature and the value of
144
Summary
the rate constant. The martensite-start temperature Ms is determined by thermodynamics and the rate constant C1 describes the kinetics, i.e. the progress of the
transformation for a certain undercooling.
For four carbon steel specimens (C50, C60, C70 and C80), the AE power is
measured during cooling using the Gleeble thermo-mechanical simulator. Based on
the energy-volume relation from chapter 5, the volume fraction of martensite as a
function of undercooling below the start-temperature can be easily derived, which
allows a direct comparison of experimental data with theoretical predictions of the
transformation kinetics. The overall results obtained for the four carbon steels
show that the fraction data as a function of temperature can be described by the
KM equation with high accuracy. For the low carbon steels a part of the specimen
is austenitized in the two-phase region. At relatively high volume fractions the
transformation in the partially austenitized region becomes predominant, which is
reflected by a decrease of the transformation kinetics caused by the relatively high
carbon content of the transforming austenite in the partially austenitized region.
The good agreement between the experimental data and the Magee model (KM
relation) indicates that the nucleation of martensite takes place heterogeneously
and that the average volume of martensite crystals is constant over the extent of
the transformation. This is different from the Fisher model, which predicts that
the average size of martensite crystals strongly decreases as the transformation
proceeds.
From the fits of the data to the KM equation the values of martensite-start
temperature and the kinetic rate constant are obtained for each steel. As expected,
the martensite-start temperature decreases with increasing carbon content. The
kinetic rate constant has a maximum value for steel C60, which means that the
transformation rate for a given degree of undercooling is the highest for steel C60.
It is argued that the dependence of the rate constant on carbon content can be
explained by the change in dislocation density with carbon content. This is based
on the fact that values of the dislocation density in martensite in similar carbon
steels observed by other researchers, also show a maximum at approximately 0.6 %
C. When a martensite crystal is formed, dislocations are created in the neighboring
austenite due to the fact that the shear stress accompanying the shape change
exceeds the yield strength of the austenite. The dislocation debris leads to extra
nucleation sites apart from the embryos initially present in the austenite. This
effect is known as autocatalysis.
Analogous to the procedure described in chapter 5, the proportionality factors
between the acoustic emission energy and the transformed volume of martensite
k are determined for the four carbon steels. Also, the values of k are found to
scale with the change in dislocation density in the formed martensite as the carbon
content is varied. The corresponding tendency of both entities suggests a close
relation between the dislocations and the actual source of acoustic emission during
the martensitic transformation.
Based on the correlation between the dislocation density and the acoustic emission energy generated per unit volume of martensite, in chapter 7 a model is pre-
145
sented that describes the generation of acoustic waves during growth of martensite
and bainite crystals. The growth of the new phase is constituted by the movement
of a glissile interface, which consists of arrays of parallel dislocations. When the interface moves, the existing interfacial dislocations have to overcome barriers, which
are smaller than the nucleation barrier, and this causes the jerky motion of the interface. After the interface has overcome a barrier, elastic strain is suddenly released
to produce vibrational waves in the lattice: acoustic emission. This dislocation
source model is adopted from acoustic emission studies of plastic deformation, and
subsequently modified to describe acoustic emission generation during displacive
phase transformations. Furthermore, the results of AE measurements during plastic deformation and displacive transformations are compared and discussed. The
results obtained for both processes show many similarities, which supports the earlier argument that it is the dislocation motion during displacive transformations
that constitutes the generation of acoustic waves.
Subsequently, the characteristics of acoustic waves generated during bainite and
martensite formation are studied. The results are discussed in a semi-quantitative
manner, since a thorough quantitative analysis of signals is not feasible because of
the limited frequency bandwidth of the system and the unknown sample response.
The frequency spectra of acoustic emission signals are interpreted using the dislocation source model discussed above. The results show that the mean frequency of
waves generated during bainite formation is significantly larger than that of martensitic waves. This difference in the spectral density of the waves can be explained
by the difference in interface motion of the two transformations and consequently
the different behaviour of the dislocations involved.
This chapter ends with an overview of the proportionality factors between the
acoustic emission energy and the transformed volume (k factors) obtained for the
steels studied in this thesis. An analysis of k factors and dislocation densities
indicates that the measured AE power during displacive transformations is linearly
dependent on the dislocation density, which is identical to the dependence found in
case of acoustic emission measurements during tensile testing.
146
Summary
Samenvatting
van het proefschrift met de titel:
Akoestische emissie tijdens martensitische en bainitische
transformaties in koolstofstaal
Deze samenvatting begint met een korte inleiding op het onderzoek, die met name
bedoeld is voor niet-vakgenoten. Daarna gevolgt een omschrijving van de belangrijkste resultaten zoals vermeld in dit proefschrift.
Akoestische emissie
Akoestische emissie (AE) is het verschijnsel dat elastische golven ontstaan in een
materiaal doordat er lokaal een abrupte spanningsverlaging optreedt. Voorbeelden
van processen die gepaard gaan met het optreden van akoestische emissie zijn
scheurgroei, plastische deformatie, oxidatie en diffusieloze fasetransformaties.
De in het materiaal opgewekte trillingen ten gevolge van de spanningsverlaging bij de AE-bron zijn voor het menselijk oor niet hoorbaar, want de frequenties
van het geluid liggen ongeveer tussen de 50 kilohertz en 10 Megahertz, terwijl
het menselijke gehoor grofweg van 20 hertz tot 20 kilohertz reikt. Deze hoogfrequente golven veroorzaken hele kleine verplaatsingen aan het oppervlak, ongeveer
1000 maal kleiner dan de afstand tussen twee atomen. Om deze verplaatsingen
te kunnen detecteren wordt een sensor gebruikt met een piëzo-elektrisch kristal.
De trillingen veroorzaken vervormingen van het kristal en zo ontstaat er een elektrische spanningsverandering over het kristal. Na versterking van het signaal kan
de geluidsgolf bestudeerd worden met behulp van een computer.
Fasetransformaties in koolstofstaal
Staal is een van de belangrijkste constructiematerialen en wordt gebruikt voor de
fabrikage van o.a. schepen, bruggen en auto’s. Het verbeteren van de eigenschappen van staal voor de verschillende doeleinden is daarom nog steeds van groot
belang. De eigenschappen van een staalsoort worden tijdens het productieproces
voornamelijk bepaald door de microstructuur die tijdens het afkoelen ontstaat.
147
148
Samenvatting
Legeringselementen spelen hierbij een grote rol, maar ook eventuele warmtebehandelingen.
Staal is een legering van ijzer (Fe) met minimaal 0.02 % koolstof (C), en heeft
verder vaak legeringselementen zoals bijvoorbeeld mangaan, silicium, chroom en
molybdeen. Bij hoge temperaturen heeft staal een andere kristalstructuur dan
bij kamertemperatuur. Boven 730 ◦ C heeft staal een kubisch-vlakken gecentreerd
kristalrooster dat aangeduid wordt met austeniet. Koelt het staal langzaam af,
dan gaat austeniet over in ferriet dat een kubisch-ruimtelijk gecentreerde structuur
heeft. Zo’n verandering van kristalrooster wordt een (structurele) fasetransformatie
genoemd. Het gevormde ferriet is de meest bekende vorm van staal: goed vervormbaar en redelijk sterk. Koelt het staal daarentegen heel snel af, dan transformeert
het austeniet naar martensiet. De kristalstructuur daarvan lijkt sterk op die van
ferriet, maar door het snelle afkoelen zit koolstof als het ware in het ijzerrooster
gevangen en ontstaat er een bepaalde interne spanning. Martensiet is daardoor
heel hard, maar tegelijkertijd ook bros waardoor het makkelijk breekt. Naast ferriet en martensiet kan austeniet ook nog naar perliet of bainiet transformeren, die
respectievelijk een gelaagde en een naaldvormige kristalstructuur hebben. Deze microstructuren ontstaan als austeniet iets sneller wordt gekoeld dan bij de vorming
van ferriet, maar minder snel dan voor martensiet nodig is.
Wat betreft de manier waarop de nieuwe fase groeit kunnen twee transformatiemechanismen onderscheiden worden. De vorming van ferriet en perliet uit
austeniet zijn diffusie gestuurd en dientengevolge relatief langzame processen. Hiermee vergeleken is de vorming van martensiet een veel sneller ofwel abrupter proces, en vindt plaats via een zogenaamd displacive mechanisme. Hierbij bewegen
de atomen gezamenlijk binnen een zeer korte tijd over een relatief kleine afstand
(d.w.z. kleiner dan de atomaire afstand). Door deze gelijktijdige beweging van grote
aantallen ijzeratomen transformeert het austeniet naar martensiet. Dit gebeurt in
het grensvlak dat het oorspronkelijke kristal (austeniet) scheidt van het nieuwe
kristal (martensiet). Een diffusieloze fasetransformatie begint doorgaans spontaan
bij een bepaalde temperatuur die wordt aangeduid als de martensiet-start temperatuur; bij deze temperatuur worden de eerste martensietkristallen gevormd. De
martensietkristallen groeien met zeer grote snelheid tot hun eindafmeting, en de
groeisnelheid is onafhankelijk van de temperatuur. Bij verdere temperatuurdaling
vindt de vorming van nieuwe martensietkristallen plaats totdat de omzetting is
voltooid.
Omdat de martensietkristallen niet passen in de austeniet waaruit ze worden
gevormd, treden spanningen in het grensvlak op. Tijdens de groei van een martensietkristal zal de spanning geminimaliseerd worden door zogeheten slipping (dislocatiebewegingen); een proces dat het kristalrooster niet verandert. De spanningverlaging die hiermee gepaard gaat heeft de trillingen in het kristalrooster tot gevolg:
akoestische emissie. De akoestische emissie die tijdens martensietvorming wordt
gegenereerd is dus een direct gevolg van de collectieve beweging van atomen. Dit
betekent dat men door een proces te monitoren met de akoestische emissie techniek te weten kan komen of het proces wel of niet een displacive karakter heeft.
149
Zo geredeneerd kan de akoestische emissie techniek van grote betekenis zijn om de
transformatie van austeniet naar bainiet beter te begrijpen. Er is namelijk tot op
heden nog steeds geen consensus tussen wetenschappers wat betreft het groeimechanisme van bainiet, d.w.z. is het een diffusionele of diffusieloze fasetransformatie.
Hoewel het reeds enkele decennia bekend is dat akoestische emissie wordt geproduceerd tijdens een martensitische fasetransformatie, is de techniek in het verleden
maar zelden toegepast voor fundamenteel onderzoek van fasetransformaties in staal.
Dit proefschrift beschrijft een uitgebreid en nauwkeurig onderzoek van de akoestische emissiesignalen tijdens martensitische en bainitische transformaties in koolstofstaal.
Dit proefschrift
Allereerst wordt in hoofdstuk 1 een korte beschrijving van akoestische emissie en
fasetransformaties gegeven, inclusief de achtergrond van dit onderzoek. Vervolgens worden in hoofdstuk 2 de grondslagen, ontwikkelingen en toepassingen van
akoestische emissie beschreven. Speciale aandacht wordt besteed aan de invloeden
van ruisbronnen op het signaal, de gevoeligheid van de techniek, en de veranderingen van de geluidsgolven ten gevolge van materiaal- en sensorrespons. Tot slot
worden in dit hoofdstuk de fasetransformaties in staal beschreven, met de nadruk
of de martensitische en bainitische fasetransformatie.
In hoofdstuk 3 staan de gebruikte methoden beschreven om de stalen proefstukken een warmtebehandeling te geven. Deze vier warmtebehandelingstechnieken
zijn: een TIG lasapparaat, een Gleeble lassimulator, een conventionele dilatometer,
en een oven in combinatie met een zoutbad. Voor elke opstelling wordt uitgelegd
hoe de akoestische emissie ten gevolge van diffusieloze transformaties wordt gemeten
tijdens continue afkoeling van de proefstukken. Tevens wordt beschreven dat het
meten van akoestische emissie tijdens fasetransformaties niet altijd even gemakkelijk
is vanwege de achtergrondruis ten gevolge van de elektronische voorversterker. Als
het elektrische vermogen van de akoestische emissie aan de uitgang van de sensor
niet groter is dan het thermische ruisvermogen van de versterker, dan is detectie van
de akoestische golven niet mogelijk. In het algemeen hangt het AE-vermogen aan de
sensor af van the soort materiaal, het volume van het proefstuk, de afkoelsnelheid
en de eventuele verzwakking van het signaal door het gebruik van een golfgeleider.
Een typische golfgeleider is een dunne stalen staaf met aan één kant een uiteinde
waarop een sensor kan worden geplaatst. Het gebruik van een golfgeleider kan
noodzakelijk zijn als het proefstuk zelf te klein is om een sensor op aan te brengen,
of als het proefstuk te heet wordt tijdens het experiment.
De resultaten van AE-metingen tijdens fasetransformaties in een aantal verschillende staalsoorten worden gepresenteerd in hoofdstuk 4. De meeste experimenten zijn uitgevoerd met behulp van de Gleeble lassimulator of het TIG lasapparaat. Voor beide methoden wordt zowel tijdens de martensitische als de bainitische
transformatie een piek-vormig AE-signaal waargenomen met een vergelijkbare intensiteit. Het gemeten AE-signaal gedurende bainietvorming geeft aan dat de groei
150
Samenvatting
van bainiet gepaard gaat met de collectieve beweging van ijzeratomen tijdens de
transformatie; dit levert dus een sterk bewijs dat de bainitische transformatie diffusieloos is. Ter vergelijking is ook de diffusie-gestuurde transformatie van austeniet
naar ferriet bestudeerd onder gelijke omstandigheden. Het blijkt dat tijdens ferrietvorming geen akoestische emissie wordt gegenereerd. Dit is in overeenstemming
met de verwachting, want diffusionele processen zijn langzaam en gaan dus niet
gepaard met het snelle vrijkomen van spanningsenergie.
Verder worden in dit hoofdstuk de simultane metingen van akoestische emissie
en dilatatie tijdens martensietvorming met elkaar vergeleken. De resultaten uit
beide metingen zijn in goede overeenstemming met elkaar wat betreft de starttemperatuur van de transformatie. De grootste beperkingen van AE-metingen
aan proefstukken in een conventionele dilatometer zijn het kleine volume van de
proefstukken en het noodzakelijke gebruik van een golfgeleider; beide reduceren de
detecteerbaarheid. Aan het eind van het hoofdstuk worden de resultaten besproken
van metingen aan proefstukken tijdens afkoeling in een zoutbad. In deze experimenten kunnen grote proefstukken worden gebruikt, en door een juiste keuze van
de temperatuur van het zout is de afkoelsnelheid van een proefstuk zodanig dat
bainiet wordt gevormd. Ook met deze warmtebehandelingsmethode wordt tijdens
het afkoelen van het proefstuk een piek in de AE-data gezien bij temperaturen
waar bainietvorming plaatsvindt, wat in overeenstemming is met het displacive
transformatie model voor bainietvorming.
In hoofdstuk 5 wordt het onderzoek naar het verband tussen het gemeten AEvermogen en de volumesnelheid van martensietvorming gepresenteerd. De theorie voorspelt dat de geproduceerde AE-energie rechtevenredig is met het volume
martensiet dat is gevormd. Om de theorie te toetsen zijn TIG-lassen gemaakt op
een werkstuk, zowel onder voortlopende boog condities als onder stationaire condities (spotlassen). Een voordeel van lassen met voortlopende boog is dat de volume
transformatiesnelheid constant en achteraf gemakkelijk te bepalen is aan de hand
van de dwarsdoorsnede van de las en de voortloopsnelheid van de boog. Een nadeel
van AE-metingen tijdens voortlopende booglassen is dat het lasproces zelf ook een
bron van akoestische emissie is; door de interactie tussen boog en werkstuk worden
akoestische golven gegenereerd. Het gemeten AE-signaal tijdens lassen heeft dus
twee bijdragen: displacive transformaties en lasruis. Metingen tonen echter aan
dat onder normale lasomstandigheden het AE-vermogen ten gevolge van de transformatie veel groter is dan dat van de lasruis. Bovendien kan de laatstgenoemde
bijdrage bepaald en dus voor gecorrigeerd worden. Gedurende de eerste seconden
van het lassen transformeert namelijk nog geen metaal en dus wordt alleen het
ruisvermogen gemeten. Tijdens het afkoelen van het eerste deel van de las kan een
duidelijke toename in het signaal worden waargenomen. Onder evenwichtsituaties
is het gemeten AE-vermogen de som van de vermogens ten gevolge van fasetransformaties en lasruis. Direct na het stoppen van het lasproces is een piek te zien,
die weerspiegelt dat de volume transformatiesnelheid toeneemt doordat het laatste
deel van de las met relatief hoge snelheid afkoelt als de boog uitdooft. Zowel de
toename aan het begin van het lassen als de karakteristieke piek na het lassen kan
151
als criterium voor martensietvorming worden gebruikt. Deze real-time detectie kan
van wezenlijk praktisch nut zijn voor het lassen van hooggelegeerden staalsoorten
omdat t.g.v. martensietvorming in een las koudschuren kunnen optreden.
Om de bovengenoemde theoretische voorspelling te valideren, zijn AE-experimenten uitgevoerd met verschillende stroomsterktes. Uit de metingen, gecorrigeerd voor
de lasruis, blijkt dat het AE-vermogen inderdaad lineair toeneemt met de volume
transformatiesnelheid. Tevens zijn ook spotlassen gemaakt om de relatie tussen de
AE-energie en het volume martensiet te bevestigen. Tijdens het afkoelen van de
spotlas wordt een pieksignaal gemeten, vergelijkbaar met de piek na het stoppen
van het lassen met voortlopende boog. Deze metingen hebben als voordeel dat
geen lasruis wordt gemeten. Daarentegen is het wel relatief moeilijk om het volume
martensiet in de spotlas te meten omdat daartoe de spotlas precies in het midden
moet worden doorgezaagd. Ook uit de meetresultaten van het spotlassen volgt dat
de AE-energie zich rechtevenredig verhoudt tot het volume martensiet. Voor verschillende staalsoorten is op bovenbeschreven manier de evenredigheidsfactor tussen
de AE-energie en het volume martensiet (k) bepaald. De kleine verschillen in de
gevonden waarden voor de evenredigheidsfactor k kunnen worden toegeschreven
aan verschillen in chemische samenstelling tussen de staalsoorten.
In hoofdstuk 6 wordt de kinetiek van de martensitische transformatie onderzocht
met akoestische emissie. Onder kinetiek wordt de voortgang van de transformatie
verstaan, ofwel de verandering van de volumefractie martensiet als functie van
temperatuur of tijd. Voor vier staalsoorten (C50, C60, C70 en C80) is tijdens de
afkoeling in de Gleeble lassimulator het AE-vermogen ten gevolge van de martensietvorming gemeten. Met het resultaat uit hoofdstuk 5 dat het AE-vermogen
proportioneel is met de volume transformatiesnelheid, kan de fractie martensiet
direct uit de AE-data worden bepaald.
Door de gevonden fractie data te vergelijken met een theoretische model voor
de kinetiek van de martensitische transformatie, de zogenaamde Koistinen & Marburger (KM) relatie, blijkt het mogelijk meer inzicht te verkrijgen in fysische parameters die ten grondslag liggen aan de martensitische transformatie. Uit het fitten van de AE-data met de KM vergelijking kan de kinetische constante en de
start-temperatuur van de transformatie worden bepaald. Het feit dat de transformatie met een kinetische constante beschreven kan worden wijst erop dat de gemiddelde kristalgrootte van martensiet niet verandert gedurende de transformatie. Ter
vergelijking, het Fisher model dat veronderstelt dat de kristalgrootte zal afnemen
als de fractie martensiet toeneemt, geeft dus geen goede beschrijving van de martensitische transformatie in koolstofstaal. Ook microscopische waarnemingen van het
oppervlak van een geheugenmetaal laten zien dat gedurende de gehele transformatie
zowel kleine als grote martensietkristallen worden gevormd.
Uit het resultaat van de metingen aan de vier staalsoorten met verschillend
koolstofgehalte blijkt dat de kinetische constante het grootst is voor staal C60
(0.6 procent koolstof). Als wordt aangenomen dat de gemiddelde kristalgrootte
niet significant verandert met koolstofgehalte, dan kan de verandering in kinetische
constante worden toegeschreven aan een verandering in het aantal nucleatiekernen
152
Samenvatting
per volume eenheid. Metingen van de dislocatiedichtheid in martensiet die zijn
uitgevoerd door andere onderzoekers op vergelijkbare staalsoorten vertonen ook een
maximum voor staal met een koolstofgehalte van ongeveer 0.6 procent. Dit duidt
erop dat dislocaties een belangrijke bijdrage leveren aan het aantal nucleatiekernen.
Deze dislocaties worden tijdens de fasetransformatie gevormd en kunnen zodoende
de transformatie versnellen; dit wordt ook wel het autocatalytische effect genoemd.
Analoog aan de methode beschreven in hoofdstuk 5 zijn ook voor deze vier
koolstofstalen de evenredigheidsfactoren k bepaald. De waarden voor k als functie
van koolstofgehalte vertonen dezelfde tendens als de dislocatiedichtheden, en deze
correlatie suggereert dat het in feite de dislocatiebewegingen zijn die de akoestische
golven genereren tijdens de martensitische transformatie.
Gebaseerd op bovengenoemde correlatie is in hoofdstuk 7 een model gepresenteerd dat het ontstaan van akoestische golven beschrijft tijdens de groei van
martensietkristallen. Tijdens deze groei beweegt het grensvlak tussen martensiet
en austeniet, waarin zich dislocaties bevinden, richting austeniet. Als het grensvlak
beweegt, dan moeten de grensvlakdislocaties barrières overwinnen, en dat zorgt
ervoor dat het grensvlak discontinu voortbeweegt. Nadat een barrière is overwonnen komt er plotseling elastische energie vrij en gaat het kristalrooster trillen:
akoestische emissie. Dit model is afgeleid van een bestaand model dat akoestische
emissie tijdens conventionele plastische deformatie beschrijft. Het blijkt ook dat er
opmerkelijke overeenkomsten zijn tussen AE-metingen tijdens plastische deformatie
(trekproeven) en AE-metingen tijdens martensitische transformaties.
Vervolgens worden in dit hoofdstuk de eigenschappen van de AE-signalen die
ontstaan tijdens de bainitische transformatie vergeleken met die ten gevolge van
de martensitische transformatie. De resultaten worden kwalitatief geanalyseerd
omdat de onbekende materiaal respons en de beperkte bandbreedte van de sensor
een gedegen kwantitatieve benadering onmogelijk maken. De frequentiespectra
van AE-signalen worden geı̈nterpreteerd aan de hand van het bovenbeschreven
dislocatiemodel. Het blijkt dat tijdens bainietvorming de gemiddelde frequentie
van de gemeten akoestische golven significant groter is dan tijdens martensietvorming. Dit verschil kan worden toegeschreven aan verschillen in de beweging van het
grensvlak voor beide transformaties.
Aan het eind van dit hoofdstuk wordt een overzicht gegeven van de evenredigheidsfactoren k voor alle staalsoorten. Analyse van deze factoren en dislocatiedichtheden geeft aan dat het AE-vermogen proportioneel is met de dislocatiedichtheden;
dit verband is identieke aan die gevonden voor AE-metingen tijdens trekproeven.
List of publications
1. J.A. Reedijk, H.C.F. Martens, S.M.C. van Bohemen, O. Hilt, H.B. Brom and
M.A.J. Michels,
Charge transport in doped polythiophene,
Synthetic Metals 101, 475-483 (1999).
2. H.C.F. Martens, H.B. Brom, J.A. Reedijk, S.M.C. van Bohemen, I. Couronne
and J. Fournier,
Dielectric study of polypyrrole/epoxy composites,
Synthetic Metals 102, 1236-1243 (1999).
3. S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden,
Acoustic emission during GTAW of Cr-Mo steel,
In: Proceedings JOM-10, Tenth Int. JOM-Jubilee Conf. on The Joining of
Materials, Helsingor, Denmark, May 11-14, 317-323 (2001).
4. S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden,
Acoustic emission monitoring of martensite formation during welding,
In: Proceedings Int. Conf. IIW, Lublijana, Slovenia, July 9-10, IIW Doc.
212-1005-01, 1-12 (2001).
5. S.M.C. van Bohemen,
Detectie martensietvorming tijdens lassen met akoestische emissie,
In: Lastechniek, Jaargang 67, no. 12, 9-12 (2001).
6. S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden,
Monitoring of martensite formation during welding by means of acoustic emission,
Journal of Physics D: Applied Physics 34, 3312-3317 (2001).
7. S.M.C. van Bohemen, M.J.M. Hermans, G. den Ouden, and I.M. Richardson,
A study of acoustic emission generated during bainite and martensite formation,
Journal of Physics D: Applied Physics 35, 1889-1894 (2002).
153
154
List of publications
8. S.M.C. van Bohemen, M.J.M. Hermans and G. den Ouden,
Acoustic emission monitoring of bainitic and martensitic transformation in
medium carbon steel during continuous cooling,
Materials Science and Technology 18, 1524-1528 (2002).
9. S.M.C. van Bohemen, J. Sietsma, M.J.M. Hermans and I.M. Richardson,
Acoustic emission as a probe of the kinetics of the martensitic transformation
in low-alloy steel,
In: Proceedings COM2003, Vancouver, Canada, August 24-28, 125-139 (2003).
10. A. Mertens, S.M.C. van Bohemen, M.J.M. Hermans, J. Sietsma, and S. van
der Zwaag,
Acoustic emission investigations on the austenite decomposition in a medium
carbon steel,
In: Proceedings COM2003, Vancouver, Canada, August 24-28, 109-123 (2003).
11. S.M.C. van Bohemen, M.J.M. Hermans, G. den Ouden and I.M. Richardson,
Acoustic emission monitoring of bainite formation during continuous cooling,
Journal de Physique IV 112, 301-305 (2003).
12. S.M.C. van Bohemen, J. Sietsma, M.J.M. Hermans and I.M. Richardson,
Kinetics of the martensitic transformation in low-alloy steel studied by means
of acoustic emission,
Acta Materialia 51, 4183-4196 (2003).
13. S.M.C. van Bohemen,
Study of acoustic emission signals generated during martensitic transformations,
In: Proceedings NWGAE, Delft, The Netherlands, November 20, 1-18 (2003).
14. S.M.C. van Bohemen, J. Sietsma, M.J.M. Hermans and I.M. Richardson,
Analysis of acoustic emission signals originating from bainite and martensite
formation,
to be submitted to Philosophical Magazine A.
Curriculum Vitae
Stefanus Matheus Cornelis van Bohemen
born January 14th , 1973 in Wassenaar, The Netherlands
1985 – 1990
Secondary School, Lucas College (HAVO) in Voorschoten,
The Netherlands.
1990 – 1994
Bachelors degree in Engineering at the Polytechnical School
in Haarlem, The Netherlands.
1994 – 1998
Masters degree in Physics, Kamerlingh Onnes Laboratory of
Leiden University, Leiden, The Netherlands.
Thesis: ’Charge carrier transport in composites’
2000 – 2004
Ph.D. research at the Laboratory of Materials Science, Delft
University of Technology, Delft, The Netherlands.
Thesis: ’An acoustic emission study of martensitic and bainitic
transformations in carbon steel’
155
156
Nawoord
Graag maak ik van de gelegenheid gebruik om een aantal mensen te noemen zonder
wie de totstandkoming van dit proefschrift niet mogelijk zou zijn geweest.
Allereerst mijn begeleiders, Marcel Hermans die er altijd was om me te helpen of
een vraag te beantwoorden, Gert den Ouden die dit onderzoeksproject heeft opgestart en de eerste twee jaar mijn promotor was, Ian Richardson die de taak van Gert
na twee jaar overnam, en Jilt Sietsma die het team van begeleiders kwam versterken
toen mijn promotieonderzoek zich voornamelijk ging richten op fasetransformaties.
Bedankt voor jullie steun en alles wat ik van jullie heb geleerd.
Graag bedank ik ook Willem Brabander en Frans Bosman bij wie ik altijd terecht
kon voor de technische ondersteuning, en Anneke van Veen voor de administratieve
hulp. Ook wil ik Herman Schoorlemmer (Physical Acoustics B.V.) noemen die mij
vertrouwd heeft gemaakt met de akoestische emissie apparatuur. Verder Bram Huis
die me heeft geholpen bij het maken van proefstukken, en Erik Peekstok die altijd
bereid was om het instellen van de optische microscopen uit te leggen.
Ik ben ook erg blij met de vele collega’s die altijd voor de nodige afleiding
hebben gezorgd. De goede sfeer binnen de vakgroep Lastechnologie heb ik tijdens
mijn promotietijd altijd bijzonder gewaardeerd.
157
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement