an acoustic emission study of martensitic and bainitic transformations in carbon steel

An acoustic emission study of

martensitic and bainitic

transformations in carbon steel

The research described in this thesis was performed in the department of MaterialsScience and Technology, Delft University of Technology, Rotterdamseweg 137, 2628AL Delft, The Netherlands.

The research described in this thesis was carried out in the framework of the Strate-gic Research Programme of the Netherlands Institute for Metals Research in theNetherlands (www.nimr.nl).

An acoustic emission study of

martensitic and bainitic

transformations in carbon steel

PROEFSCHRIFT

ter verkrijging van de graad van doctoraan 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 natuurkundegeboren te Wassenaar

Dit proefschrift is goedgekeurd door de promotor:Prof.dr. I.M. Richardson

Toegevoegd promotor: Dr.ir. J. Sietsma

Samenstelling promotiecommissie:

Rector Magnificus, voorzitterProf.dr. I.M. Richardson, Technische Universiteit Delft, promotorDr.ir. J. Sietsma, Technische Universiteit Delft, toegevoegd promotorProf.dr. G. den Ouden, Technische Universiteit DelftProf.dr.ir. M. Wevers, Katholieke Universiteit Leuven, Leuven, BelgieProf.dr. R. Boom, Technische Universiteit DelftDr. P.J. Jacques, Universite catholique de Louvain, Louvain-la-Neuve, BelgiumDr.ir. M.J.M. Hermans, Technische Universiteit DelftProf.dr.ir. S. van der Zwaag, Technische Universiteit Delft, reservelid

Dr.ir. M.J.M. Hermans heeft als begeleider in belangrijke mate aan detotstandkoming van het proefschrift bijgedragen.

Published and distributed by: DUP Science

DUP Science is an imprint ofDelft University PressP.O. Box 982600 MG DelftThe NetherlandsTelephone: +31 15 2785678E-mail: [email protected]

Keywords: Phase transformations, acoustic emission, steel, martensite, bainite

Copyright c© 2004 by S.M.C. van Bohemen

All rights reserved. No part of the material protected by this copyright notice maybe reproduced or utilized in any form or by any means, electronic or mechanical,including photocopying, recording or by any information storage and retrieval sys-tem, without written permission from the publisher: Delft University Press.

Printed in The Netherlands

ISBN 90-407-2477 -6

Contents

1 Introduction 1

2 Acoustic emission and phase transformations 7

2.1 Historical review of the AE technique . . . . . . . . . . . . . . . . . 72.2 Basic theory of acoustic emission . . . . . . . . . . . . . . . . . . . . 9

2.2.1 Material and transducer response . . . . . . . . . . . . . . . . 102.2.2 Sensors and pre-amplifiers . . . . . . . . . . . . . . . . . . . . 122.2.3 Attenuation and noise . . . . . . . . . . . . . . . . . . . . . . 14

2.3 Phase transformations in steel . . . . . . . . . . . . . . . . . . . . . . 152.3.1 Martensitic transformation . . . . . . . . . . . . . . . . . . . 152.3.2 Bainitic transformation mechanism . . . . . . . . . . . . . . . 18

3 Experimental 23

3.1 Acoustic emission system . . . . . . . . . . . . . . . . . . . . . . . . 233.1.1 Sensor mounting and noise precautions . . . . . . . . . . . . . 273.1.2 Attenuation due to waveguides . . . . . . . . . . . . . . . . . 283.1.3 Source location . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.2 Gas tungsten arc welding . . . . . . . . . . . . . . . . . . . . . . . . 323.3 Thermo-mechanical simulator . . . . . . . . . . . . . . . . . . . . . . 383.4 Dilatometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.5 Furnace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433.6 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4 Acoustic emission monitoring of phase transformations in steel 47

4.1 Study of steel C45 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484.1.1 Thermo-mechanical simulator experiments . . . . . . . . . . . 484.1.2 Welding experiments . . . . . . . . . . . . . . . . . . . . . . . 56

4.2 Study of steel 42CrMo4 . . . . . . . . . . . . . . . . . . . . . . . . . 584.2.1 Welding experiments . . . . . . . . . . . . . . . . . . . . . . . 584.2.2 Dilatometer experiments . . . . . . . . . . . . . . . . . . . . . 604.2.3 Furnace experiments . . . . . . . . . . . . . . . . . . . . . . . 61

4.3 Study of low carbon steels . . . . . . . . . . . . . . . . . . . . . . . . 644.4 Study of a high-alloyed steel . . . . . . . . . . . . . . . . . . . . . . . 65

i

ii

4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5 A study of acoustic emission energy generated during bainite and

martensite formation 73

5.1 Theoretical background . . . . . . . . . . . . . . . . . . . . . . . . . 745.2 Martensite formation . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

5.2.1 Travelling arc welding of steel 42CrMo4 . . . . . . . . . . . . 765.2.2 Spot welding of steel 42MnV7 . . . . . . . . . . . . . . . . . . 78

5.3 Bainite and martensite formation . . . . . . . . . . . . . . . . . . . . 825.3.1 Spot welding of steel C45 . . . . . . . . . . . . . . . . . . . . 835.3.2 Travelling arc welding of steel C45 . . . . . . . . . . . . . . . 84

5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

6 Kinetics of the martensitic transformation studied by means of

acoustic emission 91

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926.2 Theoretical background . . . . . . . . . . . . . . . . . . . . . . . . . 936.3 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . 946.4 Study of steels C50, C60, C70 and C80 . . . . . . . . . . . . . . . . . 96

6.4.1 Calculation of the martensite volume fraction . . . . . . . . . 966.4.2 Proportionality factors k and dislocation densities ρ . . . . . 976.4.3 Koistinen-Marburger kinetics . . . . . . . . . . . . . . . . . . 996.4.4 A different analysis of the results for steel C80 . . . . . . . . 1056.4.5 Microstructural analysis . . . . . . . . . . . . . . . . . . . . . 1066.4.6 Martensite-start temperature Ms . . . . . . . . . . . . . . . . 1076.4.7 Rate constant C1 . . . . . . . . . . . . . . . . . . . . . . . . . 109

6.5 Analysis of the results for steel 42CrMo4 . . . . . . . . . . . . . . . . 1126.6 Study of a shape memory alloy . . . . . . . . . . . . . . . . . . . . . 113

6.6.1 Acoustic emission experiments . . . . . . . . . . . . . . . . . 1136.6.2 Optical Confocal Laser Scanning Microscopy observations . . 114

6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

7 Analysis of acoustic emission signals originating from bainite and

martensite formation 121

7.1 Acoustic emission during plastic deformation . . . . . . . . . . . . . 1227.2 Dislocation dynamics during displacive transformations . . . . . . . 125

7.2.1 Nucleation and growth of martensite . . . . . . . . . . . . . . 1257.2.2 Nucleation and growth of bainite . . . . . . . . . . . . . . . . 127

7.3 Analysis of continuous acoustic emission . . . . . . . . . . . . . . . . 1277.4 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . 1287.5 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 1297.6 Discussion of proportionality factors k . . . . . . . . . . . . . . . . . 1357.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

Summary 141

iii

Samenvatting 147

List of publications 153

Curriculum Vitae 155

Nawoord 157

Chapter 1

Introduction

Acoustic emission (AE) is the name given to the phenomenon of elastic wavesbeing generated by the rapid release of strain energy from localized sources withina material [1]. As an AE event occurs at a source, elastic waves are generated andpropagate in all directions and ultimately reach the surface of the material.

Phenomena that are classified today as acoustic emission have been observedsince the beginning of technology. For example during pottery making the earlypotters learned to associate the sound of pottery cracking as it cooled with theformation of cracks in their creations. Another familiar example of audible acous-tic emissions is the so-called ‘tin cry’, heard by tin smiths during the deformationof tin, which is due to mechanical twinning [2]. These observations date back toapproximately 3000 BC. The first documented observation of acoustic emission dur-ing 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 marten-site during cooling. Around the start of the twentieth century, the martensiticmicrostructure was observed for the first time by the German metallurgist AdolfMartens (1850-1914). In 1936 Forster and Scheil reported that the martensitictransformation in steel is accompanied by ”clicks” [3]. This may be considered asthe first study of acoustic emission during martensite formation.

A unified (unambiguous) explanation of the source of acoustic emission does notyet exist. Usually the source is a process which involves a mechanism of deformationor 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 trans-formations [8]. In this thesis the acoustic emission during phase transformations insteel is discussed, mainly focussing on martensitic and bainitic transformations.

A martensitic transformation is a diffusionless first-order phase transition duringwhich the lattice distortion is mainly described by a combination of shears [9, 10].It involves a cooperative and almost simultaneous shear movement of atoms fromparent 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. Thiscauses transient surface displacements that can be detected with a transducer. Thevoltage signal from the transducer is then usually amplified in a pre-amplifier andanalyzed 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 importantrole. Due to the change in lattice structure and the fact that the transformation isdisplacive, several physical properties can be used to investigate the characteristicsof the transformation, such as the transformation-start temperature and the kinet-ics of the transformation. The most common methods used to study the martensitictransformation in-situ are electrical resistivity, dilatometry and calorimetry. Theacoustic emission technique used in this work is a rather new and promising tech-nique and has not often been used to study the martensitic transformation in real-time. Moreover, the acoustic emission technique is considered to be a good methodto investigate the displacive character of a phase transformation [12]. Since there isstill no general agreement about the mechanism of bainite formation [13, 14, 15, 16],acoustic emission measurements during bainite formation will give valuable infor-mation concerning its mechanism of growth.

Outline

The acoustic emission experiments described in this thesis have been performedmainly on medium carbon steels with the aim to study the martensitic and bainitictransformation in these materials under continuous cooling conditions. Continuouscooling of steel is achieved during/after welding, and in a thermo-mechanical simu-lator (welding simulator). One of the major merits of measurements under naturalcontinuous cooling conditions is the absence of possible external noise from a heat-ing source, which is required for isothermal transformation conditions. Moreover,the transformation rate is usually faster during continuous cooling, which resultsin a better signal to noise ratio.

In chapter 2 the basic concepts of acoustic emission are presented, including anoverview of the development and applications of the AE technique. The effect ofthe material and sensor response to the original waveform at the source is describedfollowed by a discussion of sensitivity, attenuation and noise sources. Finally, thetheory of the phase transformations from austenite to bainite and martensite isdiscussed.

The acoustic emission instrumentation, the signal processing technique and theexperimental equipment is described in chapter 3. In most experiments, a weldingapparatus or a thermo-mechanical simulator was used to apply a thermal cycle tothe steel studied. The thermal treatment of a spot weld is usually very similar to thethermal treatment of a specimen used in the thermo-mechanical simulator. In orderto investigate relatively large samples, some experiments were performed using afurnace and a salt bath. For a proper comparison between the AE technique and

3

dilatometry, simultaneous measurements of acoustic emission and dilatation wereperformed by connecting the AE system to a conventional dilatometer apparatus.The principles of these techniques are discussed and the procedure for measuringacoustic emission is given. This chapter concludes with an overview of the materialsused in this study.

In chapter 4 the results of AE experiments on various carbon steels are pre-sented, which show that not only the martensitic transformation, but also thebainitic transformation is accompanied by acoustic emission. The implication ofthis observation for the transformation mechanism of bainite is discussed. For com-parison with the AE results during the martensitic and bainitic transformation, AEexperiments were performed on a low carbon steel, which transforms from austeniteto ferrite via a diffusion controlled mechanism. Furthermore, the AE signal mea-sured during martensite formation is compared with the change in dilatation of thesample during transformation.

In chapter 5 the relationship between the AE energy released during the marten-sitic transformation and the volume of martensite formed is studied by means ofwelding experiments. The AE energy due to the released strain energy accompa-nying martensitic transformations is theoretically and experimentally studied. Itis shown that both for martensite and bainite formation a specific relation existsbetween the AE energy (rate) and the volume (rate) of the transformation.

Besides the fundamental interest in the nature of the martensitic and bainitictransformation in steels, the results of the study presented in chapter 5 may be usedto develop an AE monitoring system to detect martensite and bainite formationduring 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 ofconsiderable 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 chapter5, the volume fraction of martensite f as a function of time t during cooling canbe calculated from the measured AE signal. The results are compared with thekinetics predicted by the Koistinen and Marburger (KM) equation. At the endof chapter 6 the kinetics of the martensitic transformation in a CuAl-based shapememory alloy are studied by means of both acoustic emission and optical confocallaser scanning microscopy.

In chapter 7 the frequency spectra of acoustic waves generated during bainiteand martensite formation are studied. The change in the mean frequency corres-ponding with the transition from bainite to martensite formation during cooling ofa spot weld is attributed to differences in the interface motion of the two transfor-mations. This chapter concludes with an overview and discussion of the propor-tionality factors between the AE energy and the volume transformed, which weredetermined for a number of steels studied in this thesis.

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 Fur 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, Lon-don (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

Chapter 2

Acoustic emission and phase

transformations

The acoustic emission measurements during phase transformations in steels dis-cussed in this thesis did not only result in a better understanding of the martensiticand bainitic transformation, but also of the AE technique itself. In this chaptersome basic aspects of acoustic emission are explained, in as far as they are rele-vant for the subject of this thesis. First the background of the AE technique isbriefly discussed in section 2.1. Subsequently, the effect of the material and sensorresponse to the original waveform at the source is described in section 2.2. Further-more, a discussion of sensitivity, attenuation and noise sources is given. The theoryof the phase transformations from austenite to bainite and martensite is addressedin section 2.3.

2.1 Historical review of the AE technique

It is generally considered that acoustic emission as a technology started in theearly 1950s with the work of Joseph Kaiser [1] who monitored the emissions of(audible) sound from materials subjected to external loads. He and his coworkerswere the first to use electronic instrumentation to detect acoustic waves producedby metals during deformation. They reported that many metals such as zinc, steel,aluminium, lead and copper produce elastic waves under applied stress and thatacoustic emission activity was irreversible: acoustic emissions were not generatedduring reloading until the previous stress level was exceeded. This phenomenonhas become known as the Kaiser effect.

Investigators in the early 1960s realized extensive improvements in the instru-mentation of the acoustic emission technique. They found that many problemsconcerning background noise could be eliminated, or at least minimized, by work-ing 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 emissionand utilized this technique in studies relating to materials research, structural eval-uation and non-destructive testing.

Most papers on acoustic emission were published in the 1970s and 1980s. Inthose years, an increasing effort was devoted to the understanding of the funda-mental aspects of acoustic emission, such as the nature of the source, and theway in which the elastic waves are propagated and detected. A long-term goal ofthese studies was to learn how to calculate a description of the source event fromthe voltage signal of the sensor. To solve the problem, scientists adopted analysistechniques from earthquake engineering in an attempt to model acoustic emissionsources. Regarding the generation of waves, an earthquake is physically very sim-ilar to an acoustic emission event; it is actually only different in scale. Both theseismic and microseismic activity are initiated by a sudden release of strain energyat 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 tosome extent using this approach, in the case of a metal plate the waveform is quitecomplicated because of reflections, interference etc. Therefore, most applicationsof the AE technique were limited to a qualitative level over these years. Unsuccess-ful attempts in those years to utilize AE for investigation of the source propertiesand elementary mechanisms of martensitic transformations and plastic deformationdiscouraged researchers from these topics.

Although the application of AE for fundamental research declined in the late1980s, the AE technology became more and more popular as a tool for non-destructive testing (NDT) in industry. In these areas use can be made of AEfor in-situ detection of crack evolution and safety monitoring because of the intrin-sic nature of AE signals. Important reasons for its increasing acceptance and usewere the improvements in microelectronics and in computer-based recording andanalysis techniques to handle the high signal rates. Certain AE waveform param-eters, such as amplitude and duration became the standard quantities to describethe AE signals. The display of these parameters allowed a better analysis of resultsin comparison with the number of counts on X-Y recorders, which was used in theearly days of AE technology.

The acoustic emission technique is considered quite unique among the non-destructive testing methods. In contrast to other NDT methods the detected energyis released from within the object rather than being supplied externally by the NDTmethod, as in e.g. ultrasonics, eddy current or radiography. This has the advantagethat with the AE technique the entire structure can be monitored using only onesensor at a certain position. Moreover, with two or more sensors the location ofthe source can be determined. An inherent drawback is that the AE technique canonly detect active sources like crack growth or plastic deformation; cracks or flawswhich have formed previously do not radiate elastic waves. Another limitation ofthe AE technique is that it can be considerably influenced by ambient noise. Greatcare must be taken to distinguish AE signals from the ambient noise, especiallyduring practical application of the technique in an industrial environment.

Basic theory of acoustic emission 9

pre-amplifiersignal

sensor

AE system

sourceAE 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 pos-sibilities to use the AE technology for laboratory studies through essential im-provements in AE measuring systems and analyzing tools. For instance, the full-waveform recording and processing of individual signals became possible with theaid of powerful computers. Nevertheless, the interpretation of detected signalsshould always be performed with considerable care because they are never simplyrelated to the mechanism of the source. This and other fundamental aspects ofacoustic emission are addressed in section 2.2.

2.2 Basic theory of acoustic emission

Acoustic emission is a highly sensitive technique for detecting active microscopicevents in a material. This section gives a brief overview of the most important as-pects of the AE technique; more details about the equipment and signal-processingare addressed in section 3.1. In Fig. 2.1 the process of generation and detection isillustrated. The AE technique involves a source, which is active in a material, andAE instrumentation for the detection of the waves: a sensor, a pre-amplifier andsignal processing equipment. The event at the source causes a release of energywhich propagates in the form of a transient stress wave. This wave propagatesthrough the material, until it reaches the sensor. The sensor converts the smallsurface displacements into an electrical signal, which is transmitted to a nearbypre-amplifier and subsequently to the signal-processing equipment.

The acoustic emission waveform at the source is generally thought of as a simplepulse [3]. This is related to the nature of the generating source, and therefore theemissions contain a broad spectrum of frequencies. Depending on the source, the


frequency of the waves extends from tens of kHz up to tens of MHz [2, 4]. Ingeneral, the detected signal has a complex waveform, which depends on both thecharacteristics of the AE event and the wave propagation effects (wave modes, wavevelocity, attenuation, reflections, and interference) between source and sensor. Inaddition to the wave propagation behaviour, the waveform is further changed bythe 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, thematerial response and transducer response, can make the actual signals observedvery different from the original pulses emitted by the source. This is discussed inmore 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 theforce dipole F (t), whose time variation is pulse-like giving a step-like displacementS(t) at the source [3] as shown in Fig. 2.2. The width and height of this pulsedepend on the dynamics of the source process. During the event at a source elasticwaves are generated. In order to evaluate the surface displacements due to anevent at the source, it is important to understand the wave propagation behaviourin materials.

There are different wave modes for acoustic waves in materials: longitudinal(compression), transverse (shear), surface and plate waves. These waves travelat different velocities, which are a function of the density of the material, theYoung’s modulus and the Poisson’s ratio. In the case of steel, the velocity oflongitudinal waves is approximately 5000 m/s and transverse waves travel at a speedof approximately 3000 m/s. The wave velocity of the fastest mode in a materialcan be measured by using two sensors at different locations and is particularlyimportant for source location, because it is used in the computation of the locationof the source (see section 3.1.3).

For a semi-infinite material, it is in principle possible to relate the time variationof 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 atthe sensor has usually little resemblance to the original waveform at the source.Especially the later part of the waveform may have undergone significant changesdue to multiple reflections, interference and mode conversions.

There are usually many wave paths connecting source and sensor, as illustratedin Fig. 2.2. Waves are reflected at the boundaries of the material. The amountof energy reflected depends on the geometric (angle of incidence) and materialmismatch at the reflecting boundary. In the case that the damping of waves at re-flecting boundaries is low, the detected waveform is made up of many componentsreaching the sensor by different paths. This also implies that the signal amplitudedoes not necessarily result from the first component, but may result from the con-structive interference of several components arriving later at the sensor. The AEwave bounces around the specimen, thereby exciting the sensor each time it passes,


t

X(t

)

t

S(t

)( ) U t

( ) X t

T

( ) S t MF

F

t

U(t

)

M T

Figure 2.2: The original waveform at the source S(t) is significantly changed afterpropagation 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 specimenand the damping of the material. When the damping at the reflecting boundary ishigh, only the first component (the direct wave path) is measured. Typically, theduration of the detected waveform is much longer than the event life time at thesource. Since the measured waveform displays the response of the specimen to theinitial waves at the source, the frequency information in the waveform may there-fore be more related to the specimen geometry than to the event characteristics ofthe source.

The above described effects of the material (material properties and geometryof 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 responsefunction. 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 calcula-tion of the source description from the displacement is called deconvolution of thesignal. It can be easily verified by taking the Fourier transform of Eq. (2.1) thatan equivalent relation is valid for the source spectrum S(f), with f the frequencyof the waves. However, in Fourier space convolution is just multiplication, andtherefore deconvolution is analytically possible in Fourier space.


The conversion of the displacement X(t) to an electrical signal U(t) is madeby a highly sensitive transducer. AE transducers are typically based on a ceramicwafer of piezoelectric material (see Fig. 2.3). This material converts a mechanicaldeformation into an electrical voltage. In addition to the propagation through thematerial, the original signal is further changed during conversion in the transducer.The displacement waveformX(t) is convolved with the transducer response functionT , 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 waveformis radically different from the signal at the source. The original signal is signifi-cantly changed during propagation through the material and after conversion bythe transducer. Although the transducer response function can be measured withreasonable accuracy, it should be realized that the attachment of the transducerto the material changes the mechanical boundary conditions at the previously freesurface; the surface displacement is altered by the presence of the sensor. Further-more, the material response function is in practice difficult to determine, becausean accurate simulation of AE sources is complicated, especially inside a material.All these complications mean that deconvolution of the measured voltage signalto evaluate the source function is extremely difficult, and in general has thereforenot been pursued in the literature. Recently, some simulation studies on acousticemission have been carried out [5, 6].

2.2.2 Sensors and pre-amplifiers

Sensitivity and bandwidth are the most important factors when choosing a sensorfor AE monitoring. The sensor most often used nowadays for AE monitoring isthe piezoelectric transducer [4]. The active element in a piezoelectric transducer isusually a special ceramic such as lead-zirconate-titanate (PZT). The piezoelectriccrystal converts the displacement at its surface into an electrical voltage. It exhibitsthe piezoelectric effect: when the crystal is deformed, the electric voltage across thecrystal is changed. The design of a typical AE sensor with piezoelectric element isshown in Fig. 2.3.

The sensor housing and electronics are designed to minimize electromagneticinterference (EMI). Regarding the electronics, the sensors can be divided into twotypes: single-ended or differential. A single-ended sensor contains one crystal and issusceptible to EM noise signals. In contrast, a differential sensor has a design suchthat common noise signals due to EMI are rejected. It contains two crystal elementsof opposite polarity, and the signal outputs of these elements are transmitted to thetwo inputs of a differential pre-amplifier, where the difference of the two signals isamplified. A detected AE signal produces two voltage signals of opposite polarityand thus the difference is two times the signal output from one element. EMI signalspicked up by the two electronic circuits produce signals of the same polarity, whichcancel out in the pre-amplifier.


Figure 2.3: Schematic illustration of a typical acoustic emission sensor with piezo-electric element. After Miller [4].

Pre-amplifiers are used to provide a higher voltage, which is more usable forfurther processing. It is preferable to place the pre-amplifier close to the sensorto minimize pick-up of electromagnetic interference; sometimes the pre-amplifier isintegrated in the sensor housing. Pre-amplifiers contain a frequency filter to rejectunwanted noise signals, and have a wide dynamic range. They inevitably generateelectronic noise (thermal noise), and it is this background noise (and that of thesensor) that determines the smallest microscopic movement detectable with AE.

The sensitivity of an AE transducer (detection threshold) can be defined as theminimum level of the signal amplitude that can be detected above the backgroundnoise. Whereas in other types of experiments such white noise can be reduced bysignal averaging, this does not hold for AE experiments, because the relevant AEsignals 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 elec-tronic background noise. Typically for modern equipment, the smallest signal thatcan be well distinguished from the electronic noise is approximately 4 µV at theoutput of a typical transducer, corresponding to a surface displacement of about10−14 m. This illustrates that piezoelectric transducers are extremely sensitive.For comparison, atomic radii are in the order of 10−10 m, thus displacements of1/10000 of an atomic radius can produce well-distinguishable AE signals. The dy-namic 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 1µV corresponding to 0 dB and 100 mV corresponding to 100 dB (each 20 dB is afactor 10).

Sensitivities of sensors are typically shown as frequency response diagrams (out-put voltage versus frequency). In order to fully characterize a source in terms ofits time scale and/or frequency content, the transducer bandwidth should matchor even overlap the bandwidth of the surface displacements. This is a difficult con-dition to fulfil because the frequency range of the surface displacements typicallyextends from 500 Hz to 500 MHz [8, 9] whereas the bandwidth of the transduceris usually in the range of 100 kHz to 1 MHz [3]. Therefore, the amount of source


information that can be retrieved from the detected signal is limited.

Two other types of sensors, alternative to piezoelectric crystals, that have beenconsidered in the past are the laser interferometer and the capacitive transducer.However, their characteristics regarding sensitivity or bandwidth were not foundto be optimal for acoustic emission monitoring. Laser interferometers have a toosmall bandwidth and therefore insufficient sensitivity for the typical bandwidth ofacoustic emission. Capacitive transducers can be constructed to be sensitive overa wide frequency range with a flat frequency response. However, they are lesssensitive; the typical minimum displacement that can be measured is in the orderof 10−10 m, which is normally insufficient for acoustic emission monitoring.

Sensor coupling and reproducibility of response are important factors. Calibra-tion checks should be performed after mounting the transducer on the specimento ensure that the sensor is operating properly at the correct sensitivity. This isdiscussed 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 sensitivityof the AE instrumentation and the amplitude of the elastic waves emitted by thesource. Furthermore, the detectability of the generated AE signals depends on theattenuation and the noise over the frequency range of the detecting instrumentation.Attenuation refers to the reduction of the wave amplitudes during propagation. Themajor 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 dominantclose to the source, because due to geometric spreading the amplitude falls offinversely with distance. Due to absorption, the amplitude falls off exponentiallywith distance; thus this attenuation mechanism becomes predominant far from thesource. Also grain boundary scattering and scattering against welds may contributeto attenuation; their effects cannot be predicted quantitatively. The attenuationdue to welds in the waveguides relevant for the work described in this thesis isdiscussed in section 3.1.2.

In laboratory studies the attenuation due to damping and geometric spreadingdoes not normally limit the detectability because the specimens employed are usu-ally small. On the other hand, the use of a waveguide between the specimen andthe sensor may influence the detectability very strongly in the following way: Incase the waveguide is a rod with a cross-section that is relatively small comparedto the size of the specimen, the waveguide acts as an acoustic resistor; not all theavailable AE energy in the specimen can be transmitted to the sensor. A systematicstudy 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 acous-tic emission signals. Both relevant AE signals and noise signals can be classifiedas 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 forcontinuous AE signals amplitude and frequency variations can be observed but theduration of the signal is relatively long.

Continuous noise signals may be caused by leaking air lines in the vicinity of theset-up or electromagnetic interference (EMI). The EMI noise signals are coupledto the acoustic emission equipment by radiation or electrical conduction. Someexamples 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 potentialsource of noise. Fortunately, most mechanical noise diminishes in amplitude atfrequencies 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 absenceof external noise sources, which can be obtained for example under laboratoryconditions, 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 Ar3temperature [7]. The microstructure formed during cooling mainly depends on thechemical composition of the steel together with the cooling rate and the prior ther-mal history. The mechanical properties of steel, such as strength and toughness,are strongly correlated to the microstructure that is formed during cooling. Under-standing the various phase transformations is therefore of primary importance inorder 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 phasetransformations can be distinguished: diffusional and diffusionless transformations.Excellent overviews of the characteristics of the diffusion-controlled transformationsfrom austenite to ferrite and pearlite are given in references [7, 10, 11]. For the steelsstudied in this thesis, the final microstructure is mainly controlled by the austeniteto bainite and martensite phase transformations. In this section the most importantcharacteristics of these phase transformations are discussed.

2.3.1 Martensitic transformation

The martensitic transformation is a diffusionless first-order phase transformationduring which the lattice distortion can be described by a combination of shears[12]. It involves a cooperative and almost simultaneous movement of atoms fromparent to product phase. Sometimes this type of phase transformation is also calleda displacive or shear transformation.

Martensitic transformations can occur in many metals provided the conditionsare such that diffusion-controlled transformations are prevented [7]. The transi-


Figure 2.4: The Bain correspondence: (a) two fcc austenite cells to show that atetragonal cell can be outlined in austenite (b) Bain strain of this cell with axialratio

√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 importantmartensitic transformation because of the technological importance of hardenedsteel. About a century ago, the martensitic microstructure in steel was first ob-served with a microscope by the German metallurgist Adolf Martens. Nowadays,martensite is the term commonly used to describe the transformation product in asystem where the phase transformation occurs in a displacive manner, and corre-spondingly, the martensitic transformation is the generic name for these transitions.

Owing to the diffusionless (displacive) character of the transformation, themartensite has exactly the same composition as its parent austenite. In most prac-tical cases the amount of carbon exceeds the solubility in ferrite, and consequentlythe martensitic phase in steel can be simply described as a super-saturated solutionof carbon in the ferritic phase, in which the carbon content leads to a tetragonaldistortion of the lattice. The correspondence in lattice structure between austen-ite and martensite was first pointed out by Bain. He showed that a body-centeredtetragonal (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 unitcell into a martensite cell is known as the ’Bain strain’. There is a contraction alongthe 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 diffuseand consequently the carbon atoms are trapped in the octahedral sites of the body-


centered cubic (bcc) lattice structure. The equilibrium solubility of carbon in thebcc lattice is exceeded, and as a consequence of the transformation mechanism thisresults in the bct structure, the distorted form of the bcc structure. Accordingly, thetetragonality of the bct structure increases with increasing carbon concentration ofmartensite. Owing to the high carbon content, the martensitic crystal structure isactually a meta-stable phase. In case the temperature is increased (the martensiteis heated) the carbon atoms become mobile and will diffuse from the martensitelattice to form carbides. During this so-called tempering, martensite decomposesinto a mixture of ferrite and cementite, with concentrations according the Fe-Cphase diagram.

In steels, the transformation of an austenitic microstructure to a martensiticmicrostructure usually takes place due to a decreasing temperature rather than asa function of time, which is referred to as an athermal transformation. A necessarycondition for the transformation to start is that the free energy G of martensite(α′) is lower than that of austenite (γ). Since additional energy, such as surface en-ergy and strain energy, is required for the transformation to take place, martensitictransformations do not begin at T0, where ∆Gγ→α′

= 0, but start at a lower tem-perature, the martensite-start temperature Ms. The free energy change ∆Gγ→α′

,which corresponds to the temperature difference between T0 and Ms, constitutesthe driving force for the transformation [12].

Besides the thermodynamics, which determine the available driving force fortransformation, the occurrence of a phase change is governed by the kinetics. Thekinetics of a martensitic transformation depend solely on nucleation, because thegrowth of a martensitic crystal usually occurs rapidly. It is well known that themechanism of growth is displacive, i.e. the growth takes place by the cooperativemovement of atoms. How the phase nucleates, however, is even today not com-pletely understood. This is mainly due to the great speed of formation, whichmakes 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 duringcooling is believed to take place at structural imperfections in the parent phaseand these pre-existing embryos (defects) are stimulated to grow into martensitecrystals at different degrees of undercooling below Ms; they have different energybarriers to activation [11]. Since growth is very fast, each nucleation event almostinstantaneously leads to the formation of a certain volume of the new phase. Be-cause of the different energy barriers to nucleation the volume fraction of martensitevaries only with the degree of undercooling expressing the athermal character ofthe transformation. Although the exact nature of the nucleation sites is not com-pletely 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 ap-propriate slip planes as the interface moves [11]. A more detailed description of theinterface motion including the generation of acoustic emission during this processin addressed in chapter 7.


2.3.2 Bainitic transformation mechanism

Bainite is the transformation product that forms below the pearlite formation tem-perature and above the martensite-start temperature. Due to additional changesduring and/or after the phase transformation a strong diversity exists in the mi-crostructural appearance of bainite. Although the bainitic reaction has been studiedextensively since the discovery of bainite in 1930 by Bain and Davenport, there isstill no general agreement about the mechanism of bainite formation [13, 14]. Twoalternative models have been proposed to describe the transformation kinetics: thediffusional model and the displacive model [15, 16, 17].

The diffusional model assumes that the transformation mechanism involves re-constructive diffusion of substitutional atoms, i.e. ferrite and cementite are preci-pitated from austenite by diffusive mechanisms [16]. This mechanism thus is similarto the formation of pearlite, although the typical lamellar structure of pearlite doesnot occur.

In the displacive model, the atomic rearrangements during bainite formationare believed to occur in a diffusionless fashion as far as the substitutional atomsare concerned [15, 17]. In fact, it is assumed that a plate of bainite forms accordingto a martensite-like mechanism without diffusion, followed by a rejection of excesscarbon into the remaining austenite which subsequently forms carbides. It shouldbe emphasized that in this model the growth occurs without diffusion, whereas thenucleation at the austenite grain boundaries might still require some partitioningof carbon [17, 18].

For martensite, crystallographic analysis can be used to verify that the transfor-mation takes place without diffusion since the local compositions before and afterthe transformation are equal. Bainite, however, forms at somewhat higher temper-atures, at which the carbon can still escape from the bainitic ferrite. This impliesthat by crystallographic means it is difficult to determine the nature of the bainiticreaction mechanism.

Regarding the morphology of bainite, two main structures can be identifiedwhich are called upper and lower bainite. Upper bainite forms at a relativelyhigh temperature, usually in the range of 450 – 600 C, and lower bainite between300 and 450 C. The change in morphology with transformation temperature isa direct consequence of the change in diffusivity of carbon. In the case of upperbainite the diffusivity of carbon is relatively high and therefore carbide precipitatesfrom the carbon-enriched austenite between the ferrite plates. For lower bainitealso carbide precipitation within the bainitic ferrite occurs owing to the decrease indiffusivity of carbon in austenite at lower temperatures; the diffusivity in bainiticferrite is much higher. Therefore, two kinds of cementite can be recognized in lowerbainite: cementite particles that precipitated from the carbon-enriched austeniteand cementite particles that precipitated from supersaturated ferrite. The latterprecipitation shows a strong resemblance with the tempering of martensite. Thelayers of carbide in lower bainite are usually extremely fine compared with those inupper bainite. Consequently, a steel with a lower bainitic microstructure is tougher


than a steel with an upper bainitic microstructure. Moreover, lower bainite isstronger since the precipitates are finer.

The amount of cementite particles in bainite does not only depend on the carbonconcentration but also on the alloying elements. For example, by increasing thesilicon concentration the cementite precipitation can be greatly retarded becausesilicon has a negligible solubility in cementite. This can improve the toughnessof bainitic steels, and is also of importance in the production of TRansformationInduced 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 dis-placive (martensitic) transformations and how to define such a transition. Althoughit is still not completely understood how growth of the new (martensitic) phaseoccurs, there is general agreement between researchers that a displacive transfor-mation involves the cooperative movement of atoms that causes a shape change.Despite this theoretical agreement, it is sometimes very difficult to prove that atransformation is diffusionless, especially in the case of bainite discussed above.Usually the high transformation kinetics at the relatively low transformation tem-perature 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 tem-perature is low is open to question. The best known and well recognized proof forthe displacive character of a transformation is probably the observation of surfaceupheavals on a polished surface.

Recently, it was argued by Clapp [19] that acoustic emission is the best test toprove the displacive character of a phase transformation. This is based on the factthat the emission of acoustic energy is strongly related to the coordinated movementof atoms. In contrast to the above mentioned metallographic test, acoustic emissionoffers a relatively simple in-situ test to investigate whether or not the transformationis displacive. In this respect it is quite surprising that acoustic emission has onlyrarely been used to monitor phase transformations, especially since it has beenknown for a long time that the formation of martensite in steel is accompaniedby acoustic emission. In view of above arguments, acoustic emission monitoringduring bainite formation will give valuable information about its transformationmechanism.

A general study of acoustic emission during phase transformations in carbonsteels 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 transformedvolume. How the acoustic emission technique can be used to follow the progress ofmartensitic transformations is discussed in chapter 6. Finally, in chapter 7 the char-acteristics of the acoustic waves generated during the bainitic and the martensitictransformation are studied.

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., Nonde-structive 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).


[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 Maatschap-pij, Delft (1995).

[10] J.W. Christian, Theory of Transformations in Metals and Alloys, 3rd ed., El-sevier Science, Oxford (2002).


[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


[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 measurethe acoustic emission signals generated during phase transformations in steel. Adescription of the AE system used is addressed in section 3.1, including a discus-sion of noise suppression precautions, attenuation due to waveguides and sourcelocation. In most experiments an arc welding device or a thermo-mechanical (weld-ing) simulator was used to apply a thermal cycle to the studied specimen, whichupon cooling was monitored by means of acoustic emission. These two heat cyclingmethods are characterized by a relatively high cooling rate due to the metallic heatconduction. The set-ups for AE measurements during welding and during thermalcycling in the thermo-mechanical simulator are described in section 3.2 and section3.3 respectively. A few experiments were performed using a conventional dilatome-ter in order to facilitate comparison between acoustic emission and dilatometry. Insection 3.4 the set-up and measurement procedure for experiments with the AEsystem connected to the dilatometer is given. Furthermore, some experiments wereperformed using a furnace in order to study large-sized samples, which give a bettersignal to noise ratio. During quenching in a salt bath the acoustic emission wasmeasured as described in section 3.5. At the end of the chapter an overview is givenof the steels studied in this thesis.

3.1 Acoustic emission system

In order to measure the AE signals a PAC (Physical Acoustics Corporation) AEsystem was used [1]. This is a fully digital two-channel acoustic emission systemthat performs AE waveform and AE signal parameter measurements and stores anddisplays the resulting data. The main components are the AEDSP board, which isintegrated in a standard computer, and the Mistras software.

In addition to the two AE channels, the system has two input connections forexternal signals, which are known as parametric 1 and parametric 2. The externalsignal 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 theacquired 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 withthe magnitude and rate of occurrence of acoustic emission events [2, 3]. Dependingon the rate of occurrence of AE signals at the sensor, two types of signals can bedistinguished: burst and continuous emission. A burst-type signal is thought ofas a signal from a single, discrete event. When the rate of occurrence is high, theindividual burst signals overlap and combine to form continuous emission. It shouldbe realized that sometimes the distinction can be rather arbitrary, for instancewhen the successive individual signals are visible in the overall continuous wave. InFig. 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: timedriven 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 isperformed, 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 thebackground noise is defined by [4]

U2(t) =1

τ

t+τ∫

t

U2p(t′)dt′ − U2n (3.1)

where Up(t) is the voltage output at the pre-amplifier and τ a time constant usuallychosen as τ = 0.1 s. The amplification of signals is standard 40 dB (100 ×) andthroughout this thesis the amplified values are displayed; they are not convertedback to the voltage output at the transducer. For measurements of the rms voltagewith 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 mostimportant settings and parameters: time driven data for continuous emission andhit driven data for burst emission.

the 40 dB scale. The rms voltage is measured with a resolution of 0.02 mV relativeto a background noise level (U 2n )

1/2 of 0.24 to 0.28 mV. It should be mentionedthat early measurements were performed with a resolution of 0.2 mV.

For characterizing burst-type AE signals a threshold level is set somewhat abovethe background noise level; the chosen threshold value depends on the amplitude ofthe AE signals and the desired amount of data acquired. If the AE signal exceedsthe threshold in either positive or negative direction a so-called hit is recorded. Atypical example of a corresponding waveform is shown in Fig. 3.3. Such a waveformis produced by joining many single points called samples. They correspond to singlemeasurements at constant time intervals. The system can sample with 1, 2, 4 or8 MHz; normally a sample rate of 4 MHz is sufficient for measuring signals withfrequencies 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 amountof memory, and interpretation of the waveforms themselves is difficult. Thereforethe most important features of each waveform are determined, which are called theAE parameters. These allow an easier comparison with other results. The mainsignal parameters describing the waveform are the signal amplitude, the signal risetime and the signal duration. They are illustrated in Fig. 3.3. The time of thefirst threshold crossing is called arrival time and is needed for the calculation of thelocation of the AE event. The parameter ’counts’ gives the number of times thesignal crosses the threshold. The amplitude is the peak voltage of the AE waveformand can be a useful measure of the signal size. The time from the first threshold


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 tofilter out noise signals, since these have usually very short rise times. The durationis the time from the first to the last threshold crossing.

The above-mentioned signal parameters cannot simply be related to the charac-teristics of the source because they strongly depend on the threshold setting and thesystem timing parameters. For so-called hit driven data the measurement processbegins when the voltage signal from the pre-amplifier first crosses the threshold.This threshold crossing triggers certain timers which determine when the hit haspassed and the system is ready for the next hit, hence, this is not trivial. The timersthat 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 themeasured AE parameters described above, such as rise time, duration and peakamplitude.

The function of the PDT is to enable the determination of the true peak am-plitude and rise time of the AE waveform. The PDT circuitry is triggered by thefirst maximum after the threshold crossing and retriggered if a new maximum ismeasured within the set PDT. The function of the HDT is to enable the systemto determine the end and thus the duration of the waveform. The HDT circuitryis (re)triggered by the threshold crossing(s). When no threshold crossing occurswithin 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) separatehits will not be treated as a single hit; but also not too short to avoid fragmenta-


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 ofreflections and late-arrival parts of the AE signal. The HLT circuitry is triggeredby 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 mea-sured hit driven data cannot simply be compared with the results obtained byother researchers because they depend strongly on the system settings. In thisthesis mainly time driven data (Urms) is used, whilst in chapter 7 hit driven datais 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) differentialAE sensor (PAC model WD). The frequency response of this sensor is shown inFig. 3.4. In order to determine this sensitivity of the sensor, a calibration wascarried out according to the face-to-face technique, which is based on the voltageoutput per unit of pressure input.

The signal from the sensor is amplified by 40 or 60 dB with a low-noise broad-band (100 – 1200 kHz) pre-amplifier (PAC 1220A). The measured rms voltage dueto the electronic noise of the sensor and the pre-amplifier is 0.24 to 0.28 mV. Al-though the sensor has a differential design, the shielding to guard against externalEMI noise is not sufficient under all circumstances. For example, it was found thatin the case the pre-amplifier was positioned closer than approximately 20 cm to thecomputer monitor, the background noise level increased significantly. Positioningof the sensor close to a computer monitor also leads to an increase in backgroundnoise, typically to a value of 0.5 to 1 mV.

Usually, a noise survey will be performed before the main experiment is carried


out. Many noise problems will become apparent during setting up of the experimentand can be dealt with before data acquisition starts. Good observation and a carefuland systematic approach are very important for diagnosis of noise problems. Forexample, if the noise has a continuous character, the cause is probably an electricalproblem (e.g. bad shielding of the BNC cable). On the other hand, if the noisepattern is rather irregular, the noise source is for instance the descaling of an oxidelayer during cooling of the specimen. At best, the background noise is just theelectronic noise of the pre-amplifier and the sensor.

An essential requirement in mounting a sensor is sufficient acoustic couplingbetween the wear-plate of the sensor and the surface of the specimen. Beforemounting, the wear-plate and the surface need to be cleaned. In case the surfaceof the specimen is not smooth, it has to be polished/ground with silicon-carbidepaper. Then the couplant, e.g. vacuum grease, can be smeared on the wear-plate. Subsequently, the sensor can be pressed on the surface of the specimen. Thecouplant layer should be thin and fill all the gaps caused by the surface roughnessto ensure a good acoustic transmission. The sensor should also be attached firmlyto the mounting surface at all times during operation. This can be achieved by aholding device such as magnetic hold-down or just tape. Electrical contact betweenthe sensor case and the structure needs to be avoided.

In the case the specimen becomes very hot or very cold, a waveguide is requiredfor two reasons. The temperature range in which the sensor can operate is typically−40 – 180 C . Secondly, commonly used couplants may become unstable at veryhigh or low temperatures. Waveguides are also necessary when the size of thespecimen is smaller than the diameter of the sensor or when access to the specimen isdifficult. These reasons for using a waveguide are especially relevant for laboratorystudies. A waveguide is typically a metal rod which conducts the acoustic signalfrom the specimen to the sensor. One end is designed for acoustic coupling with thespecimen; the other end is usually conical to accommodate the mounting of an AEsensor. To minimize attenuation, the diameter of the waveguide should be as largeas possible, and the waveguide should have an acoustic impedance similar to thatof the specimen. Furthermore, it is preferred that the joints are made by weldingto obtain a good acoustic conductance.

3.1.2 Attenuation due to waveguides

Preliminary measurements showed that the use of waveguides reduces the measuredAE energy in comparison with the case where the sensor is mounted directly ontothe workpiece. The results indicated that the attenuation is primarily governed bythe diameter of the waveguide and the quality of the welded joints.

To investigate the attenuation due to the presence of a waveguide in a quantita-tive manner, three waveguides with different diameters (d = 1, 2 and 4 mm) madeof plain steel were welded onto the workpiece as shown in Fig. 3.5. The length ofeach waveguide was 100 mm; the disc-shaped mounting plates for the sensor wereidentical (diameter = 24 mm, thickness = 10 mm).


Figure 3.5: The experimental set-up used to measure the attenuation due to wave-guides: (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.

t [s]

0 1 2 3 4 5

Urm

s [m

V]

0

2

4

6

8

10

12

14

plated = 4 mmd = 2 mmd = 1 mm

Figure 3.6: The rms voltage as a function of time for a sensor on three differentwaveguides and another sensor mounted on the plate.


π d2/4 [mm2]

0 4 8 12 16

norm

aliz

ed d

etec

ted

ener

gy

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

d = 1 mmd = 2 mmd = 4 mmd = 2 mm, l = 200 mmd = 4 mm, RVS

Figure 3.7: The detected energy for each waveguide normalized to the detectedenergy on the sensor that was mounted directly onto the plate.

As a highly reproducible source of continuous acoustic emission the martensitictransformation, 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 ofmartensite in the weld was measured with two identical sensors: sensor-1 mountedon a waveguide and sensor-2 mounted on the plate. With both sensors the rms volt-age of the generated acoustic emission signals was measured as a function of time.Repeated measurements with a certain waveguide did not reveal any significantsystematic differences.

The results obtained for the three waveguides are plotted in Fig. 3.6. As ex-pected, it can be seen that the intensity of the signal decreases with decreasingwaveguide diameter. For each waveguide the AE energy detected on sensor-1 wasdetermined from the area under the peak in a plot of U 2 against t, i.e. the in-tegrated value ∫ U2dt [5, 6]. The calculated values, normalized to the AE energydetected on sensor-2, are plotted against 14πd

2 in Fig. 3.7. It can be seen that theacoustic conductance of a waveguide is proportional to the cross-section. Althoughtheoretical predictions for comparison do not exist, the relationship observed seemsreasonable in view of the fact that a similar relationship exists for electric and heatconduction.

In order to investigate the influence of the welded joint between the waveguideand the plate on the measured AE energy, the measurements were repeated withidentical waveguides spot welded on the plate. The results from this second seriesof experiments indicated that the quality of the welds have a quite strong effect onthe measured signal, which is expressed by the error bars in Fig. 3.7.


t [s]

0 1 2 3 4

U 2 /

U 2 m

axim

um

0.0

0.2

0.4

0.6

0.8

1.0

plated = 4 mmd = 2 mmd = 1 mm

Figure 3.8: The mean square voltage U 2 (normalized to the maximum value) asa function of time for the sensor mounted on the plate and the sensor mountedon three different waveguides. This shows that the attenuated signals and thenon-attenuated signal from Fig. 3.6 can be mapped onto each other by a singlemultiplication factor.

To examine the influence of the waveguide material, a measurement with astainless steel rod (d = 4 mm, RVS) was performed. The result obtained was notsignificantly different from the result obtained for the steel rods discussed above(see Fig. 3.6). Furthermore, the measured signal did not change significantly whenthe length of the waveguide rod was increased to 200 mm (d = 2 mm, l = 200mm), which indicates that the waveguide diameter is much more important thanits length. Using the same set-up with the two sensors, it was also found that athick layer of grease between the sensor and the plate has a strong attenuationeffect on the measured signal.

With respect to the power (U 2) of the signals shown in Fig. 3.6, it is interestingto mention that for all measurements, the signals obtained can be mapped onto eachother by a single multiplication factor as shown in Fig. 3.8. This indicates that theattenuated signals contain the same information of the martensitic transformationas the non-attenuated signal.

3.1.3 Source location

By using multiple sensors the position of an AE source can be determined. Com-putation of the source location is possible by using the wave velocity of the acoustic


waves and the arrival times at different sensors. An AE wave propagates in con-centric 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. InAE terminology a detected AE wave with one sensor is called a hit; a located hitis called an event.

The simplest form of source location is linear (one dimensional) location, whichcan be used on rods and long thin specimens. Linear location requires only twosensors to locate a source, whereas planar location requires three sensors. In thecase of linear location, ∆t is defined as the time interval between detections of thesame waveform at the two sensors. The distance difference between a source andthe two different sensors is equal to the product of ∆t and the wave velocity cL,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 respec-tively. The wave velocity of the longitudinal waves in steel was measured, yieldingcL = 5.2± 0.3× 103 m/s.

In order to investigate the practical use of source location, many pencil leadbreaks (used to cause an artificial AE event) were made on the middle of a 800mm long steel rod (d = 25 mm) with sensors on both ends (see Fig. 3.9). Forsuch an artificial source midway (± 2 mm) between the sensors, ∆t is theoreticallyzero; in the ideal case both sensors are hit at the same time. However, the resultin Fig. 3.9 shows a considerably broad distribution of located events. It can beseen that most events (more than 50 %) are calculated to be located within adistance 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 forlaboratory studies and therefore the source location is not pursued further in thisthesis. The accuracy of source location is mainly governed by small differences insensitivity of the transducers employed. It can be visualized that for a sensor witha better sensitivity the voltage signal U is relatively high, which implies that thethreshold is crossed a fraction earlier by the same displacement input; this leadsto an inaccurate source location. Also uncertainties in the measurement of arrivaltimes 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 producedby an electric arc, which operates at an arc length of a few millimeters between anon-consumable tungsten electrode and a metal workpiece. Under normal condi-tions the arc voltage is in the range of 10 – 20 V and the arc current ranges from20 A to 300 A.

The experimental set-up used to study the AE signals generated during GTAwelding 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 detectedas a function of position on a 800 mm steel rod.

(1)

(2)

(4) (3)

(5)

(6)

(7)(8)

Figure 3.10: Schematic drawing of the experimental set-up used for AE measure-ments 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.


unit provided the welding current. The torch can be adjusted in the vertical di-rection 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 toprotect the electrode, the arc and the weld pool from the surrounding atmosphere.During welding, arc voltage and arc current were continuously measured as a func-tion of time. The travel unit, power supply and measuring system were controlledby a computer using Labview (National Instruments), a graphical programminglanguage. The commonly used welding conditions are given in Table 3.1.

Table 3.1: Commonly used welding conditions.

electrode material tungsten, 2% thoriapolarity electrode negativeelectrode diameter 2.4 mmelectrode tip angle 60

arc voltage 10 – 12 Varc current 30 – 100 Atravel speed 0 or 2 mm/sarc length 2.0 – 3.0 mmshielding gas argonshielding gas flow rate 6 l/minshield cup inner diameter 16 mmelectrode 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 hassolidified after welding is referred to as the weld metal. Due to the heat flow inthe 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 weldmetal have a different microstructure compared to the parent material.

The high cooling rates of welds determine that the phase transformations donot usually occur under equilibrium conditions. The main variable in the weldingprocess that determines the cooling rate is the heat input H (J/m), which is afunction of welding power and travel speed v according to

H = ηV I

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 coolingrate during welding can be estimated using Rosenthal’s equation for bead-on-platewelds (the 3-dimensional heat flow case) [7], given by

dT

dt=

2πλv(T − T0)2

H(3.4)


Figure 3.11: The CCT diagram of steel 42CrMo4 determined by the Max-Planck-Institut fur 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 fromEq. 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 duringcooling. It follows directly from Rosenthal’s equation that a higher/lower heatinput 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 Cand 500 C in the range of 150 C/s to 600 C/s.

For welds made under static conditions (spot welds), the cooling rate cannotbe predicted by theory; it depends on the heat input, the welding time and thesize (heat capacity) of the workpiece. However, it can be easily visualized that forequal heat inputs the cooling rate after spot welding is higher than the cooling rateduring travelling arc welding.

To predict the solid state phase transformations that can occur in a weld whichis continuously cooled, it is common practice to use a continuous cooling trans-formation (CCT) diagram. As an example the CCT diagram of steel 42CrMo4 isshown in Fig. 3.11. In order of increasing undercooling below the Ac3 temperature,the following transformation products may be formed: ferrite, pearlite, bainite andmartensite. It should be noted that the positions of the C-curves in a CCT dia-gram, which indicate the start of the transformation, do not depend only on the


chemical composition (alloying elements) of the steel, but also on the austenitizingtemperature (the maximum temperature during welding Tp) and the austenitizingtime, which both determine the austenite grain size. This means that a numberof possible transformation products may appear in a weld at ambient tempera-ture. Since the peak temperature varies with distance from the weld center, withinthe HAZ different sub-zones can be distinguished with characteristic microstruc-tures. 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 peaktemperature Tp is between 1100 C and the melting temperature Tm. This part willbe 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 metalwill be completely austenitized. Due to nucleation and growth processes the austen-ite grain size is small in comparison with the unaffected parent metal. Because ofthe relatively low peak temperature, the austenite grain sizes remain small.

(4) Partially austenitized zone (Ac1 < Tp < Ac3): In this sub-zone a partial trans-formation takes place; only the pearlite in the base metal is austenitized. Theaustenite 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 ofcarbides and other aging effects.

The steels studied have a relatively high carbon content, and under normalwelding conditions with the cooling rates given above, usually only martensite isformed 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 dependson the austenitizing temperature, the austenitizing time and the cooling rate atthat 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 andanalyzed by the AE system. The transducer was mounted on the workpiece with amagnetic hold-down, and vacuum grease was used to achieve good acoustic couplingbetween workpiece and transducer. In normal welding situations a waveguide willbe required between the workpiece and the transducer in order to protect the trans-ducer from heating up above its operating temperature. However, in the presentlaboratory set-up the welding times were relatively short, and consequently the tem-perature 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 steelsin the form of plates with dimensions 250 × 200 mm2 and a thickness in the range


(1)(2)(3)(4)

(6)

(5)

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-grainedzone; (3) the fine-grained zone; (4) the partially austenitized zone; (5) the temperedzone; (6) parent metal.

of 5 to 10 mm. Prior to welding, the workpiece was cleaned with acetone andattached to a copper base plate to assure good electrical contact. The angle ofthe electrode tip was checked regularly during the production of the welds becauseit has a significant influence on the heat transfer. Welds were made under staticconditions (spot welding) and under stationary conditions (travelling arc).

During welding AE noise signals are generated, which are caused by the interac-tion between arc and plate. In this study an investigation of the exact origin of thisarc noise was not attempted because there are many forces involved in the weldingprocess which can contribute to the observed noise. Welds made on plain carbonsteel revealed that the arc noise increases with welding power. The rms voltage ofthe noise signals was in the range of 0.5 mV to 1.2 mV for arc currents of 20 to 120A; 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 compositionof the welded material. Tests with the sensor detached from the plate but close tothe arc showed that the high electromagnetic fields, which are present around thearc, do not give rise to EMI noise onto the sensor.

Experiments using different specimen configurations showed that the resultsobtained are independent of plate size. Moreover, the results did not change sig-nificantly when using a sample in the shape of a rod, in which case the distancebetween source and sensor was 800 mm. This indicates that in the laboratory set-up, where distances are relatively short, attenuation of the signals due to geometricspreading or damping does not play a significant role. Furthermore, the reflectionsof waves are presumably damped sufficiently before they arrive at the sensor.


(1)

(2)

(4)(3)

(6)(7)

(8)

(9)

(5)

Figure 3.13: Schematic overview of the experimental set-up for AE measurementsusing the Gleeble thermo-mechanical simulator: (1) specimen; (2) grips; (3) wave-guide; (4) transducer; (5) pre-amplifier; (6) AE system; (7) dilatometer; (8) ther-mocouple; (9) Gleeble system.

3.3 Thermo-mechanical simulator

A thermo-mechanical simulator is a combined thermal and mechanical system forphysical simulation of processes that can occur during the production or processingof steel. In the experiments described in this thesis the thermal system of theGleeble 1500 (Dynamic Systems Inc.) was used. The Gleeble heating system canheat specimens at rates of up to 10000 C per second, which is fast enough forwelding simulation. The heating system of the Gleeble is based on the electricalresistance of the specimen, and therefore this heating method is called resistanceheating (or Joule heating). Other heating methods commonly used in the laboratoryare induction heating and furnace heating, which are described in section 3.4 andsection 3.5 respectively.

For resistance heating the current density is approximately uniform throughoutthe 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 theexposed surfaces is negligible compared to the heat loss to the water cooled coppergrips via (metallic) conduction. Therefore, isothermal planes exists perpendicularto the specimen axis. The temperature profile as well as the natural cooling ratecan be controlled by the resistance of the grips and the shape of the specimen. Thenatural cooling rate is mainly determined by conduction through the water cooledjaws and to some extent also by convection and radiation through the surface ofthe specimen in the vacuum or inert gas atmosphere.

Measurement procedure

The experimental set-up used for AE measurements in the Gleeble is schematicallyillustrated in Fig. 3.13. In general, the specimen under study was subjected tocontinuous cooling from high temperatures during which the generated acoustic

Thermo-mechanical simulator 39

emission signals were measured. Optionally, the dilatation of the specimen wasmeasured simultaneously with a CCT dilatometer. The Gleeble offers the possibilityto export data, such as temperature and dilatation, to the parametric inputs of theAE system.

The specimen under study was mounted in the copper grips of the Gleeble sys-tem and a K-type (chromel/alumel) thermocouple was spot-welded onto the middleof the specimen for monitoring and controlling the temperature during the thermalcycle. With a thermocouple welder the thermocouple wires, with a diameter of 0.2mm, were welded about 1 mm apart in the same cross-section perpendicular to thespecimen axis. This is required to avoid a voltage potential along the specimen axis,which can lead to an error in the temperature measurement. During positioningof the specimen in the grips uniform tightening is required to avoid induction of amechanical stress or bending moment within the specimen. Moreover, the specimenshould be tightened enough to minimize the thermal and electrical resistance at theinterface of the specimen and the grips; this is important when a high cooling rateis desired. For further information about the mounting of a specimen, see the usersmanual [10].

A waveguide was welded to one end of the specimen to transport the acousticemission waves to the transducer and to prevent the transducer from overheating.On the conical end of the waveguide, a wideband (100 – 1000 kHz) differentialpiezo-electric transducer was mounted and vacuum grease was used in between toachieve good acoustic coupling. The transducer was connected to a pre-amplifierwith 40/60 dB gain and 100 – 1200 kHz band-pass filter. From there the AEsignals were transmitted to the analyzing system (Mistras 2001) for recording andanalysis. The dilatometer was used to monitor the radial dimension length changeof the specimen.

The specimen materials employed were medium carbon and low carbon steels.The specimens for AE and dilatometric measurements were machined from steelrods 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 con-ducted using specimens with an effective free span l1 of approximately 10 mm anda diameter d1 of approximately 5 mm (see Fig. 3.14). Due to the thermal gradientalong the axis of the specimen, the austenitized volume depends on the austeni-tizing temperature Ta, which is controlled by the thermocouple in the middle ofthe specimen: the higher the austenitizing temperature, the larger the austenitizedvolume. The thermal gradient is similar to the thermal gradient that exists in theHAZ during welding. Thus different zones can be recognized in the specimen justas in a weld, see section 3.2.

Prior to measuring, the chamber of the Gleeble system was evacuated to about0.2 mbar, twice filled with argon gas and purged, and again filled with argon gas.These precautions were taken to minimize decarburization and oxidation of thespecimen at high temperature. In a typical measurement, the specimen was electri-cally heated in 20 seconds to the austenitizing temperature Ta, austenitized for 20seconds, and continuously cooled to ambient temperature, during which the acoustic


Figure 3.14: Schematic illustration of the thermal gradient in the specimen duringaustenitizing and the design parameters that determine the thermal gradient andthe natural cooling rate when the heating stops: (1) part of specimen in betweenthe grips, indicated by the length l2; (2) austenitized volume, indicated by thestriped region.

emission was measured. The heating system of the Gleeble generates (mechanical)noise, which makes it impossible to study the acoustic emission generated duringisothermal holding. For the same reason controlled cooling, slower than naturalcontinuous cooling to obtain for example bainite formation, was not feasible.

In order to change the natural continuous cooling rate, the specimen designand free span can be changed. The free span l2, the distance between the coppergrips, has a relatively small influence on the cooling rate. The cooling rate is to alarge 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 notedthat noise resulting from oxidation may take place. Since it is difficult to obtaina high vacuum, oxidation of the specimen will always occur to some extent. Inparticular the descaling of oxidation products results in noise. The scale (ironoxide) has a different coefficient of thermal expansion than the steel, and beingbrittle it tends to fracture during cooling. This results in burst-type noise, whichis usually clearly visible as scatter superimposed on the continuous rms voltage.This is most prominent after repeating experiments with the same specimen. Asin the case of welding, high electromagnetic fields are present in the vicinity of thespecimen during heating, however, they do not give rise to EMI noise at the sensor.

In comparison with the welding experiments described before, the attenuationof the generated AE signals is much higher since a waveguide is used. The wave-

Dilatometer 41

(4)(3)

(6)(7)

(5)

(1) (2)

(8)

Figure 3.15: Schematic drawing of the experimental set-up for simultaneous mea-surements of acoustic emission and dilatation using the Bahr 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 resultsin a full fusion weld with good acoustic conductance. Although the attenuationdue 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 resultsfrom welding experiments because the shape of the specimen used for the Gleebleexperiments is significantly different, and may cause attenuation by itself.

3.4 Dilatometer

A dilatometer can measure longitudinal length changes with high accuracy (≈ 0.1µm), and can be used to monitor the volume change of a sample which is subjectedto a thermal treatment. Measuring the dilatation is a commonly used method tostudy phase transformations in steel. To facilitate comparison between acousticemission and dilatation measurements, the AE system was connected to the Bahr805A/D dilatometer in order to measure both signals simultaneously. In this sectionthe combined experimental set-up is described.

In Fig. 3.15 a schematic drawing of the set-up is given. The samples used in thedilatometer were 10 mm long with a diameter of 5 mm. The sample was clampedbetween quartz rods, and a thermocouple was spot welded onto the middle of thesample for measuring and controlling the temperature. A waveguide was weldedonto the sample as shown in Fig. 3.15 to transport the AE waves to the sensor. Inview of the attenuation a thick waveguide wire is desired, however, a wire with adiameter 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 ofthe sample.


(1)(2)

(4)(3)

(6)(7)

(5)

(8)

(9)

Figure 3.16: Schematic illustration of the experimental set-up for AE measurementsusing 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, whichis 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 thespecimen. Induction heating is usually faster than furnace heating but slower thanresistance heating. For induction heating it takes some time to have a uniformtemperature from the surface to the center of the specimen.

In the experiments, the sample is heated to an austenitizing temperature of 900C, austenitized for 2 minutes, and subsequently continuously cooled to ambienttemperature in a helium atmosphere. The use of high pressure gas spray to coolthe sample had to be avoided because it results in an additional significant noisesignal.

An important difference between dilatometry and acoustic emission is that themeasured dilatation is proportional to the amount of volume transformed, whereasthe AE signal is proportional to the transformation rate [5, 6]. Moreover, the AEsignal is measured relative to a background noise level, whereas for dilatometryonly the sample size determines the magnitude of the signal. In the case of acousticemission, a high volume transformation rate (determined by the volume and thecooling rate) is important to obtain a good signal to noise ratio. In addition, thenecessity of using a waveguide as explained above has major consequences for thedetectability of the generated signals.

It appeared that especially the small-sized sample and the small waveguidelimit the possibilities of AE measurements using the dilatometer. In an attempt tocircumvent these problems, additional experiments were performed using a conven-tional convection furnace.

Furnace 43

3.5 Furnace

With a furnace large-sized samples can be austenitized, and a relatively thick wave-guide rod can be used. In Fig. 3.16 a schematic overview of the experimental set-upis given. Typically, a sample with a diameter of approximately 15 mm and a lengthof 40 mm was employed. A waveguide was welded onto the sample; this waveguideconsisted of a stainless steel rod with a diameter of 5 mm and a disc-shaped plateon which the AE sensor was mounted.

To measure the temperature inside the sample with a special chromel/alumelthermocouple, a seven millimeter deep hole (2 mm diameter) was drilled in thesample. With a data-logger the temperature was recorded and simultaneously theacoustic emission (AE) was measured with the AE analyzing system; both with anacquisition rate of 1 Hz.

In order to austenitize a sample with the waveguide attached, the door of thefurnace was modified such that the sensor was outside the furnace. After austeni-tizing the sample at 900 C for 20 minutes, the sample was taken out of the furnaceand quenched in the fluid salt at a temperature in the range of 300 – 500 C. Duringcooling 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 anumber of steels, mainly focusing on the commercially available medium carbonsteels C45 and 42CrMo4. The compositions of all steels studied in this thesisare 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 steelreceived from Corus.

Table 3.2: Chemical composition of steels (wt%). (- not detected)

Steel C Si Mn P S Cr Mo V Cu NiC45 0.44 0.27 0.70 0.022 0.010 0.20 - - - 0.0342CrMo4 0.42 0.24 0.66 0.024 0.007 1.08 0.18 0.005 - 0.06Fe360 0.07 0.32 0.74 0.030 0.031 0.14 0.02 0 0.22 0.10St52-3 0.17 0.61 1.33 0.015 - 0.03 - - - -St50K 0.32 0.28 0.76 0.013 - 0.05 - - 0.12 0.0742MnV7 0.41 0.36 1.64 0.029 0.030 0.26 0.05 0.08 0.27 0.1175MnSiCr 0.75 1.63 1.95 0.003 0.003 1.48 0.28 0.10 - -16MnSi 0.16 1.64 1.67 0.080 0.009 0.19 - - - 0.01C50 0.5 0.44 0.51 0.024 0.046 0.20 0.01 - 0.24 0.10C60 0.6 0.39 0.50 0.020 0.043 0.23 0.01 - 0.21 0.07C70 0.7 0.37 0.68 0.027 0.042 0.29 0.02 - 0.22 0.16C80 0.8 0.41 0.61 0.012 0.049 0.28 0.02 - 0.23 0.15


The heat treatments of the steels were carried out using the methods describedin this chapter; the various heat treatments were applied in order to generate abainitic and/or martensitic structure as indicated in Table 3.3.

Table 3.3: Schematic diagram showing which steels are studied, which method isused, and which transformations occur: B = bainite, M = martensite.

Steel Welding(section 3.2)

Simulator(section 3.3)

Dilatometer(section 3.4)

Furnace(section 3.5)

C45 M + B M + B42CrMo4 M M M BFe360St52-3 M + BSt50K M + B42MnV7 M75MnSiCr B16MnSi M + BC50 MC60 MC70 MC80 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., Non-destructive 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 zurWarmebehandlung der Stahle Teil II, Verlag Stahleisen G.B.H., Dusseldorf(1956).

[10] Users Manual Gleeble 1500, Dynamic Systems Inc. (1988).

45

Chapter 4

Acoustic emission

monitoring of phase

transformations in steel

Although it has been known for a long time that acoustic emission is generatedduring martensite formation, only a few acoustic emission studies of the varioustransformations taking place in carbon steels can be found in the literature. About30 years ago Speich and Fisher measured acoustic emission during martensite for-mation in an FeNi alloy [1]. They compared their AE results with electrical resis-tivity data and results from quantitative metallography. Some years later Speichand Schwoeble presented a general study of acoustic emission during the variousphase transformations in a wide variety of steels [2]. Their results showed thatthe intensity of acoustic emission generated during martensite formation decreasedstrongly as the carbon content of the steel decreased, becoming nearly undetectablein maraging steel. Furthermore, they showed that no acoustic emission was detectedduring the formation of ferrite, pearlite and bainite, which was attributed to thediffusion-controlled growth of these transformation products. Especially the latterobservation was important because the mechanism of growth of bainite was a sub-ject of debate in those years; two alternative models had been proposed to describethe transformation kinetics: the diffusional model and the displacive model (seesection 2.3.2). Despite the ongoing debate concerning the bainitic reaction mecha-nism no attempts have been made to verify the results of the study of bainite bySpeich and Schwoeble [2].

The monitoring of acoustic emission during bainite formation in various carbonsteels is the main topic of this chapter. In section 4.1 the results of AE experimentsduring various phase transformations in steel C45 are discussed. Subsequently,the results for martensite and bainite formation in steel 42CrMo4 are presentedin 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-alloyedsteel 75MnSiCr (section 4.4) are discussed. All steels were investigated using thetechniques described in chapter 3.

4.1 Study of steel C45

In this section the AE measurements on steel C45 are discussed. The chemicalcomposition of the steel is given in Table 3.2. Section 4.1.1 contains the results ofAE measurements during the bainitic and the martensitic transformation in steelC45 specimens carried out using the Gleeble 1500 thermo-mechanical simulator.The results of AE measurements during welding of steel C45 are discussed in section4.1.2.

4.1.1 Thermo-mechanical simulator experiments

The experiments were carried out as described in section 3.3. Based on the CCTdiagram of the medium carbon steel C45 and possible cooling rates during contin-uous cooling in the Gleeble, it is expected that bainite and/or martensite may beformed. For comparison, measurements under equal conditions were performed ona steel Fe360 specimen, in which neither bainite nor martensite will be formed.

In order to achieve the required cooling rate, different specimen configurationswere tested. After some preliminary experiments two types of specimen were madewith different design parameters (see Fig. 3.14). In order to obtain martensiteformation 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 bainiteand martensite formation during continuous cooling.

For all measurements, the specimen was electrically heated in 20 seconds to theaustenitizing temperature Ta, austenitized for 20 seconds, and continuously cooledto ambient temperature. Measurements were repeated several times to check thereproducibility of the experiments. It is known that repeated rapid heating of amartensitic microstructure can result in austenite grain refinement and this wasfound 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 effectsaturates and the measurements were found to become reproducible.

After thermal cycling, the specimens were cut in the middle longitudinally andthe microstructure was analyzed using an optical microscope (Olympus). Themicro-hardness of the specimens was measured by means of a micro Vickers hard-ness 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

t [s]

0 5 10 15 20

T [º

C]

0

100

200

300

400

500

600

700

800

900

M

Figure 4.1: The temperature of steel C45 (specimen type A) during cooling in theGleeble thermo-mechanical simulator (Ta = 830 C). M: inflection in cooling curvedue to the release of latent heat during martensite formation.

∆d [

µm]

10

20

30

40

50

60

T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

4

5

(a)

(b)

Figure 4.2: Plot of (a) the dilatation and (b) the rms voltage as a function oftemperature for steel C45 (specimen type A) austenitized at Ta = 830 C, coolingaccording Fig. 4.1.


Figure 4.3: Optical micrograph of the microstructure of steel C45 (specimen typeA) after continuous cooling (see Fig. 4.1) from Ta = 830 C.

experiments is discussed below. The specimen was austenitized at 830 C, and thecooling curve is depicted in Fig. 4.1.

For comparison with the AE data, the cross-strain of the specimen was alsomeasured simultaneously. The dilatation of the steel specimen as a function oftemperature is shown in Fig. 4.2a, and reveals the formation of martensite. Thephase transformation is reflected as a change in the slope due to the difference inspecific volume and thermal expansion between austenite and martensite.

The measured rms voltage during cooling is shown in Fig. 4.2b. It can be seenthat the rms voltage measured during austenitizing is very high due to the electricalcurrent flow through the specimen. When the heating stops, the rms voltage dropsback to the background noise level. The onset of a peak in the AE data is observedat the martensite-start temperature: the martensite peak. The signal level increasesto a maximum and then tails off to the background noise level. The microstructureof 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 theexperiments were conducted using specimens of type B austenitized at 820 C. InFig. 4.4 the temperature is plotted as a function of time. In the cooling curve aclear inflection is visible, which is caused by the release of latent heat during bainiteformation; the inflection corresponding to martensite formation is less pronounced.

During cooling the cross-strain and the AE signals were measured simultane-ously. The dilatation of the steel specimen, austenitized at 820 C, is plotted againsttemperature in Fig. 4.5a. The dilatation plot reveals that besides bainite formation


martensite formation also occurs at lower temperatures.

In Fig. 4.5b the rms voltage is plotted as a function of temperature. Whenthe heating stops, the rms voltage drops back to the background noise level, andsubsequently the onset of a peak in the AE data is observed at bainite transfor-mation temperatures: the bainite peak. The signal level increases to a maximumand then tails off to the background noise level. Upon further cooling, a secondpeak is observed at temperatures of martensite formation: the martensite peak.The shapes of the bainite and martensite peaks reflect the evolution of bainiteand martensite formation. In fact, the mean square voltage is proportional to thevolume 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 thevertical lines, agree rather well with the positions in the dilatation plot where theslope changes. It should be noted that due to the inhomogeneous temperature of thespecimen an unambiguous comparison of the dilatation signal and the AE signalis complicated and therefore not attempted for the experimental data obtainedfrom this set-up. By using a different set-up a more reliable comparison of bothtechniques will be given in section 4.2.2.

Recent work [6] has indicated that the martensite-start temperature is bestdetermined by the temperature corresponding to signal maximum rather than thetemperature corresponding to the onset of the signal. The data points during therise time of the signal may be attributed to the thermal gradients in the sampleleading to localized transformations as the sample approaches the martensite-starttemperature. This results in a broadening of the peak. In relation to this, it isinteresting to note that the onset of the bainite peak occurs at an unexpectedhigh temperature of approximately 600 C. This can be explained by the thermalgradients in the sample, which are more prominent at higher temperatures. Itshould be realized that the data are plotted against the temperature in the middleof the sample. The acoustic emission measured, however, may be generated inregions with a lower temperature.

The formation of both bainite and martensite was confirmed by metallographicanalysis. The microstructure of steel C45 is shown in Fig. 4.6 and contains bothmartensite and bainite. The fine needle structure visible in the micrograph, whiteareas (M), resembles a martensitic structure with a measured micro-hardness ofapproximately 700 HV0.1, and the dark areas (B) are identified as bainite with amicro-hardness of approximately 400 HV0.1.

The observation of the bainite peak, with a magnitude of the same order as themartensite peak has major implications for the interpretation of the mechanism forbainite formation. Since only processes involving shear and release of strain energygenerate acoustic emission, the observed acoustic emission during bainite forma-tion strongly indicates that the mechanism of growth of bainite is displacive in thestudied steel. It should be noted that the mechanism of growth of Widmanstattenferrite is also considered to be displacive [7], and that this transformation productwill give rise to a peak in the AE data at temperatures just above those of bainiteformation. However, microscopic analysis did not reveal any Widmanstatten ferrite


t [s]

0 5 10 15 20 25

T [º

C]

0

100

200

300

400

500

600

700

800

900

B

M

Figure 4.4: Cooling curve for steel C45 (specimen type B) after austenitizing at Ta= 820 C in the Gleeble thermo-mechanical simulator.

∆d [

µm]

30

40

50

60

70

80

T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

4

5

(a)

(b)

Figure 4.5: The dilatation (a) and the rms voltage (b) against temperature for steelC45 (specimen type B) austenitized at Ta = 820 C, cooling according Fig. 4.4.


Figure 4.6: Optical microscopy image of the microstructure of steel C45 (specimentype B) after continuous cooling from Ta = 820 C.

in the specimen after thermal cycling. It should also be mentioned that transfor-mation plasticity caused by the density difference between γ and α might occurduring bainite formation, and the associated stress can in principle lead to acousticemission. It is believed that this does not play a significant role in the present ex-periments because the transformation plasticity occurring during ferrite formationin mild steel (Fe360), which is expected to be of the same order of magnitude, didnot result in detectable acoustic emission. The AE measurements on steel Fe360are 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 viadiffusive mechanisms to ferrite during continuous cooling, did not reveal any peakin the AE data (see Fig. 4.7). On this point it should be noted that the resultsobtained in this study are in contradiction with the results obtained by Speich andSchwoeble [2]. The fact that they did not observe acoustic emission during bainiteformation 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 relativelyslow isothermal transformation to bainite probably results in a very small signalthat does not exceed the background noise level and is thus undetectable. The factthat bainite is undetectable under the above mentioned conditions is explained inmore detail in section 4.4.

Austenite grain size effect

To evaluate the effect of the austenite grain size on the evolution of bainite andmartensite formation, AE measurements were also performed during cooling of


T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

4

5

Figure 4.7: The rms voltage during cooling of steel Fe360 after austenitizing at Ta= 880 C indicating that the formation of ferrite is not accompanied by AE. Notethat bainite and martensite formation gives a signal of Urms ≈ 4 mV (see Figs. 4.2and 4.5). The intensity at Ta = 880 C is due to noise from the heating system ofthe Gleeble.

steel C45 (specimen type B) after austenitizing at Ta = 770 C and Ta = 870 C. Itshould be noted that changing the austenitizing time ta can also alter the austenitegrain size. However, preliminary tests showed that large differences in austenitizingtime are required to obtain a significant effect on the austenite grain size, implyingthat 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 israther difficult. Therefore, quantitative metallographic analysis was carried out onthe pearlite-ferrite microstructure formed in steel C45 after slow controlled coolingfrom elevated temperatures. Microscopic observations showed that the austenitegrain 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 austeni-tizing 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 peakrelative to that of the martensite peak at each Ta gives insight into the evolutionof both phase transformations, i.e. the relative amounts of bainite and martensiteformed. In comparison with the result for Ta = 820 C, it can be seen that for thelowest austenitizing temperature, Ta = 770 C, the martensite peak is small relativeto the bainite peak (see Fig. 4.8a). Furthermore, the martensite-start temperatureMs, as determined by the maximum of the martensite peak, yields a value of 220C. This is in good agreement with the fact that in the CCT diagram of steel C45the noses move to shorter times when the austenitizing temperature decreases andthat the martensite-start temperature of the remaining austenite decreases when


Urm

s [m

V]

0

1

2

3

4

5

6

Urm

s [m

V]

0

1

2

3

4

5

Ta = 870 ºC50 µµµµm

50 µµµµm

50 µµµµm

T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

4

5

(a)

(c)

(b)

(d)

(f)

Ms

Ms

Ms

(e)

Figure 4.8: Plot of the rms voltage against temperature for steel C45 austenitizedat (a) Ta = 770 C, (b) Ta = 820 C and (c) Ta = 870 C; the correspondingmicrostructure images are given in (d), (e) and (f) respectively.


bainite is formed [8].

The plot for Ta = 870 C (Fig. 4.8c) shows that the martensite peak is dominantand the bainite peak is smaller. The martensite-start temperature in this case isapproximately 290 C, which is in line with the change of the CCT diagram forthis austenitizing temperature. It should be remembered that the magnitude of thepeaks for different Ta cannot be compared in an absolute sense since the amount ofmaterial involved is dependent on the austenitizing temperature (see section 3.3).However, the ratio of the bainite and martensite peak is proportional to the relativeamounts of both phases.

The effect of the austenitizing temperature on the bainite and martensite peakcan be well explained in terms of a model describing the relationship between theaustenite grain size and the bainitic transformation rate in this type of steel [9]. Thismodel presumes that the growth rate of a bainite sheaf into the grain is relativelysmall compared with the nucleation rate at the austenite grain boundary. Underthese conditions the overall transformation rate is governed by the number densityof grain boundary nucleation sites, which is determined by the mean austenite grainsize.

At low Ta, the austenite grain size is small and this results in a large number ofnucleation sites and thus a relatively large bainite peak. The remaining austeniteis enriched in carbon due to the partitioning of excess carbon from the bainitesheaves, and this causes a relatively low martensite-start temperature. This canbe understood by realizing that the carbon enrichment increases the strength orshear resistance of the remaining austenite, and therefore the required driving forceto initiate the shear for martensite formation is higher. Since the driving force formartensite formation increases with undercooling, the martensite-start temperaturedecreases.

Upon increasing the austenitizing temperature Ta, the austenite grain size in-creases 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 remain-ing austenite is transformed to martensite resulting in a relatively large martensitepeak. Since only a small amount of bainite has formed, the carbon enrichment ofthe remaining austenite is limited and the martensite-start temperature is relativelyhigh.

4.1.2 Welding experiments

As shown in the previous section, bainite formation can be monitored by means ofAE during continuous cooling in the Gleeble welding simulator. Regarding austen-ite grain sizes and cooling rates the transformation behaviour of a spot weld isexpected to be very similar (see section 3.3), and the AE measurements on steelC45 during/after welding are discussed in this section.

The AE measurements during welding were carried out using the set-up asdescribed in section 3.2. A typical example of the results obtained in the case ofspot welding with moderate arc currents is presented in Fig. 4.9a. In this figure


t [s]0 5 10 15 20 25

Urm

s [m

V]

0

2

4

6

8

(3)

(1)

(4)

(2)

Urm

s [m

V]

0

2

4

6

8

10

(3)

(1)

(4)

(2)

(5)

c

d

(a)

(b)

Figure 4.9: a)b) The rms voltage Urms as a function of time during and after spotwelding of steel C45 with (a) I = 55 A and (b) I = 90 A: (1) noise level dueto the arc; (2) extinction of the arc; (3) background noise level; (4) peak due tomartensite formation; (5) peak due to bainite formation. c)d) cross-sections of spotwelds 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 55A. 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-starttemperature Ms, the onset of the martensite peak in the Urms data is observed. Thesignal 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 thespot 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 themiddle and in Fig. 4.9c the cross-section is shown. Microscopic observation showsthat martensite is formed in both the weld metal and the heat-affected zone, thestructure being more pronounced in the weld metal. For both zones hardness valuesof 600 – 700 HV0.1 were found, indicative of a fully martensitic structure.

A typical example of the results obtained for spot welding with relatively highheat input (relatively low cooling rate) is shown in Fig. 4.9b. In this figure two peakscan be observed: a bainite peak (5) and a martensite peak (4). The austenite in


the zone close to the base metal, which has a relatively small grain size, transformsto bainite giving rise to the first peak (5). Upon further cooling, the remainingaustenite in the centre of the weld transforms to martensite resulting in the secondpeak (4).

The cross-section of the spot weld produced with an arc current of 90 A shows adark structure in the zone adjacent to the interface between the HAZ and the basemetal, which is identified as bainite (see Fig. 4.9d). The measured micro-hardnessvalues in this zone were in the range 300 – 500 HV0.1. The weld metal has a fullymartensitic structure with hardness values exceeding 600 HV0.1.

In summary, the results during welding of steel C45 are in good agreement withthe results obtained using the Gleeble thermo-mechanical simulator. Since it isextremely difficult to measure the temperature of a spot weld, the data shown inthis section could only be plotted against time, which complicates a quantitativecomparison between the two results. Moreover, a waveguide was used during theGleeble 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, similarexperiments were performed on steel 42CrMo4 using the techniques described inchapter 3. The chemical composition of steel 42CrMo4 is given in Table 3.2. A fewexperiments on this steel were performed using the thermo-mechanical simulatorunder the same conditions as for steel C45 (see section 4.1.1). During cooling of steel42CrMo4 in the Gleeble thermo-mechanical simulator only martensite is formed.The results obtained are very similar to the result obtained for steel C45 shownin Fig. 4.2 and are discussed in detail in Ref. [10]. In section 4.2.1 the results ofwelding experiments on steel 42CrMo4 are discussed. Owing to the relatively highhardenability of this steel in comparison with steel C45, only martensite is formedin the weld and the HAZ during cooling.

It can be concluded from the experiments described in the previous section thata quantitative analysis of the measurements using the thermo-mechanical simulatoris complicated because of the thermal gradient in the sample during austenitizing;the measured AE signals originate from all the regions of austenite, with differentgrain sizes and/or carbon content, decomposing into bainite or martensite. On theother hand, samples austenitized in a dilatometer or a furnace do have a homoge-neous microstructure as described in sections 3.4 and 3.5. The acoustic emissionexperiments on steel 42CrMo4 using the dilatometer set-up and the furnace set-upare 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 accordingthe procedure described in section 3.2. In Fig. 4.10 a typical example of the resultsobtained in the case of spot welding of steel 42CrMo4 is shown. After extinction

Study of steel 42CrMo4 59

t [s]

18 20 22 24 26 28

Urm

s [m

V]

0

2

4

6

8

10

(3)(1)

(4)

(2)

Figure 4.10: The rms voltage as a function of time during and after spot weldingof steel 42CrMo4 with I = 75 A: (1) noise level due to the arc; (2) extinction ofthe 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.


∆d [

µm]

-5

0

5

10

15

T [ºC]

200 300 400

Urm

s [m

V]

0.0

0.2

0.4

0.6

0.8

1.0

(a)

(b)

Msd

MsAE

Figure 4.12: Plot of the dilatation (a) and the rms voltage (b) against temperaturefor steel 42CrMo4 austenitized at Ta = 900 C.

of the arc, the spot weld cools down to the martensite-start temperature and theformation of martensite is reflected by a peak in the AE data. The signal observedis 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 ofapproximately 700 HV0.1 were found in both the weld metal and the heat-affectedzone, indicative of a fully martensitic structure. In Fig. 4.11b an optical image ofthe 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 com-bined AE – dilatometer (Bahr 805A/D) set-up as described in section 3.4 (see


Fig. 3.15). Steel 42CrMo4 was selected as a suitable material based on the CCTdiagram and estimated (natural) cooling rates in the dilatometer. The sample washeated to an austenitizing temperature of 900 C, austenitized for 2 minutes, andsubsequently continuously cooled to ambient temperature in a helium atmosphere.

The generated AE and change in dilatation measured during cooling are shownin Fig. 4.12. It can be seen that the transformation to martensite is reflected in boththe dilatation and the AE data. The AE signal rapidly increases to a maximumvalue, and at approximately the same temperature the slope of the dilatation curvechanges. Md

s determined from the dilatation signal yields 306 ± 8 C. For thedetermination of Md

s , the criterion is used that when the dilatation signal startsto deviate significantly from the austenite line during cooling; the correspondingtemperature is the start temperature of the transformation. From the AE signalMAEs is determined as the temperature corresponding to signal maximum, yielding

302 ± 6 C. The inaccuracy in both measurements is mainly due to the temperaturedifferences in the sample; the maximum temperature difference is estimated to be5 C based on measurements using two thermocouples. The inaccuracy in thedetermination of MAE

s and Mds 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 dur-ing the rise time of the AE signal may be attributed to the thermal gradientsin the sample, leading to localized transformations as the sample approaches themartensite-start temperature: a broadening of the peak.

A major drawback of experiments with the dilatometer described above, is thelow signal level caused by the small sample and the attenuation due to the wave-guide. This made it impossible to measure bainite formation. Although the samplesizes used are of the same order of magnitude, the waveguide in the case of theGleeble experiments has a larger diameter, and the transformation rates are higherowing to the higher cooling rates.

4.2.3 Furnace experiments

As explained in the previous sub-section, the detectability of AE signals is lowusing the dilatometer set-up, and therefore experiments were performed using aconventional furnace and a salt bath as described in section 3.5. Such experimentshave the advantage that large-sized samples and a relatively thick waveguide rodcan be used. The objective of this study was to obtain bainite formation in steel42CrMo4 and measure it by means of acoustic emission.

A sample of steel 42CrMo4 with a diameter of 16 mm and a length of 40 mmwas employed. The sample was austenitized at a temperature of 900 C for 20minutes and subsequently cooled in a salt bath at a temperature of 400 C. Duringcooling, the rms voltage of the generated AE waves and the temperature of thesample 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 50seconds before cooling. It can be seen that during austenitizing (T1 = 900 C), Urms


50 100 150 2000

2

4

6

8

10

12

14

16

Urms (T1 = 900 ºC) T Urms (T1 = 650 ºC)

t [s]

Urm

s [m

V]

300

400

500

600

700

800

900

T [ºC]

Figure 4.13: The rms voltage and the temperature as a function of time duringcooling of steel 42CrMo4 in the salt bath.

50 100 150 2000

2

4

6

8

Urms (corrected) T

t [s]

Urm

s [m

V]

300

400

500

600

700

800

900

T [ºC

]

Figure 4.14: The rms voltage against time for steel 42CrMo4 corrected for the noisedue to oxidation.


= 0.3 mV, the background noise value. At approximately t = 50 s the sample wastaken out of the furnace, and quenched directly in the salt bath. During the firstfew seconds of quenching very high signals were observed, which are attributed tooxidation and/or another chemical reaction between the steel and the salt. Duringcooling this noise signal decreases strongly and at t = 70 s (T = 550 C) the onsetof a bainite peak can be seen. This is in good agreement with the bainite-starttemperature 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 tailsoff to a constant noise level. At t = 170 s (T = 400 C) the transformation hasprobably finished, and the observed constant signal level is predominantly causedby oxidation.

To investigate the influence of the oxidation process, a dummy experiment wasperformed; the same sample was heated up to 650 C and subsequently quenchedin the salt bath. Since the sample was not austenitized, no bainitic transformationwill 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 canbe seen that the noise peak decreases relatively fast compared to the total signalUrms(T1 = 900 C) resulting from both bainite formation and oxidation. In anattempt to correct for the noise, the mean square voltage of the noise signal wassubtracted from the mean square voltage of the total signal. The result for thecorrected signal, the bainite peak, is shown in Fig. 4.14. Since the noise peak issomewhat dependent on the initial temperature of the sample, the correction forthe noise (for t < 70 s) is not very accurate. Furthermore, it can be seen that thenoise peak has a rather irregular appearance; it is not really a continuous AE signal,which complicates the correction. Nevertheless, the noise correction demonstratedabove 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 withthe acoustic emission measurements during bainite formation in steel C45 (section4.1.1) and is consistent with the displacive model of the bainitic transformation.

It should be noted that the bainitic transformation in steel 42CrMo4 (seeFig. 4.14) takes place during slow continuous cooling, and not during the isothermalpart of the treatment as was initially attempted. The lower than expected coolingrate is predominantly caused by the low heat transfer from the sample to the liquidsalt due to the oxide layer formed on the surface. Also the large sample size (largeheat 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 thesample is small and diminishes rapidly; when the temperature of the sample is ho-mogeneous the temperature difference between the sample and the salt is still 100C.

In addition, AE experiments under the same conditions were performed on steel50CrV4 and steel 16MnCr5. The observed AE signals during cooling of these steelsshowed a similar trend as found for steel 42CrMo4 (Fig. 4.13), and support theresults obtained for this steel.


T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

Figure 4.15: The rms voltage as a function of temperature for steel St52-3 duringcontinuous 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 reactionmechanism, the low carbon steels St52-3 and St50K were studied using the thermo-mechanical 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 ofresults is analogous to the analysis of the AE measurements on steel C45 describedin section 4.1.1.

In order to generate both bainite and martensite formation during continuouscooling, the experiments were performed using specimens with an effective freespan 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 secondsto the austenitizing temperature Ta = 850 C, austenitized for 20 seconds, andcontinuously cooled to ambient temperature. During cooling the rms voltage of thegenerated AE signals was measured. After thermal cycling, the microstructure ofthe specimens was analyzed using an optical microscope (Olympus) and a micro-hardness tester (Buehler Ltd.).

In Fig. 4.15 the rms voltage is plotted against temperature during cooling ofsteel St52-3; the result for steel St50K is shown in Fig. 4.16. It can be seen thatfor both steels the evolution of bainite and martensite formation is reflected by twodistinct peaks in the AE data at temperatures of 500 – 600 C and 200 – 300 Crespectively. 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 steelSt52-3 and steel St50K prove that for low carbon steels the mechanism of growthof bainite is also displacive.

Study of a high-alloyed steel 65

T [ºC]

0 100 200 300 400 500 600 700 800 900

Urm

s [m

V]

0

1

2

3

Figure 4.16: Plot of the rms voltage against temperature for steel St50K austeni-tized at Ta = 850 C.

4.4 Study of a high-alloyed steel

This section reports the AE measurements on the high-alloyed steel 75MnSiCr dur-ing cooling in the salt bath. The experiments were performed using the furnaceset-up as described in section 3.5. The chemical composition of steel 75MnSiCr isgiven in Table 3.2. This steel is characterized by a very low bainite-start tempera-ture; the bainite formed at such a low temperature results in a very high strengthof 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 foldedaround the sample (not completely sealed due to the waveguide). After austen-itizing at T = 900 C the sample was placed in the salt bath at T = 260 C. InFig. 4.17 the measured rms voltage and temperature are plotted as a function oftime during cooling in the salt bath. The cooling rate was relatively low due to theaustenitic stainless-steel bag (air layer in between the bag and the sample). Fromthe cooling curve the cooling time ∆t8/5 was determined yielding approximately200 seconds. The AE and temperature measurements were started just before thesample was quenched in the salt bath. In the beginning the background noise levelof 0.26 – 0.30 mV was measured. After 500 seconds, when the temperature of thesample was approximately T = 350 C, the onset of a peak due to bainite formationwas observed; the signal increased to a maximum at t = 900 s, yielding Urms = 2.2mV. At that point the sample is still slowly cooling (T = 300 C); the temperaturedifferences in the sample are expected to be very small. After the maximum signalwas reached, the signal tailed off to the background noise level. At approximatelyt = 2500 s the sample temperature was equal to the bath temperature and Urms =1.1 mV. After further isothermal holding the signal level slowly decreased towards


0 1000 2000 3000 4000 5000 60000.0

0.5

1.0

1.5

2.0

2.5

3.0

Urms T

t [s]

Urm

s [m

V]

200

400

600

800

1000

T [ºC]

Figure 4.17: Plot of the rms voltage and the temperature against time after quench-ing of steel 75MnSiCr in the salt bath.

the background noise level. From the AE and temperature data shown in Fig. 4.17it can be concluded that a transition of bainite formation under continuous coolingconditions to bainite formation under isothermal conditions takes place. Noise sig-nals can be identified in the AE data; these are burst signals superimposed on thecontinuous AE, which are probably caused by oxidation phenomena. In order to un-derstand that oxidation related signals are still measured when using the austeniticstainless-steel bag, it should be realized that the austenitic stainless-steel bag ismaking contact with the sample or waveguide on some points and thus acousticallycoupled to the sensor.

The result obtained for steel 75MnSiCr shown in Fig. 4.17 is in line with result ofthe measurement on steel 42CrMo4 discussed in section 4.2.3. The main differenceis that in this case the transformation takes place over a much longer period oftime. The observed peak in the AE data implies that the bainitic transformationin the high-alloyed steel 75MnSiCr is displacive, which is in good agreement withthe 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 previoussections of this chapter.

Also experiments were performed without using the austenitic stainless-steelbag. In this case the cooling rate was much higher, ∆t8/5 was approximately 40seconds, and no peak in the AE data was observed during cooling to the bath tem-perature. For steel 75MnSiCr the incubation time for bainite formation is relativelylong [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 coolingconditions. Subsequently, isothermal holding at the bath temperature T = 290 Cfor more than 12 hours did not reveal detectable acoustic emission. Experimentson the same steel by other researchers have revealed that the bainite-start temper-ature lies above 300 C [11], the martensite-start temperature at T ≈ 150 C. It istherefore expected that bainite forms at T = 290 C, but at a slow rate. The totalAE energy due to bainite formation that was measured in Fig. 4.17 is given by thearea under the curve, which is 3300 mV2s. If a similar energy is released during 10hours [11], the average level for U 2 is about 0.09 mV2, which is not distinguishableabove the background level.

Maybe with a larger sample of the studied steel or another steel with highertransformation kinetics it is possible to measure AE under true isothermal con-ditions. It should be remembered that the background noise level cannot be re-duced; the detectability of acoustic emission is completely governed by the volumetransformation rate and the amount of released elastic energy per unit volume oftransformation product. The simple calculation given below shows that it is notsurprising that bainite formation was not detected under isothermal conditions.

4.5 Conclusions

The acoustic emission measurements discussed in this chapter have resulted in abetter understanding of the growth mechanism of the various phase transformationsin steel, in particular the bainitic transformation. Based on the results obtainedusing the different techniques for thermal cycling as described in chapter 3, thefollowing conclusions can be drawn.

The chapter started with the discussion of acoustic emission measurements dur-ing continuous cooling of steel C45 using the Gleeble 1500 thermo-mechanical sim-ulator. During cooling two distinct peaks in the AE data were observed at tem-peratures of 500 – 600 C and 200 – 300 C, which are attributed to bainite andmartensite formation respectively. The occurrence of acoustic emission during thebainitic transformation implies that the bainite reaction mechanism in steel C45 isdiffusionless and is best described in terms of the displacive model. In contrast, noacoustic emission was detected during the diffusion-controlled transformation fromaustenite to ferrite in steel Fe360.

Acoustic emission measurements for different austenitizing temperatures re-vealed changes in the evolution of bainite and martensite formation in steel C45.The ratio of the bainite and martensite peak gives information about the amountsof bainite and martensite formed. It was found that decreasing the austenitizingtemperature enhanced bainite formation, as reflected by an increase of the bai-nite peak relative to the martensite peak. Moreover, the change in position of themartensite peak reflected that the martensite-start temperature decreased in caseof enhanced bainite formation due to carbon enrichment of the remaining austeniteafter bainite formation.


In addition, AE measurements during/after welding of steel C45 were performedto study the phase transformation occurring in a weld during cooling. In the caseof spot welding with moderate arc currents a martensite peak was observed duringcooling. At high heat input, the cooling rate decreased and a bainite peak followedby a martensite peak were observed after extinction of the arc. It can be concludedthat the results obtained for welding of steel C45 are in good agreement with theresults of AE measurements on steel C45 using the Gleeble thermo-mechanicalsimulator.

During cooling of steel 42CrMo4 after welding and in the thermo-mechanicalsimulator only martensite was formed. The results obtained are in line with the re-sults obtained for martensite formation in steel C45. Subsequently, the martensitictransformation in steel 42CrMo4 was studied using the combined AE – dilatome-ter (Bahr 805A/D) set-up. The simultaneous measurement of acoustic emissionand dilatation during the martensite formation in steel 42CrMo4 allowed a propercomparison between the AE technique and conventional dilatometry. The transfor-mation to martensite was reflected in both the dilatation and the AE data. Analysisof the data showed that the martensite-start temperature M d

s determined from thedilatation signal was only 4 C higher than the MAE

s determined from the AE sig-nal. This indicates that the sensitivity of both techniques in the set-up used is ofthe same order of magnitude. In general, however, it is difficult to draw unambigu-ous 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 onthe sample volume but also dependent on the transformation rate. Moreover, theintensity of the signal is strongly affected by the use of a waveguide.

To study the bainitic transformation in steel 42CrMo4, experiments were per-formed using a conventional furnace and a salt bath. Due to oxidation very highsignals were observed directly after the sample was quenched in the salt bath.However, this noise signal decreased strongly and the observed AE signal in thetemperature range of 400 – 550 C could be attributed to bainite formation. Inorder to correct for the oxidation process, a dummy experiment was performed: thesame 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 tooxidation. These data were used to extract the AE signal due to bainite formationfrom the total signal.

In addition to the study of the medium carbon steels C45 and 42CrMo4, AEmeasurements were performed during bainite and martensite formation in two lowcarbon steels, St50K and St52-3, and during bainite formation in the high-alloyedsteel 75MnSiCr. The results of the AE measurements on the low carbon steels usingthe thermo-mechanical simulator revealed two peaks in the AE data correspondingbainite and martensite formation, similar to the case of steel C45. The chapter endswith the discussion of AE measurements on steel 75MnSiCr, which is characterizedby a very low bainite-start temperature and a relatively long incubation time forbainite formation. During cooling of this high-alloyed steel in the salt bath a peak inthe 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 givesstrong evidence that the bainitic transformation is displacive.

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, ActaMat. 51, 4183 (2003).


[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. Tech-nol. 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

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 gen-erated during bainite and martensite formation. A relationship is derived and val-idated that relates a measurable AE signal parameter to the volume of martensiteor bainite formed.

As explained in section 2.3.1 a large amount of the Gibbs free energy differ-ence ∆Gγ→α′

is dissipated in the process of slip, which occurs within a martensitecrystal in order to relax the stresses accompanying the lattice transformation [1].During the transformation a part of the released strain energy is radiated as tran-sient elastic waves called acoustic emission. The dynamics of the transformation ona microscopic scale has a strong dependence on the chemical composition, and con-sequently the amount of the free energy change ∆G that is converted into acousticemission energy is expected to be material dependent.

Since many material properties of the specimen and transducer influence theAE waves during the period between generation at the source and detection at thetransducer it is impossible to determine the absolute value of the energy of an AEsource, unless a proper calibration can be carried out [2]. The energy value of theelectrical signal measured at the transducer is affected by many factors such as theacoustic impedance of the transducer and the specimen and the system bandwidth.However, for experiments using the same specimen configuration these factors maybe treated as constants, and under such conditions the results of experiments usingthe same type of sources with different magnitudes can be compared.

In section 5.1 a relation for the AE energy generated during martensite for-

73

74 Chapter 5: A study of acoustic emission energy ...

mation will be derived from basic principles, and subsequently extended to includebainite formation. The relation is tested for various steels that transform to marten-site and/or bainite during welding. Both spot welding and travelling arc weldingexperiments are performed. The validation of the relation for martensite formationin steel 42CrMo4 and steel 42MnV7 is presented in section 5.2. In section 5.3 theresults of welding experiments on steel C45 are discussed. After spot welding of thissteel only martensite is formed in the weld; under travelling arc welding conditionsbainite is also formed. Based on both sets of experiments the relationship for AEenergy during bainite formation is tested.

Besides the fundamental interest in the generation of acoustic emission duringdisplacive phase transformations, the results of this study may be used to develop anAE monitoring system to detect martensite and bainite formation during welding.Since these hard regions in the weld and heat-affected-zone (HAZ) are susceptibleto cold cracking due to hydrogen embrittlement [3], real time monitoring of thewelding process is of considerable practical importance. Although it has been knownfor a long time that acoustic emission can be detected during the formation ofmartensite, little effort has been made to study this transformation during welding.Most recently, Liu and Kannatey-Assibu performed acoustic emission experimentsduring welding of high carbon steel [4]. They showed that after welding the AEsignals originating from martensite formation could be distinguished from othersignals. However, to identify martensite formation during welding was found to bedifficult 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 weld-ing when the temperature T falls below the martensite-start temperature Ms. Afraction of the free energy change ∆Gγ→α′

at the Ms temperature is converted intoAE energy. This fraction is assumed to be constant over the temperature range Ms

to Mf , where Mf is the martensite-finish temperature. The increase in ∆Gγ→α′

atlower temperatures is necessary to start the nucleation at lattice distortions in theparent phase that are less favourable. When it is assumed that the energy dEAE ofAE waves that are generated during the formation of a volume dVm of martensiteis proportional to this volume, then

dEAE ∝ dVm (5.1)

The energy emission rate of an AE source can be expressed in terms of the meansquare voltage U2 of the acoustic waves detected at a transducer [5], or

dEAEdt

∝ U2 (5.2)

with U2 defined by Eq. (3.1). By substitution of Eq. (5.2) in Eq. (5.1) and integra-tion it follows that

Martensite formation 75

∫

U2dt ∝∫

dVm (5.3)

with the integration running over the entire transformation range. In the case ofspot welding, the integral term on the right hand side is equivalent to the volumeVm of martensite produced in the spot weld and Eq. (5.3) can be written as:

∫

U2mdt = kmVm (5.4)

with km the proportionality factor between the AE energy and the transformedvolume, which is presumably material dependent.

The integral on the left hand side can be calculated by measuring the areaunder the curve in a U2 versus t plot. Assuming that this theory is valid for everydisplacive transformation involving acoustic emission, it follows that when bainiteand martensite formation occur successively, Eq. (5.4) generalizes to

∫

U2dt = kbVb + kmVm (5.5)

with kb the proportionality factor for bainite formation and Vb the volume of bai-nite. This equation can be written in differential form as:

U2 = kbdVbdt

+ kmdVmdt

= U2b + U2m (5.6)

It should be noted that Eqs. (5.5) and (5.6) may be viewed as manifestationsof the conservation of energy (power). When bainite and martensite are formedsimultaneously, the total measured AE power at each moment is the sum of the AEpowers due to bainite and martensite formation respectively. For example, in thecase of travelling arc welding martensite and bainite may be formed at the sametime. Under steady state welding conditions the amounts of bainite and martensiteformed 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 squarevoltage of the electrical noise and the welding noise need to be subtracted from themeasured mean square voltage.

5.2 Martensite formation

In order to examine the AE energy generated during martensite formation, weldswere made on steel 42CrMo4 and steel 42MnV7. For both spot welding and trav-elling arc welding a complete transformation to martensite is achieved in thesecommercially available steels. In section 5.2.1 the results obtained for travelling arcwelding of steel 42CrMo4 are discussed. The analysis of spot welding experimentson steel 42MnV7 is presented in section 5.2.2.


t [s]

0 10 20 30 40 50 60

Urm

s [m

V]

0

2

4

6

8

10

(2)(1)

(3)(5)

(4)

Figure 5.1: The rms voltage plotted versus time during travelling arc welding ofsteel 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 mmwas used for the experiments. The chemical composition of steel 42CrMo4 is givenin Table 3.2, and the AE measurements during welding were performed using theset-up as described in section 3.2. To vary the amount of martensite formed, alarge number of welding experiments with different heat inputs was carried outunder travelling arc conditions with a travel speed v = 2 mm/s. An example of theresults obtained is shown in Fig. 5.1. In this figure the rms voltage is plotted as afunction of time for an arc current of 75 A. The plot shows that under non-weldingconditions the background noise is measured (1). When the arc is ignited and thewelding cycle starts (2), the signal abruptly rises to a higher level. This level isconsidered as the noise signal level produced by the arc during welding. After acertain time the first part of the weld has cooled down to the Ms temperature andstarts to transform to martensite, which leads to an increase in signal level (3). Thiscontinues until a steady state situation is reached (4). At the end of the weldingcycle (5) a peak is observed, which is due to the fact that after extinction of thearc the cooling rate increases and, hence, a larger volume of metal is transformedto martensite per unit time.

To test the validity of Eq. (5.6) for martensite formation the rms voltage was


t [s]

0 10 20 30 40 50 60 70

Urm

s [m

V]

0

1

2

3

4

5

6

7

8

9

10

35 A60 A75 A100 A

Figure 5.2: The rms voltage as a function of time during travelling arc welding ofsteel 42CrMo4 for different arc currents.

Figure 5.3: The cross-sections of welds produced on steel 42CrMo4 with an arccurrent of (a) 35 A, (b) 60A, (c) 75 A and (d) 100 A.


measured during 14 welding runs with arc currents in the range of 35 up to 100 Awith steps of 5 A. The results of four typical welding runs performed with 35 A, 60A, 75 A and 100 A are plotted in Fig. 5.2. In order to minimize overlapping of datapoints, the time basis is shifted for the four welding runs. It is seen that the signallevel increases with arc current and that also the peak after extinction of the arcbecomes higher and broader due to the fact that the amount of material involvedincreases.

After each welding run, the volume rate dVm/dt of martensite formation wasdetermined by measuring the area Am of the martensitic phase from weld cross-sections and multiplying this area by the travel speed v. In Fig. 5.3 the cross-sections corresponding the four welding runs of Fig. 5.2 are shown. Hardness mea-surements confirmed that in all cases both the weld metal and the heat-affectedzone (HAZ) are fully martensitic. It should be noted that the martensitic structureis more pronounced in the weld metal than in the HAZ due to the fact that in theweld metal the mean grain size is larger than in the HAZ. The results obtained forthe 14 welding runs are listed in Table 5.1.

In Fig. 5.4 the average AE power U 2m under steady state conditions (correctedfor welding noise) is plotted as a function of dVm/dt. The error bars reflect theinaccuracy in the determination of the U 2m; the value of dVm/dt can be determinedwith high accuracy. The data obtained from the welds produced with arc currentsin the range of 35 – 100 A can be fitted to Eq. (5.6) with reasonable accuracy. Theresult for km from the fit yields 0.85 ± 0.1 × 103 V2s/m3. Regarding the accuracy ofthe fit, it should be noted that for low heat inputs the welding noise has a relativelylarge contribution to the total measured AE power (see Table 5.1). Moreover, theaccuracy in the determination of the AE power due to the welding noise is limited.Therefore, the results obtained for U 2m at low heat inputs are relatively inaccurate,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 theform of plates with dimensions 250 mm × 200 mm × 8 mm, are discussed. Thechemical composition of steel 42MnV7 is given in Table 3.2. Two typical examplesof the results obtained in the case of spot welding are presented in Figs. 5.5a and5.5b. The martensite peak observed during cooling of the weld is very similar tothe peak observed after spot welding of steel C45 with moderate heat input, seesection 4.1.2.

In order to check the reproducibility of the experiments two series of spot weldswere made; first a series with arc currents in the range of 65 to 110 A, and secondlya series with arc currents in the range of 50 to 95 A. Comparing the results makesit possible to estimate the accuracy in determining the volume of martensite in aspot weld; hence, for an accurate determination of the transformed volume it isrequired that the spot weld is cut exactly in the middle.

The generated AE energy ∫ U 2mdt was determined for each weld. Microscopic


Table 5.1: Results of travelling arc experiments on steel 42CrMo4.

Arccurrent(A)

Arcvoltage(V)

Arclength(mm)

dVm/dt

(mm3/s)

Urmstotal(mV)

Urmsnoise(mV)

U2mmartensite(mV2)

35 10.1 2.1 3.1 1.4 0.8 1.340 10.8 2.5 4.4 1.5 0.8 1.745 10.2 2.8 4.9 1.7 0.8 2.350 10.2 3.0 6.3 2.1 0.9 3.655 10.1 3.0 7.1 2.3 0.9 4.560 10.3 3.0 9.9 2.8 0.9 7.065 10.1 3.0 12.0 3.1 0.9 8.870 10.3 3.0 14.4 3.5 0.9 11.475 10.3 3.0 16.8 4.0 0.9 15.480 10.6 3.0 20.3 4.3 0.9 17.785 10.5 3.0 22.1 4.5 1.0 19.390 10.6 3.0 25.8 4.9 1.0 23.095 10.5 3.0 27.3 5.1 1.0 25.0100 10.5 3.0 30.0 5.3 1.0 27.3

dVm/dt [mm3/s]0 5 10 15 20 25 30

Um

2 [m

V2 ]

0

5

10

15

20

25

30

Figure 5.4: Plot of the average AE power U 2m against dVm/dt for welding on steel42CrMo4 with arc currents in the range of 35 to 100 A.


t [s]0 5 10 15 20

Urm

s [m

V]

0

2

4

6

8

Urm

s [m

V]

0

2

4

6

8

10

(3)

(1)

(4)

(2)

(3)

(1)

(4)

(2)

(a)

(b)

(c)

(d)

Figure 5.5: a)b) The rms voltage Urms as a function of time during and after spotwelding of steel 42MnV7 with (a) I = 60 A and (b) I = 85 A: (1) noise level dueto the arc; (2) extinction of the arc; (3) background noise level; (4) peak due tomartensite formation. c)d) Transverse cross-sections of spot welds produced withan 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 casesmartensite is formed in both the weld metal and the HAZ; for both zones micro-hardness values of 600 – 700 HV0.1 were found, indicative of a fully martensiticstructure. The cross-sections were also used to measure the radius r and depth dof the martensitic zone from which the volume Vm of martensite in the spot weldwas calculated. The volume of a spot weld can be calculated from the measuredradius r and depth d of the spot weld using the illustration shown in Fig. 5.6. Thisfigure 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 =

∫ θ

0

π(r sin θ′)2r sin θ′dθ′ = πr3(

− cos θ + 1 +(cos θ)3

3− 1

3

)

(5.8)

the volume Vm is obtained. The results obtained for the spot welds on steel 42MnV7are listed in Table 5.2.


rθ

d

Figure 5.6: A typical spot weld can be approximated to fall inside a sphere withradius r. Based on this and the depth d of the spot weld, the volume of the spotweld can be calculated.

In Fig. 5.7 the values of ∫ U 2mdt (corrected for the thermal background noise)are plotted as a function of Vm. From the fit of the data to Eq. (5.4) it followsthat km yields a value of 1.15 ± 0.05 × 103 V2s/m3. Comparison of the results ofthe two series shows that the measurements are fairly reproducible, especially forrelatively high heat inputs. Only for the low heat inputs significant differences areseen between the two series. This is due to the fact that it is relatively difficult tocut 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 propor-tionality factor k. The advantages and disadvantages of each method are discussedbelow. Since the thermal background noise is much lower than the welding noise,the integral ∫ U2mdt after spot welding can be determined with relatively high accu-racy compared with the AE power U 2m during travelling arc welding. Furthermore,the value of the welding noise cannot be determined with high accuracy. On theother hand, it is in principle more difficult to determine the volume of martensiteVm in a spot weld than to determine the volume transformation rate dVm/dt duringtravelling arc welding. This is due to the fact that it is experimentally difficult tocut a spot weld exactly in the middle; this does not play a role for travelling arcwelds. The results obtained for steel 42MnV7, however, show that the volume ofmartensite in a spot weld can be determined with sufficient accuracy, especiallyfor the welds produced with high heat inputs. Overall, it can be concluded thatspot welding is a more accurate method to study the AE energy generated during


martensite formation.

In addition, a few spot welds were produced on steel 42CrMo4 in order to verifywhether a similar value of km can be obtained as in the previous sub-section. Atypical example of a result obtained for spot welding is shown in Fig. 4.10 anddiscussed in the corresponding section. The values of km were determined by mea-suring the area under the peak in a plot of U 2m against t, and the volume Vm ofmartensite formed, and inserting the values obtained in Eq. (5.4). For a few spotwelds produced with different heat inputs, a value of 1.0 ± 0.2 × 103 V2s/m3 wasfound for km. This is in good agreement with the results obtained in the case oftravelling arc welding of steel 42CrMo4 (section 5.2.1).

Table 5.2: Results for two series of spot welding experiments on steel 42MnV7.

I(A)

r(mm)

d(mm)

Vm(mm3)

∫

U2mdt(mV2s)

I(A)

r(mm)

d(mm)

Vm(mm3)

∫

U2mdt(mV2s)

50 3.2 1.3 15 22 65 3.5 1.9 31 2760 3.3 1.6 23 26 70 3.6 2.1 40 3565 3.5 1.8 30 34 75 3.7 2.4 52 5970 3.7 2.0 39 44 80 3.9 2.5 59 7175 3.7 2.3 49 57 85 4.0 2.7 73 8780 3.9 2.4 57 67 90 4.2 2.9 85 9885 4.0 2.8 74 84 95 4.4 3.1 103 11590 4.2 2.9 85 101 100 4.5 3.2 112 12895 4.4 3.0 99 113 105 4.6 3.4 129 147

110 4.9 3.6 151 178

5.3 Bainite and martensite formation

In this section the study of the released AE energy during welding of steel C45is presented. The chemical composition of steel C45 is given in Table 3.2. Theexperiments were performed using plates with dimensions 250 mm × 200 mm ×5 mm, and the welds were made using the computer controlled GTA welding unitas described in section 3.2. Based on the results discussed in section 4.1.2 it isknown that bainite and/or martensite are formed in the weld. Whether bainite ormartensite is formed at a specific location in the weld depends on the austenitizingtemperature, the austenitizing time and the cooling rate at that position.

To vary the amounts of bainite and martensite formed, experiments were per-formed with different arc currents under two conditions. Spot welds were madeto examine the AE energy generated during the martensitic transformation. Itshould be noted that after spot welding the cooling rate is relatively high, whichis required for martensite formation. The results of the AE measurements afterspot welding are discussed in section 5.3.1. Subsequently, travelling arc welding

Bainite and martensite formation 83

Vm [mm3]

0 20 40 60 80 100 120 140 160

Um

2 dt [

mV

2 s]

0

40

80

120

160 first seriessecond serieslinear fit

Figure 5.7: Plot of ∫ U2mdt against Vm for two series of spot welds produced on steel42MnV7 with arc currents in the range of 50 – 110 A.

experiments were performed to study the AE energy generated during the bainitictransformation. 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 timewas limited to 10 seconds to avoid excessive heating of the workpiece. It shouldbe noted that a high temperature of the workpiece leads to a decrease in coolingrate after spot welding. After extinction of the arc, the rms voltage Urms of thecontinuous acoustic emission was measured.

A typical example of the results obtained in the case of spot welding withmoderate arc currents is presented in Fig. 4.9a and discussed in the correspondingsection. To test the validity of Eq. (5.4), Urms was measured for a number of spotwelds produced with arc currents in the range of 35 to 80 A. Of each weld thearea under the martensite peak in a plot of U 2m against t was calculated in order todetermine the generated AE energy ∫ U 2mdt. Cross-sections were made of all welds,and microscopic observation showed that for all welds in both the weld metal andthe HAZ a fully martensitic structure, with micro-hardness values of 600 – 700HV0.1, was formed. The cross-sections were also used to measure the radius r anddepth d of the martensitic zone from which the volume Vm of martensite in the spotweld was calculated. The results obtained are listed in Table 5.3.

In Fig. 5.8 the values of ∫ U 2mdt (corrected for the background noise) are plottedas a function of Vm. It can be seen that for welds obtained with arc currents in


the range of 35 to 80 A a linear relation exists between AE energy and the volumeof martensite formed, as predicted by Eq. (5.4). The proportionality factor kmextracted from the linear fit yields a value of 2.38 ± 0.05 × 103 V2s/m3.

The results obtained for martensite formation in steel C45 are in good agreementwith 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 differencesin chemical composition between the steels. The fraction of the free energy change∆Gγ→α′

that is released in the form of acoustic waves is probably dependent onthe chemical composition of the material.

Table 5.3: Results for spot welding of steel C45.

I(A)

r(mm)

d(mm)

Vm(mm3)

∫

U2mdt(mV2s)

35 2.3 1.1 7 2240 2.4 1.3 11 2945 2.7 1.4 14 4350 3.0 1.7 23 7455 3.1 2.0 31 8360 3.2 2.3 41 10165 3.6 2.5 53 12370 3.7 2.8 66 15075 3.8 2.9 72 17280 3.9 3.0 83 195

5.3.2 Travelling arc welding of steel C45

Welding runs were made under travelling arc conditions with v = 2 mm/s anddifferent arc currents. Before, during and after welding Urms was measured. Atypical example of the results obtained is presented in Fig. 5.9a, in which Urms isplotted as a function of welding time for an arc current of 55 A. After a certainwelding time, the first part of the weld starts to cool down and transforms partly tobainite, and the remaining austenite transforms subsequently to martensite. Bothtransformations contribute to the increase in signal level (3). This continues untila steady state situation is reached (4). At the end of the welding cycle (5) a peakis observed, which is due to the fact that after extinction of the arc the cooling rateincreases.

In order to test the validity of Eq. (5.6) for the case of bainite and martensiteformation, Urms was measured during a number of welding runs with arc currentsin the range of 40 to 60 A. Under steady state conditions, bainite and martensiteformation occur simultaneously and the measured AE power U 2 is the sum of theAE 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

Vm [mm3]

0 20 40 60 80 100

Um

2 dt [

mV

2 s]

0

40

80

120

160

200

Figure 5.8: Plot of ∫ U2mdt against Vm for spot welds produced on steel C45 witharc currents in the range of 35 – 80 A.

t [s]0 10 20 30 40 50

Urm

s [m

V]

0

2

4

6

8

10

(1)

(2)

(3)

(4)

(5)

(a) (b)

Figure 5.9: (a) The rms voltage Urms plotted against time during travelling arcwelding 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 con-dition; (5) extinction of the arc. (b) Transverse cross-section of the weld producedwith an arc current of 55 A.


cross-section for I = 55 A is shown. Optical microscopy observation and micro-hardness measurements showed that the microstructure in the central part of theweld (white area) is fully martensitic whilst in the zone adjacent to the interfacebetween the HAZ and the base metal bainite is formed (dark area). In that partof 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 for-mation. The transverse cross-sections were used to measure the areas Am and Ab ofthe martensitic phase and the bainitic phase respectively. The volume transforma-tion rates dVm/dt and dVb/dt of martensite and bainite formation were determinedby multiplying the areas by the travel speed v.

For each welding run, the average AE power under steady state conditions wascorrected for the electrical noise and the welding noise to obtain the AE powerU2total 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 2b ) can be

separated from that of martensite formation (U 2m). The results obtained are listedin Table 5.4.

Table 5.4: Results for travelling arc welding of steel C45.

I(A)

Am(mm2)

Ab(mm2)

U2total(mV2)

U2m(mV2)

U2b(mV2)

40 2.1 0.4 9.9 9.6 0.345 2.3 1.1 13.2 10.6 2.650 2.5 1.6 15.4 11.5 3.955 3.4 1.7 20.4 15.7 4.760 4.2 2.3 25.0 19.3 5.7

The values of U2b are plotted as a function of dVb/dt in Fig. 5.10. It can beseen that the AE power of bainite formation is proportional to the volume rate ofbainite formation, as predicted by theory. From the least-squares fit of the data toEq. (5.6) it follows that kb = 1.2 ± 0.1 × 103 V2s/m3. It should be noted that thisis approximately a factor 2 smaller than the value of km for steel C45. This can beexplained by the fact that the bainitic transformation involves less strain energy[6], and therefore the amount of plastic deformation during the growth of bainiteis smaller than during the growth of martensite. Assuming that the AE energygenerated during growth of the new phase is proportional to the strain energyinvolved, it immediately follows that the amount of AE energy produced per unitvolume of bainite is smaller than that produced per unit volume of martensite, i.e.kb is smaller than km.

Conclusions 87

dVb /dt [mm3/s]

0 1 2 3 4 5 6

Ub2 [

mV

2 ]

0

1

2

3

4

5

6

7

Figure 5.10: Plot of U2b as a function of dVb/dt for welding runs on steel C45produced 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 anumber of steels are presented and discussed. The results obtained are interestingboth from a fundamental and practical point of view.

The results of AE measurements during travelling arc welding show that therms voltage due to martensite formation is significantly larger than the rms voltagedue to the noise of the welding process. This indicates that acoustic emission can beapplied as a real-time monitoring technique for the detection of martensite forma-tion during welding of steel. Moreover, after extinction of the arc a characteristicpeak in the rms voltage is observed during cooling, which reflects the evolution ofmartensite formation.

In the beginning of this chapter a relationship was derived that describes theAE energy generated during bainite and martensite formation. In the subsequentsections this relation was tested using three carbon steels: 42CrMo4, 42MnV7 andC45.

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 squarevoltage (U2m) is proportional to the volume rate dVm/dt of martensite formation,as predicted by theory. The relation was also validated for spot welding of steel42MnV7 and steel C45. For both steels it was found that the integrated AE signal(AE energy) during cooling of a spot weld, ∫ U 2mdt, is proportional to the volumeVm of martensite formed.


In the case of travelling arc welding of steel C45, bainite and martensite for-mation occur simultaneously and both contribute to the measured rms voltage.A combination of the spot welding and the travelling arc welding results showedthat AE power due to bainite formation (U 2b) is proportional to the volume rate ofbainite formation (dVb/dt). This indicates that the relation is generally valid fordisplacive transformations involving acoustic emission.

References


[2] Y. Berlinsky, M. Rosen, J.A. Simmons and H.N.G. Wadley, Rev. of progressin 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).


89

Chapter 6

Kinetics of the martensitic

transformation studied by

means of acoustic emission

The transition from austenite to martensite is governed by thermodynamics andkinetics as explained in detail in chapter 2. The underlying thermodynamics de-termine the available driving force for transformation; the kinetics of a martensitictransformation depend solely on nucleation, because the growth of a martensiticcrystal usually occurs rapidly. It is well known that the mechanism of growth isdisplacive. How the phase nucleates, however, is even today not completely un-derstood. This is mainly due to the great speed of formation, which makes themartensitic transformation a difficult process to study experimentally. Neverthe-less, the results presented in this chapter show that acoustic emission, which hasrarely been used in the past, is a very suitable technique to study the kinetics ofthe martensitic transformation.

In section 6.1 the techniques commonly used to study the transformation ki-netics are reviewed. The derivation of the Koistinen-Marburger (KM) equationdescribing 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 transfor-mation in the four carbon steel specimens (C50, C60, C70 and C80) is monitoredby means of acoustic emission using the Gleeble 1500 thermo-mechanical simulator;the dilatation is measured simultaneously. In section 6.4 the AE measurements dur-ing cooling of the steels are analyzed and discussed. For the analysis of the AE datathe results obtained in chapter 5 are used. In this chapter it was shown that themeasured power of acoustic waves generated during the martensitic transformationis proportional to the transformation rate. Based on this relation, the volume frac-tion of martensite as a function of undercooling below the Ms temperature can becalculated, 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 asnucleation rates and average volume of martensite crystals. At the end of the chap-ter the experiments on a model system, a shape memory alloy, are discussed. Bothacoustic emission and confocal microscopy were employed to study the martensitictransformation 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 uponfurther cooling below this temperature [1, 2]. In order to gain more insight into thekinetics of the martensitic transformation, it is necessary to investigate quantita-tively how the transformation progresses upon continuous cooling. Characterizationof 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 equa-tion that describes the volume fraction of residual austenite as a function of tem-perature below the martensite-start temperature [8]. This equation, named the KMrelation, was obtained by fitting their (and other investigators’) data from X-raymeasurements on plain carbon steels. Some years later, Magee [9] showed that theKM relation can be derived theoretically from basic principles, taking into accountcertain assumptions regarding the nucleation and growth of martensite. The mostimportant assumption of this theory is that the average volume of the martensitecrystals is constant over the extent of the transformation. On this point the KMrelation is different from a theory formulated by Fisher [10], which assumes that theaverage volume decreases strongly as the transformation proceeds. The derivationof the KM relation and the partitioning model of Fisher are discussed in detail insection 6.2.

Speich and Fisher [3] showed that electrical resistivity and acoustic emissionmeasurements during martensite formation in an FeNi alloy are in agreement withthe partitioning model of Fisher [10]. Khan and Bhadeshia studied the developmentof martensite formation in partially bainitic steel using dilatometry [6]. For volumefractions up to 0.6 they found reasonably good agreement between experimentaland calculated values using the KM model. The discrepancy in the above modellingresults can probably be explained by differences in the morphology of the martensitecrystals and the austenite grain size. In the case of FeNi the grain size is large andplate martensite is formed, whilst in the partially bainitic steel the grain size issmall and lath martensite is formed.

Theoretical background 93

6.2 Theoretical background

Theories of the kinetics of the athermal martensitic transformation attempt to de-scribe the progress of martensite formation in a sample. The overall transformationrate is governed by nucleation and growth. The nucleation of martensite duringcooling is believed to take place at structural imperfections in the parent phase andthese pre-existing embryos (defects) are stimulated to grow into martensite crystalsat different degrees of undercooling below Ms; they have different energy barriersto activation [2]. Since growth can be very fast, each nucleation event directly leadsto the formation of a typical volume of the new phase. Thus the volume fractionof martensite varies only with the degree of undercooling expressing the athermalcharacter of the transformation.

It was postulated by Koistinen and Marburger [8] that the evolution of marten-site 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 Tbelow Ms, and C1 is a constant. This fraction is defined as the volume of martensitedivided by the volume of austenite that exists in the sample prior to the formationof martensite.

Magee [9] showed that this empirical equation (the KM relation) can be derivedfrom first principles assuming that in a temperature decrease dT , the number dNof new martensite crystals (plates or laths) that form per unit volume of austeniteis proportional to the increase in driving force ∆Gγ→α′

due to the temperaturedecrease

dN

dT= −C2

d(∆Gγ→α′

)

dT(∆Gγ→α′

< 0) (6.2)

with C2 a positive constant expressing the proportionality between the increase indriving force and the consequent increase in density of activated nucleation sites.The change in the volume fraction of martensite corresponding to the temperaturedecrease dT is then given by

df

dT= Ω(1− f)

dN

dT(6.3)

where (1−f) is the volume fraction of austenite available for further transformationand Ω is the average volume per newly formed crystal. Combining the Eqs. (6.2)and (6.3) yields

df

dT= −Ω(1− f)C2

d(∆Gγ→α′

)

dT(6.4)

Assuming that Ω, C2, and d∆Gγ→α′

/dT are constant over the extent of thetransformation and integrating from Ms (f = 0) to T gives


ln(1− f) = ΩC2d(∆Gγ→α′

)

dT(Ms − T ) (6.5)

This equation is equivalent to Eq. (6.1) with the positive parameter C1 expressedby

C1 = −ΩC2d(∆Gγ→α′

)

dT(6.6)

Thus ln(1 − f) is expected to vary linearly with T when the nucleation andgrowth of the martensite crystals in a sample obeys the characteristics as proposedby Magee.

As pointed out earlier, the assumption that Ω is constant is in contradictionwith the Fisher model [10], which assumes that Ω decreases strongly as the trans-formation progresses. The decrease of the average volume of martensite crystalscan be rationalized by considering that the initial martensite crystals divide theaustenite grain into progressively smaller compartments, which is known as thegeometrical partitioning effect. This model assumes that nucleation occurs homo-geneously as a result of random fluctuations. If on the other hand the nucleationis assumed to be heterogeneous (autocatalysis) it can be shown that the averagevolume 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 musttake place heterogeneously on pre-existing embryos, which is in favour of the Mageemodel [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 transformationin the carbon steel specimens were carried out using the Gleeble 1500 thermo-mechanical simulator as described in section 3.3. The schematic diagram of theacoustic emission analyzing system in combination with the Gleeble 1500 thermo-mechanical simulator is shown in Fig. 3.13. For comparison with the AE data, thecross-strain of the specimens was also measured simultaneously.

The specimens for acoustic emission and dilatometric measurements were ma-chined from as-received steel rods with a diameter of 12 mm. The experimentswere conducted using specimens with an effective free span (l1) of 10 mm and 5mm diameter (see Fig. 3.14). The distance between the grips (l2) is adjustableand has a minor influence on the cooling rate and thermal gradient in the axialdirection. Due to the large thermal gradient during austenitizing only a part of thespecimen is austenitized; a small part is partially austenitized since its temperatureis 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 95

α+γ

Ac3Ac1

850 ºC

Vm

α+γ

T

γα α

Figure 6.1: Schematic illustration of the temperature gradient over the sample. TheAc1 temperature is fixed, whilst the Ac3 temperature depends on carbon content.

present study. Moreover, it should be noted that the austenitized volume (= samplevolume) depends on the austenitizing temperature, which is controlled by the ther-mocouple in the middle of the specimen: the higher the austenitizing temperature,the larger the austenitized volume. However, the austenitizing temperature wasthe same for all specimens, resulting in approximately equal austenitized volumesfor all specimens. The prior austenite grain size was approximately 20 µm for steelC50, 50 µm for steel C60, 40 µm for steel C70 and 30 µm for steel C80. The grainsize of the austenite formed during austenitizing is expected to be of the same orderof magnitude for all specimens.

In the Gleeble 1500 thermo-mechanical simulator each specimen was electricallyheated in 20 seconds to the austenitizing temperature Ta = 850 C, austenitizedfor 10 seconds, and continuously cooled to ambient temperature. The cooling ratewas approximately 150 C/s between 800 C and 500 C. During cooling, the rootmean square (rms) voltage of the continuous acoustic emission signals was recorded.Measurements were repeated several times to check the reproducibility of the ex-periments; comparison of the measured data revealed no obvious or systematicdifferences.

In contrast to the AE parameter used in this work, the rms voltage, Speich andFisher [3] and Malygin [7] measured the ringdown counts of the acoustic emissionsignals to investigate the progress of martensite formation. This value is the numberof times the transducer output voltage exceeds a pre-set threshold. Since thisparameter depends on instrument settings, it has a non-unique relationship to theproperties of the source and is usually not very suitable for quantitative analysis. On


t [s]

0 5 10 15 20 25

T [º

C]

0

200

400

600

800C50C60C70C80

t [s]0 5 10 15 20 25

T [º

C]

100

200

300

400

500

C50C60C70C80

Figure 6.2: Cooling curves for steel C50, C60, C70 and C80 austenitized at Ta =850 C.

the other hand, the integrated squared voltage of all events (AE waves) is directlyproportional to the source intensity and thus more suitable for quantitative sourceanalysis [11].

After thermal cycling, each specimen was cut in the axial direction and themicrostructure was analyzed using an optical microscope (Olympus). The micro-hardness of the different microstructures was measured by means of a Vickers hard-ness tester (Buehler Ltd.) using a load of 100 g. From the cross-section images thevolume of martensite in the sample after cooling to room temperature (Vm) wasmeasured.

6.4 Study of steels C50, C60, C70 and C80

To investigate the carbon dependence of the kinetics, the steels C50, C60, C70and C80 were examined under identical conditions. In Fig. 6.2 the cooling curvesof the four steels are plotted, which exhibit inflections at temperatures where thetransformation takes place, due to the release of latent heat.

6.4.1 Calculation of the martensite volume fraction

Here the results of steel C60 are used as an example to demonstrate how the volumefraction of martensite can be directly evaluated from the measured AE power. Itcan be understood from Fig. 6.3a, where the mean square voltage U 2 of acousticemission is plotted as a function of time, that the martensite formation is reflected

Study of steels C50, C60, C70 and C80 97

by a peak in the AE data. When the transformation starts, the AE power rapidlyincreases to a maximum value and then tails off to the background noise level (U 2n )as the transformation proceeds. The transformation is assumed to reach completion(f → 1) where U2 becomes equal to U2n .

It has been shown in chapter 5 that the AE energy generated between the startand the finish of the transformation is proportional to the volume of martensiteformed with a proportionality factor k. In other words, the AE power U 2 is pro-portional to the volume transformation rate dV/dt,

U2 = kdV

dt(6.7)

with V the volume of martensite in the sample. With the integrated value ∫ U 2dt,with the integration running over the entire transformation range, and the valuefor 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 V2s/m3. Usingthis value of k, the AE power as a function of time can be converted to the volumetransformation rate dV/dt. The change of the martensite volume fraction f is thenormalized volume transformation rate, which in combination with Eq. (6.7) canbe written as

df

dt=

1

Vm

dV

dt=

U2∫

U2dt(6.8)

Thus df/dt ∝ U2 and the calculated values for the transformation rate are plottedin Fig. 6.3b. The fraction of martensite f at a certain time can now be evaluated asthe area under the peak up to that specific time, divided by the total area under thepeak (Fig. 6.3c). It should be noted that f(t) is independent of the values foundfor 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 thethermal gradients in the sample leading to localized transformations as the sampleapproaches the martensite-start temperature. This results in a broadening of thepeak; at the signal maximum the thermal gradients are believed to have diminisheddue 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 ofthe signal is even shorter.

The measured AE data for steel C50, C70 and C80 were analyzed accordingto the above procedure and the results for Vm (from the cross-sections) and k aregiven 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 workand values of the dislocation densities in martensite observed in similar carbonsteels by other researchers. This correlation is important since it gives better in-sight into the origin of acoustic emission during martensitic transformations. The


t [s]

0 2 4 6 8 10 12 14 16 18

f

0.0

0.2

0.4

0.6

0.8

1.0

df/d

t [s-1

]

0.0

0.2

0.4

0.6

0.8

1.0

U2 [m

V2 ]

0

10

20

30

40(a)

(b)

(c)

Figure 6.3: The values of (a) the AE power, (b) the transformation rate and (c) thevolume fraction of martensite plotted against time during the martensitic transfor-mation in steel C60.


Table 6.1: Values for the volume of martensite Vm and the proportionality factor k(Eq. 6.7) for the studied steels.

Steel Vm(mm3)

k(103 V2s/m3)

C50 76 0.28C60 89 0.61C70 90 0.48C80 92 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 areplotted in Fig. 6.4 as a function of carbon content. It can be seen that both param-eters 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 dis-locations are involved in the slipping process to accommodate deformation straindue to the shape change. The decrease of the dislocation density beyond 0.6% Cmay be due to twinning partly accounting for the accommodation of transformationstrain; twinning involves relatively less dislocations as compared to slipping. Thecorrelation between k and ρ indicates that the AE energy generated per unit vol-ume of martensite is proportional to the dislocation density in the martensite. Thedislocations arise due to the shear transformation mechanism, which results in slip-ping to accommodate the shape deformation [13]. Thus the shear mechanism andthe movement of dislocations are strongly related and the limited number of dataavailable does not permit an unambiguous distinction between both types of AEsources. In many previous AE studies on twinned martensites the shear mechanismitself was regarded as the major source of AE during martensitic transformations[14, 15].

The exact origin of AE generated during martensite formation is important forthe analysis of the acoustic waves (events). For twinned martensites one AE eventwas 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 movementsaccompanying the martensitic transformation are the actual source of acoustic emis-sion. Because the dislocation movements keep up with the discontinuous (jerky)interface motion, the formation of a lath or plate of martensite presumably resultsin many AE events. A model describing the generation of acoustic waves duringinterface 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


0.4 0.5 0.6 0.7 0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Carbon content (wt%)

k [1

03 V2 sm

-3]

k

1.0

1.5

2.0

2.5

3.0

3.5 Disl. density ρ [10

15 m-2]

ρ

Figure 6.4: The values of k and the dislocation density ρ [12] as a function of carboncontent. The corresponding tendency of both entities suggests a close relationbetween the dislocations and the actual source of acoustic emission during themartensitic transformation.

measured AE power. For all four steels the values of df/dt, f and −ln(1− f) areplotted 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 (upto f = 0.995 at least). In comparison, the plots for steel C50, C60 and C70 areinitially (20 – 30 K below Ms) also linear. Below a certain temperature, however,a crossover to another regime is observed. Finally, as the transformation reachescompletion (f = 0.99), the dependence is linear again. The crossover to the non-linear (intermediate) regime indicates that the transformation kinetics decrease atrelatively high volume fractions (see right-hand axis in Fig. 6.5c). It is arguedthat this is due to the increase of carbon content of the transforming austenite. Itis easily verified by the KM relation that the progress of the transformation at acertain temperature is lower for higher carbon contents, since an increase in carboncontent implies a decrease of Ms. In relation with this it should be remembered thatthe 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 regionA (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 theaustenite is, according to the phase diagram, enriched in carbon. The martensitictransformation in that region starts at a lower temperature and probably becomespredominant beyond the crossover observed in Fig. 6.5c. In the intermediate regimethe measured (total) volume transformation rate dV/dt is a superposition of thevolume transformation rates from different austenite parts in region B, each with


T [ºC]

100 150 200 250 300

-ln(1

-f)

0

1

2

3

4

5

f

0.0

0.2

0.4

0.6

0.8

1.0

df/d

t [s

-1]

0.0

0.2

0.4

0.6

0.8

1.0 C50C60C70C80

0.99

0.95

0.98

0.65

0.87

f

(c)

(a)

(b)

Figure 6.5: The progress of the martensitic transformation in four carbon steelsas a function of undercooling: (a) the change in volume fraction; (b) the volumefraction and (c) the volume fraction data plotted on a logarithmic scale. The solidlines through the data are least-squares fits to the KM equation. The dashed lineindicates the second linear regime for steel C50, C60 and C70.


a specific carbon content. This transformation behaviour is consistent with thedifferences in carbon content of the formed austenite in region B based on the Fe-Cphase diagram.

After cooling the martensite volume fractions in region A and B are denotedas fA and fB = (1 − fA) respectively. To describe the progress of the martensitictransformation 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 infraction in a sample with a homogeneous composition and thus a single value forMs and for C1; it does, in its original form, not account for a sample with a rangeof carbon contents. However, region B is relatively small and the KM equation forthe entire sample can be approximated as

f = fA(1− e−C1(Ms−T )) +

n∑

i=1

fBi

(

1− e−CBi(MBi−T ))

(6.9)

for which region B is assumed to consist of n volume elements and fBi representsthe 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 temperatureof 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 austeni-

tized region A (f < fA), and calculated values for d∆Gγ→α′

dT from MTDATA.

Steel fA(-)

Ms

(C)C1(K−1)

d∆Gγ→α′

dT(J/molC)

ΩC 2(mol/kJ)

42CrMo4 1.00 302 0.051 7 7C50 0.83 317 0.054 7.1 7.6C60 0.90 282 0.067 7.2 9.3C70 0.95 248 0.055 7.0 7.9C80 1.00 211 0.046 6.9 6.7

In Fig. 6.5c, the solid lines through the data represent least-squares fits to thefirst term of Eq. (6.9), i.e. up to the crossover (0 < f < fA). The extractedfit parameters are listed in Table 6.2. The fitting results for steel 42CrMo4 arealso 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 inEq. (6.9) cannot describe all measured data the fitting of the data was executed inan iterative way by extending the fit up to higher fractions until the next data pointstarts to deviate relatively strongly from the least-squares fit through the previousdata points. From the fits it follows that the volume fraction of martensite in regionA 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 partiallyaustenitized region for higher overall carbon contents, since the Ac3 temperaturedecreases with increasing carbon content. The fitting curves in Fig. 6.5c show that


Table 6.3: Fitting results for the KM equation for the two-phase region B.

Steel fB1(-)

MB1

(C)CB1(K−1)

fB2(-)

MB2

(C)CB2(K−1)

fB3(-)

MB3

(C)CB3(K−1)

C50 0.095 247 0.052 0.072 190 0.050 0.005 110 0.046C60 0.060 222 0.060 0.035 170 0.055 0.005 110 0.046C70 0.025 180 0.053 0.010 150 0.050 0.005 115 0.046

the fits to the first term of Eq. (6.9) also describe a part of the deviation fromthe initial linear regime. It can be easily verified that when the KM equation ismodified with a pre-factor fA the data in a ln(1 − f) vs. T plot start to deviatefrom linearity before fA is reached.

For f > fA the transformation in region B becomes predominant and the mea-sured data result from the simultaneous transformation of a number of austeniteparts in this region as explained above. Consequently, comparison of experimentaldata with the KM theory is more complex in this regime. However, at approxi-mately 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, thefraction 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 asf → 1, which indicates that when the transformation in the C50, C60 and C70sample 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 elementsin region B were considered in the second term of Eq. (6.9) (n = 3). The fittingresults for C50, C60 and C70 are shown in Fig. 6.6 and the extracted fit parametersare given in Table 6.3. A physical interpretation of the MBi is not meaningfulbecause of the correlation between the fitting parameters and the limited numberof elements (n = 3). This is supported by the fact that the fit near each MBi

deviates strongly from the data. It is important to notice that beyond f = 0.99 allfitting curves show a good correspondence with the measured data.

The good agreement between the experimental data and the KM relation upto f = fA confirms that C1 (= ΩC2d∆Gγ→α′

/dT ) (Eq. (6.6)) is constant over theextent of the transformation for the steels studied. This does not necessarily implythat Ω, C2, and d∆Gγ→α′

/dT are all constant, because in the derivation of the KMequation the product of these parameters is assumed constant to solve differentialEq. (6.4); mathematically no explicit assumptions about each individual parameterare made. Using a thermodynamical database, the value of ∆Gγ→α′

was calculatedas a function of temperature in the range of 0 – 400 C for the studied steels. Overthe whole temperature range the slope d∆Gγ→α′

/dT is approximately constant foreach carbon content and the values obtained are given in Table 6.2.

On combining the above results, it follows that the product ΩC2 is constantover the extent of the transformation (Table 6.2). It should be noted that thesetwo parameters are not separable; if it would be possible for example to determine


0.99

0.95

0.98

0.65

0.87

f

T [ºC]

100 150 200 250 300 350

-ln(1

-f)

0

1

2

3

4

5

-ln(

1-f)

0

1

2

3

4

5

-ln(

1-f)

0

1

2

3

4

5

0.99

0.95

0.98

0.65

0.87

f

0.99

0.95

0.98

0.65

0.87

f(a)

(b)

(c)

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.


f 0.0 0.2 0.4 0.6 0.8

C1

[K-1

]

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

T [ºC]

100 140 180 220

C1

[K-1

]

0.02

0.04

0.06

0.08

Figure 6.7: Plot of C1 against f for steel C80 resulting from the direct evaluationof Eq. (6.4). The inset shows the values of C1 as a function of temperature. Thehorizontal line represents the average value of C1.

the average volume of martensite crystals by other experiments, information aboutthe 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 shownearlier implies that C1 is constant, which strongly indicates that Magee’s assump-tions [9] are justified. Here it is shown that the data of C80 can be analyzedwithout making Magee’s assumptions of a constant C1 in order to test the differ-ential Eq. (6.4). This equation can be validated directly over the whole extent ofthe transformation since the values of df/dT can be calculated as

df

dT=

df

dt

(

dT

dt

)−1

(6.10)

with dT/dt the cooling rate values deduced from the cooling curve (Fig. 6.2) bynumerical differentiation. With the values of (1− f), the value of C1 can be evalu-ated as a function of fraction or temperature (Eq. (6.4) and Eq. (6.6)) as shown inFig. 6.7. Despite the considerable scatter in the data it can be seen that C1 doesnot change significantly during the transformation. The scatter in the data maybe partly due to the inaccuracies in the calculated values of the cooling rate. Thisbecomes more pronounced for high volume fractions where df/dt→ 0. A constantvalue fitted to the data yields C1 = 0.047 K−1, which is in good agreement with


5 mm

2.2 mm0.2

AB

Figure 6.8: The cross-section of the steel C60 sample. The volume ratio of region Aand region B is in good agreement with the observed crossover at f = 0.90 shownin Fig. 6.5c.

the previously obtained value C1 = 0.046 K−1. With d∆Gγ→α′

/dT being constantit follows that the product ΩC2 is constant.

If the assumption of a constant Ω and C2 is called into question, from a math-ematical point of view Ω ∝ 1/C2 is a valid solution. Assuming now that Ω variesaccording to Fisher [10] as Ω = Ω0(1 − f)10, C2 would increase progressively withundercooling which is strongly doubted from a physical point of view. It should beremembered that C2 describes the distribution of effectiveness of nucleation sites asa function of undercooling. A constant value of C2 (Magee [9]) means that the num-ber density of activated nucleation sites varies linearly with activation energy. Onthe other hand, an increasing C2 with decreasing T implies that the number den-sity of activated nucleation sites increases progressively for temperatures far belowMs. The constant distribution of effectiveness of nucleation sites is considered tobe more realistic. Thus the argument by Fisher that Ω decreases progressively withincreasing volume fraction [10] seems invalid for the steels studied in the presentwork.

6.4.5 Microstructural analysis

After completing the experiments each specimen was cut in the axial direction formicroscopic analysis. The cross-section of steel C60 depicted in Fig. 6.8 clearlyshows a region A, which is fully martensitic, and a region B, which contains a


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 microstructureformed in the partially austenitized region B (middle part) and the fully martensiticmicrostructure formed in the completely austenitized region A (right).

mixed microstructure. It can be seen in this figure that the ratio of the two regionsis approximately 9:1, which is in good agreement with fA = 0.90. In Fig. 6.9the magnifications of the microstructures of the partially austenitized regions forall four steels are shown. The partially austenitized regions of steel C50, C60and C70 show a mixed microstructure of ferrite and martensite. On the left-handside the pearlite-ferrite base material of steel C50, C60 and C70 can be seen. ForAc1 < Ta < Ac3 the ferrite remains untransformed during austenitizing, whereas thepearlite transforms to austenite with a carbon concentration depending on Ta. Themicrostructure on the right-hand side corresponds to martensite with a measuredmicro-hardness of approximately 600 HV0.1. The micro-hardness of the ferrite andpearlite is approximately 250 HV0.1. The magnification of the partially austenitizedregions show less untransformed ferrite for the higher carbon steels; for C80 thereis no untransformed ferrite.

6.4.6 Martensite-start temperature Ms

As expected, the martensite-start temperature Ms resulting from the fit of the datato the KM equation decreases with increasing carbon content. The results areplotted in Fig. 6.10. From the least-squares fit through the Ms data it is deducedthat Ms varies with carbon content xC as


0.50 0.55 0.60 0.65 0.70 0.75 0.80200

220

240

260

280

300

320

Ms

[ºC

]


Figure 6.10: The values of Ms extracted from the fit for steel C50, C60, C70 andC80. The solid line represents the least-squares fit of the data to Eq. 6.11

Ms = 495− 355xC (6.11)

with Ms inC and xC in wt% carbon. As mentioned earlier, the transformation

kinetics of the four carbon steels are approximately the same as the transformationreaches completion. The last per cent of the transformation in C50, C60 and C70can be described by the KM equation with Ms = 220 C and C1 = 0.046 K−1. Thisvalue was obtained by drawing a line in Fig. 6.5c parallel to the data of C80 andthrough 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 theabove equation yields a carbon content for the last part of the transformation ofapproximately 0.77, which is in good agreement with the expected value from theFe-Fe3C phase diagram.

For comparison with the AE data, the dilatation of the samples was measuredsimultaneously. The results in Fig. 6.11 show that the changes in slope are inreasonable agreement with the martensite-start temperatures derived from the fitsthrough the AE data. It should be noted that the transformation kinetics can alsobe derived from the dilatation curves; in principle this would allow comparisonwith the transformation kinetics from the AE measurements. However, with thecross-strain dilatometer the radial length change in the middle of the specimen ismeasured, which may not be representative for the total volume change taking placeduring the martensitic transformation. Therefore, an unambiguous comparison ofthe transformation kinetics derived from the dilatation signal and the AE signal iscomplicated and not pursued in this work.


T [ºC]

0 100 200 300 400 500 600 700 800

∆d [

µm]

-20

0

20

40

60

80

100 C50C60C70C80

Figure 6.11: Plot of the dilatation against temperature for steel C50, C60, C70and C80. The arrows indicate the temperature at which the slope changes: the Ms

temperature.

6.4.7 Rate constant C1

According the KM equation the progress of the transformation from austenite tomartensite in carbon steels is quantitatively described by the Ms temperature andthe value of C1. The effect of material and process parameters such as chemicalcomposition and austenitizing temperature on the transformation can be expressedin terms of Ms and C1. The martensite-start temperature Ms is determined bythermodynamics and the rate constant C1 describes the kinetics, i.e. the progressof the transformation for a certain undercooling.

The values of C1 obtained from the fits are plotted as a function of carboncontent in Fig. 6.12. It can be seen that C1 has a maximum value for steel C60and decreases approximately linearly by 30% as the carbon content increases from0.6 to 0.8% C. This means that the transformation rate for a given degree of un-dercooling is the highest for steel C60. Now the question can be raised how thecarbon dependence of C1 can be explained in terms of the underlying physical pa-rameters Ω, C2 and d∆Gγ→α′

/dT . The carbon dependence of d∆Gγ→α′

/dT isrelatively small compared to the decrease in C1 (see Table 6.2). The values forΩC2 = 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 similarmorphology. Therefore, it is justified to assume that Ω is approximately the samefor the steels studied. Thus the carbon dependence of C1 is expected to mainlyresult from a change of C2 with carbon content.

The dependence of C2 on carbon content can in its turn be explained by the


0.5 0.6 0.7 0.8

0.04

0.05

0.06

0.07


C1

[K-1]

C1

7

8

9

10Ω

C2 [m

ol/kJ]ΩC2

Figure 6.12: The carbon dependence of C1 and the product ΩC2. With Ω assumedconstant, the rate constant C2 appears to correlate with the dislocation densityshown in Fig. 6.4.

change in dislocation density with carbon content shown earlier; both have a max-imum at approximately 0.6 % C. When a martensite crystal is formed, dislocationsare created in the neighboring austenite due to the fact that the shear stress ac-companying the shape change exceeds the yield strength of the austenite [13]. Itshould be noted that the dislocations in the austenite (strain energy) do not in-crease the driving force for transformation because the dislocations are inheritedin the martensite upon transformation. However, these dislocations have a strongeffect on C2 because the dislocation debris leads to extra nucleation sites apartfrom the embryos initially present in the austenite. This is consistent with theheterogeneous nucleation model, which Magee used to validate that Ω does notchange during transformation [9]. Martensitic crystals tend to occur in clusters;the probability of nucleation is greater in the vicinity of a previously formed crys-tal (dislocation debris) than it is for random nucleation in untransformed regions.This effect is known as strain-induced autocatalysis [13]. Thus for steels with highdislocation densities (C60) the autocatalytic effect is stronger, which is expressedby a high value of C2. A higher C2 means that the number density N of activatednucleation sites for a given driving force increases and thereby the progress of thetransformation is enhanced.


T [ºC]

200 220 240 260 280 300 320

-ln(1

-f)

0

1

2

3

4

5

f

0.0

0.2

0.4

0.6

0.8

1.0

0.00

0.01

0.02

0.03

0.04

0.05

0.06df

/dt

[s-1

]

0.99

0.95

0.98

0.65

0.87

f

(c)

(a)

(b)

Figure 6.13: The progress of the martensitic transformation in steel 42CrMo4 asa function of undercooling: (a) the change in volume fraction; (b) the volumefraction and (c) the volume fraction data plotted on a logarithmic scale. The solidline through the data is least-squares fit to the KM equation.


(1)

(3)(4)

(5)

(2)

(6)

Figure 6.14: Schematic drawing of the experimental set-up used for AE measure-ments 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 specimenduring continuous cooling in the dilatometer, presented in section 4.2.2, is analyzedin order to compare the transformation kinetics in this steel with the results for thecarbon steels discussed earlier. It should be remembered that for the dilatometerset-up as described in section 3.4 (see Fig. 3.15), the thermal gradients in the sampleduring austenitizing are negligible compared to the thermal gradients when usingthe thermo-mechanical simulator set-up. In fact, no partially austenitized regionoccurs in comparison with the case of steels C50, C60 and C70; all the decomposingaustenite has the same carbon content. This is the main advantage of the dilatome-ter 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 ispredominantly caused by the strong attenuation due to the small-sized waveguide.Moreover, the natural cooling rate of a sample in the dilatometer is low comparedto the typical cooling rates obtained using the Gleeble thermo-mechanical simula-tor. Consequently, only for steel 42CrMo4 a complete transformation to martensiteis achieved in the dilatometer; under the same conditions steels C50 – C80 do nottransform to martensite completely.

In order to facilitate comparison with the KM kinetics, the measured rms voltageduring martensite formation in steel 42CrMo4, shown in Fig. 4.12b, was analyzedaccording the procedure described for steel C60 (see section 6.4.1). The results areshown in Fig. 6.13. From the least-squares fit of the data to the KM equation itfollows that Ms = 302 C and C1 = 0.051 K−1. It is important to notice that the fitis very accurate over the whole extent of the transformation, like the result for steelC80. The overall results for steel 42CrMo4 are in line with the results obtained forsteels C50 – C80 discussed earlier, and support the conclusions drawn in that partof the study.

Study of a shape memory alloy 113

t [s]

0 10 20 30 40 50 60

T [º

C]

-20

0

20

40

60

U2 [m

V2 ]

0

200

400

600

800 (a)

(b)

Figure 6.15: (a) The mean square voltage measured during the martensitic trans-formation in the CuAl-based shape memory alloy, and (b) the cooling curve of thesample.

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 furtherinsight into the evolution of martensite crystals as the transformation progresses.Optical Confocal Laser Scanning Microscopy (CLSM) observation of the surfaceof a shape memory alloy (SMA) was carried out. This CuAl-based SMA has amartensite-start and -finish temperature of approximately 25 C and 0 C respec-tively. In addition to the optical in-situ observations, acoustic emission experimentson 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 diam-eter of 15 mm and a thickness of 1 mm (see Fig. 6.14). A waveguide was spot weldedonto the sample to transport the acoustic waves to the sensor and to protect the


sensor from excessive heating and/or cooling. The sample was austenitized (heatedto a temperature of 90 C) using an air heater. Subsequently the sample was cooledto −20 C using a Dewar filled with liquid nitrogen. During cooling, the root meansquare (rms) voltage of the continuous acoustic emission signals was measured. TheAE 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. InFig. 6.16c, the solid line through the data represents a least-squares fit to the KMequation (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 rateconstant C1 is approximately four times larger expressing that the transformationkinetics are much faster in the CuAl-based SMA. Again, the high accuracy of thefit implies that C1 is constant over the extent of the transformation, which stronglyindicates that Magee’s assumptions are justified, i.e. the average size of martensitecrystals Ω is constant over the extent of the transformation. On this point theresults obtained for the shape memory alloy are in good agreement with the resultsobtained for the carbon steels. Furthermore, optical CLSM observation (and videorecording) of the surface upheavals on the shape memory alloy was carried outin order to gain a better insight into the evolution of martensite crystals as thetransformation progresses.

6.6.2 Optical Confocal Laser Scanning Microscopy observa-

tions

The SMA sample for the CLSM-observations, with a diameter of 15 mm and athickness of 10 mm, was placed in a petri-dish and by filling the petri-dish with ice-water the martensitic transformation was induced. Prior to the measurements, thesample surface was ground and polished at a high temperature by using hot water.Although the grinding scratches could not be removed completely, the polishingwas found adequate to reveal the surface upheavals due to the transformation.

During cooling the confocal microscope acquired 1 image per second, whichrevealed interesting information about the growth of martensite plates; six char-acteristic micrographs are shown in Fig. 6.17. The transformation starts with theformation of thin plates of martensite with distinct orientations. A thickening ofthese plates is observed as the transformation proceeds, and for some plates thisthickening clearly occurs in a discontinuous manner. This thickening effect makesthe interpretation in terms of nucleation and growth complex; it is not clear whethera 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 atthe beginning and at the end of the transformation. A partitioning effect and aprogressive decrease of Ω as proposed by Fisher [10] is clearly not observed. Thisis in good agreement with the results of the AE measurements on the SMA dis-cussed in the previous sub-section. It should be noted that the characteristics of

Study of a shape memory alloy 115

T [ºC]

-20 -10 0 10 20 30 40

-ln(1

-f)

0

1

2

3

4

5

f

0.0

0.2

0.4

0.6

0.8

1.0

0.00

0.05

0.10

0.15

df/d

t [s

-1]

0.99

0.95

0.98

0.65

0.87

f

(c)

(a)

(b)

Figure 6.16: The progress of the martensitic transformation in the shape memoryalloy as a function of undercooling: (a) the change in volume fraction; (b) thevolume 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.


Figure 6.17: Observation of the change in microstructure during cooling of the SMAsample over a period of 30 seconds.

the martensitic transformation in this model system may be different from those insteel to some extent. However, the observed evolution of martensite plates in themodel system is in line with the Magee model, and supports the results of the AEexperiments on the carbon steels.

6.7 Conclusions

It has been demonstrated that acoustic emission is a very suitable technique tostudy the kinetics of the martensitic transformation. This is based on the fact thatthe volume fraction of martensite f as a function of time t during cooling can bederived directly from the measured AE power U 2, since U2 ∝ df/dt. In general, itcan be concluded that the acoustic emission measurements during the martensitictransformation in the studied carbon steels (C50, C60, C70, C80 and 42CrMo4) andthe CuAl-based shape memory alloy have resulted in a better understanding of thenucleation and average volume of martensite crystals. A more detailed summaryof the results obtained is described below.

The analysis of results obtained for the four carbon steels shows that for steelC80 the fraction data as a function of temperature T can be fitted to the KMequation with high accuracy up to f = 0.99, whilst for steel C50, C60 and C70deviations occur beyond f = 0.83, f = 0.90 and f = 0.95 respectively, due to thepartially austenitized region in the samples. The decrease of the transformation

Conclusions 117

kinetics at these relatively high volume fractions is caused by the relatively highcarbon content of the transforming austenite in the partially austenitized region.The extent of this region decreases with carbon content, which was confirmed bymicrostructural analysis. It was shown that a modified KM equation could takeaccount for the presence of the partially austenitized region.

The rate constants C1 extracted from the fits together with the values ford∆Gγ→α′

/dT calculated to be constant, imply that ΩC2 is constant over the ex-tent of the transformation. These parameters are not separable; however, from aphysical point of view it is most realistic that both Ω and C2 are constant duringmartensite formation. Furthermore, the values of Ms obtained from the fits werefound to be in reasonable agreement with the values extracted from the dilatationcurves, and with literature values.

The observed carbon dependence of the rate constant C1 can be attributed tothe dependence of C2 on the carbon content, which is consistent with the changein dislocation density with carbon content. The dislocation debris in the residualaustenite leads to extra nucleation sites (autocatalysis) which is expressed by ahigher value of C2. Also, the acoustic emission energy generated per unit volumeof martensite (k factor) has been found to scale with the change in dislocationdensity in the formed martensite as the carbon content is varied. This is consideredto be important for the interpretation of AE measurements during martensitictransformations.

In contrast to the thermo-mechanical simulator set-up used for the study ofsteels C50 – C80, no significant thermal gradients are present during austenitizingof steel 42CrMo4 in the dilatometer set-up. Consequently, the sample is completelyaustenitized (no partially austenitized regions) and therefore the experimental datacould be fitted to the KM equation with high accuracy up to f = 0.99, like steelC80.

Optical CLSM observation of the surface of a CuAl-based shape memory alloygave valuable information about the sizes of martensite crystals formed duringtransformation. It was shown that both small and large plates are formed both atthe beginning and at the end of the transformation, which is in good agreement withthe Magee model. Furthermore, AE experiments on the shape memory alloy wereperformed and the fraction data obtained could be described by the KM equationwith high accuracy over the whole extent of the transformation, like for the carbonsteels.

The overall results presented in this chapter validate the KM model that de-scribes the kinetics of the martensitic transformation. According to Magee thismeans that Ω is constant over the extent of the transformation in the studied ma-terials. It is important to notice that this transformation behaviour is differentfrom the evolution of martensite crystals in the Fisher model, which predicts thatΩ strongly decreases as the transformation proceeds.

References



[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, Ameri-can 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 progressin quant. nondestr. eval. 5, 1345 (1986).

[12] T. Furuhara, S. Morito and T. Maki, J. de Phys. IV 112, 255 (2003).


[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

Chapter 7

Analysis of acoustic emission

signals originating from

bainite and martensite

formation

In this chapter the characteristics of acoustic waves generated during bainite andmartensite formation are studied. The results are discussed in a semi-quantitativemanner, a thorough quantitative analysis of signals is not feasible because of thelimited frequency bandwidth of the system and the unknown sample response. Thefrequency spectra of acoustic emission signals are interpreted using a dislocationsource model adopted from acoustic emission studies of plastic deformation [1, 2, 3].It is assumed that the predominant source of acoustic emission during displacivetransformations is the movement of dislocations, that is the slip taking place dur-ing growth in order to relieve internal stresses. The results show that the meanfrequency of waves generated during bainite formation is significantly larger thanthat of martensitic waves. This difference in the spectral density of the waves canbe explained by the difference in interface motion of the two transformations andthe consequently different behaviour of the dislocations involved.

Before discussing the generation of acoustic waves during growth of marten-site and bainite in section 7.2, the theoretical model concerning acoustic emissionduring plastic deformation is reviewed (section 7.1). In section 7.3 the analysis ofcontinuous acoustic emission is discussed in general terms, and subsequently theexperimental procedure is described in section 7.4. In section 7.5 the measuredfrequency spectra during the bainitic and the martensitic transformation in steelC45 and steel 16MnSi are discussed. This chapter ends with a discussion of theproportionality factors between the acoustic emission energy and the transformedvolume (k factors) obtained for the steels studied in chapters 4, 5 and 6.

121

122 Chapter 7: Analysis of acoustic emission signals originating from ...

7.1 Acoustic emission during plastic deformation

The observed correlation between the proportionality factor k and the dislocationdensity ρ (section 6.4.2) suggests that the dislocation motion (slip) accompanyingdisplacive transformations is the source of acoustic emission, like in the case ofconventional plastic deformation. During displacive transformations a large amountof the free energy release is dissipated in the process of slip [4, 5], which occurs whenthe interface moves into the austenite matrix. Previous studies of acoustic emissionduring plastic deformation have resulted in a model for the generation of acousticwaves during dislocation motion, which is discussed in this section.

The movement of dislocations is the dominant mechanism of plastic deformationin most crystalline solids, especially metals. A large amount of the total deformationenergy is released as thermal energy in the specimen, which means that most ofthe energy generated by the moving dislocations is dissipated as heat in the crystallattice. This thermal energy is the cumulative effect of the generation of phonons bydislocations as they pass through the crystal lattice. The phonons with relativelylow frequencies can be considered as acoustic emissions, since they can be detectedas such.

The phonon generation process is based on the fact that there are specific posi-tions 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 Peierlsbarrier before moving. As the dislocation moves away from a minimum-energy posi-tion 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 vibrationalwave in the lattice. This process is considered as the basic mechanism by whichdislocations can cause acoustic emissions.

To explain the phonon generation, Kiesewetter and Schiller presented the anal-ogy with electrons, which emit photons (Bremsstrahlung) when they start to moveor stop [1]. They argued that dislocations behave similarly and can be considered asa source of AE (acoustic Bremsstrahlung) when they are accelerated or decelerated.

In the past several attempts were undertaken to derive quantitative informationabout the acoustic waves generated by moving dislocations. Scruby et al. used theGreen’s function to obtain the surface displacement resulting from the expansion ofdislocation loops [2]. In another approach Kiesewetter and Schiller proposed thatthe acoustic waves are produced at active Frank-Read sources where dislocationsare generated and move outwards [1]. In this work [1] the moving dislocations areconsidered to radiate acoustic waves in those instances where they are accelerated.The main difference between the above models is that the former approach takesinto account the directionality of the source, i.e. the orientation of the moving dislo-cation with respect to the surface upon which the transducer is mounted. However,for quantifying the wave energy from dislocation motion this is not required. Ac-tually, the two approaches lead to essentially the same expression of the AE energyradiated per second (E), 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 isproportional to the final radius a of the loop, or

u =c22Dc31

bav (7.1)

with b the Burgers vector, D the distance from the loop to the surface at which themeasurement takes place, and c1 and c2 the longitudinal and the transverse wavevelocities respectively. The power E radiated by one dislocation can be assumed tobe proportional to u2 [6]. Then for N dislocations moving at a velocity v, E canbe written as

E = αNc42

D2c61b2a2v2 (7.2)

with α the proportionality constant between E 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 harmonicallywith frequency ω0, is

E = NπGω0

2

5c31b2l2v2 (7.3)

with G the shear modulus and c1 the velocity of the longitudinal waves. Since thelength l can be viewed as proportional to the radius a of a dislocation loop, bothformalisms give the same dependence on the dislocation characteristics N , b, l andv. It should be noted that the measured AE power is proportional to the AE energyradiated per second: U2 ∝ E.

For plastic deformation the strain rate ε can be written, according to Orowan’sequation, as

ε = ρmbv (7.4)

where ρm is the mobile dislocation density. A change in ε can in principle be dueto a change in ρm or a change in v [7]. Many experimental results [1, 2, 8] of AEduring tensile testing showed that the mean square voltage (the AE power) of thecontinuous acoustic emission is proportional to the strain rate, which is expressedby

U2 ∝ ε (7.5)

A typical plot of U2 against ε during tensile testing is shown in Fig. 7.1. It can beseen that the peak signal increases with strain rate. Fig. 7.2 shows that AE powerU2 at a given strain is proportional to the strain rate. This linear dependenceindicates that the change in strain rate is primarily due to a change in the mobiledislocation density, because Eqs. (7.2) and (7.3) show a linear dependence on N .Therefore, v can be assumed constant. If a change in ε would be due to a changein v, a quadratic dependence would have been found, U 2 ∝ ε2, see Eqs. (7.2) to(7.4).

On this point it should be noticed that above relation (Eq. (7.5)) is very similarto the relation between the AE power and the volume transformation rate duringdisplacive transformations, given by


Figure 7.1: The solid lines represent the acoustic emission power as a function ofstrain for strain rates of 1, 5 and 20 mm/min; the gauge length LG of the specimenswas 30 mm (From Ref. [8]). It can be seen that the intensity of the signal increaseswith increasing strain rate. The dashed line is the stress-strain curve.

Figure 7.2: The acoustic emission power measured at a given strain during tensiletesting of polycrystalline aluminium with different strain rates (From Ref. [2]. Themeasured 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 tothe strain rate.

Dislocation dynamics during displacive transformations 125

U2 = kdV

dt(7.6)

which was derived from basic principles and validated for different steels in chapter5. If it is assumed that the strain rate that occurs during displacive transformationsis proportional to the volume transformation rate, Eqs. (7.5) and (7.6) are in factidentical.

The similarities between plastic deformation and displacive transformations con-cerning acoustic emission discussed above, support the argument made earlier thatit is actually the movement of dislocations during displacive transformations thatconstitutes the generation of acoustic waves. How the motion of dislocations takesplace during displacive transformations is explained in the next section.

7.2 Dislocation dynamics during displacive trans-

formations

7.2.1 Nucleation and growth of martensite

As discussed in chapter 6, the nucleation of martensite takes place heterogeneouslyon defects in the parent phase, which nuclei are stimulated to grow at differentdegrees of undercooling [4]. Since the growth is relatively fast, the overall rate oftransformation is governed by nucleation. Nevertheless it is the process of growththat is relevant for AE waves being radiated, and therefore martensite growth isdiscussed 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 mostlikely sites for such nuclei are the interfaces of inclusion particles and grain bound-aries [4]. The growth of the new phase is constituted by the movement of the glissileinterface, which consists of arrays of parallel dislocations [4, 5]. As the interfacemoves into the austenite matrix the dislocations glide on appropriate slip planes.After the nucleation barrier has been overcome, the interface motion is usuallyassumed to be discontinuous until the growth of a plate or lath is completed. In or-der for the interface to move, the existing interfacial dislocations have to overcomepossible obstacles (crystal defects) and new dislocation loops have to be generated,which requires an activation energy [9]. Both types of barriers, which are smallerthan 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 formedand has grown to a certain size at time t0. The overall rate of the discontinuousinterface 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 anotherbarrier is encountered where the interface comes to a halt again. This is illustratedin Fig. 7.3 for two jerky movements (events), with τ the mean life time of such


(a) (b) (c) (d)

t0 + τt0

l

Figure 7.3: Schematic drawing of the discontinuous interface motion during thegrowth 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 meanlife time τ .

an event and l the mean displacement of the interface in this time interval. Thisdisplacement 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. Duringsuch an event (displacement), dislocation motion takes place similar to the case ofconventional plastic deformation as described in section 7.1, and this can be re-garded 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. Conse-quently, after several events the interface comes to a rest and a plate or lath ofmartensite is formed (Fig. 7.3d).

From the frequency analysis of acoustic waves information can be obtained aboutthe time scales involved in the underlying process. The frequency fs of an acousticwave is inversely proportional to the time τ that the source operates [10], which isexpressed by

fs =1

2τ(7.8)

By inserting Eq. (7.7) into Eq. (7.8) the mean frequency of acoustic waves at thesource f s becomes

f s =v

2l(7.9)

On this point it should be remembered that the mean frequency of acoustic emissionmeasured at the transducer f may be significantly different from f s due to the

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 ofacoustic emissions during displacive transformations is different from the sourcemodels found in the literature on this subject. In many previous studies the vol-umetric 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 mod-els one AE event was usually related to the formation of one martensite crystal.Although the amount of dislocation motion is presumably strongly related to thevolume change and the shear mechanism, the time and length scales at the sourcecorresponding to an AE event are significantly different in view of the dislocationsource model. Since the dislocation motion is related to the jerky interface move-ment, 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 bet-ter framework for a suitable analysis of the acoustic waves, which carry dynamicalinformation of the source mechanism.

7.2.2 Nucleation and growth of bainite

In contrast to martensite, bainite grows at relatively small driving forces, and there-fore nucleation events are almost always confined to the austenite grain bound-aries, which contain the most potent nucleation sites (smallest activation energy).Whereas the martensite nucleation is diffusionless, the nucleation stage of bainiteprobably involves partitioning of carbon, which leads to a greater reduction in freeenergy [16]. The growth of bainite is diffusionless and the crystallographic features(slip) and surface relief effects are identical to those associated with martensite for-mation. After transformation most of the excess carbon in the bainitic ferrite willpartition into the residual austenite [17].

Compared with martensite, the overall interface motion (growth rate) is rela-tively slow, presumably because of the plastic work that is done as bainite grows[17]. Concerning the discontinuous interface motion, the interface mobility and themean displacement of the interface, no clear differences between martensite andbainite have been reported in the literature.

7.3 Analysis of continuous acoustic emission

Plastic deformation and displacive transformations are usually classified as sourcesof continuous AE. For continuous emission the recorded wave is a superpositionof a number of individual waves, with frequencies depending on the characteristictimes 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 principlemeaningful AE parameters, cannot unambiguously be related to source character-istics. Moreover, rise time and amplitude depend on the threshold setting andthe wave-recording settings, such as peak definition time and hit definition time


(see section 3.1). Therefore, results obtained in different set-ups cannot easily becompared, which makes these AE parameters less useful.

Since the actual wave can be written as the summation of the original waves dueto the individual events, the frequency information of each individual AE sourceis contained in the observed signal, regardless of the strong overlapping. It is im-portant 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 in-formation about time scales of the source processes can be obtained. However, byusing the same set-up a qualitative (comparative) study of different processes canbe undertaken, as is done in this work.

In the literature, no research has been reported about frequency analysis ofacoustic emission during displacive phase transformations. On the other hand,elaborate studies of the acoustic waves generated during plastic deformation havebeen carried out. In the early 1980s, Rouby et al. presented a thorough analysis ofthe acoustic waves generated by dislocations moving between two steady positions[3]. They showed that by using the elastic Green’s function an expression for theelastic waves could be obtained, in the time and frequency domain. Some yearslater, Schaarwachter and Ebener performed AE measurements on deforming copperusing two resonant transducers [18]. They showed that the AE power spectraldensity shifts to higher frequencies with increasing strain, which could be wellexplained by changes in the event lifetime in the course of work hardening. Recently,Vinogradov et al. [19] investigated the acoustic emission spectral density duringcyclic deformation of copper single crystals. They observed an increase in the meanfrequency of the AE power spectral density and attributed this to the decrease inthe dislocation mean free path.

7.4 Experimental details

The experiments were performed using steel C45 (plate thickness of 5 mm) andsteel 16MnSi (plate thickness of 3 mm). The chemical compositions of the steels aregiven in Table 3.2. Spot welds were made as described in section 3.2. Monitoringthe transformations in a spot weld has the advantage that no waveguides needto be used, and therefore attenuation and mode-conversion effects are minimized.Reflections of waves may play a role but they are presumably damped sufficientlybefore they arrive at the transducer (see section 3.2). Since the geometry of theplate remains unchanged during each experiment, the effects of reflections andinterference are constant, and therefore the variations in the power spectral densityshould be attributed to changes related to the forming microstructure. In otherwords, since the effects of sample response are identical for bainite and martensiteformation, it is justified to compare the wave characteristics in terms of differencesbetween 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 16MnSialso resulted in bainite and martensite formation upon cooling.

After extinction of the arc, several AE parameters were measured simultaneouslyduring cooling of the sample and recorded for further analysis. First, the rms voltageUrms of the continuous acoustic emission was measured. Second, the waves (AEevents) constituting the total AE power were recorded. As discussed earlier, thecontinuous AE signal with average amplitude Urms can be regarded as a resultantof individual events. Since the frequency of the waves is analyzed, it is not requiredto isolate each AE event, which would be practically impossible. For statisticalaveraging on the other hand it is important to record many AE events. To measurethe events with high amplitudes a floating threshold was set with a minimum valueof 32 dB; the floating threshold automatically adjusts the detection threshold inorder to obtain an acquisition rate of approximately 300 events/second. By usingsuch a threshold only the highest amplitudes are measured, which are generatedduring bainite and martensite formation.

For each wave U(t) its power spectral density P (f) (with f the frequency ob-tained by Fast Fourier Transformation, FFT) was calculated. The distributionP (f) can be most simply represented by the mean frequency f , which is related toP (f) by

f =

∫

P (f)fdf∫

P (f)df(7.10)

with the integration running over the bandwidth of the system. The mean frequencydata were averaged in each time interval of 1 second, which means averaging overapproximately 300 measurements.

7.5 Results and discussion

According to the procedure described above, Urms and f were measured as a func-tion of time during cooling of steel C45 and steel 16MnSi. The result for steel C45is shown in Fig. 7.4. It can be seen that the mean frequency of waves generatedduring bainite formation fb is approximately 460 kHz, whilst the mean frequencyof waves generated during martensite formation fm is approximately 380 kHz. Al-though 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 ofevents during bainite formation, τb, is shorter. This in its turn can be attributedto a smaller interface mean free path lb under the assumption that the interfacemobility (or velocity v) is not significantly different for bainite and martensiteformation. The limited number of data available does not permit an unambiguousexplanation for the relatively small mean free path of interfacial dislocations in


0 2 4 6 8 10 12 14 16 18 20 22 240

100

200

300

400

500

MB

W

t [s]

f [k

Hz]

f

0

2

4

6

8

10

Urm

s [mV

] Urms

Figure 7.4: The mean frequency f and the average voltage Urms of waves generatedduring bainite and martensite formation in steel C45 after spot welding (W). Therms voltage shows two peaks: B = bainite formation, M = martensite formation.The standard deviation in the value of f is expressed by the error bars.

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 300

100

200

300

400

500

600

W

M

B

t [s]

f [k

Hz]

f

0

1

2

3

4

5

6

Urm

s [mV

]

Urms

Figure 7.5: The mean frequency f and the average voltage Urms of waves gener-ated during bainite and martensite formation in steel 16MnSi after spot welding(W). The rms voltage shows two peaks: B = bainite formation, M = martensiteformation. The standard deviation in the value of f is expressed by the error bars.


case of bainite, but it could be related to the self-induced strain during bainiteformation.

In Fig. 7.5 the AE parameters measured during bainite and martensite forma-tion in steel 16MnSi are plotted. It can be seen that the results obtained are similarto the results obtained for steel C45. During the bainitic and martensitic trans-formation in this steel the values obtained for fb and fm are approximately 480kHz and 400 kHz respectively. Compared with the result for steel C45, the valuesof both fb and fm for steel 16MnSi are approximately 20 kHz higher. This smalldifference may be caused by the difference in chemical composition of the steels, orthe 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 ofthe source can be obtained due to the unknown sample response and the limitedtransducer bandwidth. This is explained in more detail below. Concerning thesource, the interface motion during the transformation, it is believed that the meaninterface velocity and the mean interface displacement have a value in the orderof magnitude of v = 100 m/s and l = 100 nm respectively. Inserting these valuesinto Eq. (7.9) yields a mean frequency of waves at the source of f s = 500 MHz. Itshould be noted that this is approximately three orders of magnitude larger thanthe measured values of f at the transducer output. This simple calculation confirmsthat no quantitative information about the source mechanism can be derived fromthe measured f . The observation that surface displacements with much lowerfrequencies are measured at the position of the transducer can be explained bythe fact that the source function S(t) is convoluted with the material responsefunction M (see section 2.2.1).

Consider a frequency spectrum of the source S(f) with a mean frequency of500 MHz as shown in Fig. 7.6. In order to evaluate the material response M onthe source function S, the theoretical work of Scruby et al. is adopted [20]. Theycalculated the frequency spectra of surface displacements generated by force dipoleswith life times in the range of 30 ns to 1 µs (f s = 0.5 – 15 MHz). The result of theircalculations is shown in Fig. 7.7. It can be seen that the width of the spectrumMS(f) is inversely proportional to the life time. For example, in case of a lifetime of 30 ns the spectrum extends up to approximately 30 MHz. Furthermore, itcan be seen that the low frequency components in the spectrum have a relativelyhigh intensity. Based on this result, the material response on the source functionconsidered in this work was evaluated. The response on a source with a mean lifetime of 1 ns, MS(f), is schematically shown in Fig. 7.6. In the inset of Fig. 7.6the low frequency spectrum of the surface displacements MS(f) is shown togetherwith 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 fb > fm implies that the mean life time of events at thebainitic source is smaller than the mean life time of events at the martensitic source.


0 200 400 600 800 1000

MS( f )

MS( f )

T

S( f )

frequency [MHz]

pow

er

0.0 0.2 0.4 0.6 0.8 1.0

frequency [MHz]

pow

er

Figure 7.6: Schematic drawing of the source spectrum S(f), the spectrum of thesurface displacements MS(f) resulting from the material response on the sourcefunction, and the frequency response of the transducer T .

Since the material response M is constant, a source spectrum with a higher meanfrequency results in a frequency spectrum of the surface displacements with strongerhigh frequency components, and thus a higher measured mean frequency.

Noise analysis

Regarding the electronic noise contributing to Urms (0.28 mV) it should be men-tioned that the noise waves have highest amplitudes of approximately 20 dB (= 1mV) and a mean frequency fn of approximately 500 kHz. They may give a smallcontribution to each measured wave, and thus the calculated mean frequency maybe affected by that of the noise to some extent; the noise signals are continuouslypresent on the background. However, by setting the minimum threshold at 32 dB itis 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 wavesfrom which f is calculated have amplitudes that are at least four times larger thanthe amplitudes of the noise signals.

In Fig. 7.8 the amplitudes of the waves measured during the bainitic and marten-sitic transformation in steel C45 are plotted. Although the rms voltage is relativelylow during bainite formation, the maximum amplitudes are of the same order ofmagnitude 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 emis-sion, is more strongly dependent on the rate of occurrence of waves than on the


Figure 7.7: Fourier transforms of the surface displacements MS(f) generated byforce 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) decreasesfor (a) to (d).


0 2 4 6 8 10 12 14 16 18 20 22 24

-20

0

20

40

60

MB

W

t [s]

ampl

itude

[dB

]ampl.

0

2

4

6

8

Urm

s [mV

] Urms

Figure 7.8: Plot of the amplitudes of waves generated during bainite and martensiteformation in steel C45 after spot welding. W = welding, B = bainite formation, M= martensite formation.

0 20 40 60 80 100

380

400

420

440

460

480

500

520

B

M

N

ηn

f [k

Hz]

100 % noise Bainite Martensite

Figure 7.9: Mean frequency data of steel C45 from Fig. 7.4 plotted against the per-centage of noise contributing to the signals measured during (B) bainite formationand (M) martensite formation; (N) represents the mean frequency for the back-ground noise. The solid line indicates that for martensite formation fm is linearlydependent on ηn.

Discussion of proportionality factors k 135

amplitudes of the waves. Furthermore, it should be remembered that with thefloating threshold setting only the waves with the highest amplitudes are measuredas shown in Fig. 7.8; the waves with amplitudes lower than the threshold, however,also contribute to the measured Urms. During martensite formation the overlappingof waves is presumably very strong. The amplitudes of the signals measured duringthe bainitic and martensitic transformation in steel 16MnSi were in the same rangeas for steel C45 shown in Fig. 7.8.

Since the amplitudes of bainitic and martensitic waves are of the same order ofmagnitude, the possible influence of noise would have a similar contribution to bothsignals. It is therefore unlikely that the noise can cause the significant differencebetween fb and fm. Moreover, increasing and decreasing of the threshold did notresult in significant different values for fb and fm.

In both the results of steel C45 and steel 16MnSi a slight increase in fm isobserved when the martensite signal tails off to the background noise. In orderto investigate whether this is due to noise or related to a change in the growthdynamics as the martensitic transformation reaches completion, a systematic noiseanalysis was carried out using the results obtained for steel C45.

In order to quantify the noise contribution to each measured wave, it is assumedthat the percentage of noise in each wave ηn is equal to the typical amplitude ofthe noise waves (20 dB) divided by the amplitude of the total wave. Based on thisassumption ηn was calculated for the bainitic and the martensitic waves using theamplitude values from Fig. 7.8. For example, ηn corresponding to the maximum ofthe martensite peak (46 dB) is 5 %, whilst ηn for the tail of the martensite peak (34dB) is 20 %. The corresponding values of fm (see Fig. 7.4) are approximately 380kHz and 400 kHz. To verify if this increase in fm can be attributed to an increasein ηn, f is plotted against ηn in Fig. 7.9, together with the result for backgroundnoise: ηn = 100 %, fn = 500 kHz. The straight line that can be drawn through thedata points for martensite formation and the data point for 100 % noise indicatesthat the increase in fm at the tail of the martensite peak is due to a larger noisecontribution.

The result for bainite formation, fb 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 throughthe martensite data points discussed above. This supports the argument madeearlier that the difference between fm and fb cannot be attributed to the influenceof noise.

7.6 Discussion of proportionality factors k

In section 6.4.2 it was reported that a clear correlation exists between the AEenergy per volume unit of forming martensite (km) and values of the dislocationdensities (ρ) in martensite observed by other researchers [21]. This indicates thatan interpretation of k factors based on dislocation densities is most suitable. Forcompleteness, the plot of k and ρ as a function of carbon content from section


0.4 0.6 0.80.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4


k m [

103 V

2 sm-3]

km

0

1

2

3

4

5

6

7

8

Disl. density ρ [10

15 m-2]

ρ

0 1 2 3 40.0

0.2

0.4

0.6

0.8

ρ [1015 m-2]k m

[10

3 V2 sm

-3]

Figure 7.10: The values of the proportionality factor km and the dislocation densityρ [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 steel42CrMo4.

All values of km and kb, obtained for martensite and bainite respectively, arelisted in Table 7.1. The km factors for steel C45 and steel 42CrMo4 are evaluatedfor both welding and thermo-mechanical simulator experiments. It can be seenin Table 7.1 that the km factors obtained in the case of the thermo-mechanicalsimulator experiments are one order of magnitude smaller, which is primarily dueto the attenuation caused by the waveguide. This suggests that the km factors ofsteels C50, C60, C70 and C80 (thermo-mechanical simulator experiments) shouldbe multiplied by approximately a factor 10 for a proper comparison with the kmfactors of steels C45, 42CrMo4 and 42MnV7 (welding experiments).

For steel C45 it was found that km > kb, and this is in line with experimentaldata from electron microscopy showing that the dislocation density in martensite ishigher than in bainite of similar composition [22]. It can be seen in Table 7.1 thatkm of steel C45 is approximately a factor two larger than km of both steel 42CrMo4and 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 differentfrom the carbon contents of the steels studied in Ref. [21]. Although only a fewvalues of ρ are found in the literature to compare with, it appears from the inset thatk is proportional to ρ, indicated by the solid straight line. Based on this and theabove discussed results of the k factors for different steels and different experimentaltechniques, it is proposed that the proportionality factor k of a certain steel can bewritten as

k = ρW (7.11)

with W the total transfer function, which is to a large extent affected by the waveg-uide (if used).

Table 7.1: Values of the proportionality factors km and kb for the studied steels.

Steel km(103 V2s/m3)(Welding)

km(103 V2s/m3)(Gleeble)

kb(103 V2s/m3)(Welding)

C45 2.38 0.23 1.242CrMo4 0.85 0.1142MnV7 1.15C50 0.28C60 0.61C70 0.48C80 0.41

7.7 Conclusions

In this chapter the measured frequency spectra of acoustic waves generated duringbainite and martensite formation are presented and discussed. It is explained thata quantitative analysis of signals is not feasible mainly because of the unknownsample response. However, the specimen configuration is the same during bainiteand martensite formation, and therefore the results can be interpreted in a semi-quantitative manner. The frequency spectra are interpreted using a dislocationsource model adopted from acoustic emission studies of plastic deformation. Thismodel is utilized because the movement of dislocations is considered to be theactual source of acoustic emission during displacive transformations; the dislocationmotion is related to the jerky interface movement and therefore the formation of asingle 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 significantlylarger than fm. This difference is attributed to the characteristics of (jerky) inter-face movement of the two transformations: the mean life time of events τ , the meanvelocity v and the mean free path l. The difference in f implies that τb < τm. Byassuming that v is not significantly different for bainite and martensite formationthis difference in τ means that lb < lm.


Through a systematic analysis of noise signals it is shown that noise signalshave only a small influence on the measured frequency spectra when the floatingthreshold is at least 10 dB above the highest amplitude values of noise signals. Thebackground noise signals give a small contribution to measured signals and have arelatively high mean frequency. They can explain the slight increase in the meanfrequency observed at the end of the martensitic transformation, however, the largedifference between the mean frequency of bainitic and martensitic waves cannot beattributed to the noise signals.

Finally, all the proportionality factors between the acoustic emission energy andthe transformed volume k obtained for the steels studied in this thesis are comparedand interpreted based on dislocation densities ρ. The analysis of values of k andρ indicates that k is proportional to ρ. This means that the measured AE powerduring 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 emissionmeasurements 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).



[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).


139


[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 industrialsectors such as ship building and the automotive industry. In order to meet therequirements for steel applications, new multi-phase steels are being developed. Themicrostructure of these steels consists of a variety of different phases, which leads tosuperior material properties - a combination of high strength with good formability.For the development of such steels research is required to gain more insight intothe underlying microstructure and the mechanisms by which it is formed.

This thesis describes unique acoustic emission experiments during martensiticand bainitic transformations in steel. The main objective of this work is to obtaina better understanding of the growth mechanism and kinetics of these solid-statephase transformations that can occur in carbon steel. In view of fact that acousticemission is an unexplored technique in this kind of steel research, this study alsoaims to give a good overview of the possibilities and limitations of acoustic emissionas a real-time monitoring technique for the evolution of bainite and martensiteformation.

A concise introduction to acoustic emission (AE) and phase transformations insteel is provided in chapter 1. Subsequently, in chapter 2 a elaborate descriptionof the basic principles of the acoustic emission technique is given, including anoverview of the developments and applications of the technique. Special attentionis paid to the influence of noise sources on the measured signal, the sensitivityof the technique and the material and sensor response to the original AE signal.Furthermore, the theory of martensitic and bainitic phase transformations in steelis discussed.

The experimental equipment for measuring acoustic emission generated dur-ing phase transformations in steel is described in chapter 3. First, some technicaldetails concerning the used AE equipment are provided. Secondly, the methodsutilized for applying a thermal cycle to the studied specimens are described. Inmost experiments an arc welding device or a thermo-mechanical (welding) simu-lator is used; in some cases a dilatometer or a furnace in combination with a saltbath is employed. For each set-up the procedure is described for the measurementof acoustic emission during continuous cooling of the studied specimens from ele-vated 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 canbecome 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 canstrongly attenuate the signal. A waveguide is typically a metal rod which conductsthe acoustic signal from the specimen to the sensor. One end is designed for acous-tic coupling with the specimen; the other end is usually conical to accommodatethe mounting of an AE sensor. The use of a waveguide is required when the sizeof the specimen is smaller than the diameter of the sensor or when the specimenbecomes very hot during the experiment.

The results of AE measurements during various phase transformations in anumber of steels are discussed in chapter 4. During cooling of a medium carbonsteel specimen (C45) in the Gleeble simulator two distinct peaks in the AE data areobserved, which can be attributed to bainite and martensite formation. Consistentwith 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 thatthe bainite reaction mechanism in steel C45 is diffusionless and is best described interms of the displacive transformation model. In contrast, no acoustic emission isdetected during the diffusion-controlled transformation from austenite to ferrite insteel Fe360.

The effect of the austenite grain size on the evolution of the bainite and marten-site formation in steel C45 is studied by varying the austenitizing temperature. Itis found that upon lowering the austenitizing temperature, i.e. with decreasingaustenite 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 in-sight into the evolution of both phase transformations, i.e. the relative amounts ofbainite and martensite formed. Furthermore, it is seen that the martensite-starttemperature decreases with decreasing austenitizing temperature. Both effects canbe well explained with the existing theory for bainitic and martensitic transforma-tions.

To facilitate comparison between the AE technique and conventional dilato-metry, the acoustic emission and the dilatation are measured simultaneously duringthe martensitic transformation in steel 42CrMo4. From both signals a similar valueof the martensite-start temperature is derived, which indicates that the sensitivityof 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 noiseratio of acoustic emission is relatively low in this set-up due to the small size ofsamples studied and the necessity of using a waveguide. At the end of chapter 4the results of AE measurements during cooling of specimens in the salt bath arediscussed. These experiments make it possible to study large-sized samples, and thetypical cooling rate is such that bainite is formed in the sample. After correctingthe raw data for the noise due to oxidation, a clear peak in the AE data is obtainedthat can be attributed to bainite formation.

In chapter 5 the study of the acoustic emission energy generated during bainiteand martensite formation is presented. A theoretical model is derived from basic

143

principles that predicts that the measured AE energy at the sensor is proportionalto the volume of martensite or bainite formed. Both travelling arc welding andspot welding are employed in order to test the theoretical prediction. Travellingarc welding has the advantage that the volume transformation rate can be easilyand accurately determined from the transverse cross-sections of the welds. Onthe other hand, determining the AE power originating from martensite and/orbainite formation during travelling arc welding is rather complicated due to thenoise from the arc. Fortunately, for common welding conditions the contribution ofthe welding noise to the total measured AE power is very small. Moreover, the noisecontribution can be determined and thus corrected for. During the first seconds ofwelding no transformations take place, and consequently only the welding noise ismeasured. After a certain time the first part of the weld cools down and a significantincrease in signal level is observed. Under steady state conditions, the measuredAE power is the sum of the powers due to transformations and welding noise. Atthe end of the welding cycle a peak is observed, reflecting that after extinction ofthe arc the volume transformation rate increases due to a higher cooling rate. Bothobservations, the increase during welding and the peak after welding, indicate thatacoustic emission can be used as a real-time monitoring technique for martensiteformation during welding. This can be of practical importance since martensite inwelds can lead to cold cracking.

In order to validate the above mentioned theoretical prediction, AE experimentsare performed during welding with different heat inputs. The analysis of the resultsshows 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 relationbetween the AE energy and the volume transformed. During cooling of a typicalspot weld a peak signal is observed similar to the peak signal after travelling arcwelding. By integrating the area under a peak in a plot of AE power versus time, theAE energy is obtained. Also the results of spot welding experiments with differentarc 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 otherhand, it is quite difficult to determine the volume of martensite in a spot weldbecause this requires that the spot weld is cut exactly in the middle. In accordancewith the procedure described above, the proportionality factor between the AEenergy 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 differencesin the chemical composition of the steels.

In chapter 6 the kinetics of the martensitic transformation in a number of car-bon steels are studied using the acoustic emission technique. By studying thekinetics of the martensitic transformation one can obtain fundamental knowledgeabout the transformation such as nucleation rates and average volume of marten-site crystals. The experimental results are compared to a theoretical model for thetransformation kinetics, the Koistinen and Marburger (KM) relation. According tothe KM equation, the progress of the transformation from austenite to martensiteis quantitatively described by the martensite-start temperature and the value of

144 Summary

the rate constant. The martensite-start temperature Ms is determined by thermo-dynamics and the rate constant C1 describes the kinetics, i.e. the progress of thetransformation for a certain undercooling.

For four carbon steel specimens (C50, C60, C70 and C80), the AE power ismeasured during cooling using the Gleeble thermo-mechanical simulator. Based onthe energy-volume relation from chapter 5, the volume fraction of martensite as afunction of undercooling below the start-temperature can be easily derived, whichallows a direct comparison of experimental data with theoretical predictions of thetransformation kinetics. The overall results obtained for the four carbon steelsshow that the fraction data as a function of temperature can be described by theKM equation with high accuracy. For the low carbon steels a part of the specimenis austenitized in the two-phase region. At relatively high volume fractions thetransformation in the partially austenitized region becomes predominant, which isreflected by a decrease of the transformation kinetics caused by the relatively highcarbon content of the transforming austenite in the partially austenitized region.The good agreement between the experimental data and the Magee model (KMrelation) indicates that the nucleation of martensite takes place heterogeneouslyand that the average volume of martensite crystals is constant over the extent ofthe transformation. This is different from the Fisher model, which predicts thatthe average size of martensite crystals strongly decreases as the transformationproceeds.

From the fits of the data to the KM equation the values of martensite-starttemperature and the kinetic rate constant are obtained for each steel. As expected,the martensite-start temperature decreases with increasing carbon content. Thekinetic rate constant has a maximum value for steel C60, which means that thetransformation 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 beexplained by the change in dislocation density with carbon content. This is basedon the fact that values of the dislocation density in martensite in similar carbonsteels observed by other researchers, also show a maximum at approximately 0.6 %C. When a martensite crystal is formed, dislocations are created in the neighboringaustenite due to the fact that the shear stress accompanying the shape changeexceeds the yield strength of the austenite. The dislocation debris leads to extranucleation sites apart from the embryos initially present in the austenite. Thiseffect is known as autocatalysis.

Analogous to the procedure described in chapter 5, the proportionality factorsbetween the acoustic emission energy and the transformed volume of martensitek are determined for the four carbon steels. Also, the values of k are found toscale with the change in dislocation density in the formed martensite as the carboncontent is varied. The corresponding tendency of both entities suggests a closerelation between the dislocations and the actual source of acoustic emission duringthe martensitic transformation.

Based on the correlation between the dislocation density and the acoustic emis-sion 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 martensiteand bainite crystals. The growth of the new phase is constituted by the movementof a glissile interface, which consists of arrays of parallel dislocations. When the in-terface moves, the existing interfacial dislocations have to overcome barriers, whichare smaller than the nucleation barrier, and this causes the jerky motion of the inter-face. After the interface has overcome a barrier, elastic strain is suddenly releasedto produce vibrational waves in the lattice: acoustic emission. This dislocationsource model is adopted from acoustic emission studies of plastic deformation, andsubsequently modified to describe acoustic emission generation during displacivephase transformations. Furthermore, the results of AE measurements during plas-tic deformation and displacive transformations are compared and discussed. Theresults obtained for both processes show many similarities, which supports the ear-lier argument that it is the dislocation motion during displacive transformationsthat constitutes the generation of acoustic waves.

Subsequently, the characteristics of acoustic waves generated during bainite andmartensite formation are studied. The results are discussed in a semi-quantitativemanner, since a thorough quantitative analysis of signals is not feasible because ofthe limited frequency bandwidth of the system and the unknown sample response.The frequency spectra of acoustic emission signals are interpreted using the dislo-cation source model discussed above. The results show that the mean frequency ofwaves generated during bainite formation is significantly larger than that of marten-sitic waves. This difference in the spectral density of the waves can be explainedby the difference in interface motion of the two transformations and consequentlythe different behaviour of the dislocations involved.

This chapter ends with an overview of the proportionality factors between theacoustic emission energy and the transformed volume (k factors) obtained for thesteels studied in this thesis. An analysis of k factors and dislocation densitiesindicates that the measured AE power during displacive transformations is linearlydependent on the dislocation density, which is identical to the dependence found incase 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 namebedoeld is voor niet-vakgenoten. Daarna gevolgt een omschrijving van de belang-rijkste resultaten zoals vermeld in dit proefschrift.

Akoestische emissie

Akoestische emissie (AE) is het verschijnsel dat elastische golven ontstaan in eenmateriaal doordat er lokaal een abrupte spanningsverlaging optreedt. Voorbeeldenvan processen die gepaard gaan met het optreden van akoestische emissie zijnscheurgroei, plastische deformatie, oxidatie en diffusieloze fasetransformaties.

De in het materiaal opgewekte trillingen ten gevolge van de spanningsverlag-ing bij de AE-bron zijn voor het menselijk oor niet hoorbaar, want de frequentiesvan het geluid liggen ongeveer tussen de 50 kilohertz en 10 Megahertz, terwijlhet menselijke gehoor grofweg van 20 hertz tot 20 kilohertz reikt. Deze hoogfre-quente golven veroorzaken hele kleine verplaatsingen aan het oppervlak, ongeveer1000 maal kleiner dan de afstand tussen twee atomen. Om deze verplaatsingente kunnen detecteren wordt een sensor gebruikt met een piezo-elektrisch kristal.De trillingen veroorzaken vervormingen van het kristal en zo ontstaat er een elek-trische spanningsverandering over het kristal. Na versterking van het signaal kande geluidsgolf bestudeerd worden met behulp van een computer.

Fasetransformaties in koolstofstaal

Staal is een van de belangrijkste constructiematerialen en wordt gebruikt voor defabrikage van o.a. schepen, bruggen en auto’s. Het verbeteren van de eigenschap-pen van staal voor de verschillende doeleinden is daarom nog steeds van grootbelang. De eigenschappen van een staalsoort worden tijdens het productieprocesvoornamelijk bepaald door de microstructuur die tijdens het afkoelen ontstaat.

147

148 Samenvatting

Legeringselementen spelen hierbij een grote rol, maar ook eventuele warmtebehan-delingen.

Staal is een legering van ijzer (Fe) met minimaal 0.02 % koolstof (C), en heeftverder vaak legeringselementen zoals bijvoorbeeld mangaan, silicium, chroom enmolybdeen. Bij hoge temperaturen heeft staal een andere kristalstructuur danbij kamertemperatuur. Boven 730 C heeft staal een kubisch-vlakken gecentreerdkristalrooster dat aangeduid wordt met austeniet. Koelt het staal langzaam af,dan gaat austeniet over in ferriet dat een kubisch-ruimtelijk gecentreerde structuurheeft. Zo’n verandering van kristalrooster wordt een (structurele) fasetransformatiegenoemd. Het gevormde ferriet is de meest bekende vorm van staal: goed vervorm-baar en redelijk sterk. Koelt het staal daarentegen heel snel af, dan transformeerthet austeniet naar martensiet. De kristalstructuur daarvan lijkt sterk op die vanferriet, maar door het snelle afkoelen zit koolstof als het ware in het ijzerroostergevangen en ontstaat er een bepaalde interne spanning. Martensiet is daardoorheel hard, maar tegelijkertijd ook bros waardoor het makkelijk breekt. Naast fer-riet en martensiet kan austeniet ook nog naar perliet of bainiet transformeren, dierespectievelijk een gelaagde en een naaldvormige kristalstructuur hebben. Deze mi-crostructuren ontstaan als austeniet iets sneller wordt gekoeld dan bij de vormingvan ferriet, maar minder snel dan voor martensiet nodig is.

Wat betreft de manier waarop de nieuwe fase groeit kunnen twee transfor-matiemechanismen onderscheiden worden. De vorming van ferriet en perliet uitausteniet zijn diffusie gestuurd en dientengevolge relatief langzame processen. Hier-mee vergeleken is de vorming van martensiet een veel sneller ofwel abrupter pro-ces, en vindt plaats via een zogenaamd displacive mechanisme. Hierbij bewegende 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 groteaantallen ijzeratomen transformeert het austeniet naar martensiet. Dit gebeurt inhet grensvlak dat het oorspronkelijke kristal (austeniet) scheidt van het nieuwekristal (martensiet). Een diffusieloze fasetransformatie begint doorgaans spontaanbij een bepaalde temperatuur die wordt aangeduid als de martensiet-start temper-atuur; bij deze temperatuur worden de eerste martensietkristallen gevormd. Demartensietkristallen groeien met zeer grote snelheid tot hun eindafmeting, en degroeisnelheid is onafhankelijk van de temperatuur. Bij verdere temperatuurdalingvindt de vorming van nieuwe martensietkristallen plaats totdat de omzetting isvoltooid.

Omdat de martensietkristallen niet passen in de austeniet waaruit ze wordengevormd, treden spanningen in het grensvlak op. Tijdens de groei van een marten-sietkristal zal de spanning geminimaliseerd worden door zogeheten slipping (dislo-catiebewegingen); een proces dat het kristalrooster niet verandert. De spanningver-laging die hiermee gepaard gaat heeft de trillingen in het kristalrooster tot gevolg:akoestische emissie. De akoestische emissie die tijdens martensietvorming wordtgegenereerd is dus een direct gevolg van de collectieve beweging van atomen. Ditbetekent dat men door een proces te monitoren met de akoestische emissie tech-niek 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 detransformatie van austeniet naar bainiet beter te begrijpen. Er is namelijk tot opheden nog steeds geen consensus tussen wetenschappers wat betreft het groeimecha-nisme van bainiet, d.w.z. is het een diffusionele of diffusieloze fasetransformatie.

Hoewel het reeds enkele decennia bekend is dat akoestische emissie wordt gepro-duceerd tijdens een martensitische fasetransformatie, is de techniek in het verledenmaar zelden toegepast voor fundamenteel onderzoek van fasetransformaties in staal.Dit proefschrift beschrijft een uitgebreid en nauwkeurig onderzoek van de akoestis-che emissiesignalen tijdens martensitische en bainitische transformaties in koolstof-staal.

Dit proefschrift

Allereerst wordt in hoofdstuk 1 een korte beschrijving van akoestische emissie enfasetransformaties gegeven, inclusief de achtergrond van dit onderzoek. Vervol-gens worden in hoofdstuk 2 de grondslagen, ontwikkelingen en toepassingen vanakoestische emissie beschreven. Speciale aandacht wordt besteed aan de invloedenvan ruisbronnen op het signaal, de gevoeligheid van de techniek, en de verander-ingen van de geluidsgolven ten gevolge van materiaal- en sensorrespons. Tot slotworden in dit hoofdstuk de fasetransformaties in staal beschreven, met de nadrukof de martensitische en bainitische fasetransformatie.

In hoofdstuk 3 staan de gebruikte methoden beschreven om de stalen proef-stukken een warmtebehandeling te geven. Deze vier warmtebehandelingstechniekenzijn: een TIG lasapparaat, een Gleeble lassimulator, een conventionele dilatometer,en een oven in combinatie met een zoutbad. Voor elke opstelling wordt uitgelegdhoe de akoestische emissie ten gevolge van diffusieloze transformaties wordt gemetentijdens continue afkoeling van de proefstukken. Tevens wordt beschreven dat hetmeten van akoestische emissie tijdens fasetransformaties niet altijd even gemakkelijkis vanwege de achtergrondruis ten gevolge van de elektronische voorversterker. Alshet elektrische vermogen van de akoestische emissie aan de uitgang van de sensorniet groter is dan het thermische ruisvermogen van de versterker, dan is detectie vande akoestische golven niet mogelijk. In het algemeen hangt het AE-vermogen aan desensor af van the soort materiaal, het volume van het proefstuk, de afkoelsnelheiden de eventuele verzwakking van het signaal door het gebruik van een golfgeleider.Een typische golfgeleider is een dunne stalen staaf met aan een kant een uiteindewaarop een sensor kan worden geplaatst. Het gebruik van een golfgeleider kannoodzakelijk 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 ver-schillende staalsoorten worden gepresenteerd in hoofdstuk 4. De meeste experi-menten zijn uitgevoerd met behulp van de Gleeble lassimulator of het TIG lasappa-raat. Voor beide methoden wordt zowel tijdens de martensitische als de bainitischetransformatie een piek-vormig AE-signaal waargenomen met een vergelijkbare in-tensiteit. Het gemeten AE-signaal gedurende bainietvorming geeft aan dat de groei

150 Samenvatting

van bainiet gepaard gaat met de collectieve beweging van ijzeratomen tijdens detransformatie; dit levert dus een sterk bewijs dat de bainitische transformatie dif-fusieloos is. Ter vergelijking is ook de diffusie-gestuurde transformatie van austenietnaar ferriet bestudeerd onder gelijke omstandigheden. Het blijkt dat tijdens fer-rietvorming geen akoestische emissie wordt gegenereerd. Dit is in overeenstemmingmet de verwachting, want diffusionele processen zijn langzaam en gaan dus nietgepaard met het snelle vrijkomen van spanningsenergie.

Verder worden in dit hoofdstuk de simultane metingen van akoestische emissieen dilatatie tijdens martensietvorming met elkaar vergeleken. De resultaten uitbeide metingen zijn in goede overeenstemming met elkaar wat betreft de start-temperatuur van de transformatie. De grootste beperkingen van AE-metingenaan proefstukken in een conventionele dilatometer zijn het kleine volume van deproefstukken en het noodzakelijke gebruik van een golfgeleider; beide reduceren dedetecteerbaarheid. Aan het eind van het hoofdstuk worden de resultaten besprokenvan metingen aan proefstukken tijdens afkoeling in een zoutbad. In deze experi-menten kunnen grote proefstukken worden gebruikt, en door een juiste keuze vande temperatuur van het zout is de afkoelsnelheid van een proefstuk zodanig datbainiet wordt gevormd. Ook met deze warmtebehandelingsmethode wordt tijdenshet afkoelen van het proefstuk een piek in de AE-data gezien bij temperaturenwaar bainietvorming plaatsvindt, wat in overeenstemming is met het displacivetransformatie model voor bainietvorming.

In hoofdstuk 5 wordt het onderzoek naar het verband tussen het gemeten AE-vermogen en de volumesnelheid van martensietvorming gepresenteerd. De theo-rie voorspelt dat de geproduceerde AE-energie rechtevenredig is met het volumemartensiet dat is gevormd. Om de theorie te toetsen zijn TIG-lassen gemaakt opeen werkstuk, zowel onder voortlopende boog condities als onder stationaire condi-ties (spotlassen). Een voordeel van lassen met voortlopende boog is dat de volumetransformatiesnelheid constant en achteraf gemakkelijk te bepalen is aan de handvan de dwarsdoorsnede van de las en de voortloopsnelheid van de boog. Een nadeelvan AE-metingen tijdens voortlopende booglassen is dat het lasproces zelf ook eenbron van akoestische emissie is; door de interactie tussen boog en werkstuk wordenakoestische golven gegenereerd. Het gemeten AE-signaal tijdens lassen heeft dustwee bijdragen: displacive transformaties en lasruis. Metingen tonen echter aandat onder normale lasomstandigheden het AE-vermogen ten gevolge van de trans-formatie veel groter is dan dat van de lasruis. Bovendien kan de laatstgenoemdebijdrage bepaald en dus voor gecorrigeerd worden. Gedurende de eerste secondenvan het lassen transformeert namelijk nog geen metaal en dus wordt alleen hetruisvermogen gemeten. Tijdens het afkoelen van het eerste deel van de las kan eenduidelijke toename in het signaal worden waargenomen. Onder evenwichtsituatiesis het gemeten AE-vermogen de som van de vermogens ten gevolge van fasetrans-formaties en lasruis. Direct na het stoppen van het lasproces is een piek te zien,die weerspiegelt dat de volume transformatiesnelheid toeneemt doordat het laatstedeel van de las met relatief hoge snelheid afkoelt als de boog uitdooft. Zowel detoename 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 kanvan wezenlijk praktisch nut zijn voor het lassen van hooggelegeerden staalsoortenomdat t.g.v. martensietvorming in een las koudschuren kunnen optreden.

Om de bovengenoemde theoretische voorspelling te valideren, zijn AE-experimen-ten uitgevoerd met verschillende stroomsterktes. Uit de metingen, gecorrigeerd voorde lasruis, blijkt dat het AE-vermogen inderdaad lineair toeneemt met de volumetransformatiesnelheid. Tevens zijn ook spotlassen gemaakt om de relatie tussen deAE-energie en het volume martensiet te bevestigen. Tijdens het afkoelen van despotlas wordt een pieksignaal gemeten, vergelijkbaar met de piek na het stoppenvan het lassen met voortlopende boog. Deze metingen hebben als voordeel datgeen lasruis wordt gemeten. Daarentegen is het wel relatief moeilijk om het volumemartensiet in de spotlas te meten omdat daartoe de spotlas precies in het middenmoet worden doorgezaagd. Ook uit de meetresultaten van het spotlassen volgt datde AE-energie zich rechtevenredig verhoudt tot het volume martensiet. Voor ver-schillende staalsoorten is op bovenbeschreven manier de evenredigheidsfactor tussende AE-energie en het volume martensiet (k) bepaald. De kleine verschillen in degevonden waarden voor de evenredigheidsfactor k kunnen worden toegeschrevenaan verschillen in chemische samenstelling tussen de staalsoorten.

In hoofdstuk 6 wordt de kinetiek van de martensitische transformatie onderzochtmet akoestische emissie. Onder kinetiek wordt de voortgang van de transformatieverstaan, ofwel de verandering van de volumefractie martensiet als functie vantemperatuur of tijd. Voor vier staalsoorten (C50, C60, C70 en C80) is tijdens deafkoeling in de Gleeble lassimulator het AE-vermogen ten gevolge van de marten-sietvorming gemeten. Met het resultaat uit hoofdstuk 5 dat het AE-vermogenproportioneel is met de volume transformatiesnelheid, kan de fractie martensietdirect uit de AE-data worden bepaald.

Door de gevonden fractie data te vergelijken met een theoretische model voorde kinetiek van de martensitische transformatie, de zogenaamde Koistinen & Mar-burger (KM) relatie, blijkt het mogelijk meer inzicht te verkrijgen in fysische para-meters die ten grondslag liggen aan de martensitische transformatie. Uit het fit-ten van de AE-data met de KM vergelijking kan de kinetische constante en destart-temperatuur van de transformatie worden bepaald. Het feit dat de transfor-matie met een kinetische constante beschreven kan worden wijst erop dat de gemid-delde kristalgrootte van martensiet niet verandert gedurende de transformatie. Tervergelijking, het Fisher model dat veronderstelt dat de kristalgrootte zal afnemenals de fractie martensiet toeneemt, geeft dus geen goede beschrijving van de marten-sitische transformatie in koolstofstaal. Ook microscopische waarnemingen van hetoppervlak van een geheugenmetaal laten zien dat gedurende de gehele transformatiezowel kleine als grote martensietkristallen worden gevormd.

Uit het resultaat van de metingen aan de vier staalsoorten met verschillendkoolstofgehalte blijkt dat de kinetische constante het grootst is voor staal C60(0.6 procent koolstof). Als wordt aangenomen dat de gemiddelde kristalgrootteniet significant verandert met koolstofgehalte, dan kan de verandering in kinetischeconstante worden toegeschreven aan een verandering in het aantal nucleatiekernen

152 Samenvatting

per volume eenheid. Metingen van de dislocatiedichtheid in martensiet die zijnuitgevoerd door andere onderzoekers op vergelijkbare staalsoorten vertonen ook eenmaximum voor staal met een koolstofgehalte van ongeveer 0.6 procent. Dit duidterop dat dislocaties een belangrijke bijdrage leveren aan het aantal nucleatiekernen.Deze dislocaties worden tijdens de fasetransformatie gevormd en kunnen zodoendede transformatie versnellen; dit wordt ook wel het autocatalytische effect genoemd.

Analoog aan de methode beschreven in hoofdstuk 5 zijn ook voor deze vierkoolstofstalen de evenredigheidsfactoren k bepaald. De waarden voor k als functievan koolstofgehalte vertonen dezelfde tendens als de dislocatiedichtheden, en dezecorrelatie suggereert dat het in feite de dislocatiebewegingen zijn die de akoestischegolven genereren tijdens de martensitische transformatie.

Gebaseerd op bovengenoemde correlatie is in hoofdstuk 7 een model gepre-senteerd dat het ontstaan van akoestische golven beschrijft tijdens de groei vanmartensietkristallen. Tijdens deze groei beweegt het grensvlak tussen martensieten austeniet, waarin zich dislocaties bevinden, richting austeniet. Als het grensvlakbeweegt, dan moeten de grensvlakdislocaties barrieres overwinnen, en dat zorgtervoor dat het grensvlak discontinu voortbeweegt. Nadat een barriere is over-wonnen komt er plotseling elastische energie vrij en gaat het kristalrooster trillen:akoestische emissie. Dit model is afgeleid van een bestaand model dat akoestischeemissie tijdens conventionele plastische deformatie beschrijft. Het blijkt ook dat eropmerkelijke 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 dieontstaan tijdens de bainitische transformatie vergeleken met die ten gevolge vande martensitische transformatie. De resultaten worden kwalitatief geanalyseerdomdat de onbekende materiaal respons en de beperkte bandbreedte van de sensoreen gedegen kwantitatieve benadering onmogelijk maken. De frequentiespectravan AE-signalen worden geınterpreteerd aan de hand van het bovenbeschrevendislocatiemodel. Het blijkt dat tijdens bainietvorming de gemiddelde frequentievan de gemeten akoestische golven significant groter is dan tijdens martensietvorm-ing. Dit verschil kan worden toegeschreven aan verschillen in de beweging van hetgrensvlak voor beide transformaties.

Aan het eind van dit hoofdstuk wordt een overzicht gegeven van de evenredig-heidsfactoren k voor alle staalsoorten. Analyse van deze factoren en dislocatiedicht-heden 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 andM.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. Couronneand 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 ofMaterials, 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 emis-sion,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 forma-tion,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 inmedium 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 transformationin 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. vander Zwaag,Acoustic emission investigations on the austenite decomposition in a mediumcarbon 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 meansof acoustic emission,Acta Materialia 51, 4183-4196 (2003).

13. S.M.C. van Bohemen,Study of acoustic emission signals generated during martensitic transforma-tions,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 martensiteformation,to be submitted to Philosophical Magazine A.

Curriculum Vitae

Stefanus Matheus Cornelis van Bohemenborn 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 Schoolin Haarlem, The Netherlands.

1994 – 1998 Masters degree in Physics, Kamerlingh Onnes Laboratory ofLeiden University, Leiden, The Netherlands.Thesis: ’Charge carrier transport in composites’

2000 – 2004 Ph.D. research at the Laboratory of Materials Science, DelftUniversity of Technology, Delft, The Netherlands.Thesis: ’An acoustic emission study of martensitic and bainitictransformations in carbon steel’

155

Nawoord

Graag maak ik van de gelegenheid gebruik om een aantal mensen te noemen zonderwie de totstandkoming van dit proefschrift niet mogelijk zou zijn geweest.

Allereerst mijn begeleiders, Marcel Hermans die er altijd was om me te helpen ofeen vraag te beantwoorden, Gert den Ouden die dit onderzoeksproject heeft opge-start en de eerste twee jaar mijn promotor was, Ian Richardson die de taak van Gertna twee jaar overnam, en Jilt Sietsma die het team van begeleiders kwam versterkentoen mijn promotieonderzoek zich voornamelijk ging richten op fasetransformaties.Bedankt voor jullie steun en alles wat ik van jullie heb geleerd.

Graag bedank ik ookWillem Brabander en Frans Bosman bij wie ik altijd terechtkon voor de technische ondersteuning, en Anneke van Veen voor de administratievehulp. Ook wil ik Herman Schoorlemmer (Physical Acoustics B.V.) noemen die mijvertrouwd heeft gemaakt met de akoestische emissie apparatuur. Verder Bram Huisdie me heeft geholpen bij het maken van proefstukken, en Erik Peekstok die altijdbereid 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 afleidinghebben gezorgd. De goede sfeer binnen de vakgroep Lastechnologie heb ik tijdensmijn promotietijd altijd bijzonder gewaardeerd.

157

an acoustic emission study of martensitic and bainitic transformations in carbon steel

Documents