Philosophical Magazine A

ISSN: 0141-8610 (Print) 1460-6992 (Online) Journal homepage: http://www.tandfonline.com/loi/tpha20

Ultrasonic spectral analysis for microstructural

characterization of austenitic and ferritic steels

Anish Kumar , T. Jayakumar & Baldev Raj

To cite this article: Anish Kumar , T. Jayakumar & Baldev Raj (2000) Ultrasonic spectral analysis
for microstructural characterization of austenitic and ferritic steels, Philosophical Magazine A,
80:11, 2469-2487, DOI: 10.1080/01418610008216486

To link to this article: http://dx.doi.org/10.1080/01418610008216486

Published online: 11 Aug 2009.

Submit your article to this journal

Article views: 94

View related articles

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tpha20

Download by: [University of Lethbridge] Date: 02 November 2015, At: 13:20

 MAGAZINE
PHILOSOPHICAL A, 2000, VOL.80, No. 1 1 , 2469-2487

Ultrasonic spectral analysis for microstructural

characterization of austenitic and ferritic steels

ANISHKUMAR,T. JAYAKUMARand BALDEVR A J t

Metallurgy and Materials Group, Indira Gandhi Centre for Atomic Research
Kalpakkam 603 102, India

[Received 19 July 1999 and accepted in revised form 2 December 19991

ABSTRACT
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

The paper discusses the use of ultrasonic spectral analysis for microstructural
characterization in AISI type 316 stainless steel and modified 9Cr-1Mo ferritic
steel (T91/F91).Several specimens of AISI type 316 stainless steel have been heat
treated at different temperatures varying from 1373 to 1623K for different time
durations (from 15 to 120min) to obtain specimens with different grain sizes.
Spectral analysis has been carried out on the rf signal corresponding to the first
back-wall echo obtained with a 25 MHz delay line longitudinal beam transducer.
The shift in the peak frequency and the change in the full width at half-maximum
of the autopower spectrum have been correlated with the average grain size
and yield stress. Various specimens of modified 9Cr-1Mo ferritic steel have
also been subjected to a series of heat treatments consisting of soaking for
Smin at different temperatures in the range 1073-1623K followed by oil
quenching. These treatments were given to obtain different microstructures and
grain sizes in the specimens simulating the different regions in the heat-affected
zones of the weldments. Spectral analysis has been carried out on the rf signal
corresponding to the first back-wall echo obtained with the 20 MHz longitudinal
beam transducer. Two distinct peaks at around 7.0 and 17.5MHz have been
found in the autopower spectrum of the first back-wall echoes of all the
specimens. The ratio of these two peaks (the amplitude of the lower-frequency
peak to the amplitude of the higher-frequency peak) has been correlated with the
prior austenitic grain size. A linear correlation between the spectral peak ratio
(SPR) and the grain size is found to be valid for a wide range of microstructures,
+
that is femte, ferrite martensite and martensite. The variation in the SPR with
the soaking temperature shows that various metallurgical events affecting the
grain size, namely the formation of martensite (AcI-Ac3 temperature range),
dissolution of NbC and V4C3 and formation of &ferrite (Ac4 temperature), can
also be determined using the SPR but, for more accurate determination of these
temperatures, on-line measurement is required. For the first time, we show that
the various spectral parameters used in this study are independent of the coupling
condition. The significance of the results for practical utility is also discussed.

0 1. INTRODUCTION
The use of ultrasonic pulse waveform analysis for material characterization takes
advantage of the easy availability of improved digital signal-processingmethodology
as a part of ultrasonic measurement instrumentation and sensitive broad-band

Email: dmg@igcar.ernet.in.

Philosophical Magazine A ISSN 0141-8610 print/ISSN 1460-6992 online 0 2000 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
 2470 A. Kumar et al.

probes. Ultrasonic spectroscopy is a further development of the pulse-echo technique

which uses broad-band ultrasound and analyses the spectra of the echo pulses. In an
analogy with light, the specimen is insonified with white sound. The defects or
microstructure alters the ‘colour’ of the wave that has travelled through the material
and the change in the ‘colour’ of the sound wave gives information about the
material. Ultrasonic spectroscopy has been used for characterization of defects
and microstructural features (Fitting and Adler 1981). Ultrasonic spectroscopy has
been widely used for measurement of the depth of surface opening cracks (Morgan
1975), quality testing of interfacial bonding (Brown 1973), etc. The use of ultrasonic
spectroscopy for grain size determination has been illustrated by Gericke (1971),
who investigated metal plates with broadband ultrasound. The ‘loop frequency’
responses of steels of various grain sizes showed that, as the mean grain size
increased from 50 to 100pm, the ratio of the heights of the two frequency humps
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

changed; the higher frequencies preferentially become attenuated by coarser-grained

material. The shift in the spectral peak frequency has been used for evaluation of
structural variations induced by tensile deformation in SUS304 stainless steel
(Honjoh 1994). In this case, the spectral peak frequency was found to increase
with increase in the tensile elongation. This was attributed to formation and growth
of the martensite structure due to tensile deformation, which resulted in smaller
crystalline grains, thus reducing the attenuation due to ultrasonic scattering. In
recent years, the applicability of ultrasonic attenuation (Papadakis 1984), ultrasonic
relative attenuation (Palanichamy et al. 1994), ultrasonic back-scattering (Hetch et
al. 1981) and ultrasonic velocity (Palanichamy et al. 1995) measurement methods for
grain size determination by non-destructive means has been reported. In the present
study, spectral analysis of the ultrasonic first back-wall echo signal has been used for
characterization of microstructural features including grain size in AISI type 316
stainless steel and modified 9Cr-1Mo ferritic steel.
AISI type 316 stainless steel, with higher elevated temperature strength and good
corrosion and oxidation resistance, is a widely used structural material for fast
breeder nuclear reactors and in chemical, fertilizers and many other industries.
Modified 9Cr-1 Mo steel, with high creep strength and thermal conductivity, low
thermal expansion coefficient and good resistance to corrosion and stress corrosion
cracking is widely used as a structural material in the power-generating and petro-
chemical industries. The alloy is recommended for use in the normalized and tem-
pered condition. Because of the transformable nature of the ferritic alloy, it is
susceptible to the formation of undesirable microstructures during the
fabrication and/or heat treatment processes. It is important to characterize non-
destructively the microstructures of these steels for quality control during fabrication
and heat treatment to ensure presence of desired microstructure and mechanical
properties.
In the present study, spectral analysis of the first back-wall echo obtained with a
25 MHz delay line longitudinal transducer and 20 MHz longitudinal beam transdu-
cer has been carried out for the microstructural characterization of AISI type 316
stainless steel and modified 9Cr-1Mo ferritic steel respectively. The shift in the peak
frequency and full width at half maximum (FWHM) of the autopower spectrum of
the first back-wall echo have been correlated with the grain size in AISI type 316
stainless steel. The change in the spectral peak ratio (SPR) (the amplitude of the
lower frequency peak to the amplitude of higher-frequency peak) has been correlated
with the grain size and other metallurgical events affecting the grain size, that is the
 Microstructural characterization of steels 247 1

formation of martensite (AcI-Ac3t temperature range), the dissolution of NbC and

V4C3,and the formation of &ferrite (Ac4$ temperature) in modified 9Cr-1Mo ferri-
tic steel.
The reliability of many of the ultrasonic methods for grain size measurement,
particularly amplitude-based measurements, strongly depends on the couplant con-
dition. Therefore, adoption of these methods may be difficult for many practical
applications such as continuous production line measurements, where a uniform
coupling condition cannot be .reliably ensured. The spectral-analysis-based meth-
odologies proposed in this paper would overcome this limitation as the spectral
parameters employed in this study are found to be independent of the coupling
condition.
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

9 2. EXPERIMENTAL
DETAILS

2.1. Sample preparation

2.1.l. AISZ type 316 stainless steel

Several specimens of AISI type 3 16 stainless steel (50 mm diameter and 15mm
thick) were obtained from bar stock. The chemical composition of the material
used is as follows: C, 0.06wt%; Mn, 1.80wt%; Si, l.OOwt%; Cr, 16.36wt%; Ni,
12.1wt%; Mo, 2.2wtY0; Fe, balance. The specimens were heat treated at different
temperatures varying from 1373 to 1623K for different time durations (from 15 to
120min) in order to obtain specimens with various grain sizes. Table 1 gives the
details of the heat treatments employed. Subsequently,all the specimens were given a
common treatment at 1323K for 30min followed by water quenching in order to
obtain a uniform structure with the same substructural features except for the varia-
tion in the grain size (Mannan 1981). Specimens of 5 and lOmm thicknesses were
prepared from these heat-treated specimens.

2.1.2. Modified 9Cr-1Mo ferritic steel

Specimens of 10 mm x 10mm cross-sectional area and 60 mm length were pre-
pared from the normalized (1333 K for 6 h and then air cooled) and tempered
(1043 K for 4 h and then air cooled) forged rounds of 70 mm diameter. The chemical
composition of the material used is as follows: C, 0.12wtY0; Si, 0.28wtY0; Mn,
0.43 wt%; S, 0.01 wt%; P, 0.01 wt%; Cr, 8.84wt%; Mo, 0.95wt%; V, 0.21 wt%;
Nb, 0.09wtY0; Fe, balance. The specimens were soaked for a duration of 5min at
different temperatures starting from the a-phase region (1073 K) to the y &phase +
region (1623 K) followed by oil quenching. These treatments were given to obtain
different microstructures and grain sizes in the specimens, simulating the different
regions in the heat-affected zones of the weldments. It may be noted that this soaking
time of 5 min was given after the sample attained the required soaking temperature.
The optical microstructure of the cross-section revealed that the microstructure was
uniform throughout the section of the specimen, indicating that the whole volume of
the specimen reached the soaking temperature. Table 2 gives the details of the heat
treatments.
t A q is the temperature at which austenite starts to form during heating and A c ~is the
temperature at which the transformation of ferrite to austenite is completed during heating.
3 A q is the temperature at which austenite transforms to &ferrite during heating.
 Downloaded by [University of Lethbridge] at 13:21 02 November 2015

ils of heat treatments given to AISI type 316 stainless steel specimens, the corresponding average grain sizes and the ultrasonic spectral
parameters.
Grain Peak frequency (MHz) FWHM (MHz)
Heat sizea ?
treatment (w ) 5 mm thick 10 mm thick 5 mm thick lOmm thick 7~
As received 30 17.58 14.91 14.94 13.16
I373 K for 0.25 h water quench 63 11.65 9.81 10.34 7.56
+
5F
1373 K for 1 h water quench
+ 78 10.92 8.87 9.72 7.23 2
1573 K for 1.5 h water quench
+ 106 9.70 7.58 8.53 6.78 Q
'c
1623 K for 0.5 h water quench
+ 121 8.85 6.37 7.95 6.31
1623 K for 1 h +water quench 138 8.44 6.57 7.62 5.90
metallography (ASTM Standard E 112-88, 1992).
 Microstructural characterization of steels 2473

Table 2. Details of the heat treatments given to modified 9Cr-1Mo steel, the corresponding
microstructures,the average grain sizes and the ultrasonic SPRs.
Soaking temperature Grain size
Specimen (K) Microstructurea (w) SPR
1 1073 u+C 48 0.821
2 1098 usc 49 0.783
3 1123 f+m+C 45 0.717
4 1148 u+m+C 30 0.521
5 1173 a+m+C 19 0.456
6 1223 m+C 19 0.49 1
7 1273 m+C 18 0.46
8 1323 m+C 29 0.473
9 1373 m+C 40 0.562
10 1423 m 92 1.027
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

11 1473 m 131 1.591

12 1523 m+6 116 1.841
13 1573 m+6 85 1.713
14 1623 m+6 55 1.645
au,ferrite; m, martensite; C, carbide; 6,&-ferrite.

Surface grinding of all the specimens was carried out to obtain plane parallelism
to an accuracy of better than f 3 pm. Metallographic examination was carried out to
reveal the microstructures in different specimens. The linear intercept method
according to ASTM Standard E122-88 (1992) was used to find the average grain
size in all the specimens of both materials for correlation with the ultrasonic para-
meters.
Because of the presence of martensite laths, it was not possible to see the prior
austenitic grain boundaries clearly in the quenched specimens. After completing all
the ultrasonic.measurements,these specimens were tempered at 1033 K for 1 h and
aged for 2000 h at 923 K in order to reveal the prior austenite grains by etching the
specimens.

2.2. Ultrasonic measurements

The experimental set-up used for ultrasonic measurements is shown in figure 1. A
100MHz broad-band ultrasonic pulser-receiver (Accu-Tron, USA) and 500 MHz
digitizing oscilloscope (Tektronix TDS524) were used for the ultrasonic measure-
ments. The 25 MHz delay line longitudinal transducer (supplied by Accu-Tron,
USA) and the 20 MHz longitudinal beam transducer (supplied by Panametrics,

I - I

Figure 1. Experimental set-up for ultrasonic measurements.

 2474 A. Kumar et al.
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

A, A1 A,
Figure 2. Photomicrographs of the stainless steel specimens A . (d = 30 pm), A2 (d = 78 pm)
and A , ( d = 138pm).

USA) were used for the ultrasonic measurements in AlSI type 316 stainless steel and
modified 9Cr-1Mo ferritic steel. The frequency range 20-25 MHz was used to keep
the operating frequency on the higher-frequency side of the Rayleigh scattering
region, so that the influence of the microstructures on the frequency spectrum can
be studied. The gated back-wall echoes from the oscilloscope were transferred to a
personal computer with the help of general-purpose interface bus (GPIB) interfacing
and LabVIEW programming system. The autopower spectrum was taken on the first
back-wall echo with full length. The data length was selected in such a way that it
covered more than the full length of the back-wall echo (data length in which the
complete energy of the ultrasonic pulse is contained). Hence the gated data length for
the 20MHz probe with longer pulse length (about 50011s) was taken as 204811s
whereas, for the 25 MHz probe with shorter pulse length (150 ns), the data length
of 102411s was found to be sufficient. The peak amplitude of the echo was kept in the
centre of the data length.

2.2.1. 25 M H z delay line transducer

Spectral analysis was carried out on the rf signal corresponding to the first back-
wall echo obtained with the 25 MHz delay line longitudinal transducer (supplied by
Accu-Tron, USA). The sampling frequency used in the experiment was 125MHz and
the gated data length was 102411s (taking the position a in figure 3(a), the zero
crossing between the highest positive to negative peak as the central point of the
data length). The peak position and FWHM of the autopower spectrum for the first
back-wall echo in all the specimens were measured on-line automatically with the
help of GPIB interfacing and the LabVIEW programming system. The software for
the on-line measurement of the ultrasonic parameters was specifically developed as a
part of this study. To see the dependence of the peak frequency and FWHM on the
coupling condition, these parameters were measured at different loads (dead weight)
applied on to the transducer.
 Microstructural characterization of steels 2475

Fig. 3a

3 -1-
P

aE
-
-2:
-3 .
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

0.12 ’
3 0.10 !
U 0.08
0.06 :
2 0.04:
g 0.02 1
a 0.00 :

2.2.2. 20 MHz contact transducer

Spectral analysis was carried out on the rf signal corresponding to the first back-
wall echo obtained with the 20MHz longitudinal beam transducer. Two distinct
peaks were found in the autopower spectrum of the first back-wall echoes for all
the specimens. The ratio of these two peaks was used as an ultrasonic parameter for
characterizing the microstructures. For the spectral analysis, the ultrasonic signal
was digitized at 500MHz. Specific software was developed for automatic on-line
peak ratio measurements from the autopower spectrum.

0 3. RESULTSAND DISCUSSION
The 25 MHz delay line longitudinal transducer and the 20 MHz longitudinal
transducer were used for both the materials for the spectral analysis. The spectral
characteristics of the 25 MHz delay line longitudinal transducer showed a single peak
with Gaussian distribution in the autopower spectrum, whereas the spectral char-
acteristics of the 20 MHz longitudinal transducer exhibited two distinct peaks
centred around 7.0 and 17.5 MHz in the autopower spectrum. Owing to the higher
scattering in the austenitic stainless steel, the two peaks in the autopower spectrum of
the first back-wall echo were found to merge in the specimens with grain sizes more
than 78 pm, making the use of the SPR using the 20 MHz transducer for grain size
 2476 A. Kumar et al.

measurement impossible. Hence, the shift in peak frequency and FWHM obtained
with the 25 MHz transducer was used for the grain size measurements in austenitic
stainless steel.
When the 25MHz delay line transducer was used for the spectral analysis in
modified 9Cr-1Mo steel, because of the lower scattering power of the ferritic steel
(Papadakis 1981) the maximum shift in the peak frequency was not more than about
4 MHz. Hence the shift in the peak frequency could not be used reliably for correlat-
ing with the grain size. However, when the 20MHz contact longitudinal beam
transducer was used, which exhibited two distinct peaks centred around 7.0 and
17.5MHz in the autopower spectrum of the first back-wall echo for all the speci-
mens, the ratio of the two peaks could be used for assessing the grain size in this
material. For the first time, this paper demonstrates the importance of spectral
characteristics of different transducers for assessment of grain size in materials
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

with differing scattering powers.

3.1. AISI type 316 stainless steel

Figure 2 shows typical photomicrographs for specimens with grain sizes 30, 78
and 138 pm. Table 1 gives the details of the heat treatments given to the specimens
and the corresponding average grain sizes obtained. Figures 3(a) and (b) show
typically how the first back-wall echo and its spectral distribution change with the
change in the grain size. The amplitude of the first back-wall echo decreases with
increase in the grain size. Table 1 also shows the values of the spectral peak fre-
quency and the FWHM for the specimens with different grain sizes and thicknesses.
These values are average of ten readings taken for each specimen. The maximum
scatter bands in the measurement of the peak frequency and the FWHM are
f0.15 MHz and f0.25 MHz respectively for all the specimens. The log-log plot of
the peak frequency and the grain size and the log-log plot of the FWHM of the
autopower spectrum and the grain size were found to show linear relationships.
It has been shown that yield strength Y S follows a Hall-Petch-type relationship
as follows, which predicts the mechanical properties to be linear with d-'I2 (Klinman
et al. 1980)

YS = Yo + k,dp"2,
where Yo depends on the composition and temperature, ky is a constant and d is the
average grain diameter. It is, therefore, advantageous to relate the ultrasonic para-
meters with d-'/* as this would enable an assessment of yield strength using the
ultrasonic parameters, provided that there is no substructural variation which influ-
ences the yield strength significant1 . Figure 4(a) shows a good linear relation
Y
between the peak frequency and d-'I for the specimens of two different thicknesses.
Figure 4 (a) also shows the effect of the specimen thickness on these ultrasonic
parameters. Figure 4(6) shows the variation in the FWHM with d-'I2. For the
higher-thickness specimens, the peak frequency and the FWHM shift towards
lower values. As the peak frequency fp and the FWHM are linear functions of
d-'I2, these parameters can be fitted with a Hall-Petch-type equation, as shown by

FWHM = Fo + kod-'/*,
 Microstructural characterization of steels 2477

16: Fig.4a
N
I
r

2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0

d-112, mm-112
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

(4
d. urn

(4
Figure 4. Variation in (a) the peak frequency and (6) the yield strength, peak frequency and
FWHM with d-'f2: PF, peak frequency.

wherefo and Fo are the intercepts, and kf and ko are the slopes for the straight lines
for the relations involving peak frequency and the FWHM respectively. These values
are obtained for a given thickness of the specimen and the transducer frequency
employed. Table 3 gives the values for the slope and the intercept and the correlation
coefficient derived from the experiments for the peak frequency and the FWHM
respectively. It is clear from table 3 that the correlation coefficient is lower (0.976)

Table 3. Values of slope, intercept and correlation coefficient for the correlations obtained
for the peak frequency f p , FWHM and yield strength with d-'l2.
Parameter Specimen thickness Correlation
(units) (mm) Intercept Slope coefficient
f (MW 5 0.42 2.94 0.997
4 (MH4
FWHM (MHz)
10
5
-1.19
1.17
2.79
2.37
0.999
0.999
FWHM (MHz) 10 -0.60 2.30 0.976
Yield strength (MPa) - 190.57 15.00 0.960
 2478 A. Kumar et al.

for the correlation between the FWHM and d-'l2 for specimens with lOmm thick-
ness and is attributed to the weak back-wall echo signal obtained for the specimen
with higher thickness.
As indicated, the yield stress also follows the linear relation with d - 1 / 2(Klinman
1980). The yield strength values corresponding to different grain sizes of the material
of the same heat as used in the present study have been reported by Mannan (1981).
The variation in the yield strength with d-'12 is plotted using the yield strength
values for different grain sizes reported by Mannan (1981) and the plot is shown
in figure 4 (b). Figure 4 (b) also shows the variation in the peak frequency and the
FWHM with d - 1 / 2for specimens of 5mm thickness. It can be seen that there is a
linear relationship for the yield strength, the peak frequency and the FWHM with
d-'12. Table 3 also shows the values of the intercept, slope and the correlation
coefficient for the correlation between the yield strength and d - ' / 2 .Therefore, these
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

ultrasonic parameters can be linearly correlated with the yield strength, if there is no
other substructure microstructural variation which could affect the yield stress.
The shift in the peak frequency with increase in the grain size can be explained
with the help of the ultrasonic attenuation theory and the spectral analysis of the
ultrasonic waves. Ultrasonic waves in a polycrystalline material are attenuated by
structural boundaries and grains. The total attenuation coefficient Q may be
expressed as Q = a, +as,where Q, is the losses due to internal friction, that is
ultrasonic absorption, and Q, takes into account losses due to scattering of the
ultrasonic waves. Ultrasonic absorption consists of thermoelastic losses, magnetic
losses and losses due to dislocation damping. For a non-ferromagnetic material such
as austenitic stainless steel, absorption is due to the thermoelastic losses and losses
due to dislocation damping only. The absorption coefficient Q, can be expressed as
+
(Bergner and Popp 1990) Q, = alf '3 ad', where al and a2 are the constants and
f is the frequency of the ultrasonic waves. The first and second terms on the right-
hand side of the equation take into account the thermoelastic losses and losses due to
dislocation damping respectively. The general theory of ultrasonic scattering
attenuation was introduced by Mason and McSkimin (1948) and by Bhatia
(1959), who showed that this attenuation is dominated by Rayleigh scattering
when the wavelength X of the ultrasonic waves is greater than the grain size d (the
condition that is valid in our study, i.e., X = 230 pm and the values of d are from 30
to 138 pm). The ultrasonic attenuation coefficient a, due to the Rayleigh scattering in
polycrystalline materials (Papadakis 1965, 1984) can be expressed as
Q, =Sd3y, (4 a>
where S is a scattering factor that depends on the elastic properties of the material
(including the sound velocity), d is the average grain size in the specimen and f is the
frequency. Hence, the total attenuation can be expressed as
Q = alp.'+ a2f' + S d 3 p . (4 b)
It can be seen that, in materials with almost the same substructural features, total
attenuation is mainly dependent on the grain size and frequency. Even if there is
some change in the substructure, because of its lower power of frequency depen-
dence, it has less influence on the frequency-dependent attenuation. In the case of
austenitic stainless steels with high elastic anisotropy, the reported contribution from
ultrasonic absorption is very low in comparison with ultrasonic scattering
(Jayakumar 1997, Jayakumar et al. 1998). At a frequency of 5 MHz, the absorption
 Microstructural characterization of steels 2479

measurements made in AISI type 304 stainless steel showed an increase in absorption
from 2% of total attenuation in the solution-annealed condition to 5% of total
attenuation even after 20% cold work with a high dislocation density, clearly indi-
cating the dominance of scattering (Jayakumar 1997, Jayakumar et al. 1998). Hence,
a can be replaced by a,,and therefore the observed influence of the grain size on the
spectral parameters may not be observed even if there is variation in the substruc-
ture, which is fortunately not the case for our investigation. When the ultrasonic
wave is attenuated in the material, it can be assumed that the oscillatory amplitude
decays exponentially as the propagation length of the ultrasonic wave increases. If
we consider the spectral distribution of the ultrasonic wave, which has an oscillatory
pulse of one or two cycles with an exponential decay envelope, the amplitude g ( t ) of
the ultrasonic wave is expressed as (Simpson 1974)
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

where a d is the damping coefficient, t is the time, fo is the carrier frequebcy, Q is the
phase and A is the maximum amplitude of the pulse wave. The spectrum power
distribution IG(f ) l is derived from the Fourier transform (Simpson 1974)

The damping coefficient a d in equations ( 5 ) and (6) can be replaced by a as shown in

equations (4a) and (4b) (Honjoh 1994). As a, depends on the fourth power of the
frequency in the Rayleigh regime (equation (4a)), the peak frequency of the auto-
power spectrum shifts to lower values owing to increased attenuation of the higher-
frequency component. As the grain size increases, a and also the frequency depen-
dence of a increase (Williams and Gobbles 1981). Because of this, the peak fre-
quency shifts towards the lower values with increase in the grain size. As the
attenuation rate increases, the spectral power at the higher frequencies decreases
and hence the FWHM in the autopower spectrum also decreases. The quantitative
verification of the experimental results with the information derived from equation
(6) is being studied separately. The peak frequency in the far field decreases as the
propagation length increases, because the ultrasonic pulse is attenuated by space
diffusion (Honjoh 1994), that is diverging beam spread. This explains the lower
values of the peak frequencies and the FWHM of the autopower spectrum of the
first back-wall echoes for the thicker (10mm) specimens of the same average grain
size (figure 4 (a)).

3.2. Modijied 9Cr-1Mo ferritic steel

Typical photomicrographs for the specimens with various microstructures and
prior austenite grain sizes obtained by heat treatment at different temperatures are
shown in figure 5. Various microstructural features in different specimens are docu-
mented in table 2. In the specimens heat treated at less than the lower critical
temperature (Ac, M 1100 K), only ferrite with carbides (tempered martensite) was
present (figure 5 (a)). As the heat treatment temperature increased beyond Ac, ,
partial austenization took place during soaking, resulting in a microstructure with
ferrite and freshly formed martensite on quenching (figure 5 (b)). A high heat treat-
ment temperature within the ferrite-austenite region (between Acl and Ac3) gives a
 2480 A. Kumar et al.
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

Figure 5. Microstructures of the specimens soaked at (a) 1073 K, (b) 1148 K, (c) 1273 K, (d)
1373 K, (e) 1473 K and cf) 1623 K for 5 min.

higher volume fraction of austenite before quenching resulting in a higher volume

fraction of martensite after quenching. The amount of martensite increased up to the
upper critical temperature (Ac3 M 1180 K), where only martensite laths with fine
prior austenitic grain size were found (figure S(c)). Because of the fresh formation
of martensite, the grain size decreased with increase in soaking temperature in the
intercritical temperature region (between Ac, and A q ) and the finest prior austenitic
grain size (about 18pm) was found in the specimens soaked at just above Ac,
 Microstructural characterization of steels 248 1

(~1180K).In the specimens heat treated at higher temperatures above 1180K, the
grain size increased slowly with increasing temperature up to about 1373K (figure
S(4). Above this temperature, there was a sudden increase in the prior austenitic
grain size (figure 5 (e)).This has been attributed to the dissolution of V4C3and NbC,
which restricts the grain growth (Orr et al. 1992). At about 1490K (just above Ac4),6-
ferrite starts to form, which restricts the grain growth, thus resulting in relatively finer
prior austenitic grain size in a duplex martensite-6-ferrite microstructure (figure 5 0).
Figures 6 (a) and (6) show typically the changes in the first back-wall echoes and
the autopower spectra respectively with the change in the grain size in the specimens
with martensitic structure. As the attenuation of ultrasonic waves increases with
increase in the grain size, the higher frequencies become scattered preferentially
and hence the relative heights of the two peaks change. Table 2 and figure 7 show
the variation in the SPR with the soaking temperature. As explained earlier, as the
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

prior austenitic grain size increases in the ferritic steel, attenuation of the higher-
frequency component increases and hence the ratio of the two peak heights in the
autopower spectrum changes. The variation in the grain size with solutionizing
temperature also followed a similar behaviour to that of the SPR, which decreased
with increase in the soaking temperature in the intercritical region (1 100-1 180K)
because of the decrease in the grain size. Above this, because of the increase in the
grain size, the SPR increased with increase in the soaking temperature. Above

-*I 0 . 500
' . lodo
' . 1500
' . a K)
Time, ns
(a)

p27k20
0.002 1 R'
0 10 20 30
Frequency, MHz
(4
Figure 6. Variations in (a) the first back-wall echo and (b) the autopower spectrum with
grain size in modified 9Cr-1Mo steel with a martensitic microstructure (A.U., arbitrary units).
 2482 A. Kumar et al.

H
-p 1.2
la41
\
a 0.8
l-01 /i
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

Figure 7. Variation in the SPR (amplitude at PI divided by amplitude at Pz,as shown in

figure 6 (6)) and grain size with the soaking temperature.

1223K, the SPR increased sharply because of the sharp increase in the grain size as
discussed earlier. Because of the formation of &ferrite above 1473K, the grain
growth was restricted and hence the grain size decreased. This is reflected by a
reduction in the SPR also.
Figure 8 shows the variation in the SPR with the grain size for the three micro-
structures: ferrite, martensite and martensite +&ferrite. Even though the SPR is
found to be a strong function of grain size, the effect of other microstructural
features could also be seen clearly. For a similar prior austenite grain size (about
50pm, indicated as the straight line OBAC in figure 8), the SPR in the specimens
with martensitic structure (about 0.6, point A) is less than that in the specimens with
ferrite (about 0.8, point B) or with martensite and &-ferrite(about 1.6, point C). This
is attributed to the presence of fine laths (figure 5 ( c ) ) , which makes the martensite
more elastically isotropic than ferrites and hence decreases its scattering power. Even
in the presence of martensite, the specimen with martensite and ferrite exhibits a
higher SPR (about 1.6, point C) than that in the specimen with ferrite and carbides

0 1.8-
'fi 1.6:

1 ;::; 1.o-

0.4-
b
Grainsize, pm
Figure 8. Variation in the SPR with grain size in specimens with ferrite and martensite (a),
only martensite ( 0 )and martensite and &ferrite (A). The vertical line OBAC shows
three spectral ratios for specimens with the same grain size but different phase combi-
nations.
 Microstructural characterization of steels 2483

(about 0.8, point B). This is attributed to the presence of coarsened laths in the ferrite
(figure 5 (a), as it is simply heavily tempered martensite), which makes the ferrite
more isotropic than the &-ferrite(figure 5 ( f ) ) . Besides this, the presence of two
phases makes the specimens with martensite and &ferrite most anisotropic and
increases its scattering power and SPR, compared with the ferrite or martensite
structure.
Figure 9 shows the variation in the SPR with grain size only for the specimens
with ferritic and martensitic structures. These microstructures were obtained by heat
treatment in the temperature range 1023-1473 K (the usual solutionizing tempera-
ture of this steel is around 1323-1373 K). A linear correlation between the SPR and
grain size is found to be valid for a wide range of microstructures, that is ferrite,
+
ferrite martensite and martensite. From the practical utility point of view, this is
advantageous as the correlation can be used for evaluation of the prior austenitic
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

grain size with various microstructures of practical utility. The linear fit between the
SPR and grain size gives a correlation coefficient of 0.98. As discussed above; this
relationship cannot be used for the specimens with martensite and &ferrite, because
of its higher scattering behaviour.
The relationship between the SPR and the microstructure (grain size and struc-
tural phases) can be explained with the help of equations (4) and (6) and the experi-
mental findings of many workers (Gericke 1971, Canella and Monti 1976, Fitting
and Adler 1981) demonstrated that, as the grain size increases, the attenuation and
the frequency dependence of the attenuation increase.
The shift in the peak frequency and the change in the SPR are dependent not
only on the frequency dependence of the attenuation but also on the total attenua-
tion. Only because of this (even thou h the frequency dependences of the attenuation
r$
are the same in the two materials ( )), was the shift in the peak frequency found to
be much less in ferritic steel having overall lower attenuation, compared with that in
the austenitic steel for the same variation in grain size. Similarly, a considerable
change in the SPR is also expected in ferritic steel, when there is a change in prior
austenite grain size or in the presence of &ferrite.
A comparison of the results obtained from AISI type 316 stainless steel and 9Cr-
1Mo ferritic steel reveals that the shift in the peak frequency in the autopower
spectrum of the first back-wall echo is more reliable for the grain size measurement
in materials with a higher scattering factor, for example stainless steel, whereas the

Grain Size, pm
Figure 9. Variation in the SPR with grain size in the specimens with ferrite and martensite
and only martensite.
 2484 A. Kumar et al.

change in the SPR is more reliable for grain size measurement in materials with a
lower scattering factor, for example ferritic steels. To the best of the present authors’
knowledge, this is first time that such an understanding could be obtained.

3.3. Eflect of the coupling condition on the spectral parameters

3.3.1. AISI type 316 stainless steel

Figures 10 (a)and (6)show the effect of the coupling condition (no load, i.e. hand
pressed and released condition, 1 kg load and 3 kg load on the transducer) on the
amplitude of the first backwall echo and the peak frequency and FWHM respectively
for one of the specimens (30 pm grain size and 5 mm thick). It can be seen from figure
5 ( 6 ) that the peak frequency and the FWHM in the autopower spectrum are the
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

same for different coupling conditions (no load, 1 kg load and 3 kg load on the
transducer), in spite of the variation in the total spectral energy because of the
change in the coupling condition. This type of behaviour could also be observed

3 Kg leak Fig. 10a

100 Fig. led

1 1 . 9 . - . . . . . . . , . I
0 1 0 2 0 3 0 4 0 5 0 6 0 7 0
Frequency, MHz
(b)
Figure 10. Effect of the coupling on ( a ) the first back-wall echo and (6) the autopower
spectrum for AISI type 316 stainless steel specimen with 30pm grain size (A.U.,
arbitrary units).
 Microstructural characterization of steels 2485

in all other specimens with different grain sizes. This clearly shows that these spectral
parameters are independent of the coupling condition.

3.3.2. Modified 9Cr-lMo ferritic steel

Figures 11 (a)and (6)show the effect of the coupling condition on the first back-
wall echo and its autopower spectrum respectively for the specimen solutionized at
1573K. It can be seen clearly that, when the coupling condition was not good, the
amplitude of the first back-wall echo and its autopower spectrum decreased. In the
autopower spectrum, the amplitude of both the peaks decreased, maintaining the
peak ratio almost constant and hence independent of variations in the coupling
condition. Similar behaviours have been found for all the other specimens used in
this study.
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

All the spectral parameters used in both the steels show their independence of
variations in the coupling condition. This has an important practical sigdcance,
that is, if these spectral parameters are used for grain size measurement, then the
error involved in the measurements due to variation in the coupling condition can be

Fig. l l a

Time, ns
(4

I
0 10 20 30 A 3
Frequency, MHz
(b)
Figure 11. Effect of different coupling conditions on (a) the first back-wall echoes and (b)
their autopower spectra (note that the heights of both the peaks are changed by the
change in the coupling condition, but the SPR remains independent of the coupling
condition) (A.U., arbitrary units).
 2486 A. Kumar et al.

minimized for on-line measurements and/or when a large number of measurements are
to be made by moving the plate or transducer. In addition to this, since only the first
back-wall echo is required for the grain size measurements, this approach can be used
for grain size measurements even in thicker and highly attenuating materials, where
obtaining multiple back-wall echoes with a good signal-to-noise ratio is very difficult.
The other parameters that affect the ultrasonic measurements are the thickness and the
roughness of the specimens. The effect of the thickness on the peak frequency and
FWHM has been studied in the present paper. It has been found that doubling the
thickness (from 5 to 10 mm) reduces the peak frequency and FWHM by only 2 MHz
(whereas the peak frequency and the FWHM decrease by more than half, i.e. from
17.58to8.44MHzandfrom 14.91 to6.57 MHzwithincreasein thegrainsizefrom30to
138 pm for specimens 5 mm thick). This shows that the slight variation in thickness will
not affect the accuracy for grain size measurement using these two spectral parameters.
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

Similar to the very weak influence observed for the thickness on the spectral para-
meters, we believe that the surface roughness would also be expected to have a mar-
ginal influence compared with that of grain size. Other than these practical variables,
the spectral peak or spectral distributions of a broad-band ultrasonic probe would be
influenced by the electrically induced condition to the pulser-receiver also but, dur-
ing calibration and testing, the same electrically induced conditions can be main-
tained by using the same pulser-receiver and electrical damping. This can also be
easily adopted under practical circumstances without any difficulty.

5 4. CONCLUSION
The spectral analysis of the first back-wall has been carried out for the grain size
measurements in AISI type 316 austenitic stainless steel and modified 9Cr-1Mo
ferritic steel. The spectral approach discussed in this paper has several advantages
over conventional ultrasonic techniques. These include firstly that the requirement of
only first backwall echo makes this approach useful for highly attenuating and
thicker materials and secondly that the spectral parameters, namely the peak fre-
quency, FWHM and SPR, are found to be independent of the variations in the
coupling conditions, and hence this technique has potential for reliable automatic
on-line measurement of grain size on the shop floor, where maintaining a uniform
coupling condition is rather difficult, if not impossible.
This study has also shown, for the first time, the influence of the frequency
characteristics of the transducers, which can be effectively utilized for grain size
measurements in steel with various scattering factors. In the case of austenitic stain-
less steel, with a higher scattering factor, the use of transducers with a single peak
frequency spectrum is ideal as a good correlation between the peak frequency and
FWHM with grain size could be established. In the case of ferritic steels with lower
scattering factors, the use of transducers with a double peak frequency is ideal as a
good correlation between the peak ratio (the amplitude at the lower-frequency peak
to that at the higher-frequency peak) and the grain size could be established.
In stainless steel, as the grain size increases, both the peak frequency and the
FWHM of the autopower spectrum of the first back-wall echo decrease. These
ultrasonic spectral parameters have been found to exhibit the Hall-Petch type of
relation with the grain size and hence can be linearly correlated with the yield stress,
if the other microstructural (substructural) features remain the same.
A linear correlation between the SPR and the grain size is found to be valid
for a wide range of practically relevant microstructures, namely ferrite,
 Microstructural characterization of steels 2487

ferrite + martensite and martensite in modified 9Cr-1Mo ferritic steel. The SPR
could also be used to identify the various metallurgical events affecting the grain
size, namely the formation of martensite (Ac1-Ac3temperature range), the dissolu-
tion of NbC and V4C3,and the formation of &ferrite ( A q temperature). For accu-
rate determination of these temperatures, on-line measurements would be required.

ACKNOWLEDGEMENTS
We are grateful to Dr P. Palanichamy, Dr K. Laha and Dr K. Bhanu Sankara
Rao for providing various specimens and useful discussions. We are grateful to Dr
Placid Rodriguez, Director, Indira Gandhi Centre for Atomic Research, Kalpakkam
for constant encouragement and support. We also thank Dr S. L. Mannan,
Materials Development Group, and Mr P. Kalyanasundaram, Division for Post-
Irradiation Examination and Non-destructive Testing Development, for their co-
Downloaded by [University of Lethbridge] at 13:21 02 November 2015

operation.

REFERENCES
ASTM Standard E 112-88, 1992, Annual book of ASTM Standarcis, Vol. 3.01 (Philadelphia,
Pennsylvania: American Society for Testing and Materials), p. 294.
BERGNER, P., and POPP,K., 1990, Scripta muter.. 24, 1357.
BHATIA,A. B., 1959, J. acoust. SOC.Am., 31, 16.
BROWN,A. F., 1973, Ultrasonics, 11, 203.
CANELLA, G., and MONTI,F., 1976, Non-Destructure Testing International, 9, 187.
FITTING, D. W., and ADLER,L., 1981, Ultrasonic Spectral Analysis for NDE (New York:
Plenum).
GERICKE,0. R., 1971, The Future of Ultrasonic Spectroscopy. edited by P. M. Reynolds
(London: British Non-Ferrous Metals Research Association), paper 1.
HETCH,A., THIEL,R. NEWMANN, E., and MUNDRY, E., 1981, Muter. Evaluation, 39, 934.
HONJOH,K., 1994, Jap. appl. Phys., 33, 1554.
JAYAKUMAR, T., 1997, PhD Thesis, University of Saarland, Saarbruecken, Germany.
JAYAKUMAR, T., MUKHOPADHYAY, C. K., KASIVISWANATHAN, K. V., and BALDEVRAJ,
1998, Trans. Indian Inst. Metals, 51, 485.
KLINMAN, R., WEBSTER, G. R.,MARSH,F. J., and STEPHENSON, E. T., 1980, Muter. Evalua-
tion, 38, 26.
MANNAN, S. L., 1981, PhD Thesis, Indian Institute of Science, Bangalore, India.
MASON,W. P., and MCSKIMIN, 1948, J. appl. Phys., 19, 940.
MORGAN,L. L., 1975, PhD Thesis, The City University, London.
ORR,J., BURTON,D., and RASCHE,C., 1992, The Manufacture and Properties of Steel 91 for
the Power Plant and Process Industries, 5 November 1992 (Dusseldorf: VDEh).
PALANICHAMY, P.,JOSEPH,A., and JAYAKUMAR, T., 1994, Insight, 36, 874.
PALANICHAMY, P., JOSEPH, A., JAYAKUMAR, T., and BALDEVRAJ, 1995, Non-Destructive
Testing Evaluation, 28, 179.
PAPADAKIS, E. P., 1965, J. acoust. SOC.Am., 37, 711; 1981, Physical Acoustics: Principles and
Methods, Vol. 46 (New York: Academic Press), 269; 1984, Int. Metall. Rev., 1, 1.
SIMPSON, W. A., JR, 1974, J. acoust. SOC.Am., 56, 1776.
WILLIAMS, H., and GOBBLES, K., 1981, Metal Sci., 15, 549.