Classification of AE Signals for Monitoring

Martensite Formation from Welding

BY X. LIU A N D E. KANNATEY-ASIBU, Jr.

Acoustical emission shows promise as a real-time monitor

of martensite formation

ABSTRACT. Welding of high-carbon and

some high-hardenable alloy steels often
can cause martensite formation, which
may lead to cold cracks. Acoustic emission
is one of the most promising techniques
for monitoring martensite formation, since
it can be implemented in real time with a
remote sensor, and is highly sensitive and
nondestructive. In this paper, martensitic
transformation occurring during welding
of AISI 4340 steel is identified from signals
from porosity and sound welds using pattern recognition analysis of the acoustic
emission signals generated.
A frequency-based pattern recognition
technique using linear discriminant functions has been implemented with successful classification. Using a binary decision
strategy and independent data testing,
classification rates of 100 and 99.1%,
respectively, were obtained between
sound welding and martensite formation/
porous weld signals, and 96.4 and 76.0%
between martensite formation and porous weld signals, respectively, were
achieved using the eight best features.
However, for a simultaneous decision
strategy and independent testing, classification rates of 96.4, 70.0 and 100% were
obtained for martensite formation, porous and sound welding signals, respectively, using all 17 features.
Introduction
Welding of high-carbon and some alloy
steels often introduces conditions that result in martensite formation. Martensite
may lead to the formation of cold cracks,
which in general are detrimental to weld
quality and cannot be permitted in most
applications. Thus, it is desirable to monitor the formation of martensite in real
time so that welding conditions ,can be
changed to prevent further damage to the
weldment. Furthermore, welding discontinuities can be more easily repaired when
detected in real time. Traditional nondestructive testing techniques such as radiography, ultrasonic testing, etc., cannot
detect the martensite formed from a
welding process. Detection by micros-

copy is destructive and time consuming.

Acoustic emission (AE) is a term describing a phenomenon whereby transient
elastic waves are generated by the rapid
release of energy from localized sources
within a material, such as plastic deformation, fracture, crack formation and martensitic transformation. It is a high-sensitivity technique for detecting active macroscopic events in a material. Since these
events give rise to elastic waves that
propagate out into the material, it is not
necessary to focus on the exact location
of the source event to detect it. Thus, AE
is a monitoring technique that, rather than
requiring 100% volumetric scanning, can
inspect using remote and contact sensors.
In-process inspection with AE allows repair to be accomplished on a pass-by-pass
basis, and either eliminates or greatly
reduces postinspection repair. This can allow a significant reduction in fabrication
costs through reduced repair cost.
This paper presents the results of experiments done to monitor martensitic
transformation during welding of AISI 4340
steel and classification of AE signals generated by martensite formation from signals
associated with porosity and sound welds.
Background
Acoustic emission monitoring of martensitic transformation has been investigated by a number of researchers. The
"click" that occurs during the formation of
martensite in high-nickel steel was re-

KEY W O R D S
Acoustic Emission
AE Signals
Martensite Formation
High-Carbon Steel
Hardenable Alloys
Weld Porosity
Martensitic Transformation
Multiple-Pass Welding
Monitoring AE Signals

ported by Forster and Scheil (Ref. 1) in one

of the first studies of AE in the literature.
However, from that time until the early
1970s, no further work was done on AE
generated during phase transformations
in solids. Liptai, ef al. (Ref. 2), studied AE
during formation of martensite in goldcadmium and indium-thallium alloys and in
cobalt and plutonium, during both heating
and cooling. He also studied the eutectoid
transformation in a tin-cadmium alloy and
showed that such a diffusion-controlled
transformation did not generate AE. Speich and Fisher (Ref. 3) have studied martensite formation in an Fe-28Ni-0.11C alloy steel using AE. It was found that an average of 15 martensite plates were
involved in each acoustic emission event,
and the volume of martensite formed per
acoustic emission event decreased with
increasing volume fraction. Phase transformation in a wide variety of steels was
also investigated by Speich and Schwoeble (Ref. 4). They found that acoustic
emission was generated during the formation of martensite but not during the
formation of ferrite, bainite or pearlite,
and the intensity of acoustic emission
generated during martensite formation
increased markedly as the carbon content
of the steel increased. Simmons and Wadley (Ref. 5) undertook the first quantitative
work relating acoustic emission to the
process of martensitic transformation.
Subsequent quantitative work was done
by Kannatey-Asibu and Dong (Ref. 6), relating the Gibbs free energy change (the
driving force for the transformation) to
the RMS value of acoustic emission signals
(a measure of the intensity of AE signals).
Many researchers have demonstrated
the potential for acoustic emission monitoring of welding. Jolly (Ref. 7) investigated the possibility of acoustic emission
for in-process inspection of welds. His ex-

X. LIU is a Project Engineer with Automated

Analysis Corp., Peoria, III. E. KANNATEY-ASIBU,
Jr., is an Associate Professor in the Department
of Mechanical Engineering, The University of
Michigan, Ann Arbor, Mich.

WELDING RESEARCH SUPPLEMENT 1389-s

perimental results showed that cracks and

porosity in the welds emit acoustic emission signals during and after solidification
of the weld, and some weld discontinuities not detected by radiography were
detected by AE. Acoustic emission measurements during electron beam welding
of nickel-base alloy were made by Dickhaut and Eisenblatter (Ref. 8). T w o different types of acoustic emission were found:
the background noise caused by weld solidification, and flaw formation indicated
by single-burst signals having a much
higher amplitude than those due to solidification. Kumar (Ref. 9) applied acoustic
emission to gas metal arc weld monitoring.
He concluded that selective filtering could
increase the signal-to-noise ratio so that
cracks could be detected during real-time
welding of A-514 steel. However, the
acoustic emission rate from porous welds
was of the same magnitude as that from
the welding arc noise and would be difficult to detect during real-time welding.

Prine has made substantial contributions

to acoustic emission monitoring of arc
welding. He studied several flaw types,
including cracks, slag inclusion, porosity,
incomplete penetration and incomplete
fusion. The results of Prine's work (Ref. 10)
have indicated the detectability of t w o
primary discontinuities using AE: cracks
and slag inclusion. However, his work did
not include the formation of martensite.
No significant research has been undertaken to monitor martensitic transformation from welding, and hence, is the
primary scope of this investigation.
Pattern recognition technique has been
used to distinguish various types of AE
signals. Linear discriminant function analysis of acoustic emission signals for cutting
tool monitoring was done by KannateyAsibu and Emel (Ref. 11) in which frequency-based pattern-recognition concepts using linear discriminant functions have been
used in analysis of acoustic emission signals from sources such as chip formation,

Welding Gun

tool fracture and chip noise. Acoustic

emission signals obtained during submerged arc welding of a plain carbon steel
were classified using pattern recognition
methods by Chan, et al. (Ref. 12). The
classifiers were able to clearly distinguish
between normal process noise and AE
signals due to internal discontinuity generation during unsound welding conditions.

Experimental
A gas metal arc welding system was
utilized for the experiments, with specimens consisting of 1.5-in. (38-mm) diameter 1018 and 4340 steel bars, 28 in. (711
mm) in length. The welding conditions included a welding current of 300 A, voltage
of 30 V, 13.1 ft/min (4 m/min) wire feed
rate, Vi6-in. (1.6-mm) mild-steel wire diameter, 17.7 in./min (7.5 mm/s) welding
speed, and a welding time of 20 s, using
argon (Ar) as the shielding gas. The piezoelectric transducer used was a wide-band
PAC WD-445, which was mounted on the
surface at the end of the bar. The signals
were first preamplified by 40 dB and then
further amplified by 15 dB during welding,
and 35 dB after welding. They were
band-pass filtered between 30 and 1000
kHz, and then recorded with a modified
video recorder for later analysis. An oscilloscope was used to monitor the signals
while the AE RMS voltage was sampled at
30 Hz using a 12-bit data-acquisition systemFig. 1. O n play back, the recorded
AE signal was passed through a gating unit
where a 16.7-ms length of the signal was
clipped for sampling, with the recorder
put in a pause mode at a desired location
of the tape. To avoid saturation, the signal
was then passed through an attenuator to
reduce the amplitude to a level between
254 and -F254 mV compatible with the
8-bit high-speed waveform digitizer,
which was set at a sampling rate of 4 MHz.
Sixteen kbytes sample data points were
used for analysis, covering a time period
of 4 ms.
Three types of experiments were undertaken promoting the following three
as-welded conditions: 1) martensite formation, 2) porous weld and 3) sound
weld. Each set of experiments was conducted under essentially the same conditions. AISI 4340 steel was used to ensure
martensitic transformation, since a nearly
complete transformation occurs at cooling rates greater than 6C/s (Ref. 4). Porosity was introduced by shutting off the
Ar gas supply during the welding of a 1018
steel. For a sound weld, 1018 steel and Ar
gas were used during welding, and no
porosity was found on the surface or inside the specimens. Since 1018 steel has
very low hardenability, no martensite
formed.

Oscilloscope
Fig. 1 Experimental setup for monitoring AF signals from welding.

390-s I OCTOBER 1990

A total of 18 experiments were conducted, with each class having six experiments. For each experiment, about 15

2.0

cn

0.6

>
co

20

32

40
Time (s)

Fig. 2 RMS voltage during and after welding of a 4340 steel with 65 dB
gain.

samples w e r e o b t a i n e d f o r martensite
f o r m a t i o n , p o r o u s a n d s o u n d w e l d s . They
w e r e s a m p l e d at a b o u t u n i f o r m t i m e
intervals.

Results and Discussion

RMS Voltage of AE Signals
Figure 2 s h o w s the AE RMS voltages
o b t a i n e d d u r i n g a n d after w e l d i n g of a
4 3 4 0 steel s p e c i m e n using an amplification
gain o f 65 dB. The signal in the first 20 s
c o m e s f r o m arc noise (during welding)
a n d has large fluctuations in a m p l i t u d e
since the w e l d i n g current itself fluctuates.
T h e average a m p l i t u d e is a b o u t 0.6 V ,
w h i c h is m u c h higher than t h e martensitic
t r a n s f o r m a t i o n signal, w h i c h has an amplit u d e o f a b o u t 0.1 V. C o n s e q u e n t l y , it is
quite difficult t o d e t e c t t h e martensitic
t r a n s f o r m a t i o n signal directly d u r i n g w e l d ing. Martensitic t r a n s f o r m a t i o n begins at
the M s (martensite start) t e m p e r a t u r e . The

48
Time (s)

Fig. 3 - RMS voltage during (55 dB) and after (75 dB) welding of a 4340
steel.

signal g e n e r a t e d d e p e n d s o n t h e cooling
rate and t h e a m o u n t of material i n v o l v e d
(Ref. 13). T h e t e m p e r a t u r e a n d c o o l i n g
rate for a w e l d e d specimen in t u r n d e p e n d o n t h e w e l d i n g c o n d i t i o n s such as
arc v o l t a g e , c u r r e n t , w e l d i n g t i m e , size o f
t h e s p e c i m e n , etc. Martensitic t r a n s f o r m a t i o n normally occurs b e l o w 4 0 0 C (752F)
in m a n y steels (Ms = 3 0 0 C / 5 7 2 F f o r
the 4 3 4 0 steel). Since t h e w e l d deposit
generally cools s l o w l y , t h e t r a n s f o r m a t i o n
can b e d e t e c t e d after w e l d i n g is c o m p l e t e d . For a multipass w e l d i n g process,
d e t e c t i o n of martensitic t r a n s f o r m a t i o n
after each pass w o u l d p r o v i d e a w a r n i n g
t o t h e w e l d i n g system, enabling p r o p e r
action t o b e taken b e f o r e the next pass.
For the 6-in. (152-mm) w e l d beads in o u r
experiments, the transformation was
found to occur for a period of about 4 0 6 0 s after c o m p l e t i o n of t h e w e l d . Signals
f r o m martensite f o r m a t i o n are c o n t i n u o u s
t y p e , giving a s m o o t h RMS v o l t a g e c u r v e .
C o m p a r i n g Figs. 3 and 4 , w e notice that

ra 0.6 -

t h e spiky RMS voltage d u r i n g t h e first 10

s after w e l d i n g is associated w i t h s o u n d
w e l d i n g signals. A l t h o u g h martensite is
t r a n s f o r m i n g d u r i n g this t i m e p e r i o d , w e
o n l y started sampling martensite f o r m a t i o n signals a b o u t 15 s after the arc w a s
s t o p p e d so as t o minimize all o t h e r signal
sources. A p o r o u s w e l d (Fig. 5) generates
signals w i t h amplitudes greater t h a n a
s o u n d w e l d (Fig. 4), a n d t h e y b o t h h a v e
burst signals s u p e r i m p o s e d o n c o n t i n u o u s
signals. The spikes in the RMS voltage
curves are f r o m burst signals, w h i c h m a y
c o m e f r o m o x i d e cracking or m i c r o c r a c k ing d u r i n g c o o l i n g o f t h e s p e c i m e n . Since
burst t y p e signals have large variations in
amplitudes, it is better t o utilize the c o n tinuous signals f r o m b o t h processes, a n d
samples w e r e t a k e n at a b o u t u n i f o r m t i m e
intervals. T h e system's total gain w a s increased t o 75 dB immediately after w e l d ing in o r d e r t o o b t a i n large e n o u g h signal
amplitudes f o r p a t t e r n - r e c o g n i t i o n analysis. By c o m p a r i n g Figs. 3 - 5 , w e can see

ra 0.6 -

>

>
CT)

40

Time (s)
Fig. 4 RMS voltage for sound welding of a 1018 steel.

Fig. 5-RMS

20

40
Time (s)

60

80

voltage for a porous weld of a 1018 steel.

WELDING RESEARCH SUPPLEMENT 1391-s

-10
^ H B Martensite
I I
Porosity
H i Normal

-16-

-20

"

-30

-18-

i * f

3
o

* \

Ooo0c

20-

-40 -

* * *

<b o f %.

0o

Q.

"

'

-50
0.0

i
0.2

0.4

Martensite

lu

o Porosity
??

' ' "

-27

1.0

0.6

| M 1 II II 1 1 | l l IL! 1 1 1 1 |

-26

Frequency (MHZ)

A p a t t e r n - r e c o g n i t i o n system requires
t h r e e essential steps (Ref. 11):
1) Sampling or scanning the input signal
t o p r o d u c e t h e p a t t e r n space.
2) Extraction of the features, w h i c h often involves t r a n s f o r m a t i o n o f the signal
f r o m the p a t t e r n into the feature space,
f r o m w h i c h useful i n f o r m a t i o n can b e o b tained, a n d subsequent r e d u c t i o n of the
data size.

Signal discrimination is f u n d a m e n t a l t o
AE m o n i t o r i n g of processes. The high
sensitivity inherent in t h e AE t e c h n i q u e
leads t o p r o b l e m s in distinguishing b e t w e e n signals o f interest a n d those that
are not desirable, i.e., noise. D u r i n g w e l d ing, martensite f o r m a t i o n , p o r o s i t y a n d
the s o u n d w e l d all g e n e r a t e d intense AE

Fig. 8 Decision
space plot for
martensite
formation-porous
weld.

3) Classification of the feature space t o

p e r m i t identification o f the individual signal sources (classes).
A total of a b o u t 300 samples of signals
f r o m martensite t r a n s f o r m a t i o n , p o r o s i t y
and s o u n d w e l d s w e r e o b t a i n e d i n d e p e n -

//

<*>

0*

>.
17>
o
o
DL

%o

\0-_

0.5-

/ x
'

0.0-

//
/
/

^&t

Vb

*
%-s,

' Martensite

**

0.0

0.5
Martensite

392-s | O C T O B E R

1990

1 | 1 1 1 . 1 1 1 1 1

H l |

-22

-21

-20

weld.

dently. T h e s u p e r i m p o s e d burst signals

w e r e used f o r p a t t e r n - r e c o g n i t i o n analysis because those are m o s t likely noise signals f r o m o x i d a t i o n .
T h e sampled data w e r e t r a n s f o r m e d
into a f e a t u r e space via Fourier t r a n s f o r m .
This t r a n s f o r m p r o v i d e d a clear physical
i n t e r p r e t a t i o n of the results and fast algorithms are readily available. In o r d e r t o
minimize distortion of t h e original signal
t r a n s f o r m , a relatively large data (16
kbytes) size w a s used. Large data blocks
increase spectral resolution b u t require
m o r e time f o r c o m p u t a t i o n s . T o i m p r o v e
the c o m p u t a t i o n a l efficiency, the data size
w a s r e d u c e d b y averaging o v e r a f r e q u e n c y b a n d o f 6 0 kHz a n d eliminating
the spectral c o m p o n e n t s that contain less
i n f o r m a t i o n a b o u t t h e system. Figure 6
s h o w s t h e n o r m a l i z e d m e a n spectra o f
the t h r e e classes of signals in the f r e q u e n c y range 30 k H z - 1 M H z . T h e t o t a l
p o w e r of a signal and the m a g n i t u d e of 16
n o r m a l i z e d spectral c o m p o n e n t s w e r e
used as features. T h e class m e a n scatter
criterion w a s used f o r f e a t u r e selection.
T h e n u m b e r of features used is v e r y
i m p o r t a n t , since in a d d i t i o n t o r e d u c i n g
c o m p u t a t i o n a l efficiency, a high d i m e n sionality requires a large n u m b e r o f experimental data sets f o r training the syst e m a n d d e v e l o p i n g the classifier. This is
because the a d e q u a t e n u m b e r of training
data sets has t o be a b o u t f o u r or m o r e
times the n u m b e r of features. H o w e v e r ,
an excessive n u m b e r of data sets adds t o
the cost w i t h o u t necessarily i m p r o v i n g the
system p e r f o r m a n c e . O n the o t h e r h a n d ,
a small training set will result in an inadeq u a t e evaluation of the classifier w e i g h t ing functions.

o Porosity

-0.5 '
-0.5

Ill ] 1

-23

Fig. 7 Feature space plot for martensite formation-porous

signals. T h e p a t t e r n - r e c o g n i t i o n t e c h n i q u e
p r o v i d e s a means by w h i c h positive signal
sorting, classification and source identification can b e o b t a i n e d . T h e b a c k g r o u n d
o n the p a t t e r n - r e c o g n i t i o n scheme used
f o r classifying the AE signals is p r e s e n t e d in
an earlier p a p e r (Ref. 11) a n d will n o t be
r e p e a t e d here.

Pattern-Recognition Analysis

-24

3 3 0 KHz

Fig. 6 Normalized mean spectra for three classes of signals.

that the f o r m a t i o n of martensite and p o rosity resulted in larger a m p l i t u d e signals.

H o w e v e r , signals f r o m the p o r o u s w e l d
decrease quickly t o the b a c k g r o u n d noise
level in a b o u t 15 s after the w e l d i n g p r o cess stops. Since martensitic t r a n s f o r m a t i o n occurs in the t e m p e r a t u r e range of
3 0 0 - 8 0 C ( 5 7 2 - 1 7 6 F ) , w e are able t o
o b s e r v e martensite f o r m a t i o n signals o v e r
a relatively long p e r i o d o f t i m e . T h e s o u n d
w e l d i n g process also generates signals
right after w e l d i n g , but their amplitudes
are smaller a n d diminish w i t h i n 15 s.

-25

1.0

1.5

T w o of t h e best features f o r martensite

f o r m a t i o n a n d p o r o u s w e l d signals are
p l o t t e d in Fig. 7, a n d it is e v i d e n t that t h e r e
is s o m e o v e r l a p b e t w e e n these t w o signals, and that can cause misclassification.

AE Signals

a)

-35

Normal
Weld

-40

AE Signals

*x

on

o
b)

Martensite Formation
& Porous Weld

Normal Weld

-45H

CO

0)

-50-

* Martensite
Porosity

Martensite Formation

Porous Weld

x Normal
-55
30

a) Simultaneous Decision
b) Binary Decision

For simultaneous decision strategy,

there is a large overlap between martensite formation and porous weld signals in
the feature space Fig. 10. However, between sound weld signals and martensite
formation/porous weld signals, the overlap is small. All 17 features had to be used
in order to achieve relatively good classification rates of 96.4, 70 and 100% for
martensite formation, porous weld and

40

45

50

Power (dB)

Fig. 9 Decision strategies for classification of AE signals.

Further transformation of the patterns

from the feature space to a decision space
using a transform that relocates the patterns such that all patterns that belong to
a specific class are positioned in a region
distinct from other classes, results in a decision space that facilitates classification
Fig. 8.
The linear discriminant function approach involves developing a set of linear
functions, one for each class of interest.
An unclassified signal or pattern is used to
calculate the value of each function, and
the signal is assigned to the class with the
largest function value. The last step, that
of classification, can be implemented in
one of several ways. A common approach is to determine all function values
at the same time and compare, i.e., simultaneous decision strategy Fig. 9A. Another way is the binary decision strategy
(Fig. 9B) in which w e first classify the AE
signals into sound weld signals, and martensite formation/porous weld signals,
and then further distinguish between the
martensite formation and porous weld
signals.

35

Fig. 10-Feature
sound weld.

space plot for martensite formation - porous

sound weld signals, respectively - Table 1.

From Fig. 10, we observe that binary
decision would result in better classification, since the overlap between sound
weld and martensite formation/porous
weld is small. Classification rates of 100
and 99.1% were obtained for sound weld
and martensite formation/porous weld
signals, respectively (Table 2), with the
eight best features, while between martensite formation and porous weld signals,
classification rates of 96.4 and 76.0% were
achieved. With additional features, classification rates between martensite formation and porous weld signals increased insignificantlyTable 2. It is interesting to
note that successful classification between
sound weld and martensite formation/
porous weld signals depends to a great
extent on the signal power, and that
increasing the number of features does
not produce a significant improvement in
classification when there are more than
eight features.
Conclusions
Experiments were done to monitor AE
signals from martensitic transformation
during welding. The signals are relatively
small in amplitude compared with arc
noise during welding. For multipass welding, the detection of martensitic transformation signals between passes will enable
appropriate action to be taken to minimize defect formation. In addition to

weld,

Table 1Simultaneous Decision

StrategyIndependent Data Test
No. of
Features

Martensite
Formation

Porous

Weld

Normal
Weld

5
8
11
14
17

80.0
67.3
100
98.2
96.4

38.0
40
60
62.0
70.0

95.6
97.8
97.8
100.0
100.0

Table 2Binary Decision

StrategyIndependent Data Test

Number
of
Features
1
2
5
8
11
14
17

Martensite
Formation/
Porous
Weld vs.
Normal Weld

Martensite
Formation vs.
Porous Weld

99.1
100
97.1
99.1
100
100
100

92.7
94.6
90.9
96.4
94.6
96.4
92.7

86.7
88.9
95.6
100
100
100
100

80.0
74.0
78.0
76.0
80.0
76.0
76.0

transformation signals, other signal sources

such as weld porosity and sound weid also
generate acoustic emission signals. Discriminant functions were obtained for
each of these sources using the magnitude
of normalized spectral components in the
frequency range of 30 KHz-1 MHz, and
the power as features formed the basis for

WELDING RESEARCH SUPPLEMENT 1393-s

classification. For simultaneous decision,

17 features have to be used for better
classification. For binary decision, eight
features are appropriate for classification,
first between sound weld and martensite
formation/porous weld, and then between martensite formation and porous
weld. The coefficients of the discriminant
functions were obtained by training the
system using signals generated by each of
the sources of interest. The results showed
a classification rate of 96.4, 70.0 and 100%
for martensite formation, porous weld
and sound weld signals, respectively, with
17 features and a simultaneous decision
strategy. For binary decision with eight
features, 100 and 99.1% between sound
weld and martensite/porosity formation
signals, and 96.4 and 76.0% between
martensite and porosity signals were obtained.

Acknowledgment
The support of this work by the National Science Foundation under grant No.

108, pp. 328-331.

7. Jolly, W. D. 1969. Acoustic emission exposes cracks during welding. Welding Journal
48(1): 21-27.
8. Dickhaut, E and Eisenblatter, J. 1975.
Acoustic emission measurements during electron beam welding of nickel-base alloys. ASME
References
J. Eng. Power, Vol. 97, pp. 47-52.
9. Kumar, A. 1980. Application of acoustic
1. Forster, F., and Scheil, E. 1936. Zeitschrift
emission to gas metal arc (CMA) weld monitorfur Metallkunde, Vol. 9, pp. 245-247.
ing. Nondestructive Evaluation in the Nuclear
2. Liptai, R. C , Harris, D. O., Engle, B. B., and
Industry-1980, ASM International, pp. 121-136.
Tatro, C. A. 1971. International Journal of Nondestructive Testing, Vol. 3, pp. 215-275.
10. Prine, D. W. 1978. A two channel microprocessor-controlled acoustic emission
3. Speich, C. R and Fisher, R. M. 1972.
monitor for in-process weld monitoring. ProAcoustic emission during martensite formation.
ceedings, 24th Annual ISA Conference, AlbuAcoustic Emission, ASTM STP 505, American
querque, N. Mex.
Society for Testing and Materials, pp. 140-151.
11. Kannatey-Asibu, Jr., E., and Emel, E.
4. Speich, G. R., and Schwoeble, A. J. 1975.
1987. Linear discriminant function analysis of
Acoustic emission during phase transformation
acoustic emission signals for cutting tool monin steel. Monitoring Structural Integrity by
itoring, lournal of Mechanical Systems and SigAcoustic Emission, ASTM STP 571, American
nal Processing, Vol. 1, No. 4, pp. 333-347.
Society for Testing and Materials, pp. 40-58.
12. Chan, R. W. Y., Hay, D. R., Caron, V.,
5. Simmons, J. A., and Wadley, H. N. C.
Hone, M and Sharp, R. D. 1985. Classification
1984. Theory of acoustic emission from phase
transformation. Journal of Research of the Na- of acoustic emission signals generated during
tional Bureau of Standards, Vol. 89, No. 1, pp. welding. Journal of Acoustic Emission, Vol. 4,
No. 4, pp. 115-123.
55-64.
6. Kannatey-Asibu, Jr., E., and Dong, P.
13. Liu, X., and Kannatey-Asibu, Jr., E. 1990.
1986. Analysis of acoustic emission signal genAcoustic emission during athermal martensitic
eration during martensitic transformation.
transformation in steels. ASME j. Eng. for Ind.,
ASME Journal of Engineering for Industry, Vol. Vol. 112, pp. 84-91.
MSM-8519980
edged.

is

gratefully

acknowl-

WRC Bulletin 336

September 1988
Interpretive Report on Dynamic Analysis of Pressure ComponentsFourth Edition
This fourth edition represents a major revision of WRC Bulletin 303 issued in 1985. It retains the three
sections on pressure transients, fluid s t r u c t u r e interaction and seismic analysis. Significant revisions
were made to make t h e m current. A new section has been included on Dynamic Stress Criteria which
emphasizes the importance of this technology. A new section has also been included on Dynamic
Restraints that primarily addresses snubbers, but also discusses alternatives to snubbers, such as limit
stop devices and flexible steel plate energy absorbers.
Publication of this report was sponsored by the Subcommittee on Dynamic Analysis of Pressure
Components of the Pressure Vessel Research Committee of the Welding Research Council. The price of
WRC Bulletin 336 is $20.00 per copy, plus $5.00 for postage and handling. Orders should be sent with
payment to the Welding Research Council, Suite 1 3 0 1 , 345 E. 4 7 t h St., New York, NY 10017.

WRC Bulletin 355

July 1990
Programming and Control of Welding ProcessesExperience of the USSR
By V. Malin
This report is an in-depth look at technical welding studies and their implementation in the USSR, a
c o u n t r y t h a t has a long history of welding automation development. More than 300 articles published in
the USSR over the last three decades were examined, and 177 are referenced in this report.
Publication of this report was sponsored by the Interpretive Reports C o m m i t t e e of the Welding
Research Council. The price of WRC Bulletin 355 is $35.00 per copy, plus $5.00 for U.S. and $10.00 for
overseas postage and handling. Orders should be sent with payment to the Welding Research Council,
345 E. 4 7 t h St., Room 1 3 0 1 , New York, NY 10017.

394-s | OCTOBER 1990