CONF-8908160 — 2 '-' !

'l )•'') , '-

DE90 001335
EDDY-CURRENT INSPECTION OF SHUTTLE HEAT EXCHANGEE TUBE WELDS'
C. V. Doddf, G. W. Scott' and L. D. Chitwoodf
ABSTRACT
The Space Shuttle Main Engine is inspected during manufacture,
after each test firing, and as part of refurbishment between
missions. The LOX heat exchanger has four tubing welds in
locations that are difficult to reach and inspect. The most
difficult to access is located approximately 30.6 in. (0.78 mj from
the inlet, through tubing 0.190 in. (4.83 mm) ID, past a 90 degree,
0.69 in. (18 mm) radius bend, and approximately halfway around a
loop 16.5 in. (42 cm) diameter. We designed a multiple property
test using computer-aided engineering software written, at OWL
earlier,, that modeled the tube and probe and enabled simulation of
instrument response in the inspection environment. Design
parameters were optimized for a single-winding, single-layer
"pancake" coil. The necessary sensitivity to small defects can be
achieved within the limits imposed by material variations. The
probe contains an array of eight coils. Four coils each are
mounted 90 degrees apart on the equators of two plastic spheres
joined at their poles by a straight section. The two four-coil
arrays are offset 45 degrees, so that as the probe passes through
the tube, one coll covers each 45 degree segment of the wall
circumference. Lift-off up to 0.008 in. (0.2 mm) is electronically
compensated by corrections developed in the design, so there is no
requirement to maintain intimate contact between the probe and the
tube wall.
Experimental multifrequency measurements performed on tubing weld
specimens supplied by Rockwell Rocketdyne Division indicated the
presence of territe in the welds, which likely accounts for
anomalous results observed in single-frequency eddy-current tests.
In addition, there was a t/eld pedestal that ranged up to 0.004 in.
(0.10-mm) in the tube samples tested. This pedestal tended to
produce signals that were in a direction opposite to that of the
defect signals.
The system response to multiple discrete frequencies was simulated
by computing the response to each Independently and combining the
results. The presence of ferrite in the welds suggested the use
of pulse excitation to reduce sensitivity to variations in
permeability. Final system tests demonstrated that a 0.005 in.
(0.13 mm) deep defect could be detected with the array probe, using
the three-frequency eddy-current instrument driving the multiplexer
coils. A 0.003 in. (0.08 mm) deep defect and a 0.006 in. (0.25 mm)

rrn,
Research Sponsored by NASA 'VfTiU I i-Il
f
Oak Ridge National laboratory, operated by Martin Marietta
Energy Systems Inc., under contract DE-ACO5-84OR21400 with the
U.S. Department of Energy
' Martin Marietta Michoud, New Orleans.
toy ii n « i a a or ite us

(USTJUfiUTION OF THIS DOCUMENT \l UKUMHED SESTETS *£* T ^

It* M U M lonm of M conrCUHn. or
•low < * n K db K. 'fan US
 DISCLAIMER

This report was prepared as an accounl of work sponsored by an agency of the

United States Government. Neither the United States Government nor an; agency
thereof, nor an\ of their employees, makes any warranty, express or implied, or
assumes any legal liabilin or responsibility for the accuracy, completeness, or use-
fulness of an> information, apparatus, product, or process disclosed, or represents
that its use would not infringe privately owned rights Reference herein to any spe-
cific commercial product, process, or service b\ trade name, trademark, manufac-
turer, or otherwise does not necessanii constitute or imply its endorsement, recom-
mendation, or favoring by the United Stales Government or any agency thereof
The views and opinicns o( authors expressed herein do not necessarih stale or
reflect those of the I. ruled States Government or an> agency thereof
deep defect at the welds were detected in the split tube samples.
A single coil was driven by the impedance analyzer at three and
then six frequencies The signal to noise ratio was best at the
six-frequency test, which indicates that the pulse test, with its
higher frequency content, will be superior to the three-frequency
test. However, the pulsed measurements were not made due to time
constraints.
INTRODUCTION
Background
The present method for manufacturing inspection the of welds in
heat exchanger tubing and combustion chamber injection posts in the
Space Shuttle Main Engine (SSME) uses single-frequency eddy
currents. There were difficulties with signal interpretation and
probe clearances through accesses when using this method. Eddy-
current tests have not been applied to the inservice inspection of
these components.
The Oak Ridge National Laboratory has expertise in multiple-3
frequency and pulse excited eddy-current inspection technology.'
The incentive for use of multiple-frequency techniques is the need
to achieve accurate detection and sizing of weld flaws while
rejecting or discriminating against the effects of variations in
other properties, such as tubing wall thickness or probe liftoff.
Single-frequency systems respond to several different property
variations, and in many cases c&nnot distinguish the anomalous test
indications from flaws, causing false positives.
Pulse Eddy-current (PEC) technology effectively examines the part
with many frequencies.'•* Measurements of the pulse shape required
for analyses, however, can be accomplished with simpler, more
reliable^ and less expensive instrumentation than that required for
multiple discrete frequencies. If the pulses are large enough,
they saturate any permeability changes in the material and reduce
their effect on the test. Nevertheless, PEC was a later
development because the analysis leading to signal processing
algorithms was more difficult.
Problem
A representation of the heat exchanger along with the three welds
chat were proposed for technique development are shown in Figure l.
The welds are listed below:
(1) the LOX heat exchanger Primary Tube-Bifurcation Joint Transi-
tion (RS003812 -#3 in Figure 1 ) ;
(2) the LOX Heat Exchanger Secondary Tube-Bifurcation Joint (so-
called "Baby Pants") welds (RS008812 -#l and #2 in Figure 1 ) ; and
(3) the Main Combustion Chamber Injection Posts which are straight
tubes in the sama size range with similar welds.
 tax
MOT
OXYRCM

TOIYTAM
ompica
FftOM BYPASS
OfttFSCf

Figure 1 Space Shuttle Main Engine Heat Exchanger Tube Assembly

The second weld from the inlet (RS00&812 -#3 in Figure 1) was
judged the most difficult and was selected for demonstration, Held
#3 joins the 0.190 in. ID x 0.0125 in, wall inlet tubing to the
expansion tubing. It is located approximately 30 in. downstream
of the inlet flange, but the probe must negotiate a sharp (0.69 in.
radius) bend to reach the weld. Alternate access is from the
outlet port, but from that direction, the probe must traverse
approximately 26 ft. of 0.330 in. ID tubing wound into a cylindri-
cal coil approximately 16 in. diameter.
The target flaw for detection is 0.003-in.-deep x 0.075-in.-long.
Depth is measured in the radial direction; length is measured
circumferentially. Flaws of this general shape may result from
lack of penetration in the original weld or cracks induced by
forming or low cycle fatigue during service.
Capability is desired for manufacturing inspection, inservice
inspection following ground test engine firing, and inspection
during refurbishment between shuttle flights.
Objective
This goal of this project was to develop the system necessary to
demonstrate in the laboratory that an eddy-current system can
inspect the tubes and welds described above, screening for the
existence of flaws equal in size to, or larger than, the target
flaw. The laboratory system was to include the probe necessary to
traverse the tubing, the electronics to dri^e (i.e., electrically
excite) the probe and receive and process signals from it, a data
display, data recording and playback devices, and microprocessor
software or firmware necessary to operate the system.
 APPROACH
Eddy-Current Database Review and Preliminary Estimates
ORNL has an extensive library of software that accomplishes for
eddy-current testing what is now commonly called "computer aided
engineering" (CAE). This software allows the insertion of various
parameter values for a complete test environment, including a
simulated specimen, probe coil, and instrument. These parameters
can be systematically varied to produce an optimized coil design
and a set of optimized system operating parameters (e.g., operating
frequencies) for a specified range of specimen properties.
From the results of years of simulations for many system designs,
a large database has been developed. This database provides a
ready reference for the rapid estimation of new problems and a
checkpoint for new results as they are computed. Approximate
curves and "rules of thumb" have been developed. These indicated
that, for defect detection in this size tubing and with these
conductivity values, a coil mean radius of about 0.025 in. and an
operating frequency of 1.5 MHz were suitable. With these
parameters, a defect about 20% of the wall thickness should be
detectable.
Probe Configuration
Rocketdyne previously developed a fluid pumping system, freon
filled, capable of propelling a string of spherical beads to pull
a cable through the finished heat exchanger. Therefore, we looked
at bead shaped carriers for the probe coils that could be adapted
for use with the existing freon drive system. However, for the
demonstration system, a rigid dumbbell shaped carrier (two spheres
connected by a straight section) was connected to a stiff cable
that could be pushed by hand through the inlet entry tube.
Figure 2 shows the probe form inside the bend in the tuba. The
coils are molded to the spherical shape of the probe form and
mounted in the recesses provided.
The freon pumping system offers limited restraint against torques
and appropriate connections can prevent continuous rotation of the
beaded inspection probe. To accomplish inspection of the entire
weld circumference, probe coils were mounted on two successive
beads in a string with an angular offset to create a circular
array, so that when the complete string has passed the weld, the
entire circumference will have been inspected. The probe contains
an array of eight coils. Four coils each are mounted 90 degrees
apart on the equators of the two spheres, in patterns which are
offset 45 degrees. The diameter of the spheres was chosen to allow
passage around the curve in the tubing and through the weld
sections of the tube. The shape of the probe allows the probe to
pass the small 90 degree entry bend without flexing the straight
section.
Figure 2 Probe form inside the small radius bend of the heat
exchanger tube

Coll Design
Dimensions and electrical parameters for individual probe coils
were estimated from reference data derived from the results of
previous computer simulations, as described above. However,
because of the introduction of variables not considered in the
preliminary estimate, confirmatory simulation by computer was
considered to be a prudent precaution. Although the coils in the
holder have spherical contours, the approximation of a flat pancake
coil was the best available. The curvature of the coil will add
some liftoff to parts of the coil. The effect of this curvature
can be approximated by making the liftoff for the flat coil case
equal to that of the curved coil, measured at the mean radius of
the curved coil.
System Options and Configuration
Probes using coil arrays similar to that described above have
previously been used with single-channel instrumentation by
including a multiplexer to switch connections to the instrument in
a programmed sequence. Sufficient switching speed can be achieved
to compensate for continuous probe motion while ensuring adequate
surface coverage. Single-channel instrumentation is significantly
less costly and awkward to operate than multichannel.
 Three instrument packages were considered for evaluation and
laboratory demonstration:
(1) A Hewlett-Packard Model 4192A Low Frequency Impedance
Analyzer, controlled through a General Purpose Interface Bus (GPIB,
IEEE-488 standard) by an existing laboratory computer {IBM System-
9000 or PC A T ) ;
(2) A modular multifrequency eddy-current instrument with
phase-sensing capability, designed and built by ORNL, controlled
by GPIB with the laboratory computer; and
(3) A pulse-excited eddy-current instrument of initially
unspecified configuration.
RESULTS AND DISCUSSION
Functional System Design. Modeling, and Simulation
A modeling program initially written and used for flat plates was
modified to accommodate the design of a so-called "pancake" coil,
which is a low profile cylindrical coil with its axis normal to the
surface against which it is used. The probe design incorporates
these coils with their axes coincident with radii of the spherical
beads, placing the major plane tangent to the sphere and ap-
proximately parallel to a tangent of the inside tube wall.
Early results from the simulations (which included full instrument
response and data processing) indicated that a pulse instrument
would be superior to the multifrequency instrument for this
application. The multifrequency instrument and the pulsed
instrument both give similar responses and collect data that
contain similar information with the following exception. The
multiple frequency inspection was more affected by the permeability
variations observed in the weld region. In cases where the tube
wall material has a large permeability value, the electromagnetic
wave generated by the probe will not penetrate the material. Thus,
the outer surface of the material cannot be inspected. However,
with high power pulses the material can be saturated and inspected
in a manner similar to that of normal eddy-current tests. Although
the permeability of the weld is not great enough to shield the
outer part of the tube, it does cause a variation in the signal
which must be compensated for and which increases both the overall
noise and the difficulty of the measurement.
Coll Design
Coil design simulations were carried out for the pancake coil. The
properties of the test were varied over the ranges shown in
Table 1. Instrument readings were computed for a total of 500
different combinations of the properties. A least-squares fit
relating the readings to the properties yielded a set of linear
coefficients. The fitting coefficients are linear but can be
multiplied by nonlinear combinations of the readings. The ras
error in the measurement of a particular property for both the fit
and instrument drifts was calculated for the entire range of
 Table l Range of property
variation of the properties variations for the NASA heat
shown in Table 1. The exchanger problem-
instrument readings Were mm^^tmm^mm^t^t^m^i^mmtm^m^^^^^^m
computed at six different
frequencies (0.2, 0.5, 1, 2, 5, Defect depth 0 - 6 roils
and 10 MHz) , and a pattern of Mall thickness
nonlinear readings was fitted 13 - 17 mils
to the properties for each
combination of frequencies taken Resistivity 70 - 90
three at a time. The frequency micro-ohm cm
and reading combination that
gives the least error was Lift-off 0 - 8 mils
chosen. This error varies as
the coil size (mean radius) is , a ^ ^ H B a H M H | ^ i a ^ a H
varied. The results of the
error calculations for the pan-
cake coil are shown in Table 2.
It appears that the optimum mean Table 2 Defect measurement error
coil radius is 0.0325 in. and (in.) due to the property
that the minimum error remains variations.
0.0G06 in. between 0.030 and ••HBHBBiBBH^aiHnMiM
0.035 in. mean radius. Typical
inner and outer radii would be Coil Mean Defect Size
0.8 and 1.2 times this value. Radius Error
The variation in the material
will generally .. irait the 0.020 0.00088
detectability to a defect depth
equal to about 5 to 10% of the 0.025 0.00076
tube wall thickness. By
increasing the mean radius 0.030 0.00062
slightly, to 0.035 in., a larger
gauge wire can be used which 0.0325 0.00062
will further ease the fabrica-
tion problems and give better 0.035 0.00062
coverage with no sacrifice in
sensitivity. With such a coil 0.040 0.00092
size, an array of eight coils
spaced at 45 degree angular m^••^^•^•^•^••^•^^^^•B
offsets will ensure sufficient
coverage to prevent any circumferentially oriented defects from
escaping detection.
Although our discussion has centered on the detection of defects,
any of the other property variations, such as liftoff (and
therefore tube inner diameter), wall thickness, conductivity, and
permeability can also be determined. These properties can be
computed from the same readings that the defect sizes are, but
using a different set of coefficients. Other properties can be
discriminated against, such as the variations caused by tube
supports. The only requirement is that the property variations be
present in an adequate number in the calculated or experimental
readings. We can then determine these properties to the degree of
accuracy that we know their input data values.
 Instrument Selection
Initial experimental work was done with the HP Impedance Analyzer,
using a single coil mounted on a holder. The impedance analyzer
was stepped from frequency to frequency using signals from a
computer transmitted over the IEEE-488 bus. Although the
instrument is rather slow (making only about one reading per
second), it is adequate for the low-speed test design readings.
A bridge circuit with a differential amplifier was added so that
the difference between a reference coil in air and the test coil
on the sample could be measured.
Three of the weld samples from Rocketdyne were split so that the
liftoff could be varied. The samples were placed in a mechanical
positioner that was also controlled over the bus. The samples were
scanned at 0.005 in. intervals and the liftoff was varied from 0.0
to 0.008 in. in 0.002 in. increments. About 50 points were taken
in the free tube and in the weld reading from each of the six
samples. A "shaped" defect value was used for the 0.003 and 0.006
in. defects that were in two of the samples. The "shaped" value
was equal to the defect depth when the coil was directly over the
defect and decreased as the coil moved away from the defect. This
decrease matched the natural response of this type of coil to this
type of defect, and is similar to the defect shapes shown in
Figure 3.
Measurements were made at frequencies of 200 KHz, 500 KHz, 1 MHz,
2 MHz, 3 MHz, and 4 MHz. Due to limitations on the frequency
response of the bridge amplifier, we used 3 MHz and 4 MHz rather
than the computed frequencies of 5 MHz and 10 MHz. Fitting
coefficients were obtained for both three and six frequencies using
these readings. The best three-frequency fit showed an error of
0.0014 in., while the six-frequency error was 0.0013 in. It should
be recognized that, although the experimental readings did include
the permeability variation, they did not include nearly as many
property variations (particularly the defect variations) as the
computed readings. In particular, we should have had defect
standards with defects in the heat-affected zone and in the bare
tubing, as well as in the center of the weld.

Since only two weld defects were available, these were the only
ones used. The samples were difficult to align properly due to the
small size of both the coil and the samples. A small error in the
axial position of the defect could result in a very poor fit.
Several attempts were required before the samples were adequately
aligned and good fits were obtained. Figures 3 and 4 illustrate
the scans of the defects using a three-frequency fit.
These scans are made at five different liftoff values, and plotted
sequentiallyc Although these defects look rather good in the
figures, the noise was also high. Scans were also made using a
six-frequency fit, and in general, the six-frequency fit was better
and less noisy than the three-frequency fit. This suggests that
 A
. _ ../i

1
Figure 3 Scan of 3 mil deep defect for five lift-off values

4,

ft

•
Figure 4 Scan of 6 mil defect for five values of lift-off

the pulse instrument, with its greater frequency content, would

t^ive better data.
Neither fit works well for the entire 0.008 in. liftoff range, but
scans of the tube samples available showed that none of the tubes
would allow more than about 0.004 in. of liftoff. The liftoff of
the probe was difficult to set exactly due to the small size of
the coils and the irregularity of the samples. The liftoff was
quite likely greater than the value assumed for these measurements.
Additional fits were run for smaller liftoff ranges and the fits
improved as the liftoff range decreased.
The measurements were repeated using the three-frequency
instrument, as shown in Figure 5. This instrument consists of
three independent oscillators to generate the three frequencies.
These frequencies are mixed and then fed to a power amplifier and
through a dropping resistor to a coil. The coil signal, which is
modified by the interaction with the eddy currents in the
conductor, is then fed to three bandpass amplifiers that separate
the individual frequencies. The magnitude and phase of each
frequency is measured, digitized, and then fed to the controlling
computer. Measurements can be made with this instrument about 4 0
times faster than with the impedance analyzer, and the results,
although similar, are not quite as good. This is probably due to
a small sample misalignraent.

MAC 4
BANDPASS
AMP 1

OSCILLATOR t

PHASE
DETECTOR
1
MIXER OEfECTSIZE
PHASE!
OSCILLATOR 2 AND
POWER AMP
MAG 2
BANDPASS DEFECT
AMP Z LOCATION

OSCILLATOR 3 DEMODULATING
COMPUTER COIL TO TUBE
| TUBE
DISTANCE
PHASE
DETECTOR
TUBE TO SUPPOR
PHASE 2 DISTANCE

MAG 3
TUBE WALL
BANDPASS
AMP3 THICKNESS

PHASE
DETECTOR
3
PHASES

Figure 5 Block diagram of a three frequency instrument

Finally, measurements were made using the array coil. The fitting
on the six samples was repeated first using a single coil in the
array. The array probe was electrically connected as shown in
Figure 6. An unbalanced bridge circuit is formed between the
selected coil and the reference coil, and the multiplexer, on
signals from the three-frequency instrument, steps from one coil
to the next. A program in the microcomputer of the instrument
controls the multiplexer module.
 EX CONNE)CTDRS
CQNHEX

t*,

Figure 6 Electrical connections for the eight-coil multiplexed

array probe.

The fitting coefficients were calculated for only one coil in the
array, and an offset function was calculated from the value of each
of the coils in air. This was done because it is very difficult
to wind the coils exactly the same, particularly when the coils are
this small. The offset is a function of the coil inductance, and
is needed to electronically "balance" the bridge with the coil in
air. This correction was then applied to each of the coil readings
as the multiplexer stepped from coil to coil. This technique saves
the time and effort that would be required to run a complete
training set on each coil in the array, compute the fitting
coefficients, and store a separate set for each coil in the array.
A better match would be obtained if a gain correction were also
applied. However, to do this requires an additional standard which
was not available at the time.
Figure 7 illustrates the scan of an independent tube sample (not
one of the standards used to obtain the fitting coefficients).
Each trace is from a different coil in the array, and the lowermost
coil is the only one positioned to pass over the defects. The
first defect is 0.005-in. deep, and the second is 0.007-in. deep.
The rising signal at the end of the scan for the upper four coils
is from the tube end. The data from these coils are offset by
about 0.240 in. from the first four coils (see Figure 2 ) . In plots
made on the CRT this offset is corrected so that data from the
 coils at the sane position across the screen represent
the same axial location on the tube.
J
,
Although the readings from the multiple-frequency —-TW-wT

instrument were much faster than the impedance J

analyzer, they are still too slow to allow the weld to
be scanned and insure that there is no data skip due
to "probe pop." This occurs when the probe hangs on —-—~^-^~ A
the weld, the cable stretches and then the probe pulls
loose. The probe can then move at a rate of 0.5 m/s
for a short time, and the examination of a short length
of tubing will be skipped. Unfortunately, this is the
section where a defect is likely to occur. The maximum
data rate for the three-frequency instrument is about
40 readings per second, but about 32,000 readings per
second are needed to insure that there are no skips — •
when the tube is inspected at 0.5 m/s. The three- Figure 7
frequency instrument uses an 8080 based microcomputer Scan u? ing
with an IEEE-488 bus, which has been named the COMP9B an array
computer.5 Both the 8080 microcomputer and the IEEE- probe.
488 bus limit the inspection speed.
A block diagram of the pulsed instrument we developed is shown in
Figure 8.' This instrument generates a pulse that drives a coil.
The resulting electromagnetic wave is of sufficient amplitude to
saturate any ferromagnetic material near the probe and also
generates eddy currents in the sample. This signal can be detected
either by a sampling coil or directly using a dropping resistor in
series with the primary coil. The signal amplitude as a function
of time can be measured, either on the turn-on portion of the
signal or the turn-off portion of the signal when a single coil is
used. The value of this signal can be offset and amplified at each
time interval. This is necessary since the total pulse response
may be quite large and the important information may be a small
part of the signal. The amplitude of the signal is sampled using
a track-and-hold amplifier, digitized, and sent to a computer.
This amplitude can be used as a reading to determine the sample
properties in a manner similar to that employed for multiple-
frequency measurements.

Although the drawing, in the interest of simplicity, includes the

components for four channels of data, the instrument actually has
eight data channels. The instrument can be set up using signals
transmitted over the bus, so that no manual adjustments are
necessary. These adjustments include the pulse amplitude, on time
and off time, the gain and offset of each amplifier, and the offset
of bandpass filter amplifiers (not shown in the block diagram).
In addition the time at which each sample is made can be controlled
independently.

Since the instrument setup is controlled by a computer program, a

large reading pattern can be programmed and then the ones that best
fit the sample properties can be selected. In addition, if a new
set of properties is encountered as the tube is being scanned, the
 OBNL-DWG B5-131S3

400 V
1250 A

t, t2 t 3 t4

M PULSE TO SATURATING COIL AND DETECTOR RESPONSE

SAMPLE

COMPUTER PULSER

A-O A-O A-O A-O

CONVERTER CONVERTER CONVERTER CONVERTER

TRACK TRACK TRACK TRACK

AND AND AND AND
HOLD HOLD HOLD HOLD

(b) A PULSE AMPLITUDE INSTRUMENT

Figure 8 Block diagram of pulsed eddy-current instrument along with
a saturating pulse

pulse shape could be changed "on the f l y . " For example t h i s

feature would be needed if a ferromagnetic region were encountered.
 The pulse power could be increased to insure saturation, and
decreased again to allow probe cooling after the region was passed.
Pulsed Instrument Design
Preliminary design of a pulsed instrument that resides on a PC-AT
bus was completed. A block diagram for this instrument is shown
in Figure 9. This eddy-current instrument can transfer data at a
rate of 4 00,000 bytes per second, compared to a rate of 1000 bytes
per second for the 8080 based COMP9B computer. The pulsed
instrument could also pulse one coil at a time and read all the
coils in parallel, rather than driving all the coils in parallel
and multiplexing the output as the three-frequency instrument does.
In addition, less heat would be generated in the probe than would
result if all the coils were pulsed every time. The time required
for the switching transients to settle out would be eliminated.
The analog-to-digital convertors would have to be replaced with
faster models, but these are readily available.
The instrument is based on the design of the pulsed instrument
described in the previous paragraph, but has been modified to fit
on two standard PC-AT bus cards. Bus interface and control
circuits were designed using programmable array logic (PAL) chips.
These circuits are faster and more compact than comparable discrete
logic chips. Tests of the interface chips and some of the pulsed
circuit on the bus indicate that some further modification of the
circuits will be necessary before an optimum configuration is
achievable. However, a transfer of 16 bits of information at a
rate of 200 KHz was demonstrated. By using a PC-AT as the
controller for the pulsed instrument, we can significantly reduce
the cost while increasing access to a large amount of support
software. Although the present "test bed" utilizes a 16 bit 80286
microcomputer, a 32 bit 80386 based system should be faster and
cost very little more.
CONCLUSIONS
Computerized designs and simulations backed by laboratory tests and
demonstrations at ORNL have shown the feasibility of detection of
specified flaw properties in tubing welds in the SSME heat
exchangers. The flaw signals using a specially designed probe and
an existing ORNL multifrequency eddy-current instrument are robust
and field interpretable. The probe and flexible cabling can be
interfaced to existing drive systems at Rocketdyne to yield a
practical working instrument for inspecting tubing welds in the
field through existing tube ports with minimal inspection effort.
Additional simulations indicate that an even better inspection with
greater confidence could be achieved using pulsed eddy-current
techniques that have recently been developed at ORNL. This systera
offers higher speed, lower cost and simpler operation with a
computerized instrument now in the design stages.
 (M.SC Otf1

LHOOHU

9TOCTCMCB

•1O VOLTS

U X t CONTICM.
LOOiC

•ALAXI O*C

G A I N CMCOOC*

TAACK fr H X O

Figure 9 Block, diagram of pulsed eddy-current instrument interfaced to a PC-AT bus

 ACKNOWLEDGMENTS

The authors would like to acknowledge the assistance of J. F. Pate

from Hendrix College in performing the experimental measurements
for this report. They further acknowledge the help of J. D. Allen
and B. E. Foster in the review of this paper.

REFERENCES

1. C. V. Dodd, G. W. Scott, J. H. Smith, W. E. Deeds, L. D.

Chitwood, R. W. McClung, "Development of Multifrequency
Eddy-Current Techniques for In-Service Inspection of 2 1/4 Cr-1
Mo Steam Generator Tubing for CRBR," ORNL/TM-8920, February 1984.
(Applied Technology)

2. J. H. Smith, C. V. Dodd, and L. D. Chitwood, "Multifrequency

Eddy Current Inspection of Seam Weld in Steel Sheath,"
ORNL/TM-9470, April 1985.

3. C. V. Dodd, D. W. Koerner, W. E. Deeds, and C- A. Pickett

"Pulsed Eddy Current Inspection of Thin-Walled Stainless Steel
Tubing," ORNL-64 08, September 1987.

4. C. V. Dodd and W. E. Deeds, "Eddy-Current Inspection of

Ferromagnetic Tubing Using Pulsed Magnetic Saturation," ORJNL-63Q5,
July 1986.

5. C. V. Dodd and L. D. Chitwood, "Three-Frequency Eddy-Current

Instrument," ORNL/TM-10663, May 1988.