ML20148A812
| ML20148A812 | |
| Person / Time | |
|---|---|
| Issue date: | 03/15/1988 |
| From: | Advisory Committee on Reactor Safeguards |
| To: | |
| References | |
| ACRS-T-1651, NUDOCS 8803210171 | |
| Download: ML20148A812 (127) | |
Text
.
A C /2 S 7~- N S /
- T3 \\A-
.O
' UNITED STATES NUCLEAR REGULATORY COMMISSION In the Matter of:
ACRS SUBCOMMITTEE MEETING ON METAL COMPONENTS 4
\\O
\\
l rs.y, j
{l hf;0 N
'u u
.t (3 l)]
y l
en Pages:
1 through 124 gj j -
jj y
[a 7 j
Place:
Charlntte, NC Date:
March 15, 1988
=.......................w...................................
HERITAGE REPORTING CORPORATION opew Reporters O
1220 L Street, N.W., Suke 660 Washinston, D.C. 20005 (202) 628-4888 8803210171 -
1 PUBLIC NOTICE BY THE
)
2 UNITED STATES NUCLEAR REGULATORY COMMISSION'S 3
ADVISORY COHMITTEE ON REACTOR SAFEGUARDS 4
5
)
3 7
The contents of this stenographic transcript of the 8
proceedings of the United States Nuclear Regulatory 9
Commission's Advisory Committee on Reactor Safeguards (ACRS),
10 as reported herein, is an uncorrected record of the discussions 11 recorded at the meeting held on the aoove date.
12 No member of the ACRS Staff and no participant at 13 this meeting accepts any responsibility for errors or 14 inaccuracies of statement or data contained in this transcript.
15 16 17 18 19 20 21 22 23 24 25 dlh Heritage Reporting Corporation (202) 628-4888
Page 1 1
BEFORE THE 2
U.S. NUCLEAR REGULATORY COMMISSION.
3 4
ACRS SUBCOMMITTEE MEETING
)
ON METAL COMPONENTS-
)
5 6
Conference Room D 7
EPRI NDE Center 1300. Harris Boulevard s
Charlotte, North Carolina 9
. Tuesday, March 15, 1988 10 11 The. meeting convened, pursuant-to Notice, at 17 8:30 a.m.
13 la PRESENT WERE:
PAUL G.
SHEWMON, Chairman of the Subcommittee-15 CKARLES.J. WYLIE,. Member DAVID A. WARD, Member 16
..AL IGNE, Cognizant Staff Member 17 18 19 20 21 22 23 24 25
/
i l'
l^
Page 2 1
.I.N.D E X 2
PRESENTATIONS SY:
PAGE 3
Steven Doctor 4
4 Albert Curtis 53 5
Donald Adamonis
-58 6
Rick Rishel 69 7
Mohamad.Behravesh 88 8
Frank Ammirato 97 9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 1
A
i Page 3 1
P R O C.g g D 1 N G S 2
MR. DAUr Good morning, I'm Gary Dau from EPRI, i t ' ts 3
a pleasure to welcome the Subcommittee to our. facility here in
'a 4
Charlotte, and-I think. we have a g vad program laid out for you 5
people to review, following both the NRC work and the utility i
6 industry.
7 I'd like to go over a few administrative items e
before the. formal meeting starts.
Bob Stone has drawn a map 9
here showing where the restrooms are and. basically it's right to acound the corner and located in our cross. hallway here.
11 We have notified the receptionist this is to be an 12 open. meeting and anybody that shows up at the desk and asks 13 for the meeting will be directed here.
Luncheon arrangements u
have been made in our cafeteria down the hallway here and there will be a demonstration at the laboratory facility is 16
.taking place right after lunch.
17 We understand the Committee would like leave by 2:30.
13 this afternoon and we.have adjusted our schedule to that.
I think that's the only items I have and with that I'd like to 19
- c turn it over to Mr. Shewmon.
- i DR. SHENMOli
Goed morning, thank you.
This is a meeting of the ACRS Subcommittee on Metal l
n jComponents.
I'm Paul Shewmon, Chairman of the Subcommittee,
- 4 i and the other members in attendance are Dave Ward and Charlie 0
3 [ Wylie on my right.
Wu are here to review the sub-status of I
I 1
l 4
{
A
/
~
g'[
7 "\\[
~
4-tt V
h,
(-
\\
~
i
[
\\ }
Page 4 f-p r
i
\\
7'~)/
the NDE tests of steel stainless -- cast stalnless steel
_j 1
e 9
P ping and other topics.
i 2
3 Al Igne on ny left.is the cognizant ACRS staff 4
member.
s 5
Since we are a government body, the rules for N
6 participation in today's meeting have bean announced as part s
7 of the notice of the meeting that was published ln the Federal n
is 8
Register of February 29,
'88.
,The meeting is being conducted 9
in accordance with provisions of the Federal Advisory io Committee Act and/the. Government and Sunshine Act.
He do have
?
it a reporter and $1 you would speak loud enough so that he can
\\
12 hear you, we'd appreciate it.
Do you have any comments at; this time?
,13 14 (No response.)
DR. SHEWMOH:
Fine, then we'll proceed and I gues;,
15 16 it.says here Steve Doctor is first on the agenda.
17 PRESENTATION BY STEVEN DOCTOR 13 DR. DOCTOR:
Good morning everyone.
I'm going to-19 give an update on the status of work going on at PNL that is c
funded by the U.S.
Nuclear Regulatory Commission.
I have 23 given a presentation on earlier progress at a previous 22 Subcommittee meeting back in June of
'86.
23 The presentation outline that I will be following 2
24 will deal with these four topics.
25 I will first start out and review under the PISC III I
1
Page 5 m(g) 1 Progras, the.centrifuga11y cast stainless steel round robin 2
test that was conducted, more post studies that we have-collected during that j
conducted on some of.the-data - that 4
round robin exercise,.some additional studies that we're doing 5
looking at fundamental transmission properties through casting 6
of the steel materials, and then.of course the final item is 7
to talk briefly about the PISC III Program that is currently a
in the plannlag stages right now and will be starting ectual 9
testing this coming fall.
io As I indicated, the first thing I'm going to discuss 11 is the centrifuga11y cast stainless steel round robir test as 12 brought our under a NUREG CR document number 4970, and also 13
-published as PISC III Report-Number 3.
This was designed as a
\\'
screening -phase with the more detailed study of reliability of ia the material to be conducted as a part of the last topic.under 15 g
my agerda.
Now this.ini lal screening phase was started, we had i7 q
15 speciscas that-currently existed at the time, several years ago when this was planned.
They were approximately 400 jg l
millimeters long by 190 millimeters wide by 60 millimeters 20 j
pi thick.
Thure was a weld in the center of each specimen, the l
77 specimen contained either equiaxed or a coluanar l
'.ntained l
23 microstructure.
There were il specimens that 24 thermal fatigue cracks and this four specimens that contained no internal cracks.
The crowns were ground but they were not 23 09b qv r
+-e v
=
e--
w-aw+,
e,y
-we-w,-r,--at
w,vy--
e-ww-----%
em--a--------m--,*.--<a-w----e t-MT-7TN 1
m Page 6 i
ground perfectly as I will show you.in this next view.
2 There was a pinte that was-placed on.the botton of 3
each of the specimens so that.no one would.know what was 4
underneath-there, whether there were cracks in there.or not.
5 Jus you can see up here, this is the crown region and it had 6
been ground so that it was.smocth but there were still-some 7
valleys left between weld. passes.
So.it was not a perfect 8
ideal surface.
9 The procedures that were followed; there were 18 io teams that participated,.they used a. variety of procedures 11 consisting.of manual.UT, they used automated UT, automated,UT 12 with some signal processing and then there was one we called 33
,non-UT or radiographic. technique that was employed.
34 The most common technique. employed a dual probe using a 1 ngitudinal mode at one megahertz.
Each team 15 16 basically spent.one week performing their inspections and then 17 they spent some time reducing the data, putting it onto report is forms and turning it in for-then the analysis.
You can summarize the various teams in this in go particular table-from the report.
Wu have used a-notation J
21 here for M being for manual, A for automated, asp for rutomated with-signal.processlng,and then of course the non-UT 72 I
23
. located here.
In addition, you'll see over here that people.
24 uend a.varjety.of different decision criteria and sensitivity 25 that they employed during the test.
Some people worked at
l l
I Page 7
/ )
1 what they called the noise level, in otbor words they would 2
determine what the noise level was ano they would set that am 3
the particular height on their scope; other. people used, for 4
example, this 50% DAC right here; other people had a 20% EAC, 5
other people used what we callsd data shape.
They looked at 6
the nature of the noise in terms of how 5t was displayed in an 7
image and they looked at the signals-from the defects and they a
tried to use-the data shape as their discriminate for 9
determining-whethee or not what they were looking et was 10 purely a grain phenomenon or whether or not it was a 11 phenomenon related to an actual defect.
17 This ls a schematic -- I don't know if you can.see that -- of a handout that was passed around.
It's shown in 33
(
t 14 two different dashed.and dotted linen, I tried to use a l
15 different approach here in which the intended defects are a
shown in red, the box shows the zone around that that was used i
17 fer scoring the results.
If you -cunt these up, you'll find i
is that 11 cracks ex3sted in these and I might point-out-that l
n number 1 and number 14 basically have cracks that.went almost l
i
- c completely across the entire defect -- or excuse me, across the entire specimen.
72 Now in terms of grading what we called the false 23 call or where you call defects that result from the grain or 2a metallurgical type of scattering, we've put in a total of 14 73 boxes that are shown here and 'his is how they are
i Page 8
/
i distributed.
Although-there are four of the specimens that 2
contain no cracks, others contain cracks that for example were 3
clear over here to this side, so there was adequate room to a
put a grading unit in there to assess false call performance 5
on that same specimen.
6 DR. SHEWHON:
Does the box represent the areas they 7
were supposed to look at, or what?
g DR. DOCTOR:
They were to look at the entire thing.
9
'that was our grading unit.
In other words, we used these to grading units to determine if they made any crack call that 11 intersected with this box, we would classify that they made a 12 false call in that box.
If they made a false call in all 14 of these grading units, then they would have gotten a 100%
o~
i rT
(/
score on false call performance.
14 i3 Now before-we get into the results, what I wanted to 16 do is to report briefly on the destructive results.
The 17 destructive evaluation was actually conducted by ISPRA, they is used the same procedures that were employed in the PISC I and l
l 1;
PISC II exercises.
Three specimens were destructed: one wan c
Specimen Number 12 which was a 'olank and they verified by l
i p
l cross-sectioning this and a numbec of repeated exam 3 nations, l!
,y, there was no defects ' hat were contained in that specimen.
iij3pecimenNumber1wasdestructed, as I indicated, it was a 74 n deu4 :t that went the majacity of the distance across that F
I 8peciman.
The maxi.r.um depth on it was 384 through wall.
25 h
lllI n
i E
l Page 9
{j's
)
Specimen Number 5 was selected because there was a high nwaber 2
of falso calls that also existed in.the grading unit that was 3
on that same specimen, and they wanted to destruct this so a
they could look at the actual cracks and ascertain its 5
through-wall extent, but also to evaluate the zone that was non-cracks to try and understand if there were any unintended 6
7 defects that might be causing that.
This is a micrograph for Specimen Number 1.
In this 3
particular Case wo had.an equiaxed structure on both sides of 9
io the weld.
You can see the weld passes in the middle zone.
In 11 i Specimen Number 12, this was the blank unit that had a i
17 columnar microstructure over here on the lefthand side and equiaxed microstructure over here on the righthand side.
And 33 u
then Specimen Number 5, as shown here, again it had a columnar microstructure on the left and equiaxed microstructure over on 15 the righthand sidu.
je u
What I would like to do now is to step through how 33 they actually did a destructive evaluation on this with the
! next few viewgraphs.
What this is a plot of right here ic 39
- c l Specimen Number 5, and this is a plot of all the ultrasonic calls that were made by the 18 teams in this particular y
4
- ll specimens.
The point here is that the calls are being made h
- !! pretty much unif orstly all across the specimen, there is no l
2.
great clustering in any one location.
'they wan ted to look at y
this to determine how best to cross-se.ction it.
H I
1 l
l 1
1
_J
I Page 10 ym
)
This shows -- now if I can keep it consistent with i
,g 2
the preceding one, this is the weld center line running 3
through here,.the intended defect.is shown here.
There were 4
two -- well x-ray indications that were found.
This is one 5
shown there, a second one was shown over here.
The second one 6
was a false en.) made by the radiographic approach.
It 7
actually predicted one as being located over here but they a
end%d up actually classifying it as not being a defect because 9
of its location relative to the weld center line because it 10 was outside -- well outside the weld zone region, but it 11 clearly came out with a crack like type property and it was 12 put into the matrix in terms of calling something as being l
13 there, and then you make a decision as to whether it's a crsck l
t',
)
l
(~ >
or not a crack.
This was called as being there but was given.
u l
15 a non-crack type call.
g This is a cross-sectional cutting then that was o
performed in which they basically sliced it into three pieces.
I They took that center piece then and did some ultrasonic scans a
l l
n on that, some radiographic examination.
They then f urther t
sectioned it as shown here by these multiple slices,. they did i
t 71 cross-sectional profiles and split things apart.
I
- - )
Looking at the specimen under the first cut, this is n
lthe kind and nature of scan that was generated using an d
- 2. p ultrasonic normal beam across the specimen from the two l
75 ll dif ferent sides.
You can see the presence of the crack h
1 ll o
i
bJ 4
m.
Page 11 located here.and here,.so-you can see where you're getting 2
sound transmission.and where you're getting back scattering of 3
the sound.
So this is the. location of the intended. defect.
4 You can see that in this particular casa..there is m. fair 5
amount of sound. transmission that,.you know, is occurring so 6
it obviously has a fairly tight crack.
7
.DR.
SHEWMON:
More on the right end, is that a
something-you're going.to ignore.or is.that --
9 DR.. DOCTOR:
You mean-these --
10 DR. SHEWMON:
Other.end -
yeah.
11 DR. DOCTOh:
I'm not.sure.
That's right.ruuur the 12 edge of the specimen and that very well may be due to soie kind of scattering sound path around there.
-I'm not sure why 13
(
34 those are occurring out there quite frankly.
They shouldn't.
be there.
Ideally you should see the outline of this and you 15 should see nothing here and nothing up here and of course you 16 i7 should just see this where the defect is located.
I'm nct 13 aare what the significance of those are.
Let's. move on to the results.
The two performance 19 20 metrics that were selected are the false call probability, 21 that is the probability that one of these blanks grading units 22 is classified as being cracked, there are a total of 14 that l
23 were in this particular study.
The probability of detection 2
and correct interpretation is the probability that u cracked 1
75 grading unit will be classified as being cracked.
So the way l
l w,,,
,,,v,_,_,_,_,,,,,,,,,_~_,,.m_,_.,.., _ _ _ _ _ _. _,,.., _.
Page 12 M_j/
.i that we-actually come up with the estimates then is to, in the 2
case of the.falso call probability,.take the number that they 3
classJfied as being cracked, divide that by the total number 4
which in this came is 14, and then multiply it by 100.
And 5
correspondingly the same thing on the PODCI.
Best performance 6
will occur when the false call probability is near zero and 7
the correct detection and interpretation is near one.
8 Now some of the things that you do before you jump 9
in and analyze the results in great depth in to do some jo studies such me this in which we've plotted-here for, in this 11 particular case, probability of either detection or the false l
i7 call probability, shown here by the two lines.
In this i
13 particular case the solid line is the detection line, the dashed line is the false call probability.
And we've done a
15 this as a function of our grading unit. tolerance.
In other l
16 words. in this particular case it's the axial width-of the 1
(
17 grading unit and in this particular case we're going 18 circumferential tolerances of zero, five, ten and fifteen l
l millimeters.
In other words, this is the amount of extension in 2c that we made for the box as a circumferential direction.
And what we wanted to see was whether or not if we selected a 21 22 particular grading unit, whether or not we're at m. point whera l
n it really didn't make any difference if 'a increased the size t
24 of it because we pretty much had all.ttw correct calls.
And 73 that's what you see here, when you're dealing with extremely l
l i
1 1
l L.-
1 Page 13 i
small values of axial tolerance down here of around five 2
millimeters, you can see that there's a fairly steep-rise to 3
the curve, but once you get out.here.to the nature of ten to a
fifteen, the curve has pretty-much flattened off whether 5
you're talking-about falso call probability or you're talking 6
about the probability of. detection of the actual defects.
And 7
you can see that there is no-real strong relationship with a
regard to the.circumferential tolerance, which.is what one 9
would expect because.the tall ends of the crack are extremely 10 hard to see and to define, so we're fairly insensitive to 11 that.
12 What we ended up using on this was a fifteen elllimeter tolerance.for the axial and I think ten millimeter 13 14 tolerance -- yeah, ten millimeter for the circumferential.
So this is the box that was eclected.
But you can see that 3
there's not a strong dependence, which is extremely important, g
n
.for the grading unit that you're using.
l 13 Dit. SHEWMON:
Would you tell me.what adds up to 100 in that?
Apparently false call and correct calls don't add up i9
- o to 100.
7j DR. DOCTOR:
That's correct, it's whether or not 73 people made a decision and put somethJng in the box.
If 23 nobody put any decisions into a cracked unit box, then 2a everything would be at zero and if they called all the blank units cracked, then that curve would correspondingly be up at 73
l Page 14 1
7)/
i 100.
If you get all the cracked ones correct, then the POD
_g 2
would be at 100 and if you made no mistakes, the false call 3
would be at zero, a
What this shows in this particular case is that, 5
let's say roughly 60% of the calls for detection were in 6
cracked grading units and correspondingly roughly about 40% of 7
the calls were in blank grading units.
These were the calls 1
g that there were cracks located and it doesn't. sum to zero --
9 or 100, it 's whatever the calls are.
l l
10 DR. SHEWMON:
Okay.
l l
11 MR. WARD:
Steve, is there any way to express what 1
12 this random calling would be?
l DR. DOCTOR:
I'll show you that a little bit later, 13
/
\\-
but this curve is not designed to look at that particular 14 15 phenomenon.
What we're trying to do here is establish what is l
a a reasonable size of grading unit that we should use with the u
data to tell us most accurately what is actually happening.
l l
19 And you can see if you select a five millimeter grading unit, l
l n
you would be biasing things down fairly low.
What that means l
1s that changes in velocity and that that occurred in the e
l 71 material create a fair amount of uncertainty, particularly in l the axial direction.
With regard to the circumferential 72 I
' direction, if you detected something, you may miss the ends of 73 24 it so you're fairly insensitive to that parameter.
25 ll One of the things we did look at was also the dB d
s o
ll
Page 15 TT) i response for the teams that reported things, with regard to a 2
dB amplitude.
We looked at the number of correct 3
classifications for the defects.
What we've plotted here 4
then is thu average in the range of the dB response as a s
function of the number of correct classifications.
And one 6
extreme we had out here, we had better than 90% of the people 7
actually detecting this particular crack, and as you can see, e
the dB response average value is not a whole lot different I
9 from these others, plus the range of variability that we found i
l io with regard to that is extremely large.
11 What we were trying to see here was whether or not 32 dB response was a prime requirement that they were using.for 13 making their determination.
You can see there is a very u
slight, minor trend that.you might classify except for the l
l 33
- occurrence down here of these two points, but with the large l
error bars associated with.the scatter of the data, there is l
g 37 no strong trend at all that the amplitude was actually being used as the prime discriminate for making that classification.
ig Another thing that was looked at was the-number of i;
calls that were made in blank grading units, and this is a
- c b summary with regard to each of the specimens.
Of course 1 and y
0 n
14 are not in there, those were not included because they did not contain any blank grading units.
You can see from this h
a y that by far thu majority of the crack calls in blank. material Fdoccurred at twice as high a frequency as they occurred in 3
t lll!I b
~
I_
(
l i
Page 16 i
equiaxed material.
,So what this says is that the signals that 2
were most difficult to discriminate against occurred more.
3 frequently in columnar material than they did.in equiaxed a
material.
5 Now we can plot the results as a function of~various k
6 specimens.
I've put in the har.dout a complete series of all 7
of the results.
In terms of this presentation, I'm only going-8 to go through about four of them, just to illustrate 9
particular points because I just don't feel it's worthy to to spend all-the tims devoting an analysis to all 15 at this 11 particular time.
12 What we've plotted here is the number of. crack calls that occurred as a function of circumferential position.
33 These large vertically dashed lines then illustrate where the ja crack.wa', actually located, so this was Specimen Number 1 15 where the crack went the majority of the distance across the 16 17 specimen.
And then we have decisions as m. function of whether is you were looking at the near side, whether you were inspecting 19 from the far side or the solid line is where you were 20 integrating the information from inspection from.both sides, p
I'll -just put this up to show you that the peak 77 value here goes up to close to 70% right in this one location, 23 drops down to about 60.
Over the majority of the crack, 24 there's probably an average value that's in the neighborhood 25 of about 40% of the people classified this zone as being
(
i
~
Page 17 e
_j 1
cracked.
2 Another example, this was Number 5 that was 3
destructed.
This one. chows no raal.particular overall general a
trend.
The classification here in the zone where the crack 5
was located is very similar to this zone out here where the 6
crack was not located.
When we destructed this, of course we-7 found a defect that was.in this location, that was 28% through 8
wall and no. unintended defects out in this zone.
9 This is Specimen Number 12, this is the blank w
grading unit.-- or blank specimen that had two grading units 11 that were blank contained in it.
You can see here that again 12 we're getting in the neighborhood of 40 to 50% of the people who looked at this particular specimen classifying it as being 13
\\#
cracked and it was verified by destructive evaluation, that it 34 j3 did not contain any cracks.
16 I've put up one final one then.
This happens to be 17 Specimen Number 11.
It was the one that the people scored ig about 93% on that you saw in that dB response plot that we 39 were showing earlier as a function of the number of people who 7e made the classification.
This happened to be a unique.
- i specimen, it had been subjected to special treatment in which lit had been both thermally relieved and then mechanically l
lstressrelieved, the crack was bent open and then the specimen n
I 24 was bent back to its original shape.
When people inspected I
75 g this, and I'll show you some results later on, the actual dB U
l
Page 18 i
response on this particular specimen you'11 find was slightly g g/
2 higher, but out in this zonu here you'll find that the actual 3
signals that they had to discriminate against were very a
typical of what they had to discriminate against in other 5
specimens.
6 This I think shows the psychology of taking a test 7
in which.this one people really found that they had very high e
competence in it and that they were able then easily Lv l
l 9
discriminate against signals that in other specimens they I
to ended up classifying as being actually-cracks.
And that's the ti reason you see this extremely low response in this zone 17 here.
Well this shows I think that the properties of the g
Ok' crack are extremely important and now what I'd like to do is u
33 nove on, and before I go into the final results, I just wanted 16 to put up an example of.one team's results to show you some of 17 the things that you have to contend with, a
What you see here is a whole series of a
classifications that a team made in terms of cracked and non-c, cracked decisions.
As you can see, there's a large number of p
)
lines that go across the entire specimen, such as this one and h
j
- ,U this one.
In this particular case down here, they classified h
i l
r p one of these as being cracked, down here they classified one V
j n [as being cracked on one side and the other being cracked on f
the other side and want the full length of the specimun.
e l
I l
l l
l l
4
I Page 19
)
i Now let me put this up to show you that because of 2
results like-this, i t's extremely difficult to extract and
- 3 understand exactly what's been going on--in terms of the-round a
robin test.
Other people, you know, in terms of their 5
Performance were.able.to classify things with.a far smaller 6
number of crack type cal 10 that coincided-with-the intended 7
defects and did not have -- when you have something that goes a
all the way across and then intersects one of your grading.
9 units and.you end.up classifying it as being cracknd, and it to also intersects one,of your. blank grading units, you end up.
11 actually nullifying it. Obviously the person was not seeing 12 the defects, so one has.to look at both falso call and the i3 detection probability-and look.at those two numbers in 14 conjunction with one another.
Now as I indicated earlier, this was a screen type 15 test and.we used a small number of specimens.
This is put up 16 4
17 to show the error bars associated with the detection or is behavior in cracked material versus the behavior in blank material; falso call probability, probability of detection and n
7e correct interpretation.
As you can see, those error bars are 21
-vsry large.
77 Now when we look at a plot such as this, you can see 33 that if I put those error bars on, for example, performance 24 right here, they would extend quite a distance in that 75 direction and in that direction, so that in essence I would be O
S
--"e,--se
,, - - -w,
,,,-----.--ye,-
,-..-.,,mm,--,e-,-%,,-,,e p-,----,---,.w-
,--,mwe
,--w.,-,-,,w-w-www n--
,,,,,-,-----,---,,,-x,,e-,,--
---m
Page 20
/
.. unable to discriminate most of these calls in this zone from
(
i l
2 one another.
I would be able to distinguish those from-these l
3 up here, however.
l 4
What we were intending to do with this was-to try 5
and find out if.there were teams that were able to perform in 6
this upper lefthand corner.
What is in that upper,lefthand 7
corner, as you note, is a square.
That.was the results from 3
the optimized radiographic exam that was performed in a 9
laboratory using a linear. accelerator under idealized 10 conditions where they had removed the vacuum plate, put-the it film in. contact with it, so those were extremely idealized 12 conditions, the performance-was.there.
When that same radiographic technique was applied 13 14 with the.. vacuum piste in place, that performance dropped down to here.
Okay?
So if you would try and take that to the 15 field where you have double wall type inspection, you have 16 17 water, you couldn't get the two meter distance that they were la using, yoa couldn't use the linear accelerator that they were ig using, you would expect this performance to probably 20 deteriorate substantially from where it is.
The rest of these dots then are all from the results 71 22 generated by the ultramonic inspections.
What's shc,wn here by 23 this solid line is a line that would be obtained if one was 74 using a random decisionmaking type process, so that you would.
75 make an equal number of correct and wrong decisions,. things u
Page 21 M_f i
would fall along this line.
2 So if you look at this, the idealized performance is 3
as far away from this.line as you can possibly get.
You can 4
see that these techniques that are located up in this zone era 5
the fartherest way, excluding of course this laboratory type 6
test.
One can see from this that clearly it appears that 7
thers are doing better than random chance even taking into e
account the large error bars that we had in this -- they are 9
doing better than random chance.
10 One looks at these up here and one recognizes that i
11 clearly in some of these cases at least you have to I think l
12 lbelieve that people were seeing the defects.
And if you go i
l l
back and look at the original data, if they-weren't calling i3
(', )
(-
defects all the way across but were simply finding defects and ja i3 classifying some of them as being associated with blank g
grading units and some with cracked units, what onu could 17 conclude I think is that one is seeing the defects but one is cannot discriminate against the non-defect type scattering.
n Now that doesn't solve your problem, but it says at 20 least that in this particular case with these thermal fatigue
- i cracks, there were a number of people that were seeing basically all of the thermal fatigue cracks.
And the thurmal 1
I fatigue cracks are tight and difficult to see and I think that 24 l!
shows promise from the point of view of being able to see i
25 'i those.
It doesn't help us for being abJe to discriminate at I
n l
I
Page 22
?
i this particular juncture in time, but that focuses in.on what x
2 the prob 1cm is, the one-we have to solve.
3 If You rank all the performances in a table such as 4
-this, what we've done.is broken.. things out -- some people 3
looked at. things from one side, from the other side and then 6
from both, and that's the reason you see.more.than.18. teams 7
11sted here.-.
What we've.done then is.to show the FCP and a
PODCI und ask the question of what is the probability that 9
this kind of performance could have been.obtained purely by l
io chance.
And what we've done then is to rank these by.that j i-
' probability, of course-rvcognizing the-uncertainty that I 12 indicated that we had earlier with regard to the. actual-l quantification of performance.
We-have large bars, so you 13 u
can't really.say that this team is.better than that team, you have too much uncertainty.
You can probably discriminate from
(
i 15 16 here half way down the list or so, but you certainly can't try 17 and do it on any finer scale.
is What we tried to do here was to look and see if n
there are any generic all around trends with regard to what l
l 71 people were doing with regard to, you know, achieving the best performance and if you look through this, you'll see that
??
there is basically an inter-mixing of ASME-levels, data shape, 73 et catera, that were actually used.
There is no clear overall 24 trend.
7s When you ask some of these teams, for example up l
l h
Page 23 1
.here that were using the ASME 50%, how did you actually make 2
a decision, they said well we just got a feeling that this was 3
a crack and this wasn't but they couldn't put their finger on 4
actually a requirement.that they had or decision process that--
5 they had for makinc.that. determination.
It was kind of a 6
feeling that they had more than anything else that.they.could-7 put their finger.on.
So even though this ASME 50% is up.here, 8
you can't transfer that to anybody else.
It was.something 9
that that person had in terms-of.an. understanding-of the i
io responses, but they really didn't understand it themselves.
i 11-Therefore it had to be considered very unreliable.
l 17 DR. SHEWMON:
.You don' t tunre frequency or wave length on there except there was some spread.in what they do, l
i3 O
14 is that normal?
(
15 DR. DOCTOR:
There was some spread but it was not l
a very much.
People did not go up much higher than one l
17 megahertz and some people went down to 500 kilohertz..
l is DR. SKEWMON:
But going down to 500 did not help?
{
ig DR. DOCTOR:
Did not help much, that's correct.
The j
70 results that we.did with SAFT were at 500 kilohertz in the shear mode and I'll talk more about those in terms of cur 73 l
22 Post-analysis in a few moments.
I'd like to talk about our conclusions.
First off, 73 l
24 I want to point out there are a few cautions. This was purely test, it contained only thermal fatJgue cracks, it 75lnscreening O
1 I
l t
Page 24 m(_j) i only-dealt with two types of.microstructures, an equiaxed 2
microstructure and a columnar microstructure.
We also 3
quantified the performance by the two parameters PODCI and the a
false call. probability.
And the point here is that both 5
Parameters must be used in. terms of describing the I
6 Performance.
You can't use just a single one.
This was a l
7 screening test and thus we had Imrge confidence limits with i
1 8
regard to those two parameters.
9 Do not-use this data for comparing the effectiveness to of different procedures, it-was not designed as such...it.was it simply trying to see whether or not techniques existed that 12 would provide the kind of detection performanew that we were trying to.look and find so.that we could have the problea 13
\\'
solved.
But in conclusinn, there are some global trends that u
do exist and what I want to do is point out.what I think are 15 a
the relevant things that can be concluded from.this.
17 In general, the inspection performance was rather la poor.
Some operator / technique combinations did show potential 19 for detecting and classifying the material, however other 7c operators using the same procedures did rather poorly.
HIgh false call rates made it difficult to 73 77 ambiguously analyze results.
As I showed you in that one 73 plot, some people called cracks all the way across.
There was 24 such a preponderance of calls in a number of cases that it was 25 extremely difficult to really understand what that teau's O
Page 25 M
]
i performance really was.
2 Specimen Number 11 that I showed you>had much higher 3
POD and lower false call probability.
We believe this is a
related to the relaxation of crack tightness because of the 5
thermal. stress relieving and the mechanical stress relieving 6
that was performed on that particular specimen.
I think it 7
points out the-properties of the defect are extremely 8
Important insofar as detection is concerned.-
9 In general, the false call probability was higher in 10 the columnar than in the equiaxed material, 40 versus 23 11 percent numbers that I showed you.
12 Let's see, the single-sided versus both-side i3 inspection effectiveness could not be resolved due to the high-a false call probability in columnar material.
This was a result of, as I showed you, the fact that we had in most i3 specimens an equiaxed material on one side and columnar on the 16 other and because of the.high false call. numbers that occurred 37 on the columnar side, you couldn't really look at the single 18 side access versus both side to effectively understand that.
39 Now the two radiographic laboratory inspection 70 gi techniques that had been applied did show very good results, 27 but it should be pointed out that these are not adaptable for 23 field inspection.
24 DR. SHEWMON:
This was the PISC group, so this was an international group done under semi-lab basis or fully lab 25 O
Page 26 1
basis --
2 DR. DOCTOR:
Fully lab basis.
3 DR. SHEWMON:
-.and presumably not the third team a
in any given country.
5 DR. DOCTOR:
That's correct..they were the best team 6
in all countries, yes.
7 DR. SHEWMON:
There are -- it seems to me there is 8
an NRC regulation about -
I see reports that people go out to 9
a plant and Byron was one, they couldn't get a signal back out jo of it.
So there is variability in what is in the field.
Is 11 this primarily grain size or more columnar, or do you know?
12 DR.- DOCTOR:
Well I'll show you some of the work that we have done in regard.to different specimens.and we're 13 u
in the process of trying to understand what actually happens in these various microstructures because when you propagate 15 16 through these different microstructures, differwnt things u
happen.
And if.you've got an intermix of different is
.microstructures, it's.like, you know, stacking up different 19 filters and one filters out and tries to pass a band.of
- o frequencies and another one. knocks out another band and if therra two don't overlap, you end up reducing the. signal very 27 greatly.
I'll show you some of the results a little bit later 33 on that, in terms of our laboratory work.
74 DR. SHEWHON:
Any questions?
75 (No response.)
O I
Page 27 1
DR. DOCTOR:
Okay, what I'd like to do then is talk 2
about the post-CCSSRRT study.that we did.
This was conducted 3
at PNL and it involved the data beme that we collected uaing 4
the SAFT system.
We digitized all the A-scans so we have that.
5 data base on storage.
6 We used a 500 kilohertz shear-wave transducer and 4
7 the reason that was selected was for several reasons; Dave 8
Cupperman at Argonne felt that was one of the best ways to 9
perform an inspection because using the lower wave length, the io shear mode should scatter and-give you a stronger corner trap 11 response and so we did some preliminary tests.and said yeah 12 that looks good, so we used it for all of the SAFT work, which 13 we got similar results to other people in terms of the u
discrimination process.
When I go through the spectrum-to 15 show you-this I think you'll understand why.
g We complied data from defect zones and from defect-i7 free zones.
What I-mean by that is in the grading unit we is went into the zone.of the grading unit where the defect occurred and we excluded the end parts of the defect to 39 go eliminate extremely weak signals from that zone.
So we dealt i
with.the center part of the cracks.
So if a crack was three 2i 22 inches long, we might take the center two inches of it for 73 example.
And then the defect-free zones were basically all i
the blank grading units.
2a 25 We performed FFTs on all the A-scans and we summed
Page 28 l
i the results together and sorted into these four classes.
We 1
1 2
looked at equiaxed, the columr.ar and both the high probability l
l l
3 detections and the low probability detections.
What I mean by 1
I 4
that is if you go back to those curves that I showed you where 5
people made classifications, we separated them out.
If at l
6 least half the people made a classification there and that was 7
twice the number of classifications that occurred in the l
defect-free zone, we considered that a high probability one, l
g 9
.if not we put it in the low probability case, And what I'd like to do is step through first off to io it show you a composite.
This is the response from a sawcut in j
12 rough stainless steel.
It's only put up here as an example so that as I step through this, you'll understand what I'm 33 ja talking about.
We plotted relative magnitude over here as a function of frequency and in this particular case two and a 15 quarter megahertz transmission was used and the response and a
17 the defect are shown here.
The non-defect response, in other is words going to a zone adjacent to it where the defect was not i;
located, you collect basically grain noise, and that's shown down here.
When we subtract the two, the diffurence is the il curve that swings through here and what you like to do is have pr i
blthat difference bet extremely high, but you would also like to 77
,-dhave it bit peaked here where the center frequency of the h
1 transducer is located.
74 l
l How if we take all the results from the spectrum for 75 g 1
e
Page 29 ym 3_;)
i all the specimens and we separate out into the defect and non-2 defect cases, what we have plotted is.the defect and non-3 defect cases right here, the difference is down here.
In this a
particular case, we're usjng m.500 kilohertz. transducer and 5
you can see the difference shows somewhat.of a-minor peak 6
occurring right in-here, but you can see that that's not much 7
different and barely above the noise level, if you will.
So a
from an overall composite standpoint, there's not much 9
difference being shown here.
tu So then what we did is we went in and we looked at 11 the results.
In this particular case, this is foc all the 12 columnar ones.
We looked at what the defect response was.
You can see that one here being slightly higher than the non-33 ja defect.
When you look at the difference you can sea a 15 predominant peak occurring here somewhat higher than what we saw when we collapsed all the columnar and the equiaxed 16 37 microstructure information.
ig So then what we did is we said okay, there's a slight trend there, how does this break out when we look at ig the ones where people did a good job versus ones where it 70 wasn't clear that they were detecting them.
When we looked at 21 the difficult ones, this is what we found.
We found that in 77 73 general, you still have that peak there but it hasn't been 23 enhanced at all, it's at about the same level as what. we saw 25 when we collapsed all-the information.
If we contrast that
Page 30
)
i with the case where there was a high detection probability 2
what you see here is an extremely large respont., fer the 3
defect versus the non-def ect call, and this down -here is the 4
difference.
You can see tNat extends up there extremely high, 5
much higher than this base line noise.
6 So from this we conclude that at least in the 7
columnar material where people had a high probability of a
calling something, in this particular case using our 500 9
kilohertz transducer it appears to us that they're getting a io much larger response off of those flaws than what they were 13 off of these flaws.
12 DR. SHEWMON:
Why is it such a iagged curve?
One 33 could almost say if you used eight-tenth negacycles you'd do
(
u as well.
It seems to be very periodic.
Is it noise or is it real?
15 DR. DOCTOR:
Yeah, our band width of our transducer 16 i7 probably extends out here, it was like a 60 or so percent band la width.
I would say from about 300 kilohertz out to around 700 l
kilohertz is the range.
Once you get beyond that, you're 19
- 0 looking simply at a noise phenomenon that's related to looking 7;
at two basically small amplitude signals.
You're amp 11 sizing 22 extremely low noise levels, low information content.
1 73 So, you know, based on this,-the results from the 74 columnar material appear to be related to the response that 25 one gets with regard to.the amplitude, it's very evident lllI l
i l
l l
Page 31
)
there's a difference there.
i 2
Now we looked at the. composite results from all the l
3 equiaxed grained material, and again we see somewhat of a 4
similar thing, however you'll notice here that we've got a 5
quite high peak located down here at about 400 kilohertz.
And 6
we still have a peak occurring right here at about 500 7
kilohertz.but this one at 400 is about twice the size of that.
8 If one were using the 500 kilohertz informational, one would 9
draw the conclusion that it's lower than thin 400 kilohertz to one, and it's obviously not the -information - to be using.
So ti again we repeated the same type of analysis, 12 We looked at the compoulte opectrum of the difficult cracks and in this particular case what's interesting is that i3 k-there's actually an inversion almost at 400 kilohertz for y
those.
so that what we were seeing before must be due to the 15 more easily detected ones and you do not see a very strong a
u response. occurring here at the 500 kilohertz for the center is frequency that one would like to have and what one was seeing in the case of columnar material.
I 19 When we went to the easy cracks, this is shown here, n
7i you can see that we do get the peak coming back in here at k
about 400 kilohertz as to be distinguished from this case 22
- 3 where we actually had almost the inversion of that looking at 24 the composite spectrum for the difficult and the easy ones.
73 You can't really conclude that that is in fact a good zone, 1
l
Page 32 out here it's 500 kilohertz for this particular case and i
2 there's not a good response occurring either.
3 So the only thing we had left to do was to look at 4
that Speciaen Number 11 to see what kind of response we got 5
from it, and that's shown here.
It was.actually on the 6
equiaxed side and that was the one where people detected it 7
with very high. reliability, 93% of the people who saw that ande a classification and in the non-cracked zone there was 8
9 only a few percent of classification calls.
And we looked at jo the spectrum of that and you can see that there's not a strong.
ii response there at 400 kilohertz but we do have a fairly sizable response right here at the 500 kilohertz. 1h2 in this.
- 2 particular case,.for whatever. reason,.this one.was giving us a g
(
very nice response at the SOO kilohertz range but if you look u
at the spectrum of this versus the others, it's not 15 16 overwhelming.
There's nothing in here to suggest that people 17 should make a classification on this of 93% when you go back is here to this case in the columnar material and have this kind 19 of response occurring, and yet none of those cracks. people 20 even approached that kind of level.
21 So I think this shows some of the difficulty in teras of trying to understand this.
It's related to the 23 signal amplitude and it's related to signal to noise ratio, 24 but that interrelationship and what people actually use in 25 teras of a decisionmaking process is not well understood, it's O
i
__________A
l page 33-i a very complicated matter.
2 So our conclusions that we've drawn from this is.
3 that we feel the detections were-probably based on signal 4
maplitude as the strongest piece of evidence that we've got.
5 It wasn't the only thing, but in terms of'the ones where 6
people tended to find those, at least in the columnar case, 7
that appears to be the largest trend, but that doesn't answer 8
the question.
1 9
There is no simple filter, as-we pointed out, in io this that you could use.
In the 500 kilohertz case that we it were using, you would want to be right near that frequency for 12 columnar, examination and if 'fou went to equiaxed it appears 13 that perhaps you might want to be down at the 400 kilohertz 14 range in order to pick up a number of those others, although even w rking there doesn't provide you with the ability to 15 detect all the cracks-in the equiaxed material.
16 17 The problem is not really one totally of signal to is noise, and that's probably due to the spectrum of defects.and 19 noise are quite similar.
As you look at these plots, they're
- o quite complicated, but thisy also show that.the spectrum coming 21 off the coherent scattering from the grain structure and the 22 amplitude of that approaches that which one gets off of defects.
73 pa So what is the botton line of this?
I think it 25 lshows that, you know, there's still a lot of work that needs i
O 1
)
Page 34 i
to be done to understand how to inspect this material and that 2
you can't use just simply signal amplitude.
YOu must try and 3
. understand when you're performing an inspection as to what's 4
happening as the sound goes through a particular 5
mictostructure and optimize the inspection with regard to that 6
particular microstructure.
7 What I'd like to go on and. talk about is the coarse a
grained material-inspection.
The objective of this work is to l
l 9
evaluate the effectiveness of ultrasonic techniques for w
inspecting cast materials, to understand the physics of the il problem of how sound propagates through that material, to 12 assess methods to provide improvements for the inspections, a
determine thy limitations of those solutions and recommend i3 ja improvements to Code / regulatory requirements.
This task is part of one on the NDE reliability 15 16 program which is going on at PNL.
Thorn are four problem 17 areas that we're baulcally addressing,.that consist of the is far-side weld inspections, cast stainless steel weld n
inspection, dissimilar metal weld inspection and weld overlay 20 inspection.
All these are similar because they all deal with pi very coarse grained material.
And our approach as part of this is to carefully map 23 the sound field transmission properties in going through this l
i 24 to try and understand what actually happens, can you get a 1
23 coherent sound field propagated through this material, till I,
u
Page 35
,_<-)
DR..SHEWMON:
Is the far-side weld stainless steel i
J 2
also?
3 DR. DOCTOR:
Yes.
4 DR. SHEWHON:
Okay.
5 DR. DOCTOR:
We've had to go to a digital signal 6
acquisition system because when sound propagates through this 7
coarse grained material a number of things happen, you get a
scattering of the sound field, you get mode conversion, you 9
get beam skewing, and all those lead to all these multiple ic modes coming through.
In trying to capture that signal that's 11 received on the other side, to understand what it means, is 12 difficult.
So what we found is that the best way to do that is to record it and then to actually map what actually happens i3 l
[,,s}
l k
at different locations, spatial locations.
From that you can u
l then determine what is the actual signal that you want to be l
3 l
tracking, because if you mode convert into a longitudinal and l
q have twice the wave length, therefore you'll get confusing 17 results.
33 What we have been studying is four different n
micrustructures; the columnar and equiaxed which I have
(
- g,
l
\\
- i l described in the previous studies, mixed modes and layered I
h l
- 2 h type structures, and I've got examples of these and I'll l explain what we mean by that terminology in just a moment.
u l
h
- p The status is we've completed some L-wave b
75j attenuation measurements using a naught degree probe.
We've
Page 36
)
I also completed our L-wave field proflies at naught degrees and 2
we have in progress the 45 degree L-wave field profilms.
I've 3
got a series.to show you which I didn't-have time to reproduce 4
to put into the handout, on the 45 degree L-wave, which.I 5
think you'll find interesting.
6 Basically the system looks like this, in which we 7
take a specimen and we attach a very small.-point receiver to a
the ID of it and then we scan a transducer driven by a tone 9
burst at a particular frequency across the surface.
We then jo gather that signal that's transmitted.through it, we amplify si it, we use a gated RF peak detector in the analog mode but now 12 we just do an A/D, conversion on that and store it in our minicomputer.
i3
(
Now what I'd like to do is show.you what kind of u
results we get from that.
I might point out that this is done 15 i
in this manner so that we're simulating what actually happens 16 17 during inspection as if there is a defect located at this 13 spot.
This is the kind of sound field that that defect would 19 actually be illuminating.
- c Let's just spend a minute talking about this one 21 Particular case because it's one that we obtained in carbon l
steel -- I'm going to show you a whole series of these and 73 we'll spend just a minute going over what actually is being 23 24 presented here.
What we've shown is an aperture in which this 23 is a circumferential pipe like this so that this direction is l
..... ~.,,, -,. -.,
Page 37 7"
[_;
1 in the circumferential direction and.this is in the axial 2
direction.
Okay?
And the dimension here, this is 75 3
. millimeters or three inches, the dimenelons here are about 115 4
millimeters or about four and a half inches.-
5 What we've shown then is ranges of amplitude, zero 6
is the reference here in red, orange then is the 1 dB contour,-
7 the green is the 2 dB.
There's two shades of blue in here at a
3 and 4, then it goes to this kind of maroon color at 6 and a-9 brown color at 10 and a black color at 14 and a white color io then.is greater than this -- 20 dB and greater.
11 So this is what happens when we go into a piece of 12 ferretic (ph.) pipe that again has a 16 millimeter thickness.
This is the kind of response that we got using a probe at one 13 14 megahertz and had a diameter of one and a half inches.- So i
l
.this is looking straight down through it.
is 16 Now let's step through some of the other 17 microstructures.
Here is an example of a columnar grain 13 microstructure material that I've shown you results of in the 39 CCSSRRT study.
When we do our zero degree profile through t
- c, that columnar microstructure, what we find in contrast --
{
23 perhapes maybe give you a little bit of an idea -- the effect 22 of going through this coarser material of course is to spread
(
73 the sound field out.
You've lot a fair amount of -- the 74 coherency has started to spread out, you can see in particular 23.;the brown level is much, much larger. If you go into this h
l
Page 38 1
maroon color, you can see it has grown.
Specifically you've 2
seen probably a little more growth in this direction than you-3 have in the transverse direction.
4 If we look at an equiaxed specimen such as shown 5
here, when.you propagate the sound. field through this, you can 6
see that in fact there isn't.as much degradation occurring 7
with regard to again the results obtained in the carbon steel.
8 YOu can see in the equiaxed structure, there's not much 9
difference.
10 If we now.go to the next, this is a.coln=n=r and 11 equiraxed-microstructure, it's.a specimen that we obtained from 12 Westinghouse, and if you look at this you can see that there's 13 a predominance of columnar. material, as we would describe it, O
14 occurring on the upper zone of it and.the rather coarse i
15 grained more equiaxed type structure occurring down here 16 towards the ID.
This is the circumferential direction and the l
17 axial view of the same thing and you can see.still they've got la long columnar grains in this direction.
Down here we have 1;
rather large, what we call equiaxed. grains, they're just large 2c grains that have equal dimensions basically in X, Y and Z.
l l
21 And we classify those as being equiaxed.
In this particular i
(
22 case you can see more of a trend from the columnar to the 1
l 23 equiaxed.
When we propagated the. sound field through this 24 specimen, the is the nature of results that we obtained.
Now l
25 this is looking straight down on it.
You can see here in this 1
i l
l l
Page 39 p
__ j 1
particular case that we're getting a.much larger smearing of
?
the energy in the circumferential direction for this material 3
versus what we had in the columnar -- excuse me, what we had 4
in the carbon steel.
But you don't see a real 1erge 5
enhancement of spreading of the energy in this direction as 6
what we had seen earlier for this case here, the columnar 7
material.
So this isn't as bad as that pure columnar form with regard to dispersion in this direction.
9 Okay, this was a specimen that we obtained f rom to Southwest Research Institute and it's a very layored type of 11 microstructure.
If you look at it you can see very definitive 12 layers going through this.
You can see a fairly complicated i3 structure, you can see some dendritic structure down here that
/
\\
l V'
la almost goes completely through all.
In this zone here, you've 15 got, you know, dendrites occurring and then it looks like it goes into more equiaxed phase.
And as you go across here it's u
a very complicated type of structure.
These are two axial is profiles, this one across this end and this one across the 1;
other end.
You can see on this and you've got again this
- [ dendritic structure in the upper part, some very large, coarse H grained type structures that we classified as being equiaxed on the lower portion of it.
Over here, it's somewhat difficult, it's hard to define it, the structure even gets
- g h
a n f airly small there apparently in certain zones.
It's hard to P
l
- 3 actually classify this in terms of, do you call this a I
L i
l Page 40 7)
- _j I
columnar and equiaxed, what kind of description do you use to 2
accurately describe it.
3 I'm going to show you a-profile tha;
_s gathered 4
and that was -gathered in this zone right in h, One of the 5
things that we will be doing and we did not.. <a a chance to 6
do prior to this meeting, is to. gather proflies as we move 7
across the surface, to try to better understand what hapFens 8
as you go through these various zones.
9 DR. SHEWMON:
Does the interface itself, the 10 circumferential interface introduce attenuation or --
11 DR. DOCTOR:
In terms of going through it at 45 12 degrees, I would suspect one would never see those.
Perhaps 13 looking down there may be something about those that's related 14 to, you know,.the power spectrum, you may be able to see those 15 and enhance them.
Probably in general at the one megahertz 16 frequency that you're using, they're going to be pretty 17 transparent unless you do something really to try and is emphasize that. particular. aspect in the spectrum.
io DR. SHEWMON:
Magnification is such that the
- o columnar diameter is a millimeter or several millimeters?
21 DR. DOCTOR:
I don't know if you can see this, this n
is basically an inch right here, this dimension right here is n
an inch.
So if you go over here you can see that these are 24 roughly about an eighth of an inch.
If you go and look at n
some of these large structures like that, you're talking O
, -... - -,-~_,- -,._. -,_.,.
I Page 41 7
things-that are in the half inch regime...This zone right here x
2 1coks like --
3 DR. SHEWMON:.The wave length is u lot shorter than 4
the grain size then.
5 DR. DOCTOR:
Nave length at one. megahertz you' re.
6 dealing with about 250/1000 -- quarterauf an inch, quarter of 7
an inch wave length.
So this is smaller but some of these 8
other things are clearly larger than the wave length.
9 This is the profile that we obtained using a zero 10 degree probe going through that, that specimen at one location 11 that I showed vou.
It's very apparent here that we're 12 starting to. pull.this energy out-and it's almost completely 13 filling up this entire aperture that was scanned showing what u
I think is, based on this kind of structure, a breaking up the 15 coherency of the sound-field, you're starting to pull in other r
16 lobes.
p What the sound field profiling does, it allows us to is look at it in and ask the question as to whether or not one l
19 can get through this particular microstructure a coherent 20 sound field which is needed, because that's + dust you rely on 21 with regard to ultrasound to make all your decisions.
If you i
22 lose that coherency, you've got nothing to work with in i
23 essence.
And we're trying to understand what happens and 2.s whether or not we can improve on this by changing frequency if 25 one, for this type of structure for example, dropped down to a I
l i
I
I Page 42 1
m I
lower frequency, you very clearly may be able to improve that, 2
but you don't have any way of quantifying it without going 3
through this kind of procedure to understand that.
4 DR. SHEWMON:
That sort of structure only shows up 5
with centrifuga11y casting where they pour in a jerky fashion c
but not in say valve bodies, or do you know?
7 DR. DOCTOR:
I guess I don't know enough about e
static casting microstructures to be able to answer that.
7 I've heard tell rumors that in those you get grains that are io incredibly large, much larger than what you see in the 11 centrifuga11y cast process, but I don't know whether or not 12 you get this kind of, you know, intermixing of different t
33 layers of dendritic and equiaxed type structures.
(,_]
's u
Well one of the ways to look at this and try and is understand in addition to the sound coherency, is what happens to attenuation wise, because the one thing I haven't shown you 17 here is at what gain levels this information was collected at.
Is What I have here is a table and the only thing I 1;
want to focus on is over here on the relative attenuation.
What I've shown you is these four cases in the same order that i
21 ! I just described them and what we have shown here then is in d
- l; decibels per centimeter, the attenuation for the longitudinal a
n and for an SV wave.
And you can see when we compare things e
! versus carbon steel, in the equiaxed we had like a.69 dB per 2:
1, 25L centimeter on propagation and roughly slightly over twice that h
r
l i
Page 43 1
or 1.5 dB per centimeter for the SV wave.
2 As you go.to columnar, you can see that you get an 3
. increase in attenuation, you go to the mixed diffuse and you 4
get a higher increase and you go to this mixed layered l
5 structure, this last one, and you can see a much higher 6
attenuation.
And this is due to the breaking up of the sound-7 field.
When you fi\\1 out this kind of a plot, you can see the a
rationale for what's happening is that you're breaking the 9
sound field up due to a variety of different processes and 10 leading to the attenuation.
But what we're trying to 11 understand from these is what kind of coherency can we get I?
through this kino of material and this tells us that we're 13 probably going to have to increase the gain considerably to 14 penetrate this material effectively.
15 Now I had indicated that we had done some work on 45 16 1 degree L-wave.
This is an example here in carbon steel.
The 17 plate is curved like this, the microprobe is sitting here and I F.
we're scanning it again.
This is the circumferential n
direction and that's the axial direction there.
you can see a 20 fairly nice sound field here, a little bit of a break up is 21 occurring here, but that's not really much of-an alternation 22 from being basically a perfect sound field.
I
- 3 Now on 25 degree L-wave going into the columnar 24 material, you can start to see some changes occurring, if I 25 can slide those to adjacent.
What we've plotted, I should l
l_
1 Page 44 P
i point out too, is we're-putting the sound field in at 45 2
degrees and things are set up very precisely to ensure that, 3
the center of our sound field is right there at 45 degrees.
4 In this particular. case when we're going into columnar 5
material, what you find is that the center part of the beam is 6
skewed up slightly higher up around say 48 degrees.
You can 7
also see that there's been slight shift, therei s a starred a
formation or what I'd call another lobe occurring down here 9
when you're trying to propagate through the columnar material.
10 When you look at the equiaxed type material at 45 11 degrees in the L-wave mode, you can see -- you're still at 45 12 degrees basically, maybe it's skewed down a degree or two, but 13 it's pretty close, but you can see that the sound field has 14 broken up quite more dramatically.
15 When you go to the layered columnar and equiaxed 16 microstructure, in this particular case this specimen was 17 thicker.
Okay?
It was like 75 millimeters thick versus the is 60 al1Jimeters, and that's the reason that we had to adjust n
the angles occurring over here.
But surprisingly for going
- c through this particular structure, we've got a very coherent
- 1 sound field quite frankly, better than tihat I'd expected based 22 on, you know, results in the naught degree penetration.
And 27 looking at the columnar and equiaxed when we went to this 1
24 layered structure, I'd expected it to be much worse than what 25,we were seeing, but the coherency is pretty good in this.
l (I) h i
Page 45 f
1 When we go to the mixed rather than the Imyered 2
structure, this is back to the same thickness, you can see a 3
skewing.here, probably down around 40. degrees with regard to 4
the center lobe, plus you're getting another fairly strong 5
response that's only six dB down, occurring down here roughly 6
in about 15 degrees, which probably means.there's some kind of 7
a referred direction located in that.
e Furthermore, since things were set up carefully you 9
can see that this beam has actually been skewed off to the 10 right.
It should have been located, you know, in the center 11 and you can actually see a skewing that has occurred in that 12 direction, so that you would think your beam was located here, 13 but the center of the beam is really located over there, so 14 it's giving you a shift.
15 Well what conclusions can we draw from this?
The 16 conclusions say that the microstructure is very important to 17 UT inspection effectiveness and that when you get into these is more complicated microstructures they create greater n
inspection difficulties due to the distortion.
And what we're i
20 trying to do is to understand that distortion so that you can 2i utill get a coherent sound field through with known properties 22 iand then be able to use that information to actually determine l
lthepresenceofdefectsandusethe information in a reliable 23 24 fashion to talk about the properties that the defect had, 1
231specifically the size of it.
But if you don't, the skewing in I
d
(
~.
Page 46
)
i effect, you're going to place things in the wrong position, 2
and that's extremely important.
In another year or so I hope 3
we'll have all these answers and be much further along in our a
knowledge than where we are right now.
5 DR. SHEWHON:
Before we get into that, you said 6
something about increasing the intensity of the beam, to what 7
extent -- there's a lot of scattering, you didn't use those a
words but I think that's what you meant.
9 DR. DOCTOR:
Yeah,-that's --
10 DR. SHEWMON:
How far can the operator do that?
11 DR. DOCTOR:
Well this is really what I was 12 referring to.
This is an attenuation measurement and the 13 difficulty is is that as you go around, as I showed -- let me 14 see if I can pull it up real quick for you -- in this kind of 15 m --
16 DR. SHEWHON:
I understand that but my question is 17 one of equipment, not one of --
la DR. DOCTOR:
Sorry, I misunderstood.
You mean can an operator just increase the gain to compensate --
i n
20 DR. SHEWMON:
Well-you said gain and I thought l
21 power.
The power is fixed and the gain he can control if 22 noise is a problem, is that it?
i 23 DR. DOCTOR:
Right.
i DR. SHEWMON:
What limits the amount of power he 24 25 puts in, crystal?
. O
, - ~...... _ _. _ _ _ _..... _ _ _,,, _ _ _... - _ _ _. _ _ _,.. _.
l Page 47 1
DR. DOCTOR:
Well it's that and what is the output 2
voltage from the pulser that he's using.
And of conrse, 3
depending upon how you want to drive it, you can do things 4
like drive it at a particular frequency with a tone burster, 5
reduces your performance in tarms of range resolution, but it 6
give6' you better lateral type spatial resolution.
7 I guess tha real question is is you have to I think a
underetend whether or not you can prspagete a sound f.ield y
through coherently Jf you up the power regardless of the fact to
-- you can increase the power as much as you want but all it 11 does is pull up the grain by the same amount.
It isn't-going 12 to improve your signal to noloe ratio, it isn't going to give 13 you a better inspection.
What you have to do is be able to 14 get the sound field through in a coherent fashion and then.
15 optimize the power co that you can work as effectively as 16 possible to reduce the coherent scattering with the mar.imum 17 signal response.
13 DR. SHEWMON:
So presumably when the people come 19 back or the word comes back from the field that there are 20 castings in this particular plant which can't be inspected i
21 because we can't get a beam through them, it's something like l
22 lthas?
l l
i 23 DR. DOCTOR:
That's right.
And I think, you know, 24 the more samples we can got and we can understand those cases, we'll be able to determine what we can do and be able to go
- 5 l
Page 48
,:)
1 out and perform --
2 THE REPORTER:
Excuse me, I can't huar you over 3
here.
4 DR. DOCTOM:
I'm sorry.
I think all this data, I 5
find it in general rather encouraging.
Looking at some of 6
thean sound field profiles, I would have expected them to be 7
much worse than what they actually have turned out to be.
I'm 8
much encouraged that we can perform effective examinations on
- ! the materials if we can understand exactly what's happening as to the sound propagates through these materials.
11 I think the results from our round robin exercise --
12 it doesn't show the problem is solved, but when I was 13 initially doing this work, I quite frankly was very possimistic that we would ever be able to perform effective 14 15 examinations, and looking at this data and the frequencies of le calls in a number of the defect zones, I was quite surprised a
that we got as high performance as we did.
1s The problem is clearly not solved, but I think n
there's a lot of hope there and I think that through some of 2'
the work that we're doing, and I'm sure what EPRI is going to 21 report on later, we're going to increase our understanding and 2:
do a much more effective job than what has been done in the n
- past, t
[d DR. SHERMON:
Thank you.
24 25 h DR. DOCTOR:
I do have a few more viewgraphs talking h
i
l Page 49 4
t 3
)
i on the PISC III.
This is Act.lon 4 and it's called Round Robin 2j Studies on Austenitic Stuuls, and it's given the acronym AST.
2 I'm a co-leader of this work with Hans Herkenrath from ISPRA 4
Rasearch Center.
What the AST Program does, it describes u l
5 program for studies to be conducted in terms of capability, 6
parametric studies and reliability, and I'll define what these 7
are in just a moment.
The program includes wrought stainless e
steel as well as cast stainless steel, and the plan describes 9
specimen sets, test protocol and analysis methods for actually to conducting this.
11 The planning for this is underway, specimens are 12 being acquired and defects are being implanted, that nature of 13 thing, for actually conducting this type of very in-depth 14 study.
15 The program has three different sets of study as u
I'll call them; one is a capability.
The objective is to u
identify procedtres that have the potential to detect and size 9
defects and to discriminate between flawed and unflawed n
materials.
These will be specimens that will be circulated
- c i
- around from laboratory to laboratory so that people will be i
- iq'able to use some of the evolving technologies that are in li L laboratories and are not ready to go out and be tested in more F detail through what we call reliability studies.
nq The parametric studies are designed to complement 1
- sU both the capability and the reliability studien und they're o
O s
1; l.
Page 50 I
really tryin7 to evaluate-the effect of important material and 2
defect variables such as microstructure, such as defect time, 3
the effects of crown and counter-bore with regard to the cast 4
auut wrought structures.
1 5
The ru11 ability study is designed to measure in-6 service inspection performance under realistic in-field 7
conditions on realistic cracks and evaluate human reliability a
factors.
This -last. one,.tnuuut reliability factors, I'll only 9
comment there that a separate action designed to try and 10 gather information on.tuman reliability and we're jnterfacing 11 with that particular t ask for the reliability studies.
12 So this-is really trying to determine capability of 13 Potential techniques, this is trying to look at the question l
\\
14 of reliability under actual field type of conditions.
So in 1
15 this particular case, what one would do is have several sites l
r 16 and bring the teams into those sites to go through an 17 inspection such as what one would encounter when going to a la plant and actually performing an inspection.
19 The kind of matrix are shown here for thm.three 20 studies, to give you an idea.
The capability studies will be a
relative small specimens, let's may maybe a foot and a half in n
axial length and circumferential length perhaps 8 to 12 23 inches.
The kinds of defects that are going to be put into 24 these will be fatigue cracks, thermal fatigue cracks and
- 5 mechanical fatigue cracks.
The F here is for fabrication type I
Page 51 m
)
)
defects.
The A down here is for artificial sharp plane 3_;
2 reflectors.
This was introduced in the parametric studies as 3
a result of the PISC II results, to do a comparison between 4
those artificial sharp planr reflectors and other type 5
reflectors.
As 1 Indicated, parametric studies are going to 6
deal with things like base material, crack characteristics and 7
weldment geometry.
a You can see it's a fairly substantial number of 9
specimens that are being complied.
When you're dealing with 10 this kind of an international study, you'll have as a result 11 of the PISC II, if you use that as a measure, there are like 12 50 teams from around the world that perform inspections on those four blocks.
So. that gives one a tremendous data base
/~
n-14 to use to look at effectiveness of applying various technology I
15 and gives one a yardstick for determining what reliability one 16 can achieve.
You introduce typically a number of teams that u
are using quite similar procedures so that you can actually la look at the variation of applying a similar procedure by 19 various qualifications and various -- and qualifications of
- c personnel au well as equipment, 2i DR. SHEWMON
Fatigue fabrication is lack of filling 22jintheweldorporosityincasting, or --
l
- 3 l
DR. DOCTOR:
What we were planning for for our a
fabrication defects were primarily lack of fusion type of 2s defects.
That was felt to be the most important, particularly 1
il il
Page 52 1
because if it occurred let's way down near the root of a weld i
2 it could be extremely luportant from the structural 3
standpoint.
It's condition should be found during pre-service 4
and if it's found in-service, it may warrant, you know, a 5
repair.
It was thought to be one of greatest interest.
6 The reliability study, there's two different, if you 7
will, groups, cast to cast and the other group is cast to 8
wrought and a wrought to wrought series of specimens.
Most-of 9
these are going to be a pipe to a component, principally an to elbow type of spercimen.
11 So you can see there's a substantial number of I
12 specimens being put together for this.
It will be a very 13 1arge data base and extremely useful in quantifying and 14 understanding inspection in these mustenitic stainless steels.
15 So what's the status?
We had a call for intent to 16 Participate that was sent out in the fall of
'87.
There was a 17 very large interest that was shown from that, which -- the is reason this was drawn upon to give guidances to how much 19 interest there was, should we go forward with this particular 20 round robin test, and there was a large ascunt of interest
- 1 shown so things are moving ahead.
22 We plan to start testing in the fall of '88 and there will be a final call for participation that will be sent n
out this summer.
There's a Board meeting next month and it 73 will be concluded there and than the final draft will be
Page 53 1
prspared and mailed.
y 2
A schedule has not been established because it will 3
rely primarily on what kind of a final participatico we get 4
that comes in, then we'll set up a schedule when people can 3
5 actually perform the inspections, because they have to work 6
around a number of other requirements.
We'll lay out a 7
schedule-I would suspect to run probably for one to two years 8
and then of course there will be instructive work and 9
reporting the results.
10 And-with that, that wraps up all the material that I-11 had to present.
I guess I'm pretty much on time.
Are there 12 any questions?
13 (No response.)
14 DR. SHEWMONr.
Thank.you.
The schedule calls for a 15 break at this point, why don't we take one.
16 (A short receos was taken.)
17 DR. SHEWMON:
Fire when ready.
IB PRESENTATION By-ALBERT E CURTIS, III 2
19 NR. CURTIS:
Well it was June 25, 1986 that we last 20 talked to you, the ACRS Subcommittee for Material Components 21 about our joint Westinghouse Owners' Group and EPRI 22 Coordinated Program on ultrasonic examinatlon of welded joints l
l 23 in centrifuga11y cast stainless steel pipes of PWR main-24 coolant loops.
We also obviously have some statically cast 25 components included in our sample set that you will be i
i
Page 54 bd i
seeing later.today and that some of you-saw before when you 2
were at our meeting in Pittsburgh-on June 25, 1986.
3 The. background, I think Steve Doctor did an 4
excellent job of setting the stage, It's almost impossible to 5
inspect some of these configurations, although we now feel o
that we have a lot more understanding, and hopefully you'll 7
see why later today that we can do a much better job.than was a
done even just a few years ago on cast austenitic stainless 9
steel components.
10 WOG and EPRI started this program back several years 11 ago, both bringing to the party if you will the aspects we 12 thought we both could contribute and make the best possible 13 program.
The program elements I'll review with you in just a 14 minute.
The major objectives we had hoped to accomplish was 15 the optimization and quantification of flaw detection and l
16 d121DQ Capabilities for the in-servlCe inspection of main 17 coolant. piping; interface improved flaw detection and sizing is procedure with automated inspectica data processing systems; 19 and then demonstrate the improved flaw detection, sizing 20 techniques and equipment and test samples representing actual 21 field conditions and indeed part of the demonstration would be J
l 22 to this group of individuals from the ACRS.
23 We are on track, we are at the final stretch of our 24 joint coordinated program and we have we feel accomplished the J
l 25 major number of our objectives, b
L
Page 55 re#3
- [_;/
1 A q'alck review of the four phased approach we.took 2
of the coordinated program was obviously we needed to 3
fabricate test samples.
And so the first thing we wanted to 4
do was to try to determine what type.of matrix flaws we should 5
include in these samples and how we should put~these samples 6
together, and that was a joint effort between Westinghouse and 7
EPRI, the Westinghouse Owners' Group people and the a
Westinghouse personnel along with Electric Power Research.
4 9
Institute people, Gary Dau and Dr. Behravesh and others, and
'l 10 some consultants.
11 Sources of pipe material, we had some material that 12
. Westinghouse supplied for us to fabricate these test samples 13 und of course we d1C fabricate the test samples and in the s-ja inst meeting, Dr. Showmon, I know that you saw some of these is samples that had been fabricated and we were actually i
fatiguing them and thermally cycling them to produce both 16 17 mechanical and thermal fatigue cracks.
l 13 Phase II was to-improve manual technique 19 development to look to see what we could do to improve the l
20 manual techniques that are being applied across the industry 2i today.
Westinghouse and EPR, iked toguther in establishing 22 technique requirements and then we had the manual technique l
development and Rick Rishel from Westinghouse will talk about 23 24 that from the Westinghouse point of view and than Dr.
I 25 Behravesh will talk about that from the EPRI point of view l
l l
It
Page 56 i_;
)
later on.
2 Phase III, automated inspection, we had to go 3
through equipment evaluation and demonstration and that has
.i l been done here at EPRI and they have been doing.that for quite 5
awhile, not only looking at what we can do with cast piping l
6 but all piping, wrought piping, BWR piping, carbon steel 7
piping, but to factor the cast
-uq material into this a
program.
So they've evaluated Jenonstrated the use of the 9
automated inspection techniques.
Now they're integrating this 10 type of approach into the inspection regime for. cast material i1 and Dr. Behravesh will talk about that.
And.then-lu;ur wo 12 hope to have some field trials with-the automated tect~_yuss 13 and Mohamad will address that.
14 And last but not least, and we are here today doing is this, is to demonstrate the capabilities that we feel we have 16 improved upon and developed.
And that's a joint effort again 17 between Westinghouse and RPRI staffs and then we hope to in 18 the near future -- near-future -- within the next four to six I
19 months, develop the protocol for howe we would go about in 20 training and demonstration utilizing samples that have.been l
21 developed in the program that has been carried out.
(
22 As far as what are the end products, I'll just l
23 review those for you again; there are 75 test samples which 24 !' represent actual field condition, potential flaw types, flaw I
I i
75 orientations, joint configurations geometry, materials for use
!lO E
i i
l
. - _ - - - - -. _ - -.. - -. -., -- - _. _,-.,., _.. - ~. - - -..-.-.,.,, - -,- -,,_ - -,, -- - -..
Page 57
--]
g_a/
1 in establishing personnel training programs and in 2
demonstrations, and then of course demonstrate and quantify 3
fimw detection and sizing techniques and equipment for in-4 service inspection-of main-coolant loop piping..
5 Now obviously we hope to factor all this into an 6
overall long-term plan and that's to not only demonstrate we-7 can inspect this pipe but.make sure from a fracture mechanic's e
pointaof view we can detect flaws before they becomo a 9
concern.
We also, as I mentioned to you last time, Dr.
l to Shewmon, I have a personal ultimate goal and that is not the 11 purpose of this meeting today but once we demonstrate we know 12 what we're doing when it comes to inspection, we'd like.to i3 look at why are we npending a lot of time and money inspecting 14 this type of material when there may be other more critical 15 components that we ought to be spending more time and money l
16 inspecting.
j 17 So rather than, as I said last time, telling you I-1 l
18 don't need to do it because I can't do it, I'm going to say I 19 can do it but I really don't need to do it for the following 20 reasons.
For instance, I don't have a problem, literature l
l 21 says there isn't a problem and experimentation says there 22 isn't a problem, so why am I wasting my time inspecting.
23 He are doing this today, we've been working on this 24 g and I think you wills be very cleased and I hope Dr. Doctor is
- s'h! pledned also with uhat he sees that we've done.
And then of i
}
L yv.
9
_y-
l Page 58 l
l 7 '
(
1 course-we've gone through and shown a leak before a break is 2
applicable for the cast austenitic stainless steel material 3
and we are -- we have completed and are working on this 4
thermal aging question from Westing'ause Owners' Group poAnt.
5 of view, which we will factor into this overall quest!cn later 6
08.
7 That's my personal and Westinghouse Owners' Group a
ultimate goa3.
That may be down-the road quite a ways, but 9
that's where we hope to be heading.
10 So wLthout any further ado, unless there's some 11 questions for;ae personally, I'd like to introduce Don 12 Adamonic, who will be speaking to you about the fabrication of 13 the samples.
Then Rick Rishel from Westinghouse will talk to
,,)
u you about the manual development work that has been done and 15 then of course Dr. Behravesh.will talk about the EPRI work and.
16 the automated development work that has been going on.
17 Thank you.
Don.
.- 18 PRESENTATION BY. DON ADAHONIS 19 MR. ADAMONIS:
Thanks, A1.
Does everyone have 20 copies of the set of overheads we'll be using here?
21 I'd like to speak briefly on the sample set that was 22 developed under the Owners' Group Program.
As Al mentioned, 23 one of the deliverables from Westinghouse under this program 24 was to provide a set of 75 crack test samples.
Those samples 25 have been completed, we've completed the fabrication and they
F
~
Page 59 7
them 1
- reside here at the ND3 Center and you'll be.able to see 2
this afternoon.
3 That. test sampic set represents a variety of.
a material combinations.
We varied some.of the welding 5
techniques, we've represented a number of joint geometries, 6
we've included.various defect types; thermal, mechanical, 7
fatigue. cracks.
We've varied the defect sizes in terms.of a
depth and length and the location along the length of the l
9 weld.
10 These samples were fabricated from nine ring.
11 weldments-representing again different geometries, pipe to 12 elbow type configuration..We've miso included inlet and 13 outlet nozzle geometries that include the bi-metallic, tri-14 metallic welds.
15 The overview of the parameters that we. varied --
16 that came out. pretty well actually -- we have a designation 17 for the various pipe to elbow configurations.. If you look at l
18 the designation for the first column, APE, it's a pipe to l
19 elbow reeld, centrifuga11y cast pipe to a statically cast elbow 20 welded with an automatic procerc.
21 Second set designation MPE, same caterial 22 combinations, these were welded manually to represent the 23 field weld.
l 24 The third set designated OPE, we varied the pipe.
25 material.
This is some of the older vintage cast pipe where l
Page 60 1
the microstructure we'll see as we go through is more columnar 2
in nature.
All of the pipe microstructures that we look at 3
are of the-mixed variety, not.necessarily layered but the 4
mixed variety that Steve Doctor mentioned.
This particular 5
vintage of pipe demonstrates more of a columnar structure than 6
equiaxed.
7 The FPE designation is a forged pipe.
Again we.
8 varied the pipe material here, this is a forged pipe to 9
represent those plants where forged pipe materials were used.
10 We've mocked up the pump outlet pipe weld where the 11 inspection-problem-is further complicated by some overlay that 12 was applied to smooth the transition, and I'll show some of
(
13 these geometries.
ia And again the inlet and outlet nozzle.
15 configurations.
We represented a. number of different heats of 16 piping material, again automatic versus manual-welding 17 processes were included in the matrix and the next to the last is column on this overhead shows the distribution of types of 19 cracks; mechanical fatigue cracks versus thermal fatigue 20 cracks and I guess in 1986 you were able to witness in the 21 labs in Pittsburgh some of the cracking process, so you're 22 familiar with the process that was used for introducing the 23 defects.
i 24 DR. SHEWMON:
The forged pipe is plate that was then 25 forged into shape?
1
Page 61
_m h
j 1
MR. ADAMONIS:
It's really an extrusion, it's from a 2
large sheet and it's actually an extruded process,-an 3
extruding process.
We call it a forged pipe essentially, but 4
you'll see a rather fine grain structure as we go through.some 5
of the macrographs that we have in this package.
6 MR. CURTIS:
By the way, Dr. Shewson, based upon the 7
ccaments at the June meeting in 1986, we did go through and 8
categorize all the microstructures of these samples,.so we 9
have an accurate assessment of those. 'And that was done after to that meeting based upon.the comments of.the Subcommittee.
11 That has been done.
12 MR. ADAMONIS:
After completion of an individual i3 ring weldment, we sectioned the ring into various samples
/
u where the circumferential length varied from 8 to 10 inches, 15 the axial length of the specimen varied from 18 to 24 inches.
16 We would introduced a stress riser in the form of a notch l
17 whether we introduced the cracking mechanically or thermally, 18 a notch was included.
We would go through the cracking n
process and based on some calibration data that we had done, 20 calibration and sectioning early on a number of cycles could 21 be correlated with actual crack depths.
So we're looking at 22 samples that vary in length from 18 inches -- foot and a half 23 to two feet -- 8 to 10 inches wide, that had cracks in them 24 anywhere from one quarter of an inch to about one and two-
,5 tenths inches deep to -- arid lengths of cracks about i
1
Page 62
)
i eight-tenths to three at.d a. quarter inches.
2 Welds in the primary loop which are represented by 3
the samples are highlighted on this particular slide.
We've 4
mocked-up the inlet and outlet nozzle to safe end welds and 5
the safe end of pipe welds in the same mock-ups.
If a 6
particular plant were to haveamain loop isclation valves, 7
there are samples.in this set that mock those particular welds e
up.
The elbow -- essentially all the elbow to pipe welds in -
9
.the plant are mocked up and I guess the last area that we have io also been able to cover is the pump to albow weld.
11 DR..SHEWMON:
Do.many plants have isolation valves?
12 That's in the primary piping, isn't it?
13 MR. ADAMONIS:
In the primary loop.
I don't think 14 there are many, I can only think off the top of my head of 15 about three.
16 DR. SHEWMON:
I thought the code prohibited it.but 17 obviously it doesn.'t.
18 MR. CURTIS:
There's about three I think.
There's 19 about three or four plants with isolation valves installed in 20 the primary loop.
Most plants are putting nozzle bands in l
21 their steam generators so they can refuel and still work on 22 their generators.
But some plants, I think there's three or 23 four of all the plants, that have the isolation.
There's not l
24 a large number, i
l 25 MR. ADAMONIS:
But when you look at the various
!lO l
1 L
1 Page 63
(_g/
i material combinations that have been included in the program, 2
the welding types and the joint geometries + hat ::e've managed 3
to cover, we've covered essentially every area that you need 4
to look at from an ISI point of view.
5 MR. CURTIS:
No one is running with them though.
I 6
mean no one runs with them, they're issues during shutdown 7
conditions, so you can refuel and work on your steam 8
generators at the same time.
9 MR. ADAMONIS:
Here are some sketches and they're in illustrative of the joint geometries, joint configurations 11 that we've managed to duplicate during the sample. fabrication 12 process.
You can see on the pipe to elbow series, on one 13 series we've primarSly concentrated on. representing the joint u
configurations, generally from earlier plants where tne x-15 thickness of the elbow was thicker by a fairly significant 16 margin than the pipe itself and the transition from elbow 17 thickness to pipe thickness was made-across the weld, making a la rather difficult joint to inspect.
19 We've also in all cases -- and again these are 2c illustrative -- but we've maintained as well as we could 21 duplicate the counter-bores and ID surface geometry such that 22 when performing inspections of these particular samples, the 22 operator would be afforded the opportunity to try to 24 discriminate between ID geometry and real defects, as 25,
operators are given that opportunity in the field.
I h
Page 64
)
1 The pump outlet to. pipe is the one that I. mentioned 2
earlier where there is an overlay on the pipe side to take 3
care of a transition that exists between the pump nozzle 4
thickness and pipe thickness, and for the safe end wells, the.
5 outlet.and the inlet safe end wells, all the material 6
combinations are duplicated, the welding-processes were 7
duplicated as was used 17 the field where you have an inconel a
butting on the face of the 508 nozzle material.
The ID of 9
that nozzle is clad with stainless steel.
We have an inconel to weld to a stainless safe end and then the safe end is welded 11 to the pipe in a siellar or identical. configuration for the 12 outlet nozzle mock-up.
13 DR. SHEWMON:
Now there have been cracks at nozzle O
\\~'
la transition, pipe transitions but only in BWRs, or have those 15 been found in Westinghouse plants too?
16 MR. CURTIS:
We have not found any in Westinghouse 17 designed PWRs to date.
Is DR. SHEWMON:
Okay.
Let's hope it all has.to do l
l 19 with the coolant chemistry.
20 MR. CURTIS:
We hope.
21 MR. ADAMONIS:
To finish up, just a few examples of 22 the types of microstructures we found in the materials we've 23 used for fabrications plants and we'll be concentrating here l
24 primarily on the centrifuga11y cast materials but you will l
2s also have an opportunity to see the microstructures of the l0 t
Page 65 7'N, g,g/
1 statically cast. material as well.
2 This is a 360 degree ring section from a typical 3
piece of pipe that we've used.
I.have some other viewgraphs 4
that show this a bit more closely.but we looked at the 5
macrostructure, if you will, and we were looking at structures 6
that are primarily, and throughout this program you'll find 7
that we're looking at the layered -- not really the layered, 8
but the mixed microstructure combinations that Steve talked 9
about.
We see equiaxed and we see columnar together, but we to don't see it in the step fashion on the layered material that I?
you 'alk about.
1.j MR. WARD:
Why?
Is that because of the casting 13 technique or is that something --
14 MR. ADnMONIS:
It probably has something to do with 15 it.
The cooling rates, and I think it's probably not the 16 worst case from an inspection point of view but it's not the 17 best case.
is DR. SHEWMON:
What is the worst case?
19 MR. ADAMONIS:
The worst case based cut the data that 20 we've looked at so far is this very rigidly layered mix.
21 DR. SHEWMON:
Okay.
When you said it isn't the 22 worst case, I wasn't sure what "it" was.
Go s, head, l
23 MR. ADAMONIS:
As you can see, the sections give us 24 a bit of a close-up on this ring section that we looked at.
25 We do have a microstructure that is primarily columnar for the l
l t
Page 66 3_f 1-outer say two-thirds of the wall thickness and we go to an 2
equiax zone on the inside.
And as one goes around 360 degrees 3
around this section, you.see pretty much the same behavior.
4 Now just to move on to some individual test-samples, 5
we're looking at a test sample here in the APE series where 6
welds were made automatically, they're pipe to elbow welds.
7 On your right you see the statically cast fitting which is a
primarily equiaxed with some tendency toward columnar at this 9
point.
io And on the lefthand side we're looking at 11 centrifuga11y cast pipe from a heat that's identified 156529.
17 Now in this particular section of pipe, it's a'.most primarily 13 equiaxed with some tendency -- some slight tendency toward ia columnar on the inside.
15 DR. SHEWMON:
How much of a signal does the operator 16 get from that kind of a transition in microstructure between j
17 the --
18 MR. ADAMONIS:
You know, in the angle beam testing I
l 19 that we've done -- and we do-primarily angle. beam testing --
l 20 you don't see a definite reflection from that transition.
l 21 DR. SHEWMON:
No, I meant the weld metal.
l l
22 MR. CURTIS:
He means the weld interface.
1 I
23 MR. ADAMONIS:
From the interface?
Not a great deal 24 in this particular case.
Where you see it most, where the 25 interface signal in the weld and base material appears to be i
1 1
Page 67 Y\\_f 1
most -- or a significant factor is on the bi-metallic, tri-2 metallic welds.
On these particular welds it doesn't seem to 3
be a factor.
You need to contend with the geometry on the 4
inside of these to some extent, but the biggest problems-on 5
these types of welds are the access limitatior.s that are due 6
to this OD configuration that you see right here.
And in 7
fact, we were rather successful, Rick will -- I don't want to 8
steal Rick's thunder, he'll go into the examination results --
9 but in many instances we were able to. penetrate the welds and to see what I'll refer to as far side defects, the defects were 11 placed on the pipe side and on the fitting side.
12 The same mate rial configuration in terms of the pipe i3 and the elbow are represented here.
This particular la macrograph represents a manual weld of those two sets of 15 materials.
16 Now the next overhead,.we'll see some of the older 17 vintage pipe and you can see-in this particular case, we do is see more of a columnar structure on the centrifuga11y-cast 1o pipe.
So you can see we do have a variety of microstructures 20 represented in the heats of cast pipe material that we used.
2:
And even on the statically cast elbow side, there's some 22 elongation of grains that you see in this particular case.
So 23 this.might start to address the question you brought up with
- 2a Steve earlier about what effect does cooling rates have on the 25 microstructure of some of the statically cast products as O
f N
Page 68 e)
'_f I
well.
Wo.didn't intensively go into that study, but I see 2
from this particular overhand some elongation there that is 3
likely to have something to do with cooling rates.
4 This particular macrograph-is from one of the 5
samples that includes the forged pipe material.
You see a 6
rather small equiaxed zone on the pipe side.
Again we used the 7
term "forged pipe" but it's actually extruded.
8 The pump to elbow weld is represented by this next 9
overhead.
We're looking at -- you can see in this particular 10 zone the weld overlay that's used to accommodate the 11 transition and thickness between the elbow and the pipe 12 material and I guess based on some of the experience in the 13 BWRs we can see how these types of overlays may further la complicate the inspection problem on an already difficult 15 situation but that particular configuration is also 16 represented in the sample set.
17 And then the last few overheads I have to show are la the safe end configurations where we have the -- this 19 particular overhead is an inlet nozzle where we have the 508 20 material and we're looking at the weld to the statically cast 21 elbow in that particular case.
22 In this case we're showing the entire configuration 23 of an outlet nozzle safe end where we have the 508 material l
24 that's clad, the inconel weld to a stainless ring and then the 25 automatic weld directly to the centrifuga11y cast pipe.
I
Page 69 A(j/
1 So-just an. overview, at your request from the last 2-meeting, wu went ahead and looked at the macrostructure, if J
3 you will,.on all.the samples and I believe we'll still be.able i
a to see some.of that.
And we do.have -- feel as though we have 5
a wide range of structure included in.this sample set which 6
makes it a good set to go ahead and proceed with our technique J
7 development and verification.
And Rick will talk about some a
of the results we've been able to obtasn in our manual studies 9
of these samples.
10 DR. SHEWMON:
Thank you.
11 MR. ADAMONIS:
Thank you.
12 PRESENTATION BY RICK RISHEL 13 MR. RISHEL:
What I'll be going over is the manual u
inspection results of this> program, which is Phase II of the 15 program.
16 In terms of the program status itself, the manual 17 inspection program is complete.
This included a literature is search as well as manual examination program using various 19 transducers and test instrument combinations on these 75 crack 20 samples.
In this particular program all 75 crack samples were 2i examined with various, as we said before, various techniques, 22 transducers and equipment.
The final report on this 23 particular inspection results-will be completed by the end of 24 March.
This final report will include the literature search, 25 a synopsis of that; manual examinations and results:
t
Page 70 i
conclusions of the program;.some recommendations or some 2
things that I.found during the program which I think is 3
relevant and should be used to help develop sone manual 4
inspectionsprocedures, some things to look out for in these.
5 particular procedures.
6 DR. SHEWMON:
If somebodv wanted to find a copy of 7
that final report ten years f rom-now, where could they do it?
8 Who will get one, is this all confidential?
No libraries 9
except what?
10 MR. CURTIS:
Well obviously all the utilities will 11 have it and it would be our intention to provide that to --
12 I'm sure the Regional Inspectors would have it available to 13 them.
u DR. SHEWMON:
So it would be available at the plants 15 or in connection with the plants where it was germane?
16 MR. CURTIS:
Oh, yes.
Hopefully it would be 17 documentation fur widely used procedures on the material that la we're inspecting, so there should be a file.
19 DR. SHEWMON:
All right.
- c MR. RISHEL
And the last thing the report will 21 includo le a brief summary of results from the vendor 1
22 qualification program that Union Electric did on these 23 particular samples
-- on a group of these particular samples.
2a In terms of improvement of inspection results; 25 basically it involves six factors.
These factors are not
(
4 Page 71
?
g i
unique in the NDE industry -but they're just more important 2
when you're talking about main coolant loop material.
It is a 3
more difficult examination, it's not as easy as carbon steel a
itself, so you have to put more emphasis on all six of them.
5 These include knowledge of the fabrication 6
materials, what kind of material you're looking at, is it 7
mixed, is it equiaxed, is it rod; adequate surface a
preparation, the best technique in the world won't find 9
anything if you can't have double side access in some cases or i
10 you have poor surfaces to exam frum or you don't have adequate 11 coupling for your UT crystal.
Knowledge of the nature of 12 defects, what kind of defects could there potentially be out 13 there, are they branched, where are they located and such as la that.
Additional operator training and experience, providing 15 the operator the opportunity to look at crack samples with his 16 procedure to gain confidence for himself.
17 Five, understanding sound beam propagation la mechanism, beams distortion, beam skewing, understanding those o
phenomena.
- o And six, proper selection of ultrasonic test 21 parameters and procedures, which in a way is associated with ltheotherfactors.
I'll be talking to you in a little more 22 i
23 detail on each one of these particular factors.
24 In terms of the knowledge of fabrication materials:
25 what's generally known, the fabricator, year of fabrication,
Page 72
-m
)
i fabrication process, material specification.
These are
_j 2
essentially known for the fabrication materials.
They're nice 3
to know but they don't tell you important information for 4
ultrasonic purposes.
5 What you need to know is the volumetric 6
metallurgical characteristics.
Is the microstructure 7
columnar, is it mixed, is it coarse, equiaxed, fine grained.
8 You don't know the actual thickness as well as the actual 9
material velocity, all which affect the UT or the ultrasonic.
10 testing.
11 Problems associated with determining these unknowns:
12 OD is typically the only accessible surface and there's a full 13 range of volumetric metallurgical possibilities out there la which are not all known.
15 In terms of samples which I don't have down there,-
16 there are increasingly a number of samples becoming available 17 with these different microstructures.
18 Solutions to the knowledge of fabrication materials; to there are programs in development, specifically funded through 2e EPRI, on determining these metallurgical characteristics and l
l 1
21 developing ways of compensating for their effects in terms of l
22 angles, frequencies, things such as that.
l i
In the particular program that I went through I made i
n f
24 four calibration blocks of the material for these -- that l
25 represented some piece of material for the individual samples.
\\
i
p Page 73 i
7< g/
1 There was differences, surprisingly enough even in the forged j
{,j 2
pipe, when you went in two 180 degree directions.cnt. the pipe, 3
the angle shifted from -- by about three to seven degrees.
So a
centrifuga11y cast isn't the only thing that can raise some 5
questions in terms of your angle beam.
You have to know even 6
on forged pipe what can-happen.
It was surprising from the 7
microstructure, you couldn't see why it was affecting it but s
there was definitely a shift when you turn the transducer 9
around 180 degrees.
10 So by having knowledge of the fabrication materials, 11 you can compensate for your examination and perhaps locate 12 your defects more -- better and improve your inspection.
13 In terms of the knowledge of the nature of defects, la the potential service-induced mechanics, chase were is essentially provided by EPRI with their particular program, 16 thermal and mechanical fatigue are potentially -- are 17 potential-mechanisms.
Stress corrosion cracking, a very, very la low probability of that.
This is why we chose to make the 19 samples of thermal and mechanical fatigue defects.
t 20 How are these important?
Well it's nice to know 21 from an ultrasonic point of view the size, position of these, 22 the nature and the orientation of these particular cracks, 23 whether they're axial or whether they're circumferential.
i 2a The next viewgraph shows a few of the typical cracks 25 involved in this particular program.
The top portion shows
Page 74 f
(_j i
thermal and mechanical fatigue cracks.
As you can see -- you 2
can' t see it much on -the right upper one,. but the left upper 3
picture shows a thermal fatigue crack in this one particular sample.
As you can see, it's highly branched, there is an 4
axial component coming out below it and there's a-series of 5l l
axial type cracks and it's also very, very meandering.
6' Whereas in the mechanical fatigue cracks down below, 7
essentially straight, no branches involved.
8 In terms of operator training and experience, the 9
operator should understand refracted longitudinal waves.
10 They're not the sase as conventional shear. There's different 11 modes going on there, reflections, mode conversions of the ID, 12 things that the operator must understand and must be fully cognizant of.
Understand material effects on beam propagation, is knowledge of some of the programs that are out on -- like Dr.
16 Doctor's on the beam profile as it goes through the 17 microstructure.
Knowing about beam skewing and beam 13 distortions.
19 Understanding the limitations of the particular 20 event in terms of sizing, locating problems that may exist or 21 perhaps can be compensated for after detection of indications.
And probably most importantly, experience at 23 practicing UT procedure on cracked samples.
Many of the 2s operators haven't seen cracked samples and it's very difficult 25 t
Pega 75 7']
for them to recognize the echo-dynamic patterns that exist 14 I
unless they see something like that.
They can.also build 2
confidence in their procedure and themselves also by looking 3
at particular samples with cracks and finding out that they 4
don't typically look like side drill holes or notches in most 5
cases.
6 And lastly would be a demonstration of such skills, 7
where they might have blind tests or whatever. But I believe B
in this particular case that practicing on cracked samples is 9
probably the most important in terms of operator training and 10 experience.
11 Understanding the sound beam propagation mechanism.
12 l
Here you have beam distortion which is essentially disintegration of the beam cross-section.
You may have beam splitting, two beams at perhaps different angles or positions, 15 au Dr. Doctor showed.
Beam skewing, you hava a deviation of 16 the beam from predicted.
These are all effects, they could 17 occur individually or in conjunction with each other.
Is And lastly, the selection of ultrasonic test parameters and procedures.
In the inspection program that I 20 I
went through, I must limit it because I'm basically the only 21 one that did the examination so we don't have a round robin 22 study or anything like that.
What I tried to do was work 2:'
through the back door per se, take the UT technique, I knew i
24 where the cracks were, try to develop a sensitivity based on 25
Pegn 76 those cracks and more or less work backward in the operation.
In terms of forged stainless steel components, I 2
found that false echo and transmit / receive probes, shear 3
weight probes, 45 and 60, were very effective.
All thermal 4
and mechanical cracks were detected getting up to signal to 5
noise differences of 20 dB or greater.
6 The better detected cracks were the near-side 7
cracks.
In other worcis, in the forged pipe itself looking 8
from the forged pipe.
The worse case were the far-side 9
cracks. As you would expect, you get more noise associated 10 with that becsuse now you have a shear wave going through the 11 weld into the, in this particular case a statically cast 12 forging.
The cracks though were detected, you did have to put
()
up with the interface problem there.
There you do have a continuous signal from the base metal to statically cast elbow 15 interface.
16 In terms of reporting sensitivities in this 17 particular case, if you looked at the signals from the cracks 18 with respect to side drilled holes and notches, 50% DAC will 17 not find some of the cracks, you had to go further down in 20 reporting levels.
In fact, for the false echo 45 degree, 1 21 think it was something on the order of 12 dB below the side 22 drilled hole response that you had to go down to in order to 23 detect all cracks.
And this was an average value.
So there 24 was some above and some bslow.
Most of those that required 25
i Pega 77 7-'s the extra sensitivity were again those on the far side of the 3_)
I weld during the statically cast elbow.
?
In terms of lengths and depths for the forged 3
stainless steel components, length sizing using 50% -- I used a
50% half backs which is close to the 50% DAC as used in the 5
field, busically undersized in all cases.
6 But in this arena you have the capabilities of doing 7
more substantial dB drop sizing, down to 12 dB, 14 dB.
You 8
can get down further.
The high signal te noise ratios of 20 9
dB allows you to do this.
10 In terms of depth sizing, fracture to fracture would 11 I
probably be used in this case.
I didn't try it, I limited 12 myself only to dD drop-and depth in. terms-of dB drop or llg) amplitude drop again is undersize typically.
As I said beforo, on the forged, I did.get an angle 15 shift going from one axial direction to the othou-of 3 to 7 16 degrees, so this should be looked at.
And you must know --
17 granted you're on forged pipe and you think well I'm going in 13 at a 45 degree angle, well it could be a 42 or a 41, so you 19 have to know your angle of that material.
20 THE REPORTER:
I'm sorry, I can't hear you back 21 here.
HR. RICHEL:
Oh, I'm sorry.
On centrifuga11y cast 22 and statically cast stainless steel components, shear wave is 24 really impractical.
Due to the literature search I 25 till r
l Pcgs 78 p)
concentrated basically on 45 refractor longitudinal waves, b 2f I
The ones that worked the best were the 45 refractor 2
longitudinal waves frequencies of.75 to one megahertz.
These 3
are transmit / receive units and they were focused near the ID J
surface.
They were successful in detecting both thermal and 5
mechanical fatigue cracks although they did not detect ' hem 6
all.
7 The biggest problem, and if you remember tha 8
previous presentation on the POP weld was the overlay side of 9
that particular weld where there was a weld overlay on the 10 pipe side.
Inspecting from that side gave us the worst 11 results and that's basically what dropped most of the 12 transducers in terms of their percentage of detection.
(
)
I found that the better detections were gathered s-y from the statically cast side of the weld.
The.75 megahertz 15 16 DR. SHEWMON:
Does that mean where the defects on 17 the statically cast material were easier to detect than those le in the centrifuga11y cast, is that what you're saying?
MR. RISHEL:
What I mean by that is the -- I'm 20 scanning from the statically cast side, so all exams from the l
21 statically cast side were better than those exams that were 22 from the pipe side or the centrifuga11y cust, 23 DR. SHEWMON:
but that would have been true if you i
l 22 had started from the centrifuga11y cast side too.
Then the 25 '
I
l Paga 79 i
7)
cast would have been easier -- the centrifuga11y cast defects 14 I
would have been easier to find?
2 MR. RISHEL:
No, this includes both far side and 3
near side welds.
So when I say all examinations from the 4
statically cast side I'm talking about transducer locations, 5
not crack location.
6 DR. SHEWMON:
You've also said that it's easier to 7
look on the near side and not the far side.
8 MR. RILHEL:
That's core.oct, in the forged.
9 DR. SHEWMON:
I'm not sure yet what you're telling to me about the statically cast material, when that's -- whether 11 that's always easier than the centrifuga11y cast material.
Is 12 that a fair statement, easier to find the cracks?
33
(
-MR.
RISHEL:
In this particular case, yes, I found g
15 l lt much easier.
DR. SHEWMON:
Okay, fine.
g MR. RISHEL:
Where I talk about the near side and 3,
far side, I should make this point, was in the forged 33 stainless steel, I found that there was a difference.
3, In terms of looking at the attenuation from the 3
statically cast -- when scanning from the statically cast side g
or the centrifuga11y cast side, there wasn't really a g
l distinction between signal amplitudes, whether the crack was g
on the near side of the weld or the far side of the weld.
g DR. SHEWMON:
Okay.
25 l
L
Page 80 j
7')
12f 1
MR. RISHEL:
Also using this particular unit, it 2
seemed that the mechanical fatigue cracks were more difficult 3
to detect in this particular case.
4 DR. SHEWMON:
More difficult than thermally?
5 MR. RISHEL:
That's correct.
But as I said before, 6
the attenuation or the level of response were more or less the 7
same.
8 In terms of just a quick percentage of the number of 9
cracks that were found with respect to the total crack 10 population, on the statically cast stainless steel using a.75 11 megahertz.two element unit, about a 94% inspection -- it was 12 able to detect 94% of the cracks in the blocks.
Whereas for
-w 13 the centrifuga11y cast it averaged around 87%, so they're
\\~')
14 relatively closu.
15 In terms of the responses with respect to the notch 16 calibration, the notch in the calibration block, they were 17 wel. within the 6 dB reporting level.
18 Okay, the safe end nozzle welds, in this particular 19 case I primarily emphasized ID examination.
This particular D
veld is accessible by reactor vessel inspection tools, so I 21 looked at it from the standpoint well we should apply the best 22 technique that we know is available.
So I applied a contact 23 70 degree L transmit / receive, two megahertz unit and found that 24 all cracks in the safe end nozzle welds were detected, getting 25 signal to noise differences of greater than 27 dB, very, very I
-4 Page 81 TD i_J' 1
little noise, that's typically associated with bi-metallic 2
welding.
3 Again we're talking about in terms of using a 6 dB d
drop technique for link sizing and underestimation of size, 5
but the 27 dB signal to noise difference gives you the-6 opportunity to go down further to 12 dB, 14 dB.
In this 7
particular case 12 dB performsd much better.for sizing.the 8
lengths.
And depths, again a 6.dB. underestimated -- typically 9
underestimated the size.
10 In this particular case I would recommend crack tip 11 sizing from the ID probably could work but you may have a 12 little difficulty.with the bi-metallic weld and seeing some 13 noise from that interface.
O-14 Some of the things that I'm recommending -- a lot of 15 it may be opinion, what I learned from this program.
In terms 16 of1 probes, the dual element probes, 45 degree longitudinal 17 dual element, one magahertz probes are large, they're roughly 18 two inches by two inches.
That's a very large footprint which 19 requires a large surface prepared area on the pipe or the 20 elbow.
Because of its large footprint, any surface 21 Irregularities could cause coupling problems, so liberal use
??
of couplant is necessary.
And you should watch out on the way 2:-
this couplant is applied because I found that if you don't l
24 have couplant under the centur portion of the beam, you don't 25 see the crack, whereas if you do have it on the center portion AV i
I L
Page 82
-y m I
of the beam you do see the crack, but the noise level on the 2
screen has not changed.
So it's something that has to be 3
watched for-in scanning blocks or in the field itself.
4 In. terms of test sensitivity, side drilled holes and 5
notches are sometimes not sufficient when you're talking about 6
a 6 dB drop technique.
You have attenuation losses perhaps 7
due to the differences between calibration block and the 8
component, perhaps within the component itself.
I found that-
?
the best method was just to run at a 54 to 10% noise level on 10 the screen, record things that are greater than two to one and 11 have some length to them.
If a crack is there you're going to 12 see it.
If it isn't and it's in the noise-level, then you 13 won't see it.
A manual operator cannot look into a noise 14 level and reliably see something in there.
It either has to 15 appear above the noise level -- that's when you'13 find it.
16 You may increase the number of reflectors that have to be 17 evaluated but your probability of finding a defect increases.
18 And again I come with the hands on training.
I must 19 emphasize this because some of the operators I now, and 20 there's operators out there that just have not used procedure i
21 on crack samples.
They should be trained on these particular l
22 samples, let them look at them, let them see the echo-dynamic 23 responses from geometrical reflectors, metallurgical l
24 reflectors, cracks, side drilled holes, notches and see if I
25 they can see the difference.
Sometimes you can't see the i
Page 83 YD h_2[
1 difference between them.
2 In terms of sizing methodologies, amplitude drop are 3
not totally sufficient for depth but for length they are.
4 Although in the forged samples and using the contact-methods 5
from the ID, you can go to smaller dB drop or greater dB drop 6
techniques.
In terms of cast pipe or statically cast and.
7 centrifugally cast pipe, you really can't go down more than 6 8
dB because you're talking about signal to noise ratios on the 9
order of 6 to 9 dB.
So once you get a crack. signal that gets 10 near tha noise level, you have a very difficult time reliably Il telling which one is the crack and which one is the noise.
12 Since -- but the lengths are a.better-estimation 13 than the depth, although there is a tendency for undersizing 14 to perhaps compensate for this.
15 In terms of depth sizing though, perhaps we may be 16 better off just taking a length to depth ratio.
We make 17 blocks in the laboratory to be a certain depth based on the la length.
If you can assume a length to depth ratio in the 19 field where you know the length better than you do the depth, 20 perhaps this is the bere way of sizing such as this until more 21 advanced techniques such as automated systems are available.
22 And lastly automated data recording, processing and 22 analysis systems.
I think this in probably the way to go in 24 terms of providing further improvements, greater signal to 25 noise ratios, greater than the manual techniques.
Some kind
[
~
Page 84 e$
/
s_d 1
of processing perhaps to filter some of the noise out, some 2
analytical software will aid in further improving the results 3
over and beyond the manual te2chniques.
4 And this leads me to my last viewgraph where we're 5
basically looking at the future improvement in inspections of 6
main coolant loop piping inspection.
Manual techniques are 7
available which can find cracks, thermal and mechanical.
If 8
you want to go further and perhaps find smaller cracks or you 9
want to improve the signal to noise ratios then you have to go 10 to the last segment here which would be automated data 11 recording and processing.
12 And I might just want to add a few comments on the
<~
13 results of a vendor demonstration program that Union Electric 14 of St. Louis, Missouri, put together.
They brought in some 15 different vendors to look at eight particular samples that we 16 shipped out to them.I believe last year sometime.
They brought 17 in vendors, Westinghouse reviewed the results.
We didn't know l
18 who the vendors, that wes kept from us.
Of the eight 19 particular cracks there were three particular groups th?
20 detected all of them.
These were masked tests so that again l
<1 emphasizes the fact that manual techniques can work if applied l
22 properly looxing at all the parameters involved.
l l
23 DR. SHEWMON:
Could you help me on what -- what I'd l
24 like to talk about some is the spread of materials out in the 25 field and the degree to which your results would depend on i
l
'l
.h
,,1
'i Page 85 t- -
1 that'.
There was -- theire apparently is some test that must *-
f, si 1,
e 2
presumably is run by the licensee on new piping when they came
's q
7
(,
/'
3 back to the NRC and talk abc.ut what inspections they will do.
[4 And at leapt some of this -- And )he Braidwood-Byron set was g,
that sticks in my head.
Word cam \\;l beck and said basically f1,
') L
^
q vds.hdttdoanAnQectionbecauds[gwe c4.n'?'cet a signal
'4 6J
/,g' w
1 q
- thrgpgh, and this 'cisused waverifor awhile.
' 3 [g 7,
T i
I 1 lCan'you help.altraighten me out on what I do remember 7
or should b Is'aying about what is the test th at the NRC
}
lrequiresthevendororlicenseetorunandwhatfractionof j
N
\\.
this and then whend are we, what r
13 the* plants do or don't pass 12 can they do if it doesn' t?
13 4
'M MR. RISHEL:
One, I don't know what tests the NRC 3
fi c
requd3s be performed.
I know they've,done in the past fuel 14 worm \\ 'ased on zero degree,L, whether sound can~get in the
-b
}
15
\\
I 16 material,. based on a straight beam.
t i
17
(
TR. SHEWMON.* Can somebody help me?
Maybe I have my l
i i
IB story mi:cmd but I didr.' t think so.
19 DR.
BEHRAV*?AH:
We have included some presentation I
k
\\onsove.of those tests of at le dt what we did at Braidwood.
t X
't
't l
(
N 21
(
DR. SHEWM3 / Gk.ny, J.t was the Braidwood site?
e
\\
/
7, st DR. Bh.HRA, WS
- 'Braidwoo6elte.
It was work that
\\
j % }vas done a% the Tro,,)n site.
l a
l P
x s,N('
,DR.
SHEWMON:
And,th q' Cwere particularly hard to
\\
s insp74t is tt why -mu got involved, or -l 25 f
4
(
9.f 1,
c,
{
.s m
'. },
Ne
\\
j
,1 g b'
41 4
\\
\\
l
. )
~
ear b
,(..
y
Page 86 g
[_j' 1
DR. BEHRAVESH:
We were asked to go in there so it 2
is correct that there was problemmatic areas.
3 DR. SHEWMON:
And what standard test or survey of 4
the piping then brought this to light?
Was this something 5
that is done on each class of pipe when it comes into service 6
or why didn't we ignore it and go on?
Yeah?
7 MR. LANCE:
Maybe I can help you a little bit.
8 DR. SHEWMON:
Would you identify yourself?
?
MR. LANCE:
Oh, I'm sorry.
My name is Jack Lance.
10 He had during the licensing of the Seabrook stations 11 requirements to shov. that we could inspect certain piping 12 systems or ask for waivers againut those inspections.
It was fm, 13 pretty much accepted that the ferretic steel and the wrought
(
]
14 stainless steel systems were not a problem and therefore we 15 didn't have to do any demonstration, but on the cast stainless 16 steel inspections we had to develop a program within the 17 licensing arena or for oar licensing submittal and then M
successfully demonstrate that we could inspect the cast li stainless steel main coolant piping to some satisfactory
- D level.
It was not Section 11 criteria but I believe we 2' h finally settled on something that was being able to detect j something on the order of 30% through-wall block or crack.
22 I
2:'
DR. SHERMON:
And how do you do this in unflawed i
2:
piping?
l MR. LANCE:
Well we had some folks from PHL come in 25
page 87 b
I as consultants, a group of them, I think approximately 12 or s_;
2 14 NRC folks and the people from PNL came in with standards or 3
with samples.
One in partJcular that had an equiaxed 4
structure on one side and a columnar structure on the other.
5 And the inspectors and~the techniques were blind -- I don't --
6 I guess they were blind tested, it certainly wasn't a 7
qualification test.
8 Then we went out on the plant and ue showed that we 9
had siellar attenuation on things like counter-bores, weld 10 roots, both through angle beam where we could get it and il straight beam attenuation.
And we convinced ourselves as well 12 as the regulators that we were involved in a program of 13 similarities.
\\'
14 DR. SHEWMON:
Is this something required on all new 15 plants?
16 MR. CHENG:
Yeah, on --
17 DR. SHEWMON:
His name is Simon Cheng.
18 MR. CHENG:
About four or five years ago I think we 19 start requiring those demonstrations for the NTOL plants.
I 20 think what Jack was talking about was Seabrook, one of the 21 NTOL, including the Braidwood.
khat we had done at that time 22 is we required the licensee, the applicant, to demonstrate on 2 :-
their pipe that at least they can penetrate through their pipe 24 and then get back reflection and perhaps, as provided by PNL, 25 they can detect flaws of mechanical fatigue or maybe thermal
(
Page 88
.L_J 1
fatigus.
We considered that one is acceptable because in 2
future certainly they can penetrate the pipe compared to some 3
of the older plants where they cannot penetrate the pipe.
4 DR. SHEWMON:
And what fraction -- over that five 5
year span, what fraction of the pipe have you had to grant 6
waivers through because they couldn't go through it?
7 MR. CHENG:
I think they went-through almost every 8
one.
I couldn't answer how many plants we granted waivers.
9 DR. SHEWMON:
But we don't know -- it wasn't 90% but 10 was it 10% or --
11 MR. CHENG:
I think.most of them could demonstrate.
12 DR. SHEWMON:
Okay, all right.
13 MR. RISHEL:
Just to add, some of that demonstration 14 is done by looking at counter-bores and things such as that, 15 using the angle beam but in not all places you can detect 16 counter-bores, so sometimes the angle beam is not useful in 17 determining whether you can get through it or not because you.
18 don't have a reflector on the back side.
And that's where 17 there's some difficulty.
Even zero degree monetimes can be a 20 little -- if you don't have non-parallel surfaces or something 21 like that, but in most cases a good angle beam examination to 22 determine that would be to look for counter-bores.
23 DR. SHEWMON:
Okay, thank you.
24 Mohamad.
l 25 PRESENTATION BY DR. MOHAMAD BEHRAVESH
_-.. _,. _.. _... _. _, - _ _..,,.,,,, -,,...., ~
7,,._,.___
Page 89 1
DR. BEHRAVESH:
I am Mohamad Behravesh from EPRI.
I 2
have a lot of sympathy for this gentleman sitting here, that 3
he can't hear.
In a different life I used to do something 4
similar to what he does, so for most part I feel I must talk 5
to him.
6 DR. SHEWMON:
Okay.
7 DR. BEHF,AVESH :
But in any event --
8 MR. CURTIS:
It's the Southern accent that gets hin 9
though.
10 (Laughter.)
il DR. BEHRAVESH:
Some four years ago when we started 12 on this activity, it really was presented to us as morething 13 insurmountable, and in the process we have really been quite 4
e 14 successful in meeting the cha31enge for the most part and-in 15 fact been able to advance our understanding of the fundamental-16 processes that take place here.
17 But more than that, we have been successful in 18 trying what we have learned in the field.
As my presentation 19 continues, you will see examples of that.
But before I get 20 into that, I want to give you a background on what EPRI's 21 overall program dealing with cast material includes and what 22 it involves.
23 The program is that of a component reliability and 24 that's managed by Gary Dau.
The questions are very general.
25 I'll go over them.
The general questions are when and under
Page 90 L_;
I what conditions do the properties of cast material make it 2
potentially limiting to be used as a piping-material in a 3
plant.
For example, what are the flaw sizes of concern and 4
establishing in-service inspection requirements.
As Al Curtis 5
mentioned earlier this morning, there is a number of people 6
who really believe that the inspection requirements may be too 7
stringent as they are currently.
8 Also, identification and possible extent of in-9 service piping degradation mechanisms.
And finally, coming up 10 with answers to are there adequate and demonstrable NDE 11 techniques for inspection of-this material.
12 The remainder of the. talk today will concentrate.on 13 the last bullet.
I want you to be aware of the other bullets i
14 because of the work that's being done on the structural 15 mechanics program.
16 But, several presentations have been made this 17 morning, I think it is a good place to present to you at least 13 what our understanding of what light water reactor experience 19 is with this type of material.
20 Of all the information we have gathered to date 21 tells us that this material has been basically trouble-free in
[
l 22 the PWR service.
Both cast stainless steel as well as --
23 centrifuga11y cast as well as statically cast components are 24 susceptible to long-term ductility loss.
That has not been a 25 secret, but the other thing is that even aged pipe material
Page 91 1
has been shown to be tolerant of significant flaws under 2
design loading.
This is all information that has been 3
gathered.
More importantly issues of stress. corrosion 4
cracking has been raised, the information that is given to um 5
is that.intergranular stress corrosion cracking and 6
interdendritic stress corrosion cracking really are not a 7
likely damage mechanism in cast material under PWR operating a
conditions.
9 And finally, flaws in cast material and welding.
10 defects are most likely areas of fatigue initiations and from 11 the limited information that exists on fabrication of this 12 material, we'll see that weld repairs during manufacturing and 13 installations are (1) very common, but more so the control and 14
-documentation of these repairs are quite scarce and not 15 adequate.
l 16 I would like to go and present to you some of the 17 elements of the programs we have at EPRI that will address the i
l 18 inspection of cast material.
l 19 MR. WARD:
Mohamad, could I go back to your last 20 comment?
21 DR. BEHRAVESH:
Sure.
22 MR. WARD:
That weld repairs are common and they're l
23 not well documented.
What's the significance of that?
24 DR. BEHRAVESH:
Well if you were to look at a place 25 that may be most likely to degrade or to have a flaw
I l
Page 92 l
-m
)
1 initiated, perhaps it would be in these repair locations.
x_;
2 MR. MARD:
Is there some experience that indicates 3
that or is that just common sense?
4 DR. BEHRAVESH:
I think combination of both.
5 MR. WARD:
But I mean you've said that --
6 DR. SHEWMON:
These are in static castings?
7 MR. WARD:
Yeah.
8 DR. SHEWMON:
Well static castings often have 9
porosity in them and you chew out what you have to to replace 10 it with sounder metal, but if there was for example more 11 porosity there, that could be a place where a fatigue crack, 12 except they're so over-designed you wouldn't expect it, but 13 the reason they did repair there wat because there were
,f 3
( /
14 weaknesses.
15 MR. CURTIS:
We have not experienced any flaws that 16 led to leakage in any of these repair areas.
17 MR. WARD:
Well that's what I was driving at.
Your 13 first comment is, you know, that you've had trouble pre-19 service, and I guesa that includes this sort of thing.
20 DR. BEHRAVESH:
Exactly, yes.
21 MR. WARD:
Okay.
22 DR. BEHRAVESH:
Now to go back and present to you 23 some of the elements of the program we have at EPRI that is 24 designed to addrecc the inspection of cast material, we have lworkonuseofwavescatteringmodals to determine the 25 i!
m Page 93 7~'N
$_J#
1 dominant grain structure in this material.
Everyone has given 2
you information how important that is.
The reason for that is 3
to be able to help with selection of the model of the sound 4
propagation that you use, whether it be shear or longitudinal, y
5 We need to find what are the most appropriate inspection 6
angles.
That comes also from knowing the structure.
There-7 are artifact arguments in there that has to do with probe 8
angles and how to come up with minimal side lobes.
There is a 9
lateral resolution argument that has to do with the width of 10 the beam, 11 And to get a handle on any of these things, you need 12 to know not only the structure but how that structure 13 influences the sound that propagateu to it.
That has been --
14 that's an ongoing program.
We know far more about this 15 subject today than we knew four years ago, but we certainly 16 are not there completely.
That is, our understanding is far 17 from complete.
18 We have been using Rayleigh and Lamb waves to detect 19 deep cracks, particularly those that may propagate close to 20 20% -- to 20% of the outer wall if ever such a thing becomes l
21 problemoatic.
Rayleigh waves can be used for an ID 22 inspection.. as Rick mentioned to you in the nozzle casn if you 23 can get inside as well as on the outside of the pipe.
And i
24 also we have had modeling of ultrasonic beam to tell us what 25,happens in anisotropic material and how it affects the crack
(
Page 94 1
and-echos that we get from the cracks.
2 These are soro of the fundamental studies that are 3
ongoing.
4 From the outset, we knew that a lot has been learned 5
and developed as part of the BWR inspection technology, so we 6
have been trying to adopt most of what we have learned from 7
that and to use it in this area.
For example, we have had B
several inspection systems that'have been quite successful, 9
they are commercially available and we have been putting them 10 to use on this problem.
You will this afternoon as you go in 11 the high bay, you will see a demonstration of this system that 12 was basically developed for BWR inspection, it's called 13 intraspect, it's commercially available, has been used in a 14 lot of other fields besides NDE.
15 We have done considerable work in using ultrasonic 16 feature e.nalysis; that is, looking at a signature of a flaw 17 and extracting features from it and trying to understand from 18 those features what are the flaw characteristics.
And this 19 ties in with signal processing and actually there is hardware 20 out there in the field that are no more complicated than what 21 you see here.
This is an entire system that can get a 22 signature from a flaw, process it and give you far more 23 information than was available before.
You are basically 24 looking at a compact PC with a pulser receiver board and a 25 transducer that is coming and getting the result and you can
Page 95
- _j 1
do the entire analysis on that.
These are all commercially 2
available and are being used and you see some examples of them 3
this afternoon.
4 We have done considerable work in characterization 5
of this material and as a result being able to optimize some 6
of the parameters.
You will see more of this presented, and 7
also field application of technology for both cases of pre-a service and in-service inspection und capability 9
demonstration.
10 Now more details on all of these will be presented 11 to you next by Frank Ar.mirato of the NDE Center, who will be 12 giving you details of a lot of these because these are at the 13 heart of our activity.
,s
)
14 In summary, to give you a snapshot of where we are, 15 I believe that our experience with this material is still 16 limited.
We know far more than we did before but it is still 17 limited, but is improving fast.
We are getting lots of good IB information which is helping us.
M We see all kinds of variations in characteristics of
,this material, from plant to plant, from material to material, I
lfromcomponent to component or even along the same component.
?t i So that should be no secret that what you know that works i
2'
!here, there is no guarantee that it will work in the next i
] place.
24 25L We now know that we need to have very good reference a
n
o Page 96 1
mater 2a1 in order to be able to see -- to determine what 2
-sensitivity we nesd and it is -- proper reference material is 3
essential for the calibration and inspection of this material.
4 As I mentioned, an a priori knowledge of the 5
material is very necessary in order to optimize the parameters 6
and most of our work now and in the months and years to come 7
will concentrate on characterizing this material before we 8
attempt to test it.
The more we know about the specifics of 9
this material, the better chance we have in doing the credible 10 examinations.
Not knowing the material characteristics is 11 almost like walking into a dark room and attempting to see 12 what you can find.
13 The information that we have to date -- and I should O
14 emphasize that all the information that we have to date is 15 limited to the samples we have worked with.
So on the basis 16 of samples that we have, we are finding out detection 17 sensitivities of between -- good detection sensitivities exist la for flaws that are somewhere between 10 to 404 through-wall, 17 and that can be readily demonstrated.
20 DR. SHEWMON:
When you talk about characterizing the l
21 material non-destructively, do you have any techniques aside from ultrasonic probes as you go into this dark room?
22 DR. DEHRAVESH:
Not quite yet, no, we don't.
We l
24 still like to make some ultrasonic measurements th d Will tail l
25 us about the material properties rather than whether there is
i Page 97 I
a flaw in there or not.
a_;
2 And also, most of the work that has been done are 3
now published in five EPRI reports that I list in here but you 4
have them in your handouts.
I have included the front page of 5
these reports in your handout so you can get a glimpse of what 6
the reports are about and what is the concentration of-them.
7 So at this time I would like to turn it over to 8
Frank Ammirato of the NDE Center to give you details of the 9
work that is done and pretty much set the stage for some of 10 the experimental work that you will see this afternoon.
11 PRESENTATION BY FRANK AMMIRATO 12 MR. AMMIRATO:
Thank you, Mohamad.
13 My outline this morning,-I'll very briefly go over f
14 the background that has been covered already, I won't dwell on 15 it.
I'll talk about the NDE Center activities, some 16 theoretical and experimental work done here to try to 17 understand wave propagation in cast stainleus steel.
I'll 18 talk about signal processing efforts to improve the quality of 19 NDE data, the sample acquisition and characterization, 20 particularly the Westinghouse Owners' Group samples that were 21 just made available to us last year.
I'll talk about some l
22 field trials that we've done over the last three years and l
23 then I'll talk about the demonstration that you're going to l
24 see this afturnoon.
t f
25 A little bit of an overview:
The objective of NDE i
l f"
l l
Page 98 V
1 activities here at the-Center for cast stainless steel is t
2 really two-pronged.
One is to improve the effectiveness of 3
NDE, but in order to dc that you really have to be able to 4
evaluate the capability of NDE.
If you make an improvement, 5
you have to be able to measure what you did to make the 6
improvement, particularly the influences of individual joint 7-characteristics.
We've heard several times this morning that 8
that's very important, each joint is quite different and its 9
influence on NDE is quite distinct.
10 It's a difficult problem as we all know.
There's no 11 general solution, NDE solution.
By that I mean there's no one 12 fixed procedure that works in every case.
You have to know a 13 lot about the particular joint.
^ O 4
14 Some results I'll show you later on I think will-15 bear out that NDE can be effective in some specific kinds of 16 grain structures and some specAfic kinds of joints.
17 l
-The approach here at the Center can be characterized l
18 as three-pronged; theoretical and experimental work to try to 19 understand both how waves travel through cast stainless, how 1
20 you can use that information to figure out what the grain l
i 21 structure is and once you do that, pick out the best technique l
l 22 for the grain structure.
j l
l 23 Signal processing and pattern recognition to improve I
l 24 the quality of the data.
A lot of the data that you see from 25 the field is noisy, difficult to interpret.
Signal processing
l Page 99 I
can in some cases improve that and I've got rose examples of s_;
2 that.
3 Field trials are very important.
What works in the 4
lab doesn't always work in the field and furthermore you learn 5
a lot by going out to the field to find out what is the real 6
situation.
I'll talk about those too.
7 you've already heard quite a bit about what happens 8
in cast stainless, beam skew, distortion, attenuation.
There 9
has been theoretical work done here at the Center and also 10 other EPRI contractors to try to understand these effects.
II And each grain structure is very specific and we want to try 12 to use that specific effect to try to identify the grain 13 structure from measurements.
If you know the grain structure, 14 maybe you can have a compensation technique to correct your 15 data and I'll talk about that a little bit later.
Ray tracing 16 is a useful example to take some of this knowledge and try to 17 predict how the beam passes, behaving in the weldsent.
And I l
18 have an example of that also.
l l
19 On the experimental side, we started about in 1985 l
20 some basic measurements of attenuation, velocity, beam skew, l
21 really trying to understand how bad the problem was.
We 22 worked a little bit with detection of machined reflectors but 23 starting last year the Westinghouse Owners' Group samples l
l 24 became available and that gave us a chance to work with cracks l
l 25 in a large variety of grain structures, geometries, O
I
Page 100 7"
1 configurations and so forth.
You've heard about them already.
na 2
We've used them here at the Center for several things.
3 Transducer optimization; again, if we know the grain structure 4
we can optimize the transducer but you need to do a lot of 5
work to figure out what is the best combination of techniques 6
for that configuration.
Excellent signal processing test bed.
7 You can try lots of candidate signal processing procedures and 8
see what happens.
Lately we've been going through our results 9
and trying to come up with some detection performance data, 10 how well did we detect each kind of crack and each kind of 11 grain structure.
I've got some preliminary results I can talk 12 about a little bit later.
13 These are typical joints.
This is an example of the 14 ray tracing that I'm talking about and t?'s is a model 15 developed by Dr. Jung here at the Center just to illustrate 16 what happens in a complex joint.
This is one of the nozzle to 17 safe end to pipe specimens from the Westinghouse Owners' Group 1B samples and this is a calculation that Dr. Jung did, and each 19 grain direction is represented by these little short straight 20 lines and at each point the beam deflection is calculated, at 21 least the new velocity is calculated and what you see is you 22 think the beam would go this direction, it doesn't, it goes 23 someplace else.
So this is an illustration of what was I
24 mentioned earlier, the beam doesn't go where you think it's l going.
25
~
Page 101 1
DR. SHEWMON:
Do you use Snell's law or something to 2
may they'll always bend the same way instead of some bending 3
the opposite way?
4 MR. AMMIRATO:
Well they bend according.to the 5
slastic constant at that particular point and that is 6
determined by the grain orientation, and that's what you have 7
to know.
8 DR. SHEWMON:
You show a fair amount of rotation 9
inside that V-shaped gray area but it's always clockwise.
10 MR. AMMIRATO:
It depends on the wave mode,-depends 11 on the wave mode and on the horizontal shear wave it bends the 12 other way.
13 DR. SHEWMON:
Okay.
O 14 MR. AMMIRATO:
This is three but they're all 15 different.
16 This is an example of location errors.
We talked
)
4 17 about beam skew and beam distortion.
This is a simple i
1 18 experiment, a side drilled hole and it-was located by just a l
l 19 conventional angle beam at two points.
The calculated point l
20 was over here for this case and the calculated point for this l
l 21 was over here, there was considerable error.
But if you knew, l
I 22 again the grain orientation, you could correct that data.
l t
?
23 I'll have some examples of how that can be done later.
24 Location is not the only problem of beam 25 redirection, it's noise.
We have lots of samples of noise.
A
.e
-,,n
,,,r-_me
,-w
,,n,n.--,,,
w w ww-p
.,,,_--,,m-
-e
Page 102 1
Over here we have detection of a simple side drilled hole
_a 2
target.
In carbon steel and forged stainless steel there's 3
not much difficult at all, very strong signals.
If you go to 4
-- one of the worst cases, centrifuga11y cast course grain, 5
you see the signal to noise ratio is not as good.
In fact 6
there's a factor of eight difference in gain from here to here 7
and we still haven't really sharply detected that drilled 8
hole.
9 More grain structures.
I think it was mentioned to earlier that this is really a tough one, the mixed kind, columnar grain and a rather sphere sharp layered boundary, but 12 all of these except for this are represented in the I3 Westinghouse specimens, columnar and equiaxed, fine equiaxed O'
14 of course we have.
15 What we're trying to do is we're trying to make some 16 incremental improvements.
We know it's a tough problem and 17 we're not going to solve it right now today.
Typical IB performance for manual UT might be here -- this ja a crack 17 detection versus false call.
This is the random call line.
l 20 Typical manual guide might be here today and you saw some l
21 examples with Steve Doctor.
We're just trying to go in this l
l l
22 direction.
If you just increase the gain, increase the pulse 23 power, you'ra probably going to go that way, you're going to l
24 increase detections but also false calls, you want to go this
'5 way.
Some of the waves are trying to go that way.
l l
1
+
Page 103 1
Automated systeas, you'll see an example this 2
afternoon during the demonstration where the automated systems 3
just themselves are going to help view a more global picture 4
of your data instead of manually scanning across the sample, 5
you can get individual A-scans.
Just looking at the pattern 6
of those signals might help.
7 Optimize your technique.
If you use a columnar 8
grain structure, maybe you can pick out particular beam angle 9
and frequency that would do the best job.
10 Signal processing, it definitely helps in snee 11 situations.
12 Training, it was mentioned before a lot of operators 13 don't see cracks every day.
These samples are now available O'
14 so now that training is very useful.
And field experience, 15 trying to make small steps in the,right direction.
16 An overall block diagram, and I don't want to go 17 through the steps, I just want to make a few points.
This has 18 to do with defect location, those errors that I showed you 19 earlier where you get the wrong location.
This side has to do 20 with the detection of defects, noise problems.
Both have the i
21 same kind of approach, understanding the grain structure, pick i 22 the best technique for that grain structure and then 23 compensate your technique to give you the best results.
It's P
24 a pretty general approach.
25 Some examples of experimental measurements that have C
Page 104 D]
t bean going on here at the Center over the last few years.
s_
2 There's an EPRI report that Mohamad mentioned that's got.
3 hundreds of these kinds of graphs in it and I'll leave it here 4
if anyone wants to take a look at it.
This is just one 5
exas In an equiaxed grain sample with a side drilled hole 6
and ilng at just a zero degree transducer going along this 7
surface, you can see the amplitude trace and it peaks at the 8
right place, right over the hole.
9 On the columnar grain structure it doesn't do that, 10 it skews over.
Expect a peak here but it peaks over there.
M This was gone through as a function of angle, as a function of 12 frequency, as a function of grain structure.
These kinda of 13 beam redirection and beam skew data was collected.
\\'
id Another parameter is the velocity.
I already 15 mentioned that the velocity changes as a function of angle 16 relative to the grain, so that has to be known in order to 17 make these calculations and corrections.
18 Skew angles are measured and plotted.
You can see 10 for each of the different grain structures, different 20 frequencies, different transducer sizes, different transducer 21 types, it's all collected 1,' detail.
And the reason this is 22 l all done is to make a parameter study.
You want to make a I
23 I parametric repreac2 station of the wave propagation.
That's bow 24 that ray tracing was done.
Given the angle of the grain, you 25 can then predict the
Page 105 1
DR. SHEWMON:
What is the grain size on the static 2
cast, how does it -- where does it fall between your fine 3
grain / coarse grain --
4 HR. AMMIRATO:
Oh, I don't think I have the numbers 5
for you, I really don't know.
6 DR. SHEWMON:
Anybody ever looked at one?
Why is --
7 is it just pure chance that the static cast looks either 8
better than the fine or coarse centrifuga11y?
Is the 9
centrifugal always more mixed or is it -- yeah?
10 MR. JUNG:
The centrifuga11y cast fine grain --
11 DR. SHEWMON:
Identify yourself please.
12 MR. JUNG:
My name is Peter Jung from ND Center.
13 The reason why we made it as a fine grain is although we quote 14 it as a fine grain compared with the other centrifuga11y cast 15 or static cast grain size, but still it is considerably larger 16 than conventionally --
17 DR. SHEWMON:
But my question is a comparison with 13 static cast which should also be pretty big grain size, 17 shouldn't it?
20 MR. JUNG:
yes, static cast that we examined was 21 approximately comparable size with CCSS fine grain in terms of 22 mmount of attenuation or some shape of the grain, et cetera.
23 In that case, it appeared that it was just strictly size of 24 the grain but it is a comparable but we tried to classify it 25 Ion CCSS static cast.
Normally those static cast stainless f~
(,
L--
~.
c Page 106 1
could have some partially mixed type grains.
2 MR. AMMIRATO:
This is an example of using this kind 3
of data to correct location errors.
This is an example of a 4
columnar grain specimen, drill hole in the middle and an 5
experimental measurement of location done all areund the 6
periphery of the sample and that's what these experimental 7
predictions are, that's where you would have predicted the 8
defect to be.
With this parameterized analysis you can then 9
go back and correct these points back to the true location, 10 but you need to know the grain structure and grain 11 orientation, it can be done.
12 We saw some beam plots earlier today and I just want 13 to show you a few more.
This was part of.the experimental 14 measurements to characterize grain structure, what happens in 15 each kind of grain structure as a function of incident angle 16 with the grain.
Here you see a relatively uniform beam, the 17 kind that Steve showed this morning, and here's a rather 13 coarse example of beam splitting, two beams and over here this 19 very severe attenuation.
Again each grain structure has its 20 characteristic, that's what we're trying to find out.
21 Each kind of grain structure has a particular effect 22 in the frequency domain.
Here you see four different 23 frequencies, each kind of grain structure, these are just i
24 frequency spectra met.surements to get an idea how to 25 characterize each kind of grain structure, t
i i
i i
Page 107 This can all be summarized in a table, which I don't 1
2 want to go into the details of but just to mention that we 3
have each kind of grain structure; centrifuga11y cast, carbon 4
steel, all the various parameters, velocity, skew, amplitude, 5
beam profile.
Each entry has a characteristic behsvlor.
6 I'd like to get into now the Westinghouse owners' 7
Group samples, what we've oeen doing with them in the last 8
year.
9 We're using them for transducer studies, defect to detection evaluation.
The defects in these samples range from 11 about 54 of wall thickness up to about 40% of wall thickness 12 and a very large array of configurations, trying to figure out 13 what can be detected.
14 As we mentioned before, signal processing test bed, 15 training.
The possibility later of performance demonstration 16 or capability demonstration.
We would like to get other 17 industry teams to cone in here and work with us with these 13 samples to add to our data base of crack detection and just 19 try to learn some more about it.
2' You'll see this this afternoon, this is all 75 21 specimens laid out in the high bay.
One o' each kind has been 22 selected and put on the table for the demonstration this 23 afternoon, so you'll be able to see some of these effects that 24 I've already talked about in each kind of specimen, right i
25 after lunch.
4 g,
Page 108 I
- 7"g
$_J/
1 Our characterization of the samples was the first 2
job; both phyoical, weld profile, take a photograph of the 3
microstructure, exhaustive manual UT, automated system UT and 4
that's all put in a documentation folder which is kind of a 5
euphemism, it's not very much of a little folder.
All 75 6
specimens are catalogued in here.
This does not include the 7
automated data, that's on magnetic tapes now about this high.
8 I've copied one example packet of docuaentation 9
which I think you can see later this afternoon, if you're 10 interested.
11 The sample description you've already heard about, 12 the different kinds of configurations; forged on one side, 13 static cast on another or static or centrifuga11y cast on one 14 side or the other, 15 Again, this is an example of some of the pictures 16 that are in the documentation folder of the grain structure.
17 We took an etched edge of each specimen, photographed it and IB that's in that folder for each specimen.
I think you've 17 pretty much seen all of these.
20 I This is one of the samples that I showed the ray 1
tracing model done on.
These grain directione uere all 22 measured and then used for that ray tracing calculation.
22 DR. SHEWMON:
What is FGSS?
24 MR. AMMIRATO:
Forged stainless or extruded.
25 DR. SHEWMON:
And the transition is a weld then?
/(
1 f
I i
Page 100 s
7"}
1 MR. AMMIRATO:
Carbon steel nozzle weld, forged 1s 2
stainless steel grain, weld, und then the pipe --
3 centrifuga11y cast pipe.
4 Some of the specimens have cracks on this weld, some 5
of the specimens have cracks in this weld.
So there's lot of 6
different opportunities.
7 MR. WARD:
When you talk about that pipe that's B
extruded, is that really extruded to final dimensions or is it 9
extruded --
10 MR. AMMIRATO:
I really don't know.
t il MR. CURTIS:
I believe it's extruded to final 12 dimensions, it's just cleaned out.
It comes up nice.
13 DR. SHEWMON:
Is it a two foot diameter extrusion?
O 14 MR, CURTIS:
Yeah, it's thick.
15 DR. SHEWMON:
It's a big guy.
16 MR. CURTIS:
It's big.
17 MR. AMMIRATO:
This is a reproduction of the cover IS sheet that's in this documentation folder for each specimen.
17 A photograph, a sketch of 11gu3d penetrant result, a
?u photograph of liquid penetraut result, a typical automated UT 21 scan of a crack, an etch of the specimen, some typical manual 22 UT signals from the crack.
23 DR. SHEWMON:
You point at that typical automated UT l
24 there. how -- does that show up on somebody's CRT or what?
i 25 MR. AMMIRATO:
No, this is a processed image.
This
(
c
l l
Page 110 1
is looking at the crack, this is the-transducer position X and 2
Y, sort of a plan view.
I'll show you other displays like l
3 this, but this is the crack here.
i 4
DR. SHEWMON:
I've lead such an under-privileged i
5 life that I've never even seen a typical one like that.
6 MR. CURTIS:
Made your day.
7 MR. AMMIRATO:
It's a typical picture you'll find in t
8 this book.
I'll whow you lots more of those.
By the end of 9
the day, everything will be typical.
10 What we would like to do -- don't know if we can --
l 11 is try to get these crack detection rate curves.
We're pretty 12 sure it's going to depend.on the microstructure, forged is 13 going to be easiest, static next and centrifuga11y cast maybe O
14 somewhere down there.
But what you're going to see in the 15 rest of my presentation is just composite results for various 16 techniques.
We've tried two, three, sometimes four and five l
- 17 techniques on one specimen and our results are going to be 2
4 composite data.
A3mo we have the data to do a technique 19 specific, but we haven't done that yet.
We also haven't found 20 very much correlation with crack depth yet, as I'll show you.
21 We measured crack length very -imply, just when it 22 exceeded the noise level, that's where we started counting and l
23 when it dropped back we lost it.
We scanned the entire l
i 24 specimen and just recorded the coordinates.
25 For the truth in our evaluations, we used the depths
[
1
--.--.~-,-,-.-4.-w-,---,.
,..1
--v--.--.w w w----r
,-------,~,,~,_--.--.,-.4.,,&.
.-,-n
Pago ill 7"T k_/
1 supplied by the Omners' Group and we just plotted these just 2
to see how they looked.
This is the crack length.versus depth 3
and as I think Pick. mentioned, that's how they measured depth l
4 by a correlation with length.
My point,ir, that the thermal i
5 and mechanical fatigue have a different length to depth aspect 6
ratio.
That's all that means.
t We 11, talk a 11,ttle bit mbout our detection 7
t l
statisti,ca'and how We define the numbers just so we all a
?
understand.
This is a rather specific definition, if the true 10 crack length went from here to here, we looked at that with 11 say four different ultrasonic techniques; technique 1 might go 12 from here to here, technique 2 might be here, technique 3 13 might straddle the, indication, technique 4 might be over here.
O 14,He made a very specific definition of crack detection, did not 15 allow for many tolerances at this point.
You either got it or 16 you didn' t.
If your indication that you measured, which is 17 from here to here, we called that much a hit, that much a miss 13 and that much an over call.
And it's just the definition.
10 Now for example, if you detected the crack but 20 because of beam skew or some measurement error or whatever it 21 appeared over here, that was a miss.
We can go back and 22 redefine these any way we want, including some kind of 23 tolerance in our grading unit but that has not been done yet.
24 That's why it's still preliminary.
1
?5 To show you what specimenu have been looked at so O
l
- - - a
Page 112 I
far, we're not finished with all the specimens, these are the 2
nine types, a mort of qualitative measure of inspectability or 3
the difficulty of inspection.
One is the easiest and five is E
the toughest we think.
You'll see we've done ten of the 5
toughest.and 24 of the easiest.
So there's still quite a bit T
s 6
of difficult specimens yet that are.not in our analysis yet.
7 That'll be done in the next couple of weeks.
I' 8
Results.
We plotted our crack detection rate as definkd-by the little cartoon I showed you earlier, a very 9
10 specific definition of crack detection versus crack depth and il not too much correlation -- none.
You see that crack
?2 detection rate went from 10% to about 100%, all over the 13 board.
And this is a mixture of the easy specimens and the O
14 tough _specimena in here.
15 I plotted that a little bit differently again on one 16 of these perf,ormance curves.
The crack detection rate versus 7,
y d'
f 17 the falae' call rate.
This is the type 2, this is the over 18 call rate, not the crack miss, this is the over call rate.
19 closed clrcles are the specimena that have cast stainless on 20 both sides, the or a circles are the nozzle specimens that 21 have some kind of forging in them someplace.
i
- 22 DR. SHEWMON:
What was the mean depth or aree. of J
~
23 these flaws?
i 1
24 MR. AMMIRATO:
Nean depth?
25 ;
DR. SHEWMON:
If you have a three inch wall thick, is i
l
.e ll.c t N 4
l
>m ay[g\\
I(k 1
(f
[i y
4 1,
l. l '.$
,,o ' d * \\
j
.4 Page 113.
l' lQ
' )h
\\
s 3{i I
1
%.,,(, ;
itontheaveragehalfwaythroutyor--
,\\ f '
e.a -
\\
\\
1
?
MR2 1 AMMIRATO:
Oh,': they were di a t.ributed f rom 5% ';o
\\
,4 5;
4
>i,
7 3
40%, it was a mixture.
2
^
4 DR. SHEWMON:
Okay.
/
5 MR.,AMMiRATO:
Nmg the nozzle specimens --
(
t
,6 DR. SHEWMON:
Did the probability of detection rise i
{
k x
i<-
7 to 100 on thy 40% onen or wan./ it independent of size?
- \\
\\
m 8
MR. AMMIRATO:
We hLven t seen any correlation with 9p(i.
thepointIwasgoQ.q;40 make is that a lot of ize yet.i Now
^
10 thio data was takenonthenozz3/sfecimens that have some A
y y
t forgith someplace on iti, so y,M n'get to the crack through v 11
/
(
a fortp'r.h, which I meant to.say d somposite result.
12
\\
/
' im Ii
'k13 Aga $n.v the crack detec tion rate ranges from pretty l
y
- /,
.1%
+J h Landom to 100%,, but 11) general it's sort of up in this.,
s V,
5 side which is good, butlyou kr}ow, it's a very limited
, y 1
s
-.f).
~
y evaluationeq.dlatoratoryextbfment.
But we have.not 16 t
j
\\
.s included a lot of s,he t<3ugher segadments I expest these'to i
]'
/
17 S
g
. -)
show up down in LIha.
\\.
l-1
, '( (
1,
'!s l
gl p
2 i
{'t9 I'?.1 talk a little bit about afgna.1 processing, witich you will ese demonstrated K61s af terncon, I'lljustgivtl..
f 20 l3
',?..y V\\,
'l 11 you an introduction to it.
The aim ci It is to increase the l
a r
i
A i
l 22 '
sigtvd' to noise to help you.with detection, and once y:au
- i 23 [ detect it to Iaprove the classification; is li a crack or is l
25 it an interfed' nignal or in it grain noise, what 1s it.
i M-Three waya are being lodkqd at and l've already li l
f
\\
v L
I i
n4 g
3 t
page 114
-- g L_s/
1 mentioned imaging, just making a picture of your data and use 2
that to better interpret what you are seeing.
Spatial 3
averaging, it's a simple signal processing technique that 4
takes advantage.of some simple geometry which I'll maplain in 5
the next few slides.
Feature based approaches, Mohamad 6
mentioned this already, which uses the signal itself to try to 7
understand what that signal is coming from, the signal rise 8
time or its width or its shape or symmetry, those kinds of 9
things.
That's very well applied in the BWR case but we have 10 not gotten to this very much for cast stainless but we're 11 working on these first two up to now.
But it's on the list.
12 The principle of spatial averagina, very simple 13 concept.
You have a crack that you scan in a direction
/-s) 14 parallel to it and the idea is that the grain noise is going
~
15 te be -- the grain is going to be smaller than the crack 16 length as you move across, and add up signals, average them 17 together, the grain noise or other noise will smear out and 18 the crack signal oeing at the same location will sustain and 19 be reinforced.
So you would just make a scan, average, make 20 sort of a rolling average of the scans and do some 21 reinforcement that way.
22 We tried this out on the Westinghouse Owners' Group 23 specimens to see if it worked and here is one of these scans.
24 This is an individual A-scan and this display is tranuducer 25 position versus depth so it's taking these, turns them on edge n
t
I Page 115 7-,,
l and stacking them up so you're looking down on top of a set of s_
2 scans.
And scan across the crack that way, parallel to it 3
that way.
And what you see on this side is the before and 4
this is the after signal processing.
And you can see in the 5
A-scan the noise -reduction, the crack signal is still here and 6
here, but nm; you've eliminated a lot of these other signals 7
which move around.
So in the averaging process they get 8
smeared out and you can see now this indication was here but 9
it's pretty difficult to pick up from the rest of it and over 10 here, it's sharpened up.
So it worked in this case.
13 This is a technique that we applied to our field 12 exams which I'd like to --
13 DR. SHEWP,0N:
Do you have to know the sensitivity or
,s la the orientation of the crack for that or --
15 MR. AMMIRATO:
yes, it works best when you can scan i
16 parallel to it because then it's at a fixed time.
Scanning 17 perpendicular to it works also too but not anywhere near as l
18 well.
In fact, even scanning parallel to it doesn't work all 19 the time.
I remember one example where it didn't help too 20 much because the signal was already fairly strong, so it I
2' didn't really add anything.
It's always going to help some.
I l
W Our field applications I want to talk about.
I've l-l
??
been working on these since 1985 and these are really very l
l 24 important.
It's already been mentioned that things like l
l 25 surface finish are going to kill you as happened here at the l
l0 l
l l
l L
Page 116 C
l_a/
1 Arkansas Power Plant.
There was a stainless steel casting to 2
be examined and we thought we knew what was the best technique 3
to use and put it all in o'ur tool kit and when we get there 4
the surface is scalloped out by grinding and repairs and we 5
couldn't do anything.
So that's a simple thing that's just 6
going to shoot you down.
7 To start off at Vogtle and took sumo automated UT 8
collecting data and I'll show you one example later, again 9
trying to scope out the problem.
Trojan and Braidwood were 10 done last year and we applied signal processing to data 11 collected by commercial q 1entists and you'll see examples of 12 that this afternoon.
We'll go through again what was done.
13 The data was collected by Intraspect 98 system by the
\\'
14 utility's vendor, it was turned over to the NDE Center and we 15 applied signal processing technique, this is the spatial 16 averaging, to try to improve the data.
17 The joints that were examined at Braidwood, this 18 area of this valve to elbow, there was a radiographic 19 indication and the utility wanted some verification and 20 confiruation with ultrasonics so they asked us to go in and i
l 21 apply some of the signal processing on their ultrasonic data l
i 22 as collected by someone else.
l 1
23 This is an example of the Braidwood data that was l
24 looked at with our signal processing system.
This is an l
25 example, the signal was already pretty good, applying our l
t_-
F
]
Page 117
'D i_]
I signal processing to smooth it out, sharpen it up.
It was 2
already reasonably a good signal.
3 DR. SHEWMON:
Now there certainly wasn't a flaw like d
that in the Braidwood piping.
What does this mean, Braidwood 5
-- or was there?
6 MR. AMMIRATO:
This is actual data from Braidwood.
7 DR. SHEWMON:
Okay.
How big was that flaw?
8 MR. AMMIRATO:
This particular one looks like eight-9 tenths of an inch long.
10 DR. SHEWMON:
Okay.
Il MR. AMMIRATO:
But what's important is the next time 12 when you go back there, you can look at this again.
13 MR. CURTIS:
This is pre-service.
Okay?
O 14 DR. SHEWMON:
Pardon?
15 MR. CURTIS:
It's pre-service inspection.
16 DR. SHEWMON:
What is the post-service?
17 MR. CURTIS:
That's yet to be seen.
IB DR. SHEWMON:
Okay, so if the flaw is still there, 19 we'll just watch it for awhile.
20 MR. AMMIRATO:
And cleaning up the data helps you 21 look at it next time a little bit better.
22 At Trojan we were asked to go in and do the same 23 kind of thing on this ultrasonic data.
There was a 24 possibility of a snubber problem and some imposed strain on 25 this hot leg elbow joint so we did an ultrasonic exam and O
Page 118 D)
L_
I again the data was noisy and this technique helped to clean it 2
up and helped them with their analysis.
We did the same thing 3
again, again the before and after.
The indication here and 4
here and you can see this noise, this noise is sort of -- a 5
couple of causes, one is electronic pickup in the signal 6
cables but that data is already there.
So that averaged out 7
as we moved across.
Also there's the usual grain scattering 8
and here's the indication in the before original data and 9
here's the indication in the processed data.
You can see just 10 a cleaner image, cleaner picture.
11 DR. SHEWMON:
You've shown that as a sharp crack.
12 Do you have any idea whether 3t is that kind of shape or 13 whether it's just a bunch of porosities there?
O-14 MR. AMMIRATO:
No, you can't interpret this that 15 way.
This is a spec scan, so just as you move across the flaw 16 you're going to see different locations and depths as you-scan 17 across it.
l 18 MR. CURTIS:
I don't even think this is near a weld.
l l
19 I think this is in base material.
Don, do you want to address i
i 20 that?
l 21 MR. ADAMONIS:
I'm not sure, I think the confusion 22 arises though from the sketch above.
l l
l 23 MR. AMMIRATO:
We scanned toward the crack so the l
24 Indication is it appears as different depth as you get closer l
25 to it.
l O l
l
Page 119 7
1
-MR.
ADANONIS:
That's right, but the configuration 2
that you've drawn leads one to the conclusion that it is a 3
planc defect and I don't think that's --
4 MR. AMMIRATO:
Oh, no, that's just standard 5
cartooning.
There's an indication --
6 MR. CURTIS: -Standard cartoonist can lead us the 7
wrong way.
8 DR. SHEWMON:
Thank you.
9 MR. AMMIRATO:
This is that pump casing at Arkansas 10 Unit 1.
The area was in here and we just really couldn't do il much with it, it was just too rough of a surface, and as 12 mentioned before these transducers are quite large and you can 13 see those this afternoon.
You need a relatively smooth area
'\\"
14 over a larger region.
15 This is an-example of really one of the first ones 16 that was done at Vogtle, again examining a lot of welds trying 17 to understand the field problems of this kind of work, what I
la does it take to bring a system into service, that was an l
19 interesting thing, running cables, clamping scanners on pipes, l
20 that was a very useful experience.
l 21 I'd like to make some conclusions.
I still believe 22 there's really no single general ter,nnique for cast stainless 23 steel.
There's a logic tree you nave to go through to pick 24 the technique.
l 25 There are some specific conditions, I think I've O
i 1
1
~., _.. _ _ _ _.... -. -.. - ~ _. - - -,
(
Page 120 I
shown you some examples, where it does work in certain 7
conditions.
3 The EPRI work and the work here at the Center over 4
the last three or four years has I think lead to a very sound 5
experimental and theoretical basis for understanding wave 6
propagation in anisotropic material..The trick is to use that 7
to first of all figure out what resources you have and then 8
work backwards and compensate some of your measurements.
That 9
work is still in progress.
10 Signal processing can improve results.
I've shown 11 you some examples of field data that this was applied to and 12 it did improve the quality of the data.
13 Field trials are valuable.
O 14 Preliminary detection statistics on the Westinghouse 15 Owners' Group samples, and I say preliminary for several 16
. reasons, they're not finished yet and we don't have some of 17 the more difficult specimens in there yet.
It's a restrictive la definition of crack detection and false call rate.
There's no 19 tolerance that has been allowed.
It was done in the 20 laboratory by someone who knew there was a crack someplace in 21 this specimen.
And it's composite result, three or four 22 techniques thrown altogether.
23 Two sided access, if there is a forged part 24 someplace in there, forged stainless or forged cast -- forged 25 stainless or forged carbon steel, crack detection averaged
--->--,,,--,-...e,,..-
m
,,.,.,,--w,r.-,-n,
.,n,-,-
,,.m-.v-v,
.-w
, - ~, -,,, - -. ~,.,,,, -
I Page 121 7'T l
k_)
I about 80%, false call of about 10%.
2 Two sided access where there's cast steel on both 3
sides, the detection rate went down to about 60% and false 4
calls came up to about 15.
5 The one sided access question is important.
WE have 6
that data but it has not been analyzed yet.
7 Back to our chart and I think Mohamad closed.with 8
the same comment that I think we are making small steps in 9
this direction.
10 DR. SHEWMON:
Thank you.
11 MR. AMMIRATO:
I'd like to just tell you what we're 12 going to see this afternoon if you're interested.
13 DR. SHEWMON:
Let me stay with one thing, is it my 14 understanding then that there is a pre-service inspection done 15 on all welds and in the Braidwood and Trojan case they found 16 indications and that's what's being followed or where EPRI 17 came in to do additional work?
18 DR. BEHRAVESH:
Not in Trojan, in Braidwood.
19 MR. CURTIS:
Why don't you explain Trojan.
23 DR. BEHRAVESH:
Trojan is an operating BWR and --
21 MR. CURTIS:
PWR.
22 DR. BEHRAVESH:
I'm sorry, PWR, And data that was 23 collected, there was no credible pre-service data at Trojan to 24 compare this to, so whatever data was collected last year was 25 decided to let this constitute a base line now and in fact the O
r Page 122 7'}
L_]
1 plant plans to look at the same region again next month.
2 DR. SHEWMON:
Well was this part of a five-year 3
inspection or ten year at Trojan?
d DR. BEHRAVESH:
We're talking about probably the ten 5
year inspection.
6 MR. AMMIRATO:
But there was a problem too in this 7
particular joint because the pipe was displaced because there 8
was a snubber problem.
There was a symmetric displacement of 9
the pipe, so they went to look at that joint to see if some 10 damage had been done.
11 MR. CURTIS:
And.if I remember correctly, this is 12 more of a casting type flaw in the body of the material itself 13 and not a crack near a weld.
Okay?
When they were doing the O-s 14 exam, they saw this indication and further evaluated it and 15 then called EPRI and so.it doesn't even appear to be 16 associated with a flaw in connection with a weld.
It's l'
something they found as they were doing the exam in the base la material.
The problem is the cartoon, the character if you 19 will, kind of implieu that there's a crack there and that's 20 not the case I don't believe from the other data.
21 MR. WARD:
But this was found just incidentally away 22 from a weld but --
23 DR. SHEWMON:
Something had bent it out of shape.
24 MR. CURTIS:
Well they had gone through a thermal 25 cycle on the piping, it had been a strut if I remember 0
(
Page 123
-m k_)
I correctly and because of that, they said this weld could have
)
2 had some higher than normal stresses on it so because of that 3
we'll do some augmented inspection.
So they went in to do an 4
inspection and while they were doing the inspection of the 5
base metal they found this indication.
6 DR. SHEWMON:
Was this in the feed water system?
7 MR. CURTIS:
Yes, I think it had to do with the high 8
cold water striation and the sparger line and that put the 9
bending moment across that weld.
So they wanted to inspect to that weld and while they were inspecting the weld, they found 11 in the base metal this indication which has been further 12 evaluated.
13 So I'm not sure, you know -- I don't want to get you I\\)
14 thinking there's a crack in that weld because it's not there.
15 DR. SHEWMON:
Thank you.
16 MR. CURTIS:
For them -- for their sake.
17 DR. SHEWMON:
Okay.
Go ahead.
IB MR. AMMIRATO:
What you're going to see after lunch 19 out in our high bay.
You're going of course to be able to see 20 all of our samples, get a close up look at them.
We have 21 three demonstration positions set up, one is manual UT and if 22 you so care and have time you can actually do some scanning 23 yourself if you want to try to illustrate some of the effects 24 I've mentioned earlier today.
25 We'll move on to the automated UT, the Intraspect l
/
Page 124
.yn 1
98, a commercial inspection system, and we'll see some of the a__
2 benefits of that data.
3 The next stop will be a signal processing system, 4
personal computer system which is data acquisition, imaging 5
and signal processing.
This is the system that was used to 6
analyze the data from Trojan and Braidwood that I showed you 7
before.
And you'll see some examples of the field data from 8
Trojan and Braidwood and how this signal processing was 9
applied.
10 That's after lunch.
Il DR. SHEWMON:
Somebody want to tell us what we do 12 for lunch?
13 VOICE:
Yes.
Through that door and down the hallway
,s
(\\~')
14 at the end is the cafetoria.
15 DR. SHEWMON:
We're scheduled for an hour.
Why 16 don't we aim at half an hour and end up with 40 minutes from 17 now or something if that sound credible, and we'll see how it 18 goes.
19 VOICE:
We'll lead you down the hallway to the 20 laboratory, we'll not need to come back here right after 21 lunch, the demonstration takes place in a different part of 22 the building.
23 DR. SHEWMON:
All right.
a (Whereupon, the Subcommittee meeting was 25 adjourned at 12:02 p.m.)
fs v
.~
1 CERTIFICATE 2
O V
3 This is to certify that the attached proce.edings before the 4
United States Nuclear Regulatory Commission in the matter of:
5 Name:
ACRS SUBCOMMITTEE MEETING ON METAL COMPONENTS 6
7 Docket Number:
8 Place: Charlotte, North Carolina 9
Date:
March 15, 1988 10 were held as herein appears, and that this is the original 11 transcript thereof for the. file of the United States Nuclear 12 Regulatory Commission taken stenographically by me and, 13 thereaf ter reduced to typewriting by me or under the direction 14 of the court reporting company, and that the transcript is a
]
15 true and accurate record,.of the foregoing proceedings.
fx.//w-Ykvun.)
16
/S/
l 17 (Signature typed):
William L. Warren 18 Of ficial Reporter 19 Heritage Reporting Corporation 20 21 l
22 23 l
24 25
)
O seritese Regortins Correretion (202) 628-4888 l
l l
l
.J