ML20151H767
| ML20151H767 | |
| Person / Time | |
|---|---|
| Site: | McGuire  |
| Issue date: | 04/26/1983 |
| From: | Tucker H DUKE POWER CO. |
| To: | Adensam E, Harold Denton Office of Nuclear Reactor Regulation |
| Shared Package | |
| ML19268B539 | List: |
| References | |
| TAC-49164, NUDOCS 8305050283 | |
| Download: ML20151H767 (28) | |
Text
.
DuxE Powna Goxnm l'.o. Ilox 33189 CIIAHI.OTTE, N.C. 28242 II AL II. TUCKER trixenose (7*H "'"'* "
,,;"*,"/,"*"L Apri1 26, 1983 3 R0PRI ~~~A RY
.Mr. Harold R. Denton, Director O
Office of Nuclear Reactor Regulation U. S. Nuclear Regulatory Commission Washington, D. C.
20555 Attention:
Ms. E. G. Adensam, Chief Licensing Branch No. 4 Re: McGuire Nuclear Station Docket No. 50-369
Dear Mr. Denton:
Please refer to our letter of March 14, 1983 concerning a proposed change to McGuire Unit 1 Technical Specifications to reduce the measurement uncertainty for Reactor Coolant System flow rate.
It was requested that information which is proprietary to Westinghouse Electric Corporation be withheld from public disclosure.
In support of that request, enclosed is one copy of Application for Withholding CAW-83-25 (Non-Proprietary) and the supporting affidavit. Also enclosed are two copies of the attachment to our March 14, 1983 letter -- one copy is proprietary and one copy is non-proprietary.
Very truly yours, G
d b (.c Hal B. Tucker REH:jfw cc:
Mr. James P. O'Reilly, Regional Administrator U. S. Nuclear Regulatory Commission Region II 101 Marietta Street, NW, Suite 2900 Atlanta, Georgia 30303 pf bgl}
M Mr. W. T. Orders NRC Resident Inspector f
I McGuire Nuclear Station eM Y
.a%y %*,
d ev4 *N[6 B305050283 830426 PDR ADOCK 05000369 P
Non-Proprietary Version cf the Attachment to the March 14, 1983 letter from H. B. Tucker to Harold R. Denton l
l t
l 1
j e.
s s e
1.
- ues: ion
- able 2 of your s:hmi :a5 (ie::er E.3. :ucker :o E.3. :en:on, da:ed :oved er 23, 1382; Lis:a :he uncer:cin:y value for the feedua:er Ven:uri S ou coeffkien: (X) as 0.25%, uhich uas obicined from Alden Research Labora:ory Standard accuracu.
~
(a) Whar is :he range of flou Reynolds number tes:ed in the 16cra:oru calibrazion?
(b) Is the Reynolds number of the McGuire feedJa er flou uithin the 52nge of idora:ory calibration?
(c)
Does the 0.25% uncertainty also include che Ven:uri ins:cita: ion aliavance?
(d)
What is the drift affect of the Venturi fouling?
?rovide a detailed component breakdcun and justification of each component uncer:cin:y assoc'ated with the overait uncertainty of the Venturi flou coeffician:.
Resconse (a) The range of flow pipe Reynolds numbers tested in the Alden Research Laboratory
' calibration was 800,000-3,350,000.
(b) The McGuire f eedwater flow pipe Reynolds number is 14,400,000.
However, ASME is constant and and other references indicate that the Venturi flow coefficient independent of Reynolds nu=ber above 300,000 (c) No, the 0.25% flow coefficient (K) does not include an installation allowance.
The entire flow element was calibrated is a unit including 6.5 diameters of This length of straight pipe is straight pipe upstream from the Venturi encr.ance.
greater than that recommended by ASME for a Venturi flowmeter.
the (d) Conceivably fouling could occur such that crud accumulation could af fect the Venturi throat pressure caps in a nanner that static pressure distribution at in a higher flow for a specified aP, however, the reduction in throat would result area resulting in a lower flow at the specified aP is a stronger effect.
If foul-in an error in a nonconservative ing occurs and is undetected it would result Fouling has a bias effect direction in precision calorimetric RCS flow seasurement.
increase the since it shifts the flow measurement and does not not a drift effect If Venturi fouling is detected, the Venturi will be cleaned prior to random error.
Fouling has never been a problem in the the precision heat balance measurement.
Duke Power system (Fossil or Nuclear); the une of All-Volatile Chemistry precludes the build-up of crud, which is associated with Trisodiumphosphate water chemistry.
Detailed component breakdown of the Venturi flow coefficient uncertainty is providec in Attachment 1.
2.
Quession Tale 2 Lis:s the feeduater temperature and secondary side pressure measuremen:
uncercainties of to.5*? and 25 pai, respectively.
of comvonen:s and uncertainty value of each component (uith justification) associated uich feedua:er temperature and pressure measuremen:s such as RTD cali l
bra ion, transmitter calibration, drift, and precision register, conver:or and l
computer accuracy, etc.
Response
The component used to measure feedwater temperature during the precisian heat balance is a calibrated continuous lead type J thermocouple with an icebach The feedwater thermocouple MF is measured by a L and N-914 reference junction.
A breakdown of the. components follows:
Numatron 0-40 mv range.
4 a
Thermocouple Calibration
- .25'F Readout Calibration
- .03*F Standards Lab Calibration Uncertainty-(USL);
USL = /Ee' - /(0.25) + (.03)' = =0.25'T Additional conservatism is added to this measurement uncertainty.
2 x USL = 2 x 0.25'T = 0.5'F The component used to =easure feedwater pressure during the precision heat balance is a 0-2000 psig bourdon tube gauge with an accuracy of 20.25% of span.
20.25% x 2000 psig = 25 psig Neither feedwater temperature nor pressure is seasured by the computer during the precision heat balance, both are =easured using test instru=ents.
3.
Quesrion Table 2 also lis:s :he R:D c=Librc: ion and DVM ccouracu errors far :he ecli :eg
~
crd ho: leg :empera:ure measurements.
>n::: cre :he un:errain:ies for the rr ~:s-nirter calibra: ion, drift, resister ard : mpu:er?
Resoonse Hot and cold leg temperature =easurements are obtained during the precision heat balance RCS flow measurement by a Digital Ohmmeter, (Fluke-8375A Digital Ch=eter 10.002: + 1 digit).
Using a Digital Ohmmeter attached directly to the RTD leads ell =inates the drif t due to the process racks, since all process instrumentation is removed from the instrument train.
4.
Question i
?rovide justifica: ion for assigning t1.2*? (Table 2) for the ho: leg tempera:ure s:reaning error.
I
Response
A process measurement error has been incorporated into the reactor coolant system calorimetric flow measurement uncertainty to account for the steady-state tem-l perature gradient in the hot leg, caused by incomplete mixing of the coolant l
flowing out of different regions of the core at different temperatures. Measure-l ments obtained at a Westinghouse three loop plant established hot leg te=perature l
gradients of (
]+e c in one loop and [
]+e.c in another loop while at full
(
power.
To offset the effect of this temperature gradient, the hot leg te=perature on subsequent plants (including McGuire 1) is measured on a bypass loop connected to the hot leg at three locations around the pipe circumference as shown on Figure 1.
Each connection is provided with a probe, or scoop, which samples the coolant over a distance of 7 inches into the 29 inch inside diameter of the pipe With this arrangement, the potential for a difference between actual average hot leg temperature and measured temperature is minimized.
L
Il 1
0 b
1 w5
/
'/M///f f U/////
A }T e
b4 bs rlow 1
i o
- a
)
s,o no s
Section SB Y
To RTD Section.4A i
, Hot t.eq Steam
?eaC:0" Generator' I
c el _
7 A
n
~
1 l
(c Cold Le9 Fiaure 1 RTD Bypass Loco Scoop l
Arrancement for tieasurement l
of Hot Leo Temnerature
=
i Twa fcctors ara contidsrcd in tha t.nalys:s of tha mencurement arror:
ths i
temperature distributions that could be present at the scoops, and the deviation from balanced sample flows into the three scoops.
With perfectly balanced sample flows, an evaluation of several possible hot leg temperature distributions has shown that the scoops will limit the measurement error to less than [
]+a,c of the temperature gradient (i.e., ~ (
]+a C for a maximum gradient of (
]+a,c.
With a conservative sample flow imbalance (50% flow in,1 scoop, 25% flow in 2 scoops) the evaluation has shown that the scoops will limit the measurement error to less than [
]+e c of the temperature gradient.
Calculations of the scoop branch line flow imbalances for several plants has shown that the estimated flows in most cases will range between perfectly balanced and a distribution of 40%-30%-30% resulting in smaller errors.
In most plants with calculated flow imbalances, the upper scoop is expected to have the highest flow.
When the results of the three loop plant test are considered (top of hot leg was hotter than the bottom), the measured hot leg temperature is more likely to be hotter than the average hot leg temperature, leading to a conservatively low calorimetric flow measurement.
Since there are uncertainties in the temperature streaming distributions and magnitudes, the allowance for the temperature streaming measurement uncertainty has been set conservatively at [
]+a c regardless of the scoop flow distribution.
For McGuire Unit 1 the analysis of the scoop branch lines has shown that the flows should be reasonably balanced.
Therefore, the [
]+a,e allowance for temper-ature streaming is additionally conservative.
5.
- ues icn Tabies 3 ard 4 Liar :he uncer: in y of each parame:er for the CAC ard %N e2cu
(a) Provide che uncer in y value (ui:h jus:ifica: ion) in :e=a of :he percen: age of measurement span of each componen: ard the effec: fac:or of each componen:
to the RCS ficu.
(b) ?rovide jus:ifica: ion for assigning 0 value on the senscr calibration uncertain:y, sensor pressure and temper 1:ure effect, ard rack :empera-ture effect.
Response
+a e (a) Process Measurement Accuracy (PMA) - A (
) u,ncertainty for the RCS tempera-I ture has been assessed for the Automatic Rod Control.
The Automatic Rod Controller is placed in manual during the precision heat balance when the elbow tap instru-mentation is normalized and held within a very tight tolerance to keep RCS te=pera-ture steady. When the unit is returned to process control,the cold leg tempera-ture may fluctuate by (
]D*rhis affects the density of the water in the elbow meter, thus the flow measurement by the elbow cap instrumentation.
The simplified equation for the elbow flow meter is:
We = K/aP/o
+n,c The term /1/p was evaluated for the ranges of T = 550 2 [ j, P = 2200 and P = 2200 1[
fT=550*F.
The temperature and pressure fluctuation in this range have a (
]"Ind [
] * "'"
effect on flow.
The overall effect of the process control is:
- %,c
=
...r--'-
- * - vr
-w----
vw--*---w---- - - --
-weww v------
i i
o Process Measurement Accuracy, PMA, has been assessed to be
_ +,'e
. *a,c Process Measurement Accuracy (PMA)
Primary Element Accuracy-(PEA)
Sensor Drift (SD)
Rack Calibration Accuracy (RCA)
Rack Drift (RD)
Isolator Drift (ID)
Analog to Digital Conversion (A/0)
Readout (RO)
These component uncertainties are standard Westinghouse numbers for the process instrumentation.
These have previously been reviewed and approved by the NRC.
Refer to:
NUREG-0717 Supplement No. 4, Safety Evaluation Report related to the operation of Virgil C. Summer Nuclear Station, Unit No. 1 Docket 50-395, August, 1982.
(b) Sensor Calibration Accuraev (SCAi - The precision heat balance and flow normalization will be used to determine the elbow flowmeter coefficient. The standard transmitter calibration will ini:ially set the transmitter output within an acceptable tolerance for the precision heat balance.
The precision heat balance will then normalize the trans=1tter output which will be used for surveillance during the fuel cycle.
Sensor " pressure and temperature effects account for any shif:s that say occur due to changes in static pressure on the a? cell and a=bient temperature respectively.
The precision heat balance will be perfor=ed while the elbow tap flow transmitter is at its normal operating temperature and pressure.
Sensor Temperature Effects (STE) or Sensor Pressure Effects (SPE) do not need to be assessed since the transmitter normalization is performed at normal operating conditions.
The process contr'ol racks at McGuire are located in the control room environment which has a FSAR Design Criteria temperature limir of 75:5'F.
Therefore, no widely varying tarperature effects exist to affect the process racks.
Effect Factors will be quantified in the response to Question 6.
6.
Question quantify the value of the effect factor of the RCS ficu uncerrainty vi:h respear to each parameter uncerrainty listed in Tchte 2 as well as Tchtes 3 and 4 required in the question S.
Response
The effect factors listed in Table 2 were determined by incrementing each parameter required in the analysis within the computer program which is used to arrive at :he RCS flow value.
This rathod of computer iteration returns a more conservative value than the Westinghouse ITDP analysis since it arrives at an integrated value for flow changes.
The effect factors in Tables 3 and 4 are derived from conver:ing %4P span to % flew for the elbow flowmeter transmitters.
Refer to Attachment 2.
o 7.
Ques:icn Zhe rco: sum acuare (255) :echnicae of combining :he uncer:ain:ies recuires :ha:
each uncer:ain:y con:ribu:icn be independen:.
If : hey are no: independen:, -1:eir combined effec: shcuid be assessed :hrough de:erminis:ic me: hod.
There are some uncer:ain:y con:ribu:icns in Table 2 which are no: independen:.
For e:::: cts, an$) steam en:halpy (hg) are bo:h dependent on s:eam and enthalcy (h ~ are c!.i dependen: upon :he feed.sacer :empera:ure:
he feec There x:y be a:her non-independen: parame:ers.
Jus:ify your use of the 295 :achnicue to ecmbine :he uncer:cincies of :hese parameters.
Resoonse Technically, the feedvater te=perature and pressure uncertainties are co==on to several of the error co=ponents.
However, they are treated as independent quanti:ites because of the conservatis= assu=ed in the co=ponents.
The arith = etic s" - tion of their uncertainties has no significant effect en the final result.
Treatingtheerrorco=ponentsoftheSecondarySideLoop[PowerUncertaintyas[*Cc=bin dependent and su==ing all of the error co=conents.Usec =
with :he !ctal Pr:.=ary ah Uncertainty, Uah pri: =[
]%c
- he Pri=ary Si:ia Lacp Flow Uncertainty equals:
U pri=ar/ loop flow = v'(Usec) t f.U ah pri=)
= [
]%c U total pri=ary flov =
(Usec) + (Uah ori=)7 =[
]\\c 4
There are no other dependent para =eters in the analysis.
8.
Cuestion For each comycnen: associa:ed :,ri:h :he measured parame:er, uha: is the na:ure of error, i.e., randem or biased? Wha: is :he error distribucion func: ion, i.e., nor=al or uniform?
\\
Response
Rando= error is associated with the =easured para =eters.
I The error distribution function is nor=al.
9.
Ques:icn Does :he :ocal RCS uncer:cincy value derived in Table 5 represen; a 95% probabili:y a: 95% ccnfidence value? If so, can the i.:plici: assur:p: ion :ha: each uncer:ain:y value be a: its 2a (s:andard devi.ation) Limi: be,ius:ified?
Response
The value derived in Table 5 represents a nor=al, two-sided 95+% probability distribution. All instru=ent and =easure=ent uncertainties are consistent with or conservative with respect :: the Westinghouse ITDP analysis.
The probability justification is contained in Attach =ent 3.
0 ATIACIDiCTI 1 Breakdown of the Venturi Flow Co=ponent Coefficient 9
9
o P
SUBSTANTIATION OF Tile ACCURACY OF Tile CAllBRATIO i
AT ALDEN RESEARCil LABORATORIES USING 100,000-POUND WElGlilNG TANK - FOR 14",16" AND 18" TUBE l
l i
l LEGIBillTY % ON "C" ACCURACY % ON "C"
(%)2 ITEM BASIC DATA
(%)2 NO.
ITEM 0.0190 0.0004 Error on 80,000-lb is 15 lb, l
1 Weighing scale is marked in Tonk 10-lb. increments Essor on interpolation at 0.0025 0.0000063 80,000 lb. (12 lb) 0.0080 0.000064 Electric Timer 2
Time 0.0020 0.0000040 Assuming 50 sec. run with millisecond read as digit.
i Scale graduated in 0.01 ft.
3 Monometer l
Scale and assuming 0.0I ft. reading reading error (including Reading error on scale) and minimum 30 nanometer readings the l
following typical errors should be considered:
0.1177 0.0138533 0.1177 0.0138533 0.05885 0.0034633 0.05885 0.0034633 At 0.5 ft differential l.0 f.
0.02942 3.0000655 0.02942 0.0008655 a
" 2.0 f t.
0.01471 0.0002164 0.01471 0.0002164 4.0 ft.
4
c_
r 9
SUBSTANilATION OF Tile ACCURACY OF Tile CAllBRATION AT ALDEN RESEARCil LABORATORIES USING 100,000-POUND WEIGillNG TANK - FOR 14",16" AND 18" TUBES ITEM ACCURACY % OH "C" LEGlBILITY % ON "C" BASIC DATA NO ITEM
(%)2
(%)2 4/A Fluctuation Assuming 30 monometer Effect on readings and max. peak Wnometer to peak amplitud3 of:
1% of differential 0.0300 0.0009 0.0300 0.0009 0.0590 0.0035 0.0590 0.0035 0.1180 0.0139 0.1180 0.0139 2% "
4% "
With 99% confidence limit.
4/8 Fluctuation Assuming fifteen Effect On monometer readings Wnometer and max. peak to peak omplitude of:
0.0498 0.00248 0.0498 0.00248 1% of differential 0.09954 0.00991 0.9954 0.00991 0.19908 0.03963 0.19908 0.03963 2% "
4% "
With 99% confidence limit.
fhmnJ2 aff 5 - CALC-322/8
b t
SUBSTANTIATION OF Ti1E ACCURACY OF Tile CAtlBRATION i
AT ALDEN RESEARCil LABORATORIES USING 100,000-POUND WElGlilNG TANK - FOR 14",16" and 18" TUBES ITEM 8ASIC D TA ACCURACY % ON "C" LEGIBILITY % ON "C" NO.
ITEM
(%)2
(%)2 0.03 0.0009 At every run full tank shall 5
Chuting be collected 0.015 0.0023 6
Specific Three ports per 10,000 Wat. Of Water and l
Mercury 7
Thermometer 0.029 0.0000084 0.0024 0.0000057 Accuracy is 0.lf Graduation 0.5F 8
Effect of 70 ft. Useful length is ovoilable which repre-Piping sents -
Lgth in Piping, Dio.
D Dio.
Upstream 12.125 59 6*dio.
63 0
')
0 0
s 0
0 0
0 l
12.50 67 - 6* a 61 15.016 56 - 6* "
50 0
0 0
0 16.126 52 6* "
46 0
0 0
0 l
- 6 Dio. laying length of tube and downstream piping Page 3 of 5 - CALC-322/8
A SUBSTANTIATION OF Tite ACCURACY OF Tile CAllBRATION AT ALDEN RESEARCil LABORATORIES 4
USING 100,000-POUND WElGillNG TANK - FOR 14",16" AND 18" TUBES
\\
LEGIBILITY % ON "C" l
ITEM ACCURACY % ON "C" NO.
ITEM
(%)2
(%)2 BASIC DATA
SUMMARY
FOR WORST CONDITION 0.0190 0.0004 0.0025 0.0000063 4
1 Tank O.0080 0.000064 0.0020 0.00004 2
Time 0.1177 d.0138533 0.!177 0.0138533 3
2nometer Scale Read.
4/A Fluctu4 tion
'O.1991 0.03963
- 0. 91 0.03963 4/B of Wnom.
0.0300 0.0009 5
Chuting 0.0150 0.00023 6
Specific Wgt. of W. and lig.
0.0029 0.0000084 0.0024 0.0000057 7
Thermometer l
0 0
.0 0
8 Piping 0.3237 0.53499 l
0.3910 0.055086 TOTAL 0.2313 0,2347 RSS
% Accuracy
% Precision 0.3910 = 4 x a = 0.3237 0.0978 = 1 x o = 0.0809 0.1955 = 2 x o = 0.1619 0.2932 = 3 x o = 0.2427 Page 4 of 5 - CALC-322/8
SUBSTANTIATION OF Tile ACCURACY Of Tile CALIBRATION AT ALDEN RESEARCil LABORATORIES USING 100,000-POUND WEIGillNG TANK - FOR 14",16" AND 18" TU8ES t
LEGIBILITY % ON "C" ACCURACY % ON "C"
(%)2 ITEM BASIC DATA l
NO.
ITEM
~
(%)2
SUMMARY
FOR BEST CONDITIOri 0.0190 0.0004 0.0025 0.0000063 1
Tonk 0.0080 0.000064 0.0020 0.000004 2
Time 0.01471 0.0002164 0.01471 0.0002164 Monometer Scale Read.
f 3
4/A fluctimtion 0.0300 0.0009 0.0300 0.0009 4
4/B of Monom.
0.0300 0.0009 5
Chuting 0.015 0.00023 Specific Wgt. of W. and H 0.0029 0.0000084 0.0024 0.0000057 g
6 7
Thermometer 0
0 0
0 i
8 Piping 0.001132 TOTAL.
0.1196 0.002719_
0.0516__
i 0.0337 0.0521 RSS s
% Precision
% Accuracy 0.1196 = 4 x o = 0.0516 0.0299 = 1 x o = 0.0129 0.0598 = 2 x o = 0.0258 0.0897 = 3 x o = 0.0387 Page 5 of 5 - CALC-322/8
2.00 Substantiation of tha cecuracy On this subject are calculations CALC-322/B - pages 1 through 5.
Pages 1 thieugh 3 are listing the sources of " inaccuracy"
(
values to facilitate fur-2.01.
" imprecision" indic= ting % and (%)2 ther calculations and to give insight of how the accuracy statement is influenced by different calibration components.
I Pages 4 and 5 are summaries indicating worst and best c tions as picked from extremes listed on pages 1 through 3.
1 2.02 The error effect of fluctuating differential was considered in i
2.03 the following manner:
Half of the mcx, emplitude (peak to peck) is equel to plus or minus Icrgest error on differential.
Hcif of this largest differentici error ecucis to the plus er min largest coefficient error (due to squcre rect relcrion.)
This plus or minus largest coefficient errer equels to 4 x (c).
99% confidence limit:
2.585 a 4 N N = number of differentia! readings Sample calculation for 4% peak to peck emplitude:
4 Max. error on diff. = 3 = $ 2%
2
.. " C"
= - = 21%
o 2
e
(
q 1
I C
c = - = t 0.25%
(
4
'. E a
99% confidence limit.
3 2.58 x 0.25
=
- 0. l l 89'o
~
( 30 CALC-335-!
g
V Meaning th:t 99% of tha cv:mga differ:ntials cstablish:d in
^
this mann:r shall be within 10.236% (2 x 0.118 duo to sau root relation) of the tna cna corresponding to tha fisw reto of which they were taken.
i Page 4 indicates a largest possible error level of 10.391% e 2.04 well as it shows that -
f 99.7% of all "Cs" taken under such condition sho within 20.2932% (3o) of their true value and t'
95.45% should fall within 10.1955% (2a).
Page 5 indicates a best possible largest error level of
- 2.05 cs well cs it shows that under these circumstances -
99.7 'o of all "Cs" shouid fell within *0.0897% (
their true value and 95.45% shcIl fall within to.0598% (2o).
Assuming 12 "Cs" (whose mean shall be the final "C" the 2.06 confidence level for this final "C" shall be At Worst Condition o = 0.0978 I
V = 12-1 = 11 N = 12 t.995 = 3.11 0.0978 e
0.0917 %
= ! 3.11
=
t.995 x V i1 V N-1 Meoning that 99% of the final "Cs" established in this menn should fall within 0.0917% of the true value.
e 4_
At Best Condition i
o = 0.0299 V = 12 - 1 = 11 3:
N = 12
-a t.995 = 3.11 3.11 x 0.0299 o
= 0.0233%
=-
t.995 x V N-1 (12-1 CALC-335-2
Meaning that 99o of the final "Cs" esroblished in this manner should fall within 0.0280% of the true value.
I Based on the calculations and considerations presented above, 2.07 the following questions should be answered, I-2.071.
According to what rule should bad calibration points i
I be discarded.
2.072.
How should they be replaced, How many of those con be rerun without re-examining 2.073.
i the eclibration procedure, 2.071.
Points in the constant region of the coefficient i
shall be considered " bod" if they fall further then to.2*'a of the mecn coefficient value.
2.072.
Bad points should be replaced by two new points taken at about the some R -
D 2.073.
If more then two " bod" points should occur et the e=libration of any meter, the eclibretion procedure should be thoroughly onelyzed to re-veel the cause of the error and the eclibrction i
should be repected.
o i E 3&
CALC-335-3 e
\\
t
1 ATTACHMENT 2
%a? expressed in % Flow 1
l 1
f I
i l
l l
l l
l l
l-I t
1 f..
TABLE 3-30 AP MEASUREMENTS EXPoESSED IN FLOW UNITS The P accuracy expressed as percent of scan of the transmitter app 1'ies througneu:
the measured span, i.e.,1 5% of 100 inches t.P = 11.5 inches anywhere in the 1
span.
Because F2 = f(A?) the same cannot be said for flow accuracies.
When it is more convenient to express the accuracy of a transmitter in flow terms, the followini, metnoe is used:
-n :
A i
n 8
e 3-43
t:,c Equation 3-30.8 is used to excress erecrs in ::er:en: full s;:an in :nis cc :: e-:.
3-44
o Effec: Facecrs lis:ed in Tables 3 and 4.
- b a,c.
4MW t
h j#m l
3 l
I
[
[
l
}
e 7
ATTACHMENT 3 Probability Justification h
e
(
a
o.
t IV.
MROSABILITY JUSTIFICATION As noted in Section III, it is Westinghouse's belief that the total uncertainty f or Pressurizer Pre.ssure, Tayg, Reactor Power, and RCS Flew are normal, two sided, 95+% probability distributions.
This sec-tion will substantiate that position with a ecmoarison between three approaches, the first being that noted in Section II, the second involves determination of the variance assuming a uniform probability distributien for each uncertainty and then determinaticn of the 95%
probability value assuming a one sided ncrmal distribution, and the third involves determination of the varianca assuming a. normal, two side probability distribution for each uncertainty and then detar=ina-tion of the 9E% probability value assuming a two sidec nor=al distribu-tion.
Table 7b lists the results of the three approaches. Column.1 lists the values noted for CSA on Tab'e Ib which are determined through the use of equations 1, 2, or 3, whichever is applicable to that particular func-tion.
Column 2 lists the variance for each function assuming the uncer-tainty for each of the parameters listed in Section 2 is a uniform prob-ability distribution.
For this assumption, 42b I
., = =
2h Eq. 9 e =
where R equals the range of the parameter. The variance for the func-tion equals the aMthmetic sum of the parameter variances. Fmm a safety peint of view deviation in the direction of non-censervatism is important. Therefore, Column 3 Ifsts the one sided 95% probability values based on the vaMances provided in Column 2, i.e., the one sided 955 probability value for'= near normal distribution can be reasonably approximated by: 1.64F Column 4 lists the variance for each function assu=ing the uncertainty for each of the parameters listed in Section 2 is a near nor=al, two sided pmbability distribution. Efforts have been made to conserva-tively determine the probability value,for each of the parameters, see Table 8.
For example, [SCA is noted on Table 8 as having a probability of 99%,1.e., Westinghouse has detamined that SCA will have a value of 0.5% span or less 99% of the time. This is known to be conservative in that a sensor /transmittar must be calibrated to within + 0.55 span or the calibration is re,fected. Thus, in reality SCA has a probability value of IC0% but for this analysis 995 was assumed.7"'C The corre-spending I value listad on Table 8 is from the standard nomal curve where:
I = (x - v)/e Eq.10 The vaMance for a parameter is then the square ef the uncertainty divided by its I value:
,2,
[ uncertainty }2 Eq.11
\\
I.
j 44b
,,, a The variance for the function equals the arithmetic sum of the parameter variances. From the variance the two sided 95Mrobability value for a nomal distribution can be calculated: 1.96 fa2 To sumari:e; Column.1 is the results of Equations 1, 2, and 3.
Column 2 is the total vaMance assuming unifom probabilty distributions, i.e.,
+... = II
""C ) +(
unc )
2 1
+R2 R
1 2
+...
Eq.12 e=
Column 3 is 1.645 Column 4 is the total variance assuming near nomal pmbability distM-butions, i.e.,
(unc f func f 3
+<
g g = i 'l )
I I +...
Eq. la_
e
(
( '2 )
Column 5,is 1.96 i
A c:moariscri of Columns 1, 3, and 5 will show that the approach used in Section 2 results in values more conservative than those of Columns 3 and 5.
Thus, it can be concluded that the msults presentad in Section l
3 are total uncertainties with probabilities in excess of 95*..
I confidence limits are applicable only to a particular data sat, which in this case not available. Therefore, based on the mlatively small _nu=-
ber of reports indicating large values of deviation, i.e., the number of instances where a channel fails a functional test is very small as com-pared to the many thousands of functional tests per#cmed, Westinghousa believes that the total uncertainties presented on Table Ib are 95% preb-ability values at a high confidence level.
455
Y.
CONCLUSIONS The preceding sections, provide what is believed to be a reasonable means of accounting for instrt=:ent and measurement errors for four parameters used in the ITDP analysis. The assumptions used in this response are generic and conservative.
It is the intent of this response to generi-cally resolve any concerns with the measurement and control of Reactor Power, RCS Flow. Pressuri:er Pnssure and T,yg as they are applied to ITDP. As such, plant specific responses will provide only that infoma-tion which indicates that,1) the instrument and measurement uncertain-ties for that plant are consistent with or conservative with respect to those presented here, or 2) specific instrument and/or measurement uncertainties for that plant are not consistant wist th'ase pmsented.
In the second case the impact of the inconsistency on the four param-eters will be provided with corresponding new total uncertainties if the impact is sufficiently large.
4Eb
'M 6 h. s
___mm._
C- - s = %
at TAlllE 7h CCHPARISDN OF STATISTICAL FEill0DS 1
2 3
4 5
l Variance 95% Probability Variance 951 Probability i
Method 1 Hethod 2 Method 2 Method 3 Method 3 I
6a.c j
Pressurizer Pressure - Control T,,, - Control i
Steam 1f ne Pressure - Computer i
j Feedwater Temperature - Computer f
"p, Feedwater Pressure - Computer j
feedwater ap - Computer Pressurtzer Pressure - DVH l
Steamline Pressure - DVH Feedwater Temperature - DVH Tg - DVH Tg - DVH I
Hates for Table 7 b I
1.
Uncertainties presented in columns 1, 3, and 5 are in i span.
2.
While values noted are listed to the second decimal place, values are accurate only to the first decimal place. Second place is noted for round-of f purposes only.
i
A* e.
qo TABLE 8
. UNCERTAINTY PROSABILITIES TWo Sided Two Sided Nomal Prebability (5)
Nomal, Z Value
+a c PMA PEA SCA SD STE SPE RCA i
i RD I
RTE DVM ID A/D CA 4
l 40b