ML20141F595

Forwards Articles Re Turbine Missiles Cited in Memorandum & Order LBP-83-48,per Request.Related Correspondence

ML20141F595
Person / Time
Site:	Perry
Issue date:	01/06/1986
From:	Silberg J CLEVELAND ELECTRIC ILLUMINATING CO., SHAW, PITTMAN, POTTS & TROWBRIDGE
To:	Johnson W, Rosenthal A, Wilber H NRC ATOMIC SAFETY & LICENSING APPEAL PANEL (ASLAP)
References
CON-#186-683 LBP-83-48, OL, NUDOCS 8601090456
Download: ML20141F595 (44)

Text

...

~'

• n (y

u en,,,,

h ac-W SHAW, PITTMAN, PoTTs & TROWEbRIDGE A PARTNE ASMep OF PROFr$$aONAL CORPORAfiONS

'86 JMt -8 N0:27 1800 M STREET. N. W.

WASHINGTON. D. C. 20036 tt ttcopia n sacan saa eoes & saa-nee QGCL;$lha& ?ifi tetta 33/.NCd es-aeos isMAwtAw wswi Casts -sMAWLAW" JAY C. SILDERO, P C.

saca saa Oes January 6, 1986 Alan S. Rosenthal, Chairman Atomic Safety and Licensing Appeal Board U.

S. Nuclear Regulatory Commiccion Washington, D. C.

20555 Dr. W. Reed Johnson Atomic Safety and Licensing Appeal Board U. S. Nuclear Regulatory Commission Washington, D. C.

20555 Mr. Howard A. Wilber Atomic Safety and Licensing Appeal Board U. S. Nuclear Regulatory Commission Washington, D. C. 20555 Re:

The Cleveland Electric Illuminating Company (Perry Nuclear Power Plant, Units 1 and 2)

Docket Nos. 50-440 and 50-441 O b Gentlemen:

As requested, I am enclosing the following two articles cited in the Atomic Safety Licensing Board's Memorandum and Order (Summary Disposition of Turbine Missile Issue),

LBP-83-48, 18 NRC 218 (1983):

1.

S. H. Bush, "A Reassessment of Turbine-Generator Failure Probability," 19 Nuclear Safety 681 (1978);

n

(

~]

G}J pejP %8an Be8ajho o

h SHAW, PITTMAN. PoTTs & TROWBRIDGE A pantstmSM89 Or enor tssiONAL CompO1AfiCNs January 6, 1986 Page Two 2.

Patrick G. Heasler, " Missile Generation Rates From Historical Data," presented at Electric Power Research Institute Seminar on Turbine Missile Effects in Nuclear Power Plants (October 25-26, 1982).

Very truly yqurs, f,

G U, L1(

1 f

E ', Silberg o nsol for Applicants JES L Enclosures cc:

Service List (Enclosures only to (Ms. Woodhead and Ms. Hiatt) 4

\\

UNITED STATES OF AMERICA NUCLEAR REGULATORY COMMISSION BEFORE THE ATOMIC SAFETY AND LICENSING APPEAL BOARD In the Matter of

)

)

TNE CLEVELAND ELECTRIC

)

Docket Nos. 50-440 ILLUMINATING COMPANY, ET AL.

)

50-441

)

(Perry Nuclear Power Plant,

)

Units 1 and 2)

)

SERVICE LIST Alan S. Rosenthal, Chairman Atomic Safety and Licensing Atomic Safety and Licensing Appeal Board Panel Appeal Board U. S. Nuclear Regulatory Comm4ssion U. S. Nuclear Regulatory Commission Washington, D. C. 20555 washington, D. C. 20555 Dr. W. Reed Johnson Docketing and Service Section Atomic Safety and Licensing Office of the Secretary Appeal Board U. S. Nuclear Regulatory Commission

'U.

S. Nuclear Regulatory Commission Washington, D. C. 20555 Washington, D. C. 20555 Mr. Howard A. Wilber Colleen Woodhead, Esquire Atomic Safety and Licensing Office of the Executive Legal Appeal Board Director U. S. Nuclear Regulatory Commission U. S. Nuclear Regulatory Commission Washington, D. C. 20555 Washington, D. C. 20555 James P. Gleason, Chairman Terry Lodge, Esquire 513 Gilmoure Drive Suite 105 Silver Spring, Maryland 20901 618 N. Michigan Street Toledo, Ohio 43624 Jerry R. Kline Ms. Susan L. Hiatt Atomic Safety and Licensing Board 8275 Munson Avenue U.S. Nuclear Regulatory Commission Mentor, Ohio 44060 Washington, D.C.

20555 Glenn O. Bright Donald T. Ezzone, Esquire

. Atomic Safety and Licensing Board Assistant Prosecuting Attorney U.S. Nuclear Regulatory Commission Lake County Administration Center Washington, D.C.

20555 105 center Street Painesville, Ohio 44077 Atomic Safety and Licensing Atomic Safety and Licensing Appeal Board Board Panel U.S. Nuclear Regulatory Commission U. S. Nuclear Regulatory Commission Washington, D.C.

20555 Washington, D.C. 20555 John G. Cardinal, Esquire Prosecuting Attorney Ashtabula County Courthouse Jefferson, Ohio 44047

I jZl 681 1

. ~......

..r g g

.e _,

.ug r

Accident

~

86 'm a N :27

~

Analysis

}.

Edited by P. M. Haas h

A Reassessment of Turbine-Generator i

Failure Probability i

By S. H. Bush' Abstras.t: A prewous article' sn Nuclear Safety assessed the pertained to the assumptions used in the developmei.t overall probabahty (P.I of nuclear plant damage due to turbine of the failure probability (Pn) pertinent to nuclear fadures as a function of the combined probabdities of turbine fadure and elecnon of an energette massule (P,). a mistsle and disagreements, it was decided to expand on the strukmg a crttical component EP,). and srgmficant damaer occurring to the component IP,I. Due to questions raired items relevant to P, including experience since 1972.

concernmg the methodology used. the ralue of P, has been In essence, this article is limited to an assessment of P.

i reassessed, usmg a somewhat broader data base and other Failure probability in terms of Pn is defined as the methods of data analysts. The range ofinstantaneous turbine statistically determined probability of the generation fadure rates consnJered relevant to nuclear systems is of one or more missiles that penetrate the turbine 3.3 x 10** to 3.1 x 10-* per turbine year in the current article Compared to a Value of l % $0~' pe! tu!bine year in the SEiOg sob h3YO Sh0 pU$eOlbSi O AM3g ng Col $cS1

{

previous article reactor coniponents if the missiles strike them.

t On the basis of the information available, the instantaneous hazard function Z/T) at any time Twill A previous article' considered the general problem of damage to nuclear reactor components critical to be calculated rather than P. Depending on the data i

available, an alternative is to determine the cumulative safety due to the failure of the large turbine generator.

The damage probability (Pa) was determined from the hazard function #/T). The function 2/T) can be combined probabilities of turbine failure (P ),a missile obtained by differentiating #/T), or #fT) is the i

sinking a structure containing components entical to integral of 2/T). The value of 2/T)is derived from a safety (P:), and penetration or significant damage collection of those reported turbine generator failures occurring to the structure and component (P ),or considered relevant to nuclear reactor operation condi.

3 tions divided by the turbine years of operating turbine Pa = P x P x Ps generators. As indicated later, the hazard value is i

2 Since the publication of the previous article' in

• Spencer H. Bush is a senior staff consultant at Battelle 1973, several studies have been completed which agree Pacific Northwest Laboratories. Ite was a member of the with the ballistics missile strike probab,lity model used Adytsory Committee on Reactor Safeguards from 1966-1977 and served as chairman in 1971, ite has had considerable previously '* in calculating P. With regard to the "penente in the reactor Geld. including work in the physicat 8

3 damage probability Pa, any further modification awaits and mechanical metatturgy of nucleas materlats, effects of completion of jet sled missile tests, sponsored by the irradiation on metals and alloys, and stress corrosion. Ite is Electric Power Research Institute (EPRI).

active on severat national and international committees devel-The misunderstandings or disagreements arising opins safety codes and standards. He has done substantial werk from the original article' have almost exclusively on failure mechanisms of components in pressurised systems.

NUCLE AR SAFETY, Voi.19, No. 6. November-December 1978 J C.

3s2 ACCIDENT A**ALYSIS essentially equal to the failure rate, and therefore the The second option-confining the analysis to terms " hazard value," or " hazard rate" and " failure nuclear turbines or to nuclear turbines for light. water rate,"are used interchangeably here.

reactors (LWRs)-results in a small population in The Nuclear Regalatory Commission (NRC) has terms of units and turbine years of operation,particu.

not issued a "probabilistic" standard for turbine-larly if the population is limited to reactors within the generator failure. flowever, there is a greater goal of United States. The following data illustrate how few 10 ' as the probability for any specific major accident data exist.

during any year. This implies that 2/T) should not be greater than 10-* per turbine year. The three ap-For the United States proaches available to the utility are:

Number of operating 1.WRs through 1977 63

1. Orientation and placement of the turbine genera-Number of turbines 63 tor (s) to minimize Ps (e.g., peninsular orientation).

Total turbine years 367

2. Installation of missile barriers where protective Range of turbine sizes MW(e) 50-1200 orientation is not feasible, as is the case with platform-For other countries exclusive of Russia and satellites

• mounted nuclear plants.

3. Justification that the failure rate from all causes Number of operating reactors 30 Number of turbines 123 is sufficiently low. Such an approach has been reviewed Total turtine years

~1200 by NRC for acceptance of other types of failure.

Range of turbine sizes. MW(e) 30-1200 Several new plants have oriented the turbine

,Eluninated due to tack of data on turbines.

generators to minimize the possibility of missile strike on the critical components. The second and third approaches also have been investigated; however, the With regard to the first option-the one developed NRC has not accepted the third, namely, the proba.

further in this report-it should be recognized that bility approach for turbine failures at this time. The there are factors that could bias the statistical analysis:

importance NRC places on the turbine failure issue is

1. The turbines listed represent a large spectrum of apparent if one examines the development of their ages, sizes, and operating conditions.

positions pertinent to turbine generators.' The posi-2.The data from some manufacturers concerning tion of the Advisory Committee on Reactor Safeguards turbine operating histories were inadequate or non.

(ACRS) was given in a letter

• dated Apr. 18,1973,to existent.

Dixy Lee Ray, who was then Chairman of the Atomic

3. In the list, there is a mixture of both nuclear and Energy Commission.

nonnuclear turbines with a spectrum of operating conditions.

STATISTICAL ANALYSIS: PROBLEMS AND failu ALTERNATIVES 5.There is no assurance that the list of turbine llaving determined that a statistical analysis was failures is all. inclusive.

the most effective approach available, at least three 6.There is considerable subjectivity in deciding options existed:

whether or not a particular nonnuclear failure is I. Use as much data as possible in the statistical relevant to nuclear reactors; there is similar subjectivity analysis, recognizing that there are a number of concerning the applicability of degraded components limitations as noted iater.

in nuclear turbines where severe cracking, but not

2. Limit the analysis to turbines in nuclear plants, failure, has occurred.

recognizing the very small population and consequent Recognizing the above limitations, an effort was increase in the widths of confidence interval estimates.

made to establish the turbine population (nuclear as

3. Apply extreme value theory to specific compo.

well as nonnuclear) and to document failures.

nents of large steam turbines to assess the probability of generation oflarge missiles.

TURBINE YEARS OF OPERATION Although the third option appears quite attractive, probabilities based on as built quality do not cover it was necessary to make the following assumptions degradation mechanisms, such as environmentally in-when developing the body of data pertinent to turbine duced stress corrosion or malfunction of turbine valves.

years of operation:

NUCLE AH S AFETY, Vol.19. No. 6, Novemte-December 1978

ACCITENT ANALYSIS 643

l. In a few instances, data on operating years were Table I contains data that tend to substantiate available both before and after 1950. Such data were several of the assumptions made in inferring trends. In used with or without suitable modifications to cover this instance the number of units and the service years retirement of units.

were known. Twenty.seven new units were assumed to

,2.It was recognized that a substantial fraction of come on.line every year based on summing total new turhgie. generator units produced by some manufae.

units and. dividing by the number of years. Retirement turers were less than 50 MW(e). No attempt was made data are compared on the basis of arbitrarily retiring to eliminate such smaller units from the data.

units after 30 years or assuming that 5 units per year 3.In some cases the number of units fabricated are retired. Figure I presents the data in Table I both before 1950 was known, but not the operating years.

with and without retirement of units. In general, the The pre.1950 data were obtained by extrapolation of assumptions used appear to result in reasonably good trends in these cases.

agreement with the actual data. All data were from one i

4.Where data were limited to total units and total manufacturer.

operating years, an attempt was made to infer turbine Table 2 illustrates another trend in the data.

years per year, using trends observed in other data sets.

Manufacturer B produces turbine generators over the S. Arbitrary reductions in numbers of turbines entire range of sizes from < 10 MW(e) to the largest were made on the basis of an assumed turbine life of available. The data are presented in terms of units 30 years.

larger than 100 MW(e) as well as all units, regardless of

6. Data available before 1972 were used to infer size. In addition, new units are incorporated into.the turbine years of operation during the period 1972 to data set on the assumption that they operated either 6 1977.

months or 12 months in the first year. It appears, for

7. In the absence of knowledge relevant to turbine manufacturer B, that about 457 of the units are generators manufactured before 1950, the turbine

>toogw(e),

years began with operation of the first known unit (s).

Tables 3 and 4 represent a synthesis of the available 8.Where there was a clear delineatien between the information for new turbine generators and for cum t.

number of large turbines [> 50 MW(e) or > 100 MW(e)]

lative turbine years of operation. Table 3 covers the and small turbines for a given manufacturer, the data case of no units retired, whereas Table 4 assumes were examined,but the total population was used.

retirement after 30 years. Without re tiremen t, the

9. For those turbine manufacturers where data on population is about I x 10' turbine years. Assuming turbine years were totally absent, some arbitrary retirement, the total population represents almost assumptions were made to expand the turbine years 8 x 10' turbine years. An obvious consequence of and to factor in failures of units produced by these such a lirmted population is the broadening of confi.

manufacturers. Specifically,it was assumed that known dence interval estimates. These data are multiplied by manufacturers produced three. fourths of the units and 1.33 to include other turbine manufacturers. This that "other" manufacturers produced the rest; thus the figure may be too large or too small;however,it is felt turbine. years curve for known manufacturers was that the error is not too great. The maximum number multiplied by 1.33.

of turbine years without retirement by the end of 1977 10.Certain simplifying assumptions were made is about 1.33 x 10' years and with retirement about concerning new units operating in a given year and the 1.1 x 10' years.

fraction of a year assigned to such new units during One other trend is considered significant-the their first year of operation. These trends were increase in size of turbine generators with year of generally validated on the basis of deliberately synthe.

geration or order. Table 5 illustrates the change in sizing such data for manufacturers with known his.

size for all commercia! LWRs in the United States.

4 tories for new units.

It should be recognized that there are inherent ll. Data varying substantially in degree of com.

limitations in the turbine population given in Tables 3 pleteness were available from the following turbine and 4. An obvious trend is the increase in size of manufacturers: Allis Chalmers, Brown Boveri Com.

turbine generators with time. Units produced through.

pany, G.E.C. Turbine Generators, Ltd. (formerly En.

out the period 1930 to 1950 were relatively small, with o

glish Electric), General Electric Company, Kraftwerk newer units being larger. A second limitation was the Union (originally A.E.G. and Siemens), and Westing-degree of interpolation or extrapolation in the popula-house Electric Corp. No data were available for other tion of turbine generators. A third limitation was the turbine manufacturers.

lack ofinformation relevant to retirement of units.

NUCLE AR SMETY, Vol. 19. No. 6. Novemt>er-December 1978

.g j

g4 ACCIDENT ANALYSIS Table I Compag of Actual, Interpolated, and Extrapolated Turbine Years for Manufacturer A Assumes 27 new w-Assumes 27 new Actu al e xpe rience, Actual experience, units per year, Actual experience, units per year, with retirement with retirement of retirement of without retirement without retirement after 30 years 5 units per year 5 units per year No, Service No.

Service No, Service No, Service No, Service

'Y Year units years units years units years units years units years Pre 1950 166 2,012 166 2,012 166 2,012 166 2,012 166 2,012 k

1950 186 2,198 193 2,205 182 2,194 181 2,193 188 2,200 1951 224 2,422 220 2,425 220 2,414 214 2,407 210 2,410 1952 249 2,671 247 2,672 245 2,659 234 2,641 232 2,642 1953 293 2,964 274 2,946 289 2,948 273 2.914 254 2.896 1954 348 3,312 301 3,247 340 3,288 323 3,237 276 3,172 1955 385 3,697 328 3,575 372 3,660 355 1,592 298 3,470 1956 410 4,107 355 3,930 390 4,05 0 375 3,967 320 3,790 1957 445 4,552 382 4,312 418 4,468 405 4,372 342 4,132 1958' 494 5,046 409 4.721 464 4,932 449 4,821 364 4,496 i

l 1959 526 5,572 436 5,157 484 5,416 476 5,297 386 4,882 1960 562 6,134 463 5,620 506 5,922 507 5,804 408 5,290 1

1961 576 6,710 490 6,110 518 6,440 516 6.320 430 5,720 1962 593 7,303 517 6,627 533 6,973 528 6,848 452 6,172 1963 612 7,915 544 7.171 551 7,524 542 7,390 474 6,646 1964 637 8,552 571 7,742 576 8,100 562 7,952 496 7,142 1965 652 9,204 598 8,340 591 8,691

$72 8,524 518 7,660 1966 671 9,875 625 8,965 608 9,299 586 9,110 540 8,200 1967 687 10,562 652 9,617 621 9,920 597 9,707

$62 8,762 1968 710 1I,272 679 10,296 638 10,558 615 10,322 584 9,346 1969 729 12,001 7')6 11,002 652 11,210 629 10,951 606 9,952 1970 747 12,74H

' 13 11,735 664 11,874 642 11,593 628 10,580 1971 765 13,513

' 50 12,495 673 12.547 655 12.248 650 11,230

• 1972 785 14.298 787 13,272 680 13,227 670 12,918 672 11,902 1973 805 15,103 814 14,086 690 13,917 680 13,603 694 12,596 1974 825 15,928 841 14,927 711 14,628 705 14,308 716 13,312 1975 845 16,773 868 15,795 732 15,360 725 15.033 738 14,050 1976 865 17,638 895 16,690 756 16,116 745 15,778 760 14,800 1977 885 18.523 922 17.612 780 16,896 765 16.543 788 15,588

• Extrapolated below line. To correct for partial years of sersice, subtract 357 turbine years.

In the case of manufacturer A,the data on number 2 and Fig. I to permit extrapolation. In essence, the of units and years of operation were quite good. In data available censisted of a given number of units with addition, the set was more homogeneous in size since the cumulatise turbine years cited at one point in time; all units were larger than 50 MW(e). The quality of theafore it was necessary to extrapolate both back.

data for manufacturer B is similar to that for manu.

ward and forward to develop the data given in Tables 3 facturer A; however, the unit sizes include both small and 4,

[-10 MW(e)] and la ge [>$0 MW(e)], The same is true for manufacturers D and F. Manufacturer C FAILURES AND FAILURE MECHANISMS discontinued manufacturing turbine generators in 1968; however, the data set for C is similar to A in Turbine generator failures during the past 25 years quality, Data from manufacturer E was quite poor, are presented in Tables 6 and 7. Table 6 covers failures necessitating the procedures developed in Tables I and of large and medium steam turbines at or near NUCLE AR SAFETY, Vol.19. No. 6, November-December 1978 E

M; ACCIDENT ANALYSIS 805 1&MO Actu.i espo,6ence, udthout todroment

/

14.C10

.,(

Assume. 27 new units per ve.r. without retirement

,,fl, *,/ e4 w.i.x,o_.,ee,em.nt.n., m,o-

/

3,

1. .jp Ac,u.,es,.,_

ee,em,so,. -. -,es,.

p/

/

j

,, 12,000 Assumes 27 new unats per yo.r, retirement of 5 units per ye.r

= * - -

/*

8

,/

/

~

3

./

y tcLo00

$f,/j-f

/ /*

~

hooaa -

gk

~

0000 gI

. f Entreporated 4000 -

,M

/

~

I I

I f

1 I

t i

I i i t

1 I

i t

I t

i i f

I t

i g

te6o

'52

'54

'56

'58 to 12 14 to 18 70 72 74 78 YEAR Fig. I Comparison of actual, interpolated, and estrspolated turbine years with and without retirement of units for rnanufuturer A.

operating speeds. Table 7 covers cases of turbine heat. treatment procedures so that failures due to this overspeed with and without failure.

mechanism are less likely to occur in modem plants.

Although there are several failures cited in Tables 6 Several failures elsewhere in the turbine have and 7, the critical question is how many are relevant to occurred because of some aspect of generator failure.

turbine generators used in nuclear power plants. Any Events such as abrupt braking, running of the generator decision concerning relevance will be subjective. As as an induction motor, and out.of. phase have caused indicated in Tables 6 and 7, such a subjective judgment severe damage to turbine generators, but only rarely placed the minority of the failures in the relevant have missiles been generated extemally. This has been category. The following discussion develops the ratio-due to the massiveness of the stator and the shell of the nale for division into relevant and irrelevant failures.

genera tor.

An additional subset pertinent to relevant failures A number of failures have occurred in the test pit covers those failures generating energetic external or preoperationally. Considering the causes of such missiles-the only condition of significance in assess-preoperational failures, it is improbable that these ing the damage probability, turbine generators would have survived the tests and then failed in service.

Failures irrelevant to Nuclear Units or Not Capable of Major Missile Generation Failures Flefevant to Nuclear Units One class of failures occurring 20 to 25 years ago The remaining turbine generator failures can be was the brittle fracture of turbine or generator rotors.

considered marginally or directly relevant to nuclear Seven such failures occurred over a relatively short p' ants. The Tanners Creek stress rupture and the time. All were characterized by high nil ductility Gillatin creep fatigue failures are considered to be temperatures and hydrogen present as fisheyes, etc.

marginal, since nuclear service conditions are below the These failures led to changes in melt, fabrication, and range considered relevant for stress rupture.

NUCLE AR SAFETY. Vol.19. No. 6, November-December 1978

1*

000 ACCIOGNT ANALYSIS Table 2 Tuebine Years foe Two Casesocies of Turbine Generatoes foe Manufacturer 3 TurWne pnerators larps than g'

100 MW(e)

All turWne pnerstors TurWne years TurWne years Asasmes Assumes Assumes Aansmes Cume.

% mewin all new Cume.

% new in all new Year No.

lathe first year one year No.

lative nest year one year Pro.1950 1

1 2

2 56 500 1950 2

3 4

5 39 95 575 595 1951 4

7 9

12 31 126 686 721 1952 7

14 19 26 26 152 825 873 1953 14 28 40 54 33 185 993 1,058 1954 24 52 80 106 49 234 1,203 1,292 1955 21 73 143 179 35 269 1,455 1,561 1956 12 85 222 264 15 284 1,732 1,845 1957 7

92 310 356 26 310 2.019 2,155 1958 14 106 409 462 32 342 2,345 2,497 1959 21 127 525 589 30 372 2,702 2,869 1960 12' 139 658 728 24 396 3,086 3,265 1961 18 157 806 885 26 422 3,495 3,687 1962 17 174 972 1,059 28 450 3,931 4,137 1963 10, 184 1,151 1,243 19 469 4,390 4,606 1964 8

192 1,339 1,435 18 487 4,868 5.093 1965 12 204 1,537 1,639 25 512 5,367 5,605 1966 11 215 1,747 1,854 28 540 5,893 6,145 1967 19 234 1,971 2,088 31 571 6,450 6,716 1968 18 252 2,214 2,340 26 597 7,034 7,313 1969 19 271 2,485 2,611 23 620 7,643 7,933 1970 22 293 2,757 2,904 22 642 8,274 8,575 1971 11 304 3,058 3,2 0 8 _ __.

20 662 8,926 9,237

• 1972 14 318 3,369 3,526 1 23 685 9.599 9.922 1973 14 332 3,694 3,858 20 705 10,294 10,627 1974 14 346 4,033 4,204 25 730 11,012 11,357 1975 14 360 4,386 4,564 25 755 11,754 12,112 1976 14 374 4,753 4,938 25 780 12,522 12,892 1977 14 388 5,134 5,326 25 805 13,314 13,697 d

• Extrapolated below hne, This leaves the following failures as relevant:

2 The overspeed incidents leading to failure l.The initial liinkley Point A brittle. stress.

(liskmouth, Calder Hall, and Bold) are considered corrosion failure and the Duquesne Shippingport fall.

relevant even though Calder Itall occurred during the ure are both relevant and directly applicable to nuclear startup phase, ne possibility exists for a mechanism, g

units since both plants are nuclear. The other two such as the occurrence of foreign bodies in a system, Ilinkley Point failures occurred during pit testing; which could prevent valve operation, it should be hence they are marginal at best and probably should noted that several nuclear plants have overspeeded f

not be used in the body of statistics. The Oak Creek without damage. The reasons advanced by Splitt.

failure is considered a relevant brittle failure The disk gerber,' lluppman,' and Carson et al.' are considered j

cracking at Rancho Seco and Arkansas Nuclear One relevant to nuclear turbines. Carson et al.' cite,but do were not failures, but such cracking should be con.

not identify, cases of modern turbine generators going sidered a waming that mechanisms exist which, if into overspeed due to rust in valves resulting frorn undetected, could lead to failure.

water in the hydraulic fluid. Other causes may lead to k

NUCLE AR 8AFETY, Vol.19, No. 6, Novemtwr-Decomtwe 19FJ

ACCs00887 A8 sat,Y888 es?

Table 3 Total Years Synthesised-No Cervecsion fee Iteelseauet of TurWae43enerosses*

teamm8menen A

B C

D E

F Teset Year Unies Yeass Unem Veen Unem veen Unem Yeen Unie veen Unem Yean Unie Yeen j

Fw.1990 166 2,012

$6 300 40 160 233 I,$07 ISO l.000 600 2,400 1,247 7J79 i

1950 184 2,198 93 393 48 200 23 3 I,760 173 1,173 412 3,012 1.349 8,948 19$1 224 2,422 126 721

$6 264 281 2,048 200 1,373 620 3,632 1,307 10,433 1932 249 2,671 132 873 el 32$

303 2,34 6 223 1,600 H2 4,274 1,634 12.089 1953 293 2.964 183 1,038 78 396 329 2.673 250 1,850 639 4,933 1,787 13,876 1954 Me 3,312 2 34 1.292 78 474 M0 3,033 273 2,125 689 3,422 1,9H 13.860 1953 38$

3,697 269 IJ61 87

$68 349 3,424 300 2,423 fl0 6,332 2,144 18,004 1956 410 4,107 2H I,M $

to 637 408 3,832 123 2,750 732 7,084 2,276 19.280 1957 44$

4,$32 310 2,0$$

102 739 443 4.277 350 3,100 791 7,873 2,444 22,728 1958 494 3,046 342 2,497 122 881 466 4,H 3 373 3,473 823 0,700 2,423 23,34 3 1939

$26 SJ72 372 2,849 130 1,018 494 3.237 400 3,873 831 9J31 2,774 28,123 1960

$62 6,134 396 3,245 148 1,152 312 3,749 430 4,303 844 10,433 2,926 31,049 1961 376 6,710 422 3.687 131 1,303

$37 6,286 460 4,743 904 11.339 3,031 34,100 1962

$93 7,303 430 4.137 134 1,459 343 6,849 490 3,233 9H 12,373 3,187 37,287 1963 612 7,913 469 4,604 168 1,620 393 7,444 320 3,77$

953 13,228 3,313 40,600 1964 637 SJ32 487 5,093 167 I,787 424 8,070

$4$

6,320 976 14,204 3,445 44,043 196$

652 9,204 312 3,603 168 1,933 633 8,723

$70 6,890 997 13,204 3Js3 47J83 1964 471 9,873 340 6,145 148 2,123 677 9.402 600 7,490 1,018 16,219 3,673 31,263 1947 687 10J62

$75 6,716 168 2.191 699 10,101 623 8,113 1,0$ $

17,274 3,406 33,081 1968 710 11.272

$97 7,313 164 2,459 719 10,820 630 8,763 1,002 18,336 3,927 60,013 1969 729 12,001 620 7,933 168 2,627 73')

11J70 673 9,440 1,113 19,471 4,058 64,063 1970 747 12,748 642 8,375 ist 2,795 774 12,344 710 10,155 1,139 20,610 4.181 64,244 1971 763 13J13 642 9,237 168 2.943 798 13,142 740 10,890 1,150 21,764 4,292 72,334 1972 785 14,298 685 9,922 168 3.138 829 13,978 770 11,640 1.180 22,948 4.418 77,054 1973 ISOS 43,103 703 10,627 168 3,299 H4 14,813 000 12,460 1,210 24,158 4J33 81J90 1974 825 13,928 730 11.337 168 3.447 868 13,433 830 13,270 1,240 23,398 4J47 86,104 1975 MS 16,773 735 12.112 let 3.433 890 16,373 860 14,130 1,270 26,648 4,703 90.883 1976 MS 17,638 780 12,892 168 3,803 890 17J45 900 13.030 1,300 27,948 4,831 95.786 1977 885 18J23 803 13,697 168 3,971 090 18,433 930 13.980 1,330 29,298 3,000 99.924

• To correct ror time of startup in a year, subtract tels trom turbine years.

tEntr*904*i+4 6*Ia lm.

failure of valves to close with the potential of Electric Corp., etc.), ne second covers failures for

,l destructive owrspeed.

which operating histories are not known (Charles A, If one assesses the relevant cases in Tables 6 and 7, Parsons, etc.),IUustrating the reason for expanding the I '*I d

one note $ two cases at or near operating speed where failure extemal missues were generated. Both occurred within

! l tidered umhauene ln Fanum ha the past to years. Dere were Aw overspeed event 8 ng the resulting in extemal missiles, and all occurred more An assessment of the failure deta in Tables 6 and 7 y

anism, than 10 years ago. However, several cases of overspeed reveals several problems with their use in terms of their

ystem, without damage have occurred, and all were within the relevance to failures in nuclear plants, nree of the ild be past 10 years. Additionally, two nuclear plants have more obvious are (1) there may be a significant number m oed experienced phosphate buildup on the turbine valves, of failures in units produced by manufacturers other o

splitt.

which could influence closure and result in overspeed, than those included in Table 3;(2)the unit si2es are dened A8 noted in Table 7, there were other causes of valve sometimes much smauer than units uwd in nuclear but do malfunction.

plants; and (3) operating pressures and temperatures s going Table 8 presents failures within two sets. De first are not always typical of LWRs.

I I'*

co wr8 failures where turbine operating times are A further limitation 18 that the ilsting is not lead to known (General Electric Company, Westinghouse complete. Additional failures are known by hearsay; NUCLE AR SAF87Y, Vol.19, No, 6, November-Deeenter 197e

l ACCIDCT AAALYS88 egg s

.s

?

bTable 4 Total Years Synthesized: Corrected for Retirement of Turbine Generators

• i Manufacturers Year A

B C

D E

F years Pre-1953 2,012 500 160 1,507 1,000 2,400 7,579 1950 2,193 591 207 1,760 1,170 2.992 8,913 1951 2,407 709 262 2,041 1.361 3,572 10,351 1952 2,641 849 320 2,346 1,570 4,154 11,880 1953 2,914 1,019 387 2,675 1,800 4,733 13.358 1954 3,237 1,233 460 3,035 2,050 5,322 15,337 1955 3,592 1,478 541 3,424 2,320 5,912 17,267 1956 3,967 1,734 630 3,832 2,610 6J22 19,395 1957 4,372 2,012 724 4,277 2,920 7,153 21,458 1958 4,821 2,318 837 4,743 3,250 7,798 23,767 1959 5,297 2,650 937 5.237 3,600 8,449 26,170 1960 5,304 3,002 1,087 5,749 3,975 9,113 28,730 1961 6,310 3,378 1,226 6,286 4.375 9,777 30.362 1962 6,848 3,776 1.369 6,839 4,800 10,451 34,082 1963 7,390 4,184 1,516 7,414 5,250 11,126 36,880 1964 7,952 4,716 1,668 8,010 5,720 11,802 39,918 1965 8,524 5,164 1,820 8,625 6,210 12,479 42,222 1966 9,110 5,636 1,971 9,252 6,725 13,157 45,851 1967 9,707 6,135 2,121 9,821 7,260 13,852 48,966 1968 10,322 6,656 2,270 10.540 7,815 14,554 52,157 1969 10,951 7,196 2,418 11,210 8,390 15,269 55,434 1970 11,593 7.754 2,565 11,894 8,995 15,988 58,789 1971 12,248 8,328 2,711 12,592 9,625 16,706 62,210 1972 12,918 8,921 2,856 13,311 10,280 17,426 65,712 1973 13,603 9,530 3,000 14,035 10,960 18,156 70,284 1974 14,308 10,160 3,143 14,773 11,665 18,896 72,945 1975 15,033 10,811 3,285 15,525 12,395 19,646 76,695 1976 15,778 11,483 3,426 16,291 13,150 20,406 80,534 1977 16,543 12,176 3,566 17,071 13,940 21,176 84,472

• These data were obtained from Table 3 by utilizing actual experience and retiring units after 30 years when such data were available. When the data were not available, the trends from earlier years were used to estimate retirement.

however, the racessary data to permit their use are resulting from a combination of material properties and environment such as corrosion fatigue. In recogni Beet lacking.

A less apparent limitation has to do witti the tion of these limitations, one should question the data heterogeneity of the failure set,which should influence absolute validity of the failure probabilities.

extr-the validity of the statistical techniques used.The two estir methods that were examined,i.e., the Duane leaming-curve model and the Weibull failure model, probably STATISTICAL EVALUATION OF FAILURE

Pres, are valid for subsets of failures but not for the total RATES population. Some clear-cut failure subsets include popt 8

brittle fractures due to meh practice (1953-1956),

The approach used in the previous article in exar overspeed failures due to valve malfunction determining the cumulative and current failure rates mari (1956-1960), and generator failures due to field utilized the Duane growth model.8' Questions were lunit failures, etc. Another subset includes high. temperature raised conceming the use of the model, selection of tech creep fatigue, This leaves a residuum of failures, usually data points, and the lack of standard error values,8 8 PainI NUCLE AR SAFETY, Vol.19, Now 6. November-December 1978 i

ACCIDE!T ANALYSIS ggg Table $ Exemple of Increase la Turbine 4enerstoe Output [MW(e)]

)

with Time Based on Evaluation of U. S. Commercial Nuclear Power Plants and on Year of Initial Operation size, uw(e)

Year

<100 101-200 201-300 301-500 501-800 801-1000 >t000 6

1957 1

I 1958 1959 1

1960 1

1961 1962 1

1 1963 1

1964 1965 1966 1967 I

I I

1968 1969 1

2 1970 1

3 1

1971 3

1972 1

4 2

1973 1

2 4

4 1974 1

6 5

2 1975 2

2 1976 5

2 1977 6

1 1978 1

4 2

n.

1979 1

7

}

1980 2

12 1981 1

1 1982

.{

1983 1

3 8

j 1984 1

2 1985 1

4 1986 i

e 1987 2

1

't 1988 1

1989 4

' l'l nrties cogni.

Beeth and Hobbs,8 8 by appropriate selection of the ne approach discussed by Nelson'* has been used

.l

.1 the data, obtained higher values of alpha and lowr in the plotting of data. Nelson points out that the I

extrapolated failure rates.

cumulative probability value F(T) and the cumulative

-l A logical starting point is to calculate the global hazard function H(T) are essentially equal estimates of failure rate using the total turbine years or

[F(T)a'H/T)] for small probabilities (<!%), and the IRE the total population of turbines. Dese data are hazard function 2/T) can be described as the instanta.

ll presented in Table 9. He need for an expanded neous failure rate at time T for these small proba.

population of turbines should be apparent after an bilities.

e' in examination of Table 8. Rese data serve as bench The failure model believed to yield the most rates marks for the time. dependent analyses. Reir value is meaningful values of reliability, cumulative failure rate limited in that neither improvement in manufacturing

[H/T)], and instantaneous failure rate [2(T)] is the were on cf techniques nor in operation are apparent from such Weibull. It has been used extensively in the evaluation ses.' '

Point values.

of both large and small populations of pressure NUCLE AR SAFETY, Vol 19, No. 6, Nowmtw-Oecember 1978

ese ACCIDENT ANALYSIS

.f' Table 6 Known Fauures et or Near Operating Speeds (Medium or large Stearn Turt>lnes)

Amoeg Manufacturer

Sise, Year of Eaternal manufactusers (if known)

MW(e) failure Type of faawa Cause of fadure*

missues Comments A-F

l. Semens 63 1958 Low pressure turbine Bnttle fauvre (M)

Yes Factory test Yes retor burst

2. Eacher Wyss (ElectricitJ

$0 1911 Yes Yes de France Dieppedalle)

(54?)

3. General Doctric 100 1953 First-stage disk broke Hist > temperature No Yes (Tanners Creek 1) rupture (M)

4. General Electric ( Arnona 100 1954 Rotor burst Bnetle fadure (M)

Yes Factory test Yes Public Services)

5. General Electric 150 1954 Rotor burst Brittle frseture through No Yes (Cromby l) repair (M)

6. All*Chalmers(Common-150 1954 Spindle burst Brittle fracture (M)

Yes Yes westrh Edison)

7. Charles A. Parsons 100 1954 Generator retaining Brittle fatture through Yes umited missiles No (Hearn 1) nas burst vent holes (M)

8. Charles A. Parsons 100 1954 Generator retaining Bnttle fauure through Yes No i

(Hearn 2) ring burst vent holes (M)

9. General Dectric 125 1956 Generator rotor burst Brittle fracture (M)

No Yes (Pittsburg 1. Pacaric Gas & Electric)

10. Escher.Wyss (Pegun.

45 1959 Rotor fanure Brittle fracture (M)

Yes Yes Utrecht)

18. General Dectric (Cutler 6 125 1969 Generator field Out of step (0)

No Yes florida Power & Light) windmg

12. G.E.C. Turbme Generstors.

87 1969 Duk fadure Brir'te fauure (M.E)

Yes Nuclear Yes Ltd.(Hmkley Point A 5)

13. G.LC. Turbme Generstors.

87 1969 Disk faGure Brittie fanure (M.E)

Yes Factory test Yes Ltd. (H nkley Pant A4)

14. G.E.C. Turbine Generators.

87 1970 Disk faGure Bnttle failure (M.E)

Yes Factory test Yes Ltd.(Hinkley Point A4)

15. Mstsubisha (ENES A) 330 1970 Rotor fauure Flawed? (M)

Yes Factory test No 86 General Electric (Northers 63 1971 Generstor fadute Braking (O)

No Yes States Power)

17. General Electric (Essez 1.

105 1972 Generator field fauure Abrupt braking Yes Coupling as misage Yes Public Service Doctric

& Cas)

18. General Doctric (Sendai) 1972 Generator No Yes

19. Mitsubishi(Kainan) 600 1972 Generator rotor faDure Design?

No Preoperational No

20. Charles A. Parsons 500 1974 Generator ring Plastic strain plus No No (Nanticoke) hydrogen

21. Westinghouse (Duquesne 150 1974 Disk failure Brittle fagure stress No Nuclear Yes Shippesport) corrosion (M.E)

22. Westinghouse (TVA Gallatin) 1974 Rotor fsDure Fatigue (M)

Yes Yes

23. Bro==Boven Co, 1975 Generator failure (0)

No Yes

($kserbaek. Denmark)

24. General Electric (Utah 1976 Gensrator failure Ran sa induction No Ye:

Power & Light) sector; operator W

ener

25. AttieChalmers(Oak Creek 130 1977 Last-stage disk. low-Probebly brittle faaure Yes Twe large places Yes g'v Power Co. 3. Wisconsin pressure turbine 51 Dectric)

26. Alstrom-Rateau (Dectricite 600 1977 Generator rotor locked Abrupt braking Yes Only couplings No ca de Francs, Porchev9le, during no load over-g France) speed test fo Westinshouse ($ MUD, Rancho Seco)t 900 1975 Cracking turbine disks Stress corros6on (M.E)

No Nuclear A.E.G. (Wurgassen)t 670 1976 Cracking shaft Fatigue plus stress No Nucles' las corrosion (M.E)

Westinghouse AN,01. Na, it 900 1977 Cracking disks Stresa corrosion (M.E)

No Nuclear a1

• (M) = metallurgical;(E) = environmental; and (0)

• operational Of 1 Cracking only; not considered fsDure, fa.

Of Wt NUCLEAR SAFETY, Vol.19. No. 6. Nowmber-December 1978

ACCIDCT ANAL.YSIS tel e

Table 7 IncWests of Oeuropeed ConsWered Italewet to Nuclest Plants with or Without Fallwe*

Asmoeg Monstecturert

Shee, Yeme of Cause of Emterunt unanufacsueers

g (if known)

MW(e) faamse oveespeed fauere missSes Comments A-F

. asus Ovesarmed with Pasues

' ~

Fraser and Dalmers (CEGB, 60 1956 stuck valves;magneute buedup Yes No j

Uskmouth)

Charle: A. Persons (UKAEA, 23 1958 Valves pf40ged with foreign Yes No Calder Hau) material; shot f'om shot blasting Unknown 100 1958 Valves stuck No No Unknown 16 1958 Operation Yes No Generel Doctric (Morenci 3) 12 1959 Out of phase Yes 150% overspeed Not GEC-CEGB (Bold) 30 1960 Stuck walves; salt bulldup Yes

>l50% overspeed Yes Owrapeed Without Failure Westmshouse (Now Castle) 100 1952

<!30% overspeed Yes Obrigheim 320 1952

<l50% overspeed nuclear

?

SEN A, Chaos 200 1952

<l50% overspead, nuclear

?

Wesunghouse (San Onofre 1) 1952

<l50% overspeed, nuclear Yes FsMure of Valve to Function with or Without Overspeed Westir shouse (Turkey Point) 1974 Phosphate bugdup; two stop valves faBed

i Westinghouse (H. B. Robinson) 1974 Phosphate buildup;stop valve failed

s Westinghouse (ladian Pomt) 1974 Operator error; stop valve faBed

- s Westinghouse (Pomt Beach 2)

.1975 Packms too tight; stop valve faded s

General Destric (oyster Creek) 1970 Power transient; control valve failed s

General Doctric (Mdletone 1) 1971 Defective control volve General Doctric (Dresden 2) 1972 Faulty solenoid; two control

- a valves faded a

General Dectric (Dresden 2) 1974 Controlvalve faded Westinghouse (Turkey Point 4) 1974 Spring bolt faced;controi

s valve faaed General Cases of Overspeed with oe Without Fatures 4

ss 17 cases 1951-1961 e

11 cases 21 of 29 are German 1961-1965

' o I case 1965-1970

- es eData are primardy from Raft 7 and 8.

tCEGB, Central Dectrietty Generating Board;GEC, G E.C. Turbine Generators, Ltd.;UKAEA, United Kingdom Atomic Energy Authority.

' fee iNot in General Dectric large steem turbme data.

fes fee vessels.e s-is Other time.to. failure models, such as cumulative hazard function H/T), and time to failure

-fas exponential, gamma, and log-normal, are more re.

77#). 'the failure rate at any time Tis given by:

stricted. Unlike the exponential model, the Weibull is 4.

capable of representing hazard rates that vary with 2(TJ =1 8-8 T

time. It was necessary to place the equation in linear a8 form to permit a regression analysis. 'Ih3 procedure used is cited elsewhere.

This equation permits one to determine the time-Table 10 contains all the data necessary to conduct dependent failure rate for any time during operation a linear regression analysis for any specific combination and to extrapolate to end oflife. Figure 3 covers the of failures, e.g., all failures, all missiles, relevant two cases of all failurel and failures with missiles. On failures,'and relevant missiles (Fig. 2). The terminology the basis of these curves, there is noindication of wear of Ref.12 is used in this table. Table 10 includes the out near end of life. Using the equation in Table 10.

Weibull functions for reliability R(T), failure rate 2(T),

2(T)= 1.79 x 10-' T-o.se, the instantaneous (time-NUCLEAR SAFETY, Vol.19, Na 8, Novernber-Decorrd>er 1970

ACCIDENT ANALYCS g33 Table 8 IJeting of Turbine-Generator FaBures Divided into Manufacturers for

'Whom Operating Experience Is Known and Those for Whom Operating Experience is Not Known* (Failures Relevant to Nuclear Operation Are Noted)

Experience known Experience not known Total Extemal External External Year No.

I missiles z

No.

I missiles E

No.

E missiles I

1951 1

1 1

1 1-1 1

1 1953 1

2 1

2 1954 3

5 2

3 2

2 2

2 5

7 4

5 1956 1

6 It 3

1*

3 2

9 1

6 1958 3t 6

23 5

3 12 2

8 1959 It 7

1*

6 1

13 1

9 1960 It 7

12 4

1 14 1

10 1969 3t 10 23 6

3 17 2

12 1970 1

11 1

7 1

8 1

7 2

19 2

14 1

20 1971 1

12 1972 2

14 1

9 3

23 1974 2t 16 1

8 1

10 3

26 1

15 1975 1

17 1

27 1976 1

18 1

28 1977 2t 19 13 9

1 11 2

30 1

16

'Above does not include cases of disk or shaft cracking without failure (Rancho Seco,Wurgassen,and ANO I No.1) or overspeed without failure (New Castle, Obrigheim, SENA, and San Onofre 1).

tFailure considered relevant to nuclear operation.

$ Relevant failure, e xternal missiles.

dependent) failure rate 2/T) for the relevant missile two. thirds for failures early in life and approximately case varies from about 1.5 x 10-3 to I.S x 10-2 per one. half for failures later in life to eliminate smaller turbine year. Since these are expressed in percent, a units and those with different operating conditions conversion to rate yields 1.5 x 10-s to 1.5 x 10-*

than experienced by nuclear turbines. Again, this per turbine year.

approach is quite arbitrary, but it does have the effect Table 10 includes Weibull distributions covering of shifting the regression line t'o higher values of H and relevant missiles in the context of relevance to nuclear changing the slope 1/0. The relevant missile case, corrected for turbine population as defined in reactors. This relevance, as cited earlier, is a highly Table 11, yields failure rates [2/T)/100] varying from subjective judgment. (Because of the subjectivity in 3.3 x 10-s to 3.1 x 10-* per turbine year compared selecting data points,it was not considered appropriate to values of 1.5 x 10-s to 1.5 x 10-* per turbine to provide estimates of the confidence intervals for the Weibull parameters a and 0.) Even if the selection is year for the relevant missile case without correction for accepted, a valid question can be raised conceming the turbine population.

number of turbines used as a denominator in calcu.

lating the hazard function in Table 10. If one is CALCULATION OF FAILURE RATES selective in the numerator values, it follows that one BY TURBINE MANUFACTURERS may need to be selective in the denominator. This approach of adjusting the denominator to reduce the Reports concerning the probability of turbine.

effect of the nontelevant portion of the population was generator failure at design speed and overspeed have explored for the relevant missiles case. These revised been prepared by turbine manufacturers and are often data are presented in Table 11. The population after incorporated in utility safety. analysis reports for retirement was considered as a base line, and then licensing purposes. The Allis Chalmers-Kraftwerk expanded by 1.33 to cover those failures outside the Union reports are proprietary and will not be discussed known population. This population was reduced by other than to note that the approaches used in NUCLE AR SAFETY, Vol.19. No. 6. November-December 1978 L

ACCIDENT ANALYSIS 693 Table 9 Global Estimates of Failure Rates for Various Assumptions as of the End of 1977 g

i Number Total Total f

of turbine units in Cases considered failures years service Failure rate, A All available information.

population x 1.33 without retirement All failures 30 133,000 6680 2.3 x 10-* *

,h 4.5 x 10-'t All missile-generating failures 16 133,000 6680 1,2 x 10-*

2.4 x 10-8 Relevant failures 9

133,000 6680 6.8 x 10-'

l.3 x 10-8 Relevant missiles 7

133,000 6680 5.3 x 10-8 1.0 x 10-8 All available information, population x I.33 with retirement All failures 30 112,600 5250 2.7 x 10-*

5.7 x 10-s All missiles 16 112,600 5250 1.4 x 10-*

3.0 x 10-s Relevant failures 9

112,600 5250 8.0 x 10-8 1.7 x 10-s Relevant missiles 7

112,600 5250 6.2 x 10-'

l.3 x 10-'

Available in data set without retirement All data set failures 19 100,000 5008 1.9 x 10-*

3.8 x 10-'

Data set missiles 9

100.000 5008 9.1 x 10-s lately 1.8 x 10-8 I"

Relevant data set failures 4

100,000 5008 4.0 x 10-8 tions 8.0 x 10-*

this Relevant data set missiles 3

100,000 5008 3.0 x 10-8 6.0 x 10-*

ffect Available in data set with

fand retirement E**#'

All data set failures 19 84,470 3938 2.2 x 10-*

d in 4.8 x 10-'

from Data set missiles 9

84,470 3938 1.1 x 10-*

2.3 x 10-'

pared Relevant data set failures 4

84,470 3938 4.7 x 10-'

!rbine 1.0 x 10-8 t

3" IO' Relevant data set missiles 3

84,470 3938 3.6 x 10-8 7.6 x 10-*

• Per turbine year.

tPer turbine unit.

developing overspeed probabilities were quite similar to similar approach was used for brittle failure near rbine-the approach of General Electric and Westing-operating speed. Values were obtained by either Monte I have house.8 '

Carlo or importance sampling.

cftes Both General Electric and Westinghouse used a Basically, the General Electric model of the event is ts f;r fault tree approach utilizing the available data on a sequence of simple events using failure rates from

ftwerk functional reliability of components to calculate proba-electronic components, control valves, stop valves.

cussed bility of overspeed in nuclear turbine generators. A overspeed trips, etc. Additionally, the sensitivity of td in NUCLEAR SAFETY, Vol 19, No. 6 November-Decernber 1978 Imm

[

5

\\

P I

l

$m Table 10 Data Arranged for Regression Analysis in Calculating Weibull y

Distribution (Maximum Population Assumed Without Retirement x 1.33) m

-4

"8 Failures llazard, %

Cumulative harard, %

g r-greater than failed unit All Relevant Relevant All Relevant Relevant All Relevant Relevant j;;

Failure, year failure time before failure failures Missiles failures missiles failures Missiles failures missiles failures Misiles failures missiles z9 Arizona Public

?

Service (1954) 6675 0.08 1

0.015 0.015 z Calder 11211 (1958) 6675 0.08 I

I I

I 0.015 0.015 0.015 0.015 0.030 0.015 0.015.

0.015 3 Siemens(1951) 6650 0.17 1

1 0.015 0.015 0.045 0.030 3 Cromby I (1951) 6625 0.25 1

0.015 0.060

(

f Ridgeland (1954) 6575 0.42 1

1 0.015 0.015 0.075 0.045 2 Uskmouth (1956) 6525 0.58 1

1 1

1 0.015 0.015 0.015 0.015 0.090 0.060 0.030 0.030 2

Kainan (1972)*

6500 0.67 1

0.015 0.105 g

k Unknown (1958)*

l 6475 0.75 1

1 1

1 0.015 0.015 0.015 0.015 0.120 0.075 0.045 0.045 I

' ENESA (1970)*

6450 0.83 1

1 0.016 0.016 0.136 0.091 E

y Brown-Boveri Co.,

2, Denmark (1975) 6375 1.30 1

0.016 0.152

>E Utah Power (1976) 6350 1.50 1

0.016 0.168 Tanners Creek (1953) 6325 1.70 1

0.016 0.184 E

Nanticoke (1974)*

6325 1.70 1

0.016 0.200 g

Pittsburg I (1956) 6315 1.80 1

0.016 0.216 IIcarn 1 (1954)*

6225 2.30 1

1 0.016 0.016 0.232 0.107 IIcarn 2 (1954)*

6150 2.70 1

1 0.016 0.016 0.248 0.123 Alstrom-Rateau (1977)*

6125 2.80 1

0.017 0.265 Ihnkley Point A-5 (1969) 6000 3.40 1

1 1

1 0.017 0.017 0.017 0.017 0.282 0.140 0.062 0.062 Ilinkley Point A-6 (1970) 6000 3.40 1

1 0.017 0.017 0.299 0.157 Ilinkley Point A-4 (1969) 5925 3.80 1

1 0.017 0.017 0.316 0.174 Unknown (1958)*

5800 4.50 1

1 1

0.017 0.017 0.017 0.333 0.191 0.079 Bold (1960) 5725 5.90 1

1 1

1 0.017 0.017 0.017 0.017 0.350 0.208 0.096 0.079 Cutler 6 (1969) 5250 9.50 1

0.019 0.369 Sendai(1972) 4500 13.5 1

0.022 0.391 Northern States (1971) 4350 14.5 1

0.023 0.414 Shippingport (1974) 4150 15.5 I

1 0.024 0.024 0.438 0.120 Gallatin (1974) 3850 17.5 1

1 0.026 0.026 0.464 0.234 Morenci 3 (1959) 3870 17.8 1

1 0.026 0.026 0.490 0.260 Oak Creek (1977) 2875 22.0 1

1 1

I 0.035 0.035 0.035 0.035 0.525 0.295 0.155 0.!!4 Essex 1 (1972) 2075 25.5 1

0.048 0.573

(

3

f Table 10 (Continued) Pertinent Equations for Weibull Distributions Covering All Known Turbine Fagures, Failures Generating Missiles, Relevant Failures, and Relevant Failures Generating Missues All failures Relevant failures Equations All failuses generating missiles Relevant failures generating missiles R(T)t =exp exp (-1.2 x 10-8 7* ") exp (-7.36 x 10-* 7*.s :) exp (-4.36 x 10-* P * * ) exp (-4.02 x 10-* P ")

Z(T) = h* 2 -'

4 6.18 x 10-8 T-* *

• 3.8 3 x 10-' T-* *

• 1.79 x 10-* T-* "

1.49 x 10-8 T-* *

• t

1. (T) =

0.119 7* "

7.36 x 10-* 7* "

4.36 x 10-* T* * '

4.02 x 10-8 7* "

T(#) = a N'14 41.9 #' "

15i N' "

2076 N' "

5900 N' "

• Not in known population.

tR(T) multiplied exponent by 10-* to get value.

A 6

E M

8 Table 11 Sensitivity Study Varying Turbine Population to Determine 2

Effect on WeibuB Distribution for Relevant Missiles

• m 5

Time to Turbine Hazard Cumulative Revisedt Hazard Cumulative E

on

failure, population value,2(T) hazard turbine

value, hazard, Failure cause years (Table 10)

(Table 10)

(Table 10) population g

Overspeed 0.08 6675 0.015 0.015 3400 0.036 0.030 Overspeed 0.58 6525 0.015 0.030 3200 0.030 0.060 j5 Overspeed 0.75 6475 0.015 0.045 3100 0.031 0.091 Stress-corrosion crackinal P

brittle 3.4 6000 0.017 0.062 2700 0.037 0.128 2

P Overspeed 5.9 5725 0.017 0.079 2500 0.040 0.168 z

Brittle ?

22.0 2875 0.035 0.114 1250 0.080 0.248 3

3

• Parameter used in calculating the values in the table:

f a = 651; 1/d = 2.59; A = 0.386;r = 0.98 8

p R(T) = cxp (-8 x 10-* 7* ")

g Z(T) = 3.1 x 10-8 T-* **

• f

1. (T) = 8 x 10-8 7* "

T(#1 = 651 #' "

tRevision based on using population with retirement x 1.33,which removes many smaller units, then takmg % to % of e

i$

I ii$

this population as being t' levant to nuclear.

N,,

e

ACCIDE%T ANALYSIS l

ses 14 electrohydraulic systems to common-mode failure due to sitting or rusting comparable to that occurring at p,ii,,,,,,,,,oci.,,,,,,,,

Uskmouth is examined in relation to the mechanical-

$12 w.6nuit nive di.iribution

_I zm =ht*-8 hydraulic systems used at Uskmouth. Some rates used g

in the General Electric study ' are shown in Table 12.

gj 'O '

8 General Electric believes that the preceding ap-Et; a proach permits an evaluation without the limitation of M

the extremely small nuclear turbine populatlan being a

!E I * '

~

factor and without the need to base the evaluation on CE overall fossil turbine experience which General Electric

$5 xii,,,,,,,,

judges to be not applicable to nuclear turbines. In the EU,

'!, ~

~

appendix to their report,8' they point out that

~

probability values much higher than those in Table 13 (and near the values appearing in the summary of this article) are inherent in the use of direct statistical o

5 to is 20 25 30 35 40 methods based on past experience with alarge number OPERATING YEARS of fossil units whether zero or six failures are assumed.

The application of a Duane growth model to these Fis. 3 Typical weibutt failure rate = 2fr/ bathtub curve for statistics does not substantially change the results.

turbine generator failures and failures with inissiles.

The General Electric report develops several con-vincing arguments as to why the sequence of simple events using nuclear data is preferred to a direct Table 12 General Electric Failure Rate Data statistical estimation. For example, older fossil units Failures, Confidence, differ markedly from nuclear units with regard to Nuclear turbines 108 hr material properties, stresses, rotor design, and control systems, in this vein the report dismisses the Hinkley Experience Sticking rates, control valves 0.42 50 Sticking rates, stop valves 0.26 50 CONDITIONAL pro 8 ABILITY (%)

All turbines 0 01 o at 1.0 10 Failure rates, overspeed trips 0.0087 50

,,o,

, 'i

i Domestic turbines, electrohydraulic 100

}

}

Failure rate, valve silting 0.00036 50 Failure rate, valve rusting 0.0008 50 8

o a+y i

l Point and Mitsubishi.ENESA failure as due to material j 10 o

g w

properties not typical of General Electric fabrication

• 8 practice.

j j

3 g

Although the General Electric arguments are per-

~

suasive (and the same may be said of Allis.Chalmers o

k [

.a o

]

and Westinghouse), it is the author's opinion that o

N Alt failures, T(H) = 41.9 H.77 :

factors not yet revealed during the limited experience 1

o All missiles, TtH) = 151 H.92 :

with nuclear turbines to date may not be properly 1

a 0 Relevant failures.

accounted for in the General Electrie and Westinghouse

,,,,,,[,,

models and therefore will cause their projections to be i

TlH) = 5900 H.70 3

over]y optimistic. Even though the statistical estimate 2

ai contained herein is not truly representative of nuclear

,,,...t a

practice, it includes conservatisms based on long.temi noi at i.o io experience, and values of 2/T) in the range of CUMULATIVE HAZARo t%)

3.3 x 10-s to 3.1 x 10-* per turbine year (for a turbine population relevant to nuclear reactors) are Fig. 2 Reyession analyses, Weibull hazard plots of turbine.

more realistic.

generator failure data, NUCLEAR SAFETY, Vol 19. No. 6, November-December 1978 Q

r t

Table 13 Probability of Wheel Burst (1800 rpm TC6F,43 in.)*

Operating Mode A (Start-up)

B (Overspeed testing)

C (l.oss of load)

Condition Start-up Set / check emergency trip (2)t Generating (3)t lead Unloaded Unloaded Fullload Electrical status Unsynchronized Unsynchronized Synchronized Initial rotational speed,%

100 (11t i10 (4)t 100 Worst wheel-temperature condition Cold = 40*F llot = 100*F Hot = 100*F Probabilities for Three Ranges of Speed Al A2 A3 Bl B2 Cl C2 C3 k

Running speed. percent of normal 0-100 100-112 Il2-runaway 110-112 Il2-runaway 109-119 119-127 127-runaway Lifetime speed-Icvel probability 1

3.5 x 10-8 3.2 x 10-8 8 1.4 x 10-8 1.3 x 10-' 8 1.0 x 10-'

9.3 x 10-'

1.5 x 10-'

g Single-wheel-failure probability 2.3 x 10-'

3.0 x 10-*

I 5.5 x 10-'

I 9.7 x 10-'

6.6 x 10-'

I g

Number-of-wheels factor 6

6 1

'6 1

6 6

1

-e Wheel-failure probability given g

P speed 1.4 x 10-'

l.8 x 10-*

I 3.3 x 10-'

I 5.8 x 10-'

4.0 x 10-*

I y

9 Lifetime-wheel-failure probability 1.4 x 10-'

6.3 x 10-*

3.2 x 10-' 8 4.6 x 10-' ' l.3 x 10-*

• 5.8 x 10-8 3.7 x 10-'

• 1.5 x 10-'

G

)

Probabaity of Wheel Failure Total lifetime-wheel-failure probability

• Q low 4 peed (brittle) failure = sum of cases Al, A2, B1, Cl, and C2 = 2.6 x 10-' (last-stage wheel only)

Runaway failure

= sum of cases A3, B2,and C3

= 1.5 x 10-' (any wheel, equal probability)

E-Total

= 4.1 x 10-'

.E Average annual-wheel-failure probability = totallifetime probability /hfetime (= 30) = 1.4 E-8 2

• Source: Ref.16.

t(1) F00 start-ups assurned over 30-year lifetime; duration of each, I hr; 100 = 3 per year + 10 extra the first year; 3 per year allows for start-up following i refueling-and-insper; ten shutdown,1 loss of externalload, and I forced outage of plant.

(2) 40 tests at 100% speed assumed over hfetime; duration of each, I hr;40 = 1 per year + 10 extra the first year.

3g (3) 30 full-load losses assumed over lifetime; 30 = 1 per year.

(4) Debberate operation with cold whccis at i10% speed disallowed by operating instructions.

R h,

E E.

6 O

ACCIDENT ANALYS48 ges i

Honorable Dimy lee Ray, AEC Chairman, Report on SWERY Turbine Missiles Apr. 18,1973.

)

7.E. Splittaerber Overspeed Damage to Steam Turbines, Data ate presented and methods of analyses are M#^'"'"h'd'n. 35(1/2): 1 18 (February 1962).

given to permit the calculation of turbine. generator

8. H. Huppmann, Frequency and Causes of Failure to cumulative reliability and time-dependent failure rate.

Components of large Steam Turbines, in hoceedingr, The Weibull distribution clearly delineates the failures CREST Meeting on the Reliabihty of Mechanical Compo-nents and Systems for Nuclear Reactor Safety. Riso, early in life as well as permitting determination of Denmark, Sept. 24-26. 1969. Danish Report RISO-214 failure rates once conditions approximating steady pp.171187,1970. NTIS: also published in Maschinen.

state are achieved.

schaden. 43(l): 16 (1970).

A study limited to missile failures considered

9. R. L. Carson, C. A. Bucci, and R. J. Airhart Periodic relevant to nuclear reactors yields values of 2/T) of Operational Tests Help Keep Unit Availability at High Levels. Power, 120(7): 5944 (July 1976).

1.5 x 10-* to 1.5 x 10-s per turbine year for the 10.E. O. Codier, Reliability Growth in Real 1.ife, in Pro.

total turbine population and 3.3 x 10-s to endings IEEE Annual Symposium on Reliability, Boston.

3.1 x 10-* per turbine year for a turbine population Jan.16-18,1968, pp.458469, Institute of Electricaland corrected to be relevant to nuclear reactors.

Electronics Engineers, New York,1968.

The preceding values using more sophisticated

11. D. R. Beeth and S. H. Hobbs. Analysis of Bush Approach to Turbine Missile Generation Probability, Brown and Root techniques compare favorably to the value predicted Technical Memosandum. Brown and Root inc., December for 1977 in the earlier report,8 namely, a failure rate 1976.

2/T)of about 7 x 10-s per turbine year.

12.W. Nelson, Hazard Plotting for incomplete Failure Data,J.

Quality Technol., (1): 27-52 (January 1969).

13. S. H. Bush, Pressure Vessel Reliability, J. Pressure Vessel Technol. Inant ASME, Ser. //, 97(1): 54-70 (February REFERENCES 1975).

l. S. II. Bush Probability of Damage to Nuclear Components 14.0. A. Kellerman et al., Progress and Results of the Due to Turbine Failure Nuct Safety,14(3): 187-201 Reliability Study of Pressure Vessels, in Performaner of Nuclear Power Reactor Components. Symposium Pro-(May-June 1973).

2. A. K. Bhattacharya and S. K. Chaudhuri,'the Probability ceedings, Pngue, Nov. 10-14,1969, pp. 391-403, Inter-of a Turbine Missile Hitting a Particular Region of a national Atomic Energy Agency, Vienna, 1970 (ST!/

Nuclear Power Plant, Nuct Technol., 28(2): 194 198 PUB /240).

15.G. Slopianka and G. Mieze, Failure Rates of Pressure (Fek..aary 1976).

Vessels, Part I: Evaluation of VdTUV Statistics, Getman

3. B. Johnson et al., Analysis of the Turbine Missile Hazard to Report IRS-I 34, Institut fGr Reaktorsicherheit, Cologne, the Nuclear Thermal Power Plant at Pebble Springs.

Federal Republic of Germany,1968.

Oregon, Report PGE.2012 Portland General Electric

16. General Electric Company, Hypothetical 7hrbine Mis-Company, January 1976 (prepared by Sciena Applica-sites-Probability of Occurrence, General Electric Memo tions, Inc.).

Report, Mar. 14,1973.

4.S. W. Swan and M. Meleis A Method of Calculating

17. Houston Lighting and Power Co., Analysis of the Proba-Turbine Missile Strike and Damage Probabilities, Nucl.

bility of the Generation and Strike of Missiles from a Safety, 16(4): 443451 (July-August 1975).

Nuclear Turbine, Section 3.5 in Amendment 27 to the

5. Nuclear Regulatory Commission, Regulatory Guide Preliminary Safety Analysis Report, South Texas Project RG-1.115, Rev.1, Protection Against Lownalectory Units 1 and 2,

July 18, 1975. NRC Docket Turbine Missiles. July 1977.

STN 50-498-116, pp. 3.51 through 3.510b.

6. Letter from H. G. Mangelsdorf, ACRS Chairman, to NUCLE AR SAFETY, Vol.19, No. 6, November-December 1978 m

.y s-C%WID

'86 JM!-8 DO :28 p?l.Q.725;f ELECTRIC POWER RESEARCH INSTITUTE '"

~'

1 SEMINAR ON TURBINE MISSILE EFFECTS IN NUCLEAR POWER PLANTS C

HOLIDAY INN PALO ALTO, CALIFORNIA OCTOBER 25-26, 1982 f.

Jm m

page 1 e

i MISSILE GENERATION RATES FROM HISTORICAL DATA PATRICK G. HEASLER PACIFIC NORTHWEST LABORATORY

-ee*

s page 2 OBJECTIVE:

TO ESTIMATE ROTATING COMPONENT FAILURE AND MISSIL3 GENERATING RATES FROM HISTORICAL DATA r

pnge 3 ROTATING COMPONENT FAILURE TYPES:

overspeed failure operating speed failure burn-in failure s.,

missile generating event s

.c V..

J

page 3.1 There were several types of failures that were found to be

~

important while reviewing the historical data. These failure types are listed on the opposite page and defined below:

ROTATING COMPONENT FAILURE - One of the major rotating components (Discs or Rotor) breaks apart during turbine operation. It should be emphasized that generator failures are not included in this failure type.

OVERSPEED FAILURE - A rotating component failure that occurs when the turbine speed.is greater than 110 percent.

OPERATING SPEED FAILURE - A rotating component failure that occurs when the turbine speed is less than 110 percent.

BURN-IN FAILURE - A failure that occurs in the first year of turbine operation.

MISSILE GENERATING EVENT - A rotating component failure that results in the production of missiles. (The turbine housing is penetrated.)

i i

-.e.

o page 4 EVENT DIAGRAM FOR ROTATING COMPONENT FAILURES t

,p.

BOP EVENT:

BOV EVENT:

R.C. Failure c; curs R.C. Failure occurs at operating speed in overspeed during during ' Burn-in'

' Burn-in' Period Period. (1st Year)

(1st Year)

OP EVENT:

OV EVENT:

R.C Failure occurs FRAGMENT R.C. Failure occurs

?

PENETRATES in overspeed during at operating speed during remainder of TURBINE remainder of turb-turbine lifetime.

HOUSING ine lifetime.

i NO YES S

M EVENT:

Missile generating incident occurs dur-ing turbine lifetime l

t 9

page 4.1 l

The relationship between the different types of rotating compo-nent failures and a missile generating event are illustrated by the event diagram on the opposite page. Burn-in failures are distinguished from failures occuring later in turbine life because the failure rate at the begining of turbine life is much higher than that for the remainder. Most burn-in failures can be attributed to deficiencies in design or construction, therefore the causes of burn-in failures can be considered to be different also.

O j

e

.. _ _,. - -.,., _... ~

Ptge 5 s

i FORMULA FOR THE AVERAGE MISSILE GENERATING RATE R,y(M) = Pr(M/OP)*R,y(OP) + Pr (M/OV) *R,y (OV) where:

R,y(M) = Average rate of occurrence for Missile Generating Incidents.

i R

(OP) = Average operatin~g speed failure rate.

R y(OV) = Average overspeed failure rate.

1 Pr(M/OP) = Probability of a missile during an operating speed failure.

Pr(M/OV) = Probability of a missile during an overspeed failure.

(All rates are measured in incidents / year) i M..

page 5.1 4

The relationship described in the event diagram can also be expressed as a simple rate-of-occurrence equation as illustrated on the opposite page. This equation will be used to estimate the average missile generating event rate. The terms on the right-hand side of the equation will be estimated from the historical data.

page 6 4

FORMULAS FOR AVERAGE OPERATING AND OVERSPEED FAILURE RATES 1

29 R(BOP) +

R(OP)

R,y(OP)

=

30 30 and 1

29 R(BOV) +

R(OV)

R,y(OV)

=

30 30 where:

The burn-in operating speed failure rate.

R(BOP)

=

The burn-in overspeed failure rate.

R(BOV)

=

R(OP) = The operating speed failure rate excluding the first year of operation.

R(OV) = The overspeed failure rate excluding the first year of operation.

30 = Assumed length of the turbine lifetime.

(in years)

~.

m.

.__-.-,7.m._.r.,

._y,

.,-m_.

y

ptge 6.1 a

The average operating and overspeed failure rates (for a particular turbine lifetime) should not be calculated directly from the data, because the failure rate is not constant over lifetime. The formulas on the opposite page show the relationship between the average failure rates and the Burn-in rates. It should be emphasized that the average failure rates are the most relevant quantities to use for safety calculations. However, it is extremely important to recognize the existence of a Burn-in period when considering other aspects of turbine operation such as inspection and maintenance strategies.

i

)

i l

l

PEgo 7 i

TURBINE POPULATION USED FOR NON BURN-IN OPERATING AND OVERSPEED FAILURE RATE CALCULATIONS Relevant Non Burn-in Failures Year Operating Years Operating Over Comm.

NN Size Burn-in Remaining speed speed Total 1950 30 to 100 900 22621 1

1 2

to 1960 100 & up 521 12575 3

0 3

1960 30 to 100 524 7532 1

0 1

to 1980 100 & up 2219 23113 3

1 4

Total 4164 65841 8+17 2

11 i

• These catagories also include any Nuclear tubines smaller than 30MN.

? Incidents that can not be assigned to a cell because of missing information.

I

,--n--

ptge 7.1 1

The non burn-in operating and overspeed failure rates were calculated from the data presented on the opposite page. This data was gathered from 10 turbine manufacturers and represents essentially all turbines in operation in the designated cata-gories. The turbine failures were divided into those that were considered to be relevant to nuclear plant operation and those that were not. For example, about 80 percent of the overspeed failures were not considered to be relevant to a nuclear plant operating environment. This catagorization allowed us to compute failure rates that were relevant to nuclear power plant operation.

One of the most important conclusions to be drawn from this table is that there does not seem to be any st'rong relation between these failure rates and year of commission or megawatt size.

Because no strong relationship exists, it is most reasonable to pool the data together and compute failure races for the total turbine population.

l

paga 8 x

v/Y' LIST OF RELEVANT NON BURN-IN FAILURES M'

Mg

,r'

,gs" &

Plant Missiles r

Plant Name Type Date produced Failure Cause Operating Speed Failures:

Rinkley Point Nuclear 1969 yes Brittle Fracture h'gC6 Shippingport Nuclear 1974 no Cracking Gallatin Fossil 1974 yes Cracking 1

Oak Creek Fossil 1977 yes Cracking Porcheville Fossil 1977 yes Other, Generator Rotor locked.

Aberthaw Fossil 1972 yes Other, Water induction.

Yankee Rowe Nuclear 1980 no Cracking 40R Wangi 42 Fossil 1357 yes Operator Error Pittsburgh Fossil 1968 no Other Overspeed Failures:

Control Sy' stem Bold Fossil 1960 yes

% $ $ W & jLad Control Systeme("~ g Mountain Creek Fossil 1977 no W-4 9

page 8.1 a

All the failures used to calculate non burn-in operating and overspeed failure rates are listed on the opposite page along with a few details of the failures including its cause. Notice that the cause of failure has been divided into the 5 categories Brittle Fracture, Cracking, Operator Error, Control System and other.

/

page 9 ESTIMATES OF NON BURN-IN PAILURE RATES FROM RELEVANT TURBINE FAILURES 4

~

R(OP) = 9/65841 = 1.37 X 10 Failures / Year

-4 95 Percent Conf. = [0.61, 2.60] X 10 withs 11 percent of rate due to Brittle Fracture.

percent of rate due to Cracking.gc6 11 percent of rate due to Operator Error.

33 percent of rate due to Other.

~..

4

~

R(OV) = 2/65841 = 0.30 X 10 Failures / Year

-4 95 Percent Conf. = [0.03, 1.09] X 10 With 100 percent of rate due to Control System [

.--,e

,n.

w --

ptge 9.1 The calculations on the opposite page use the data discussed in previous tables to calculate the relevant non burn-in failure rates. These failure rates could also be divided up by failure cause. For example, the operating speed failure rate due to 4

cracking is 0.62 x 10 failures / year.

~

S l

O

-~

--,,--,n,.

page 10 LIST OF RELEVANT BURN-IN FAILURES

,[h Plant Missiles Plant Name Type Date produced Failure Cause Operating Speed Failures:

Siemens Fossil 1951 yes Brittle Fracture Ridgeland Fossil 1954 yes Brittle Fracture comfostilla Fossil 1970 yes Brittle Fracture Kainan Fossil 1972 yes Other, Missassembly 2

of turbine bearings.

Overspeed Failures:

Uskmouth Fossil 1956 yes Control System, Oxide buildup on valves.

Calder Hall Nuclear 1958 yes Control System, valves plugged with shot.

Tavazzano Fossil 1961

??

Brittle Fracture.

4164 = Total number of Burn-in years of operation in population.

CALCULATION OF BURN-IN FAILURE RATES

~4

+(yy#{#

gd R(BOP) = 4/4164 = 9.61 X 10 Failures / Year.

y

-4 g@Y 95 Percent Conf. = [2.40, 24.5] X 10

~4 R(BOV) = 3/4164 = 7.20 X 10 Failures / Year.

~4 95 Percent Conf. = [1.44, 21.1] X 10 s

h h'

t paga 10.1 The opposite page outlines the calculations necessary to estimate the burn-in failure probabilities. All burn-in failures in the table are considered relevant to nuclear plants.

These probab-111 ties indicate that there is approximately one in a thousand chance of a new turbine in a plant failing soon after it goes into operation.

o i

l 1

paga 11 CALCULATION OF THE PROBABILITY OF MISSILES / FAILURE Missiles Yes No

??

Total OP. Speed 10 3

0 13

,Overspeed 3

1 1

5 V

p.,

L

/

Total 13 4

1 18

\\Y[

~

Probability of a Missile / Operating Speed Failure:

Pr(M/OP) = 10/13 = 0.77 95 Percent conf. = [.47,.94]

Probability of a

'N L.

(

Missile /Overspeed Failure g,,

i 6d Pr(M/OV) = 3/4 =.75 5

95 Percent Conf. = [.20,.98]

's..

'A "me

s 1

page 11.1 1

The failure data also provide some information concerning the effects of a rotating component failure. The table on the opposite page shows how many relevant failures produced missiles.

The information in the table can be used to calculate the conditional probability of a missile given a failure and these calculations are detailed below the table.

For one of the failures, it was not possible to determine whether missiles were produced or not. It is listed under the ?? column in the table and is not used in the probability calculations.

5 o

n---.-

..n--

+.-

s PEgo 12 i

ESTIMATES FOR THE AVERAGE MISSILE GENERATING EVENT RATE USING THE FORMULAS:

1 29

-4 R**(OP) =[

9.61 +

1.37] X 10 30 30

~#

= 1.64 X 10 Failures / Year with 36 percent of the rate due to cracking.

1 29

~4 Rav(OV) =[

7.20 + - 0.30] X 10 30 30

~4

= 0.53 X 10 Failures / Year

-4 R,,(M) = t.77*1.64 +.75*.53] X 10

~4

= 1.66 X 10 Incidents / Year Due to overspeed failures: 0.40 x 10~4/ year

-4 Due to operating speed failures: J.26x10 / year with 36 percent due to cracking.'

DIRECT CALCULATION FROM NUCLEAR DATA:

-4 R,y(M) = 2/2467 = 8.11 X 10

~4 95 Percent Conf. = [0.80,29.20] X 10 (There are 2467 nuclear turbine-years of operation in the population.)

4

,m,

4 pagt 12.1 f

r All the failure rates calculated on previous pages can now be combined to produce an estimate for the rate of major concern, R,y(M), the average missile generating event rate for a 30-year turbine lifetime. The calcula-tions on the opposite page present two different ways to make this estimate.

The first calculation plugs the estimates into the formula obtained pre-viously, while the second uses data from nuclear turbines only and estimates the. rate directly.

The first estimate is lower than the last but since the error bounds on the last estimate are relatively large, the two calculations do not necessarily contradict each other. The last estimate does show that adding fossil fuel experience to nuclear experience does not unjustly inflate nuclear failure rates; if anything, it deflates them.

(

..,.,_._.e-