ML17215A630
| ML17215A630 | |
| Person / Time | |
|---|---|
| Site: | Saint Lucie  |
| Issue date: | 10/12/1984 |
| From: | Dawson T, Gillerlain J, Niederberger R FORENSIC TECHNOLOGIES INTERNATIONAL CORP. |
| To: | |
| Shared Package | |
| ML17215A631 | List: |
| References | |
| NUDOCS 8411120138 | |
| Download: ML17215A630 (114) | |
Text
THERMAL SHIELD I-"AILURE AT ST. LUCIE NUCI EAR
'ENERATING UNIT NO. 'I by Thomas H. Dawson'h.D.
Robert B. Nfederbergeri P.E.
Joseph D.,G1llerlafni Jr.~
Ph.D.
Forensic Technolog)es International Corporation 107 Rkdgely Avenue Annapolfst MD 21401 October 12'984
'Reviewed and Approved:
nneth E. Court. P.E.
Engineering Manager
(
o ph R. Reynoldsi Jr.i
.E.
nidor Vlcc Presfdentr gfneerfng FTI Case No. 5275.51
0 FIai use
ABLE 2E QMRILi 1.0 Introduction and Summary 1.1 Introductf on 1.2 Summary
2.0 Background
2.1 CE Involvement 2.2 Forensic Technol og$ es Involvement 2.3 Basic Design Features 2.4 Damage to Thermal Shield and Core Support Bar rel 2.5 Thermal Shield Removal and Repairs to the Core Support Barrel 2.6 Operational History of St. Lucre Unft No.
1 2.7 Related Problems at Other Reactor Plants 3.0 Summary of CE Act)vasty 3.1 Scope of CE Work 3.2 CE Failure Mechanism Analys)s Progr am 3.3 CE Metallurgical Examination 3.4 CE Findings on Failure Mechanism 4.0 FTI Investfgat1oni Part 1:
Analysfs of Loss of Preload 3n the Pos)t)onfng Pfns 4.1 Summary of Stress Analys)s 4.2 Geometry of Thermal Shield and Core Support Barrel 4.3 Net Loading of Positioning Pins
0
4.4 Possible Causes of Complete Loss of Preload 4.5 Stresses
$ n Posftkonfng Pfns 4.6 Comparison of Calculated Stresses with Yield Strength 4.7 Compar <son of Calculated Stresses to Design Code Requirements 4.8 Photog aph$ c Evidence of Pin Damage by Combined Un)form and Bend)ng Loads 5.0 FTI Investigation.
Part 2:
Analysfs of Thermal Shield Failure Mechanism 5.1 Summary of FTI Failure Analys)s 5.2 FTI Fracture Analysfs 5.3 Startup and Shutdown Transients 5.4 Hydraul fcs Transient Fl ow
~
~
6.0 Discuss<on and Summary of Thermal Shield'Failure 7.0 Conclusions References Appendix A - Photographs Appendix B - Fulgur es Appendix C - Posktkonfng Pin Imprfnts on Bearing Pads 22 26 28 29 31 32 36 39 41 46
LIST OF FIGURES 1.
Reactor~
Showing Thermal Shield Support System 6 Details lA.
Reactor Coolant System Arrangement 2.
Sketch of Damage:
Thermal Shield and Core Support Bar rel 3.
Thermal Shovel d:
Imprfnts on Upper/Lower Positioning Pin Pads 4.
Core Support Barrel and Thermal Shield 5.
Sketch of Positfon)ng Pfn 6.
Bending of Posft1onfng Pin Due to Relative Thermal Displacement of Core Barrel 7.
Stress Dfstrfbutfon on Ends of Upper Pos[tfonfng Pfns During Normal Power Operat)on (fr$ct$ on coefficient = 0.29)
Appendix B Appendix B Append)x 8 Appendix B Page 19 Page 21 Page 24 Page 26
TABLES 1.
Dates of Each Fuel Cycle (St. Lucre No. 1) 2.
Loose Parts Recovered frao Bottan of Reactor Vessel 3.
'Net Pfn Loads and Load Contrfbutfon 4.
Maximum and Minimum Stresses for Two Different Frfctfon Coeff1cfents Page 9
Page 13 Page 22 Page 25 5.
Changes
$ n Number of Reactor Coolant Pumps Operating by Month Page 37A
0
During a
refueling outage at the end of the fifth fuel cycle in March>
1983.
extensive structural damage was discovered in the thermal shield and thermal shield suppor t system of Florida Power and Light (FPL)
Company~ s St.
Lucie Unit 1 nuclear power plant.
On April 12~
1983 Forensic Technol ogies International (FTI) was requested to investigate this structural fail ur e.
This r eport presents the results of that investigation.
There had been no indication of a problem with the thermal shield or core support barrel during prior operation of the reactor.
The first indication was found during the refueling operation at the end of the fifth cycle~
when a fuel assembly would not seat properly on the core support plate.
Inspection of the core support plate showed debris (later identified as a lock bar from a thermal shield positioning pin) on its upper surface.
A decision was made to unload the fuel and liftthe core support barrel and thermal shield out of the reactor vessel for further investigation.
Subsequent inspection using a remote controlled underwater TV camel a revealed the extensive damage to the thermal shield and its support system.
LZ ~)marx The St.
Lucie No.
1 reactor was designed by Combustion Engineering>
Inc.
(CE) under contract with FPL signed in 1968>
and went on line in
1976.
After the thermal shield failure was discovered>
CE conducted an analytical study of the possible causes for the failure.
The ma$ n conclusion from their study
$ s that damage to the thermal shield was caused by sel f-excited v$ brat1oni made poss)ble by deter ]orat)on of the thermal shield support system.
The deter$ orat$ on fs presumed by CE to have been preceded by loss of preload on pos$ t$ on$ ng p1ns fixing the thermal shield relative to the core barrel.
The spec$ f$c reason for the loss
$ n preload of the posft1on1ng pins was not tdentff)ed by CE> although hydr aul $ c pressure loads ngi weight transfer during installation and rad)ation-Induced stress rel axat$ on were shown to be capabl e of contributing.
CE regarded these factors>
together with postul ated but unfdentff1ed 1nstallatfon errors>
as the most reasonable explanation for the loss.
Under authority fry FPLi the role of FTI 1n 1nvestkgat)ons of the thermal shield failure has been to review CE~s, analys1s and conduct
\\
addftfonal analysfsi as necessary>
$ n order to provide an unbiased th$ rd-party op$ n$ on as to the technical aspects of the failure.
The present report describes results of that
$ nvest$ gat$ on.
FTI concurs with CE that loss of preload in the posftfon1ng p$ ns was the
$ n$ t$ at1ng event 1n the failure mechanisms and that this was followed by wear and deter)oration of the thermal shield support system due to vibration of the shield.
FTI also concur s with CE that damage to the thermal shield resulted from large-amplitude motions of the shield.
but
does not agree that these were 1)mfted to sel f-excited v$brat$ ons during steady-state>
4 pump operation of the reactor coolant system.
On the bas$ s of its independent analyses>
FTI concludes that larger asymmetric hydraulic loads acting on the shield during startups and shutdowns of the reactor cool ant system were probably an addftfonal cause of the 1 arge-ampl etude motions and ultimate cracking of the thermal shield and its support system.
FTI also concludes that the loss of pt eload
$ n the pos$ t$ on1ng p1ns was the result of factors beyond those cons)dered by CE.
Independent calculations by FTI of the stresses
$ n the pos$ t$ on1ng p$ ns reveal that combined un<form compression and bending loads on the pins during power operation can cause stresses exceeding the yield strength of the materhal by more than 100%.
Thts overstress could have caused permanent plastic d$ stort$ on of the p$ ns and could have contr fbuted to loss of preload during zero-power operation by not allowing the p$ ns to spring back fully to their fn)teal conf$ gurat$ ons.
The overstress could also cause damage to the threaded connection of the p$ ni leading to loosening and mfsal $ gnment dur 1ng zero-power operation.
The hazard of overstress on threaded connectors.
as described above~
$ s recognized
$ n standard eng$ neer$ ng practice.
The 1974 ASME Code for.
design of nuclear power plants specifically requires that stresses 1n non-
<ntegral (threaded) connectors of pressure confining components never exceed the yield strength of the matetfal and~
$ n non-pressure conffn1ng
components>
never exceed the yield strength by more than 20%.
This code represents a body of engineering experience supporting the conclus)on of the pt esent report that overstress of the pos$ t$ on$ ng p$ ns could have contributed sfgnkf1cantly to
$ n$ t$ atTon of the thermal shield damage.
Nothing
$ n the above sequence of events leading to damage of the thermal shield can be attributed to operating procedures employed by FPL.
In facti fndfcatfons at e that the failure
$ s generic
$ n nature since at least two other similarly designed units have also experienced thermal shield support problems.
The St.
Lucie No.
1 reactor was designed by CE under contract with FPL signed in 1968>
and went on line in 1976.
Imnediately after the reported failures CEi at FPL's requests undertook an investigation of the extent of the damage and devel oped extensive video and photographic documentation during Aprili 1983.
Under agt cement with FPLi CE also conducted an analytical study of the possible cause.
or causes'or the failure.
This analysis is contained in Chapter 7 of Ref. (1).
in addition to the failur e analysis>
the overall analysis of the extent of damage.
repair methodsi effect on future operation and safety were carried out.
and the necessary presentations to the Nuclear Regulatory Canmission (NRC) were supported by CE.
By May 1984i all repairs to the reactor were completedi and with NRC concurrence the unit was placed on line at 10% power.
Under authority from FPL.
the role of FTI has been to review the CE failure e analysis and conduct additional analysesr as necessary.
The purpose of the investigation by FTI is to provide an unbiased third-party opinion as to the technical aspects of the failure and its probable cause.
The results of this investigation are described in the present repor t.
St.
Luc)e Unft I 1s a CE-des)gned pressurized water reactor (PWR)
$ n which nuclear fuel contained 1n a large cylfndrkcal pressure vessel heats r eactor cool ant water pumped through
$t.
The reactor cool ant system transfer s the heat to a secondary system where steam
$ s produced 1n two steam generators.
The steam 1s then used to generate electric power through the use of turbines and generators as
$ n fossil fuel plants.
A system layout typical of Untt 1
$ s included
$ n Appendix B as Figure lA.
A cross section schematic illustrating the reactor Internals is prov)ded
$ n Figure 1 (Appendix B).
The core support barrel (CSB)
$ s the prfncfpal member of the core support assembly'hich supports the reactor Internals.
It
$ s a right circular cylinder approximately 27 feet highs 148 inches I.Di with a minimum 1 3/4 Inch wall thickness.
An upper flange on the CSB rests on.a ledge of the pressure vessel.
It $ s restrained from lateral movement at its lower end by s$ x snubber assembl)es attached to the CSB and the pressure vessel.
The core support assembly of St.
Lucre Unft 1 also included a thermal shield (TS).
This was a 156 3/4 Inches I.D.
x 137 3/4 inches high cyl)nder with a wall thickness of 3
Inches which was located around the lower portion of the core support barr el.
N)ne lugs welded onto the core support barrel held the thermal shield>
mating with support p1ns welded to the thermal shield.
Radial pos1t1onfng of the thermal shield was obtained with 9 upper and 17 lower pos1t$ on$ ng p$ ns.
threaded through the thermal sh)eld and contacting the core support
barrel.
The upper pins were located under each of the nine support lugs>
approximately 15 inches bel ow the top of the 1 ug.
The 17 1 ower pos$ t$ one ng p$ ns were equal ly spaced around the c$ rcumference approx)mately 21 1/4 Inches above the bottom of the thermal shield.
All major components of the core suppot t assembly were Type 304 statnl ess steel.
Examfnatfon of the thermal shield and Its support system
$ n April 1983 by remotely controlled video cameras revealed extensive damage to the thermal shield and to the core support barrel at several of the support lugs.
The thermal shield was cracked and p$ eces were broken off at s everal support lug pos$ t$ onsi and the shield was canted about a
North/South axles relative to the core support barrel.
Damage to the core support barrel was apparent also.
Photograph l. Appendix A shows cracking of the thermal shield at Lug No.
1 as well as the breaking away of the lug from the core support barrel.
Photographs 2 and 3
show loss of portions of the thermal shield at Lug Nos.
4 and 9i respectively>
while Photograph 4 shows the locat)on of a missing pos1t$ on$ ng p1n (Pkn Z) below Lug No. 9.
A systematic
$ nspect$ on of the entire thermal shield support system.
including upper and lower pos$ t1on$ ng p1nsi was carried out with video and 35mm cameras to provide tape and photographic documentation of the damage.
These data were used to produce Figure 2 which summarizes the extent of damage at each suppor t 1 ug and pos$ t$ on$ ng p$ n 1 ocatfon.
A compl ete
record of results of Inspect)on of the thermal shield and core support barrel
$ s provided 1n Chapter 5 of Ref. (1).
The damage to the core support barrel involved tearing loose of four of the n$ ne lugsi Nos.
1.
2i 3i and 6i with the fracture at Lug No.
1 leaving a hole through the core support barrel.
All lugs except Lug No. 7 had cracks 1n the core support barrel.
and at six locations~
Lug Nos.
1>
3-6i and 8.
one or more cracks penetr ated the core support bar rel wall.
~
IlmrmalZd~ Bmmml mal Bauaim M~~ 2euurf; 33uz.eel Analyses by CE to evaluate the reactor ~s operational capabilities for the r ema$ ndet of Its design l 5fe without a
thermal shield
$ nd$ cated repl acement of the thermal sh)el d was not necessary.
Effects of both Increased neutron flux l evel on tr ansft1on temperatures of the reactor vessel and
)ncreased gamma ray heating on the vessel response to pressurized thermal shock type transients were considered
$ n the analyses.
~
Thus.
since refurbishment of the thermal shield support system was fudged to be Impractical~
and since the shield was found to be unnecessaryi a
dec$ s$ on was made to remove the thermal shield and to repair the core support barrel.
The removal and repair hav'e been completed.
but are not the subject of th1s report.
St.
Luc)e No.
1 had been operating
$ n a normal fashion for seven years
when as noted prevfouslyi the damage to the thermal shield and core support barrel was discovered during the planned refueling outage at the
end of the fifth fuel cycle in March>
1983.
The first fuel cycle began April 22'976 when the reactor went critical for the first period of operation.
Dates for the beginning and ending of each fuel cycle are shown in Table l.
I~ 1 ~m M Each Eual Qzle L2 Eual Q~
Zazi~
Xmi~
1 lA 2
3 4
5 4-22-76 12-4>>76 5-26-78 6-8-79 5-7-80 11-29-81 7-10-76 3>>28"78 4-1-79 3-15-80 9-8-81 2-26-83 Commercial power production began in Decemberi 1976 during the second portion (Cycle 1A) of the first fuel cycle.
On March 2li 1983~
problems were encountered in attempting to shift fuel assemblies.
Inspection of the reactor to determine the cause of blockage led to further examination which revealed the damage to the thermal shield and core support barrel.
There was no indication during prior operations of any problems with the thermal shield or core support barrel.
The only abnormality noted involved the Loose Parts Monitor (LPM) system which exceeded the alarm level in February 1978>
March 1980'nd September 1982.
Several investigations were carried out to determine the significance of the LPM signals.
These analyses indicated the alarms were caused by a very small part (1 to 8 ounces) in the bottcm of the containment vessels which did
0
not pose a hazard to continued safe operation of the reactor.
Remote v)sual 1nspect$ ons of the lower portion of the reactor vessel
$ n 1978 and 1981 fangled to locate loose parts.
With information then available~
a more extensive Inspection would not have been warranted.
~ Eel~ ZzulMm M 2th@. Bmcfzz Zlnn&
FPL has expel fenced an unexpected problem which may be generic 1n nature.
The St. Lucre No.
1 reactor 1s one of three skmflar CE-des)gned units 1n this country that have experienced very similar problems with thermal shield and/or pos1t$ on$ ng pin performance.
Three of the upper posit)onfng p1ns were dislodged from the thermal shield of the Maine Yankee plant of Ma1ne Yankee Atomic Power Co.
The loss was d)scoveredi and the three p$ ns were retrieved~.
$ n Octoberi 1982.
A fa)lure of the thermal shields markedly similar
$ n pattern to that of St.
Locke Unft 1>
was found at Millstone Unit 2 of Northeast Nuclear Energy Co.
1n June'983 aftet the St.
Lucre Unft 1 damage had been reported.
There had been no prior 1nd$ cat$ on of a problem at Millstone.
It $ s pertinent to note that the thermal sh)eld of each of these three units was installed by a
different contractor~
and that CE d$ d the 1nstallat)on of the Mafne Yankee 10
The moor activ1ties of CE following discovery of the extent of damage to the thermal shield and 1ts support system concerned repa1r of the unit to return it to operation.
This involved removal of the damaged shield; a complete and th'orough exam1nation of the core support barrel by visual and nondestructive test procedures to assess and document the II extent of damage; devising and carrying out means of repairing the holes and cracks in the core support barrel; determining the capability of the unit to operate without a thermal shield> including analyses of the effect of increased neutron flux on pressure e vessel properties and of gamma ray
~
~
heating on performance; and safety analyses.
The scope of work is shown by Ref.
(1).
The failure mechan1sm analysis program>
covered in Chapter 7 of Ref.
(1)>
was conducted concur ent with the repair activit1es.
This effort was of primary interest to FTE and provided the basis for further investigation of fa1l ur e mech'anisms.
Thereforei the CE pr ogram and results are reviewed briefly in this section.
~ K Zd1!um Haahm1za 3~~ Zzagzm The program devised by CE to identify the mechanism responsible for the thermal sh1eld fa1lure consisted of investigation in six areasi as follows:
Hydraulic Loads~
Per iodic and Random Structural
Response
Characteristics of Varying Support System Conditions Designs Fabrication and Installation Data Pressure Vessel Monitoring Procedure (PVMP) and Damage Visual Inspection Data Loose Parts Monitor (LPM) and Internal Vibrations Monitoring (IVM) Data Metallurgical Examination The information developed in these six ar eas was then integrated in further analysis to develop candidate mechanisms for initiation and propagation of the cracking free which,a failure mechanism was identified.
CE~~
CE conducted a metallut gical examination comprising an analysis of broken pieces of the thermal shield and cor e support barrel and an evaluation of the components~
material properties.
Because of radioactivity of the thermal shield and core barrel material. it was not possible for CE to perform any direct examination or tests on samples.
All work had to be done in a hot cell.
The facilities of Battelle Columbus Laboratories (BCL) were engaged for the metallurgical examination of samples ft om St.
Lucie No.
1.
The only work done outside of the hot cell was Scanning Electron Microscope (SEM) fractography on plastic replicas.
12
Ten loose par ts retrieved from the bottan of the reactor vessel at the beginning of the recovery program were sent to the BCL hot cell facility for examination.
Those parts are listed in Table 2.
Table 2.
Loose Parts Recovered from Bottom of Reactor Vessel One lock bar from a positioning pin Two positioning pins (upper pins X and Z
missing fran the thermal shield at lug locations 7 and 9).
One complete support pin with a piece of thermal shield attached.
One partial support pin.
Two small unidentifiable pieces (probably lockbar pieces).
Three pieces of thermal shield.
The work performed in the hot cell consisted of visual examination of the parts. tensile tests of specimens cut and machined from thermal shield samples.
and optical microscopic examination of small specimens cut from thermal shields support pin and locking bar samples.
The SEM fractography on plastic replicas was done at the CE facilities in Windsor'onnecticut.
At a later date~
Lug No. li which had broken out of the core support barrel>
was sent to BCL.
Additional fractography was performed on fracture surfacesi and tensile and fracture toughness specimens were machined from attached core support barrel material to determine material pr operti es.
CE concluded that damage to the thermal shield was caused by sel f-excfted vibration.
made possible by deter$ orat$ on of the thermal shield support system.
The deter for at]on
$ s presumed by CE to have been preceded by loss of preload on the positioning p$ ns f$x$ ng the thermal shield relative to the core-support bar rel.
Otherw)se>
no freedom of movement of the thermal shield would have been poss)blei no wear on the support pins and lugs would have occurredi and no sel f-excited mot'fons would have resulted.
According to CE~ loss of preload
$ n a suff$ c$ ent number of p$ ns would cause deter1orat$ on to begin by relat)ve motion and
)mpact1ng between the pos$ t1on$ ng p$ ns and cor e barrel.
The r esult)ng wear and loosening of the p$ ns could then lead to loss of radial restraint of the thermal shield>
with subsequent wear on the support p$ ns and lugs and>
ultimately>
self-exc$ ted v1brat$ on and failure under the action of the water flowing around Although the specific reason for the loss
$ n preload of the positioning pins has not been Ident'led by CEi hydr aul fc pt essure loadfngi fnftfal weight transfer and radfatfon-induced stress relaxatkon have been shown to be capable of contr$ but$ ng to th$ s situation.
In this contexts pressure 1 oad1 ng refers to the 1 oad$ ng resul t$ ng from dffferentkal pressure between the fnside and outside of the core barrel.
Initial weight transfer
$ s the fnftfal tr ansfer of thermal shield weight 14
to Its supports after the pos$ t$ on$ ng pins are preloaded and auxkl fary supports are removedi and rad$ at$ on-Induced stress relaxatkon
$ s the long-term modkffcatkon of the stress-stra$
n relationship
$ n the p$ n material resulting from radiation effects.
In exam$ n$ ng
- the, pos$ t1on$ ng-pin loadfngi two cases have been considered:
(1) full power operation and (2) isothermal zero-power operation.
In the f$rst case>
thermal expansion causes an Increase
$ n the compressive loading of the pos$ t$ on$ ng p1ns which
$ s much greater than any calculated reductions refer red to
$ n the previous par agraph.
Thus>
wear and loosening of the p$ ns are not to be expected during power operation.
In the second case~
the thermal loading
$ s absent and the fnkt)al preload
$ s subject to some reduction.
Howevers according to CE~s analysfs all but four p$ ns stall have s$ gn1f$ cant pr eload after taking Into account the losses due to pressure loadfngi weight transfer and stress relaxatfon.
CE recognizes this fact and speculates that unfdentkffed fnstallatton errors could have caused further reductions not amenable to direct
. calculation.
In their postulated sequence of events
$ n Chapter 7
of Ref.
(1)i CE accordingly notes:
"The reasons for the loss of preload have not been spec)focally fdentfffedi but several factors have been examined and found capable of contributing.
It 1s believed that a
comb<nation of the detrimental factors 1s the most reasonable explanation of the loss."
15
FTI concur s with CE that failure was
$ n$ t1ated by mechanisms that caused the positioning p$ ns to lose their preload.
Howevers
$ n view of the fact that the CE analysts did not demonstrate any exp11cft cause for complete loss
$ n pr el oad~
FTI has 5 nvest1 gated add[ t1onal posse bl e causes'ncludkng the.effect of thermally Induced stresses
$ n the
'p$ ns during the full power operation.
The p$ n stress analys)s
$ s presented 1n this section.
The stress cal cul at$ ons descry bed 1 n th 1 s sects on show that s$ gn$ f$cant stress~
exceeding the yield strength of the pos$ t1on$ ng pfns>
can occur during full power operation.
Thfs overstress can cause plastic d$ stort$ on of the pins and can contribute to loss of preload during zero-power operation by not allowing the p$ ns to spring back fully to their 1n)t)al confkgurat1on when the thermal loads ate removed.
The 1 arge str esses 1n the posft$ on$ ng p1ns during power operation arise ma)nly from compressive preloadi thermal expansion of the p$ nsi and radial and axial thermal expansions of the core support barrel relative to the thermal shield.
The pr eload produces uniform compressive stress
$ n the pins.
The r clat)ve radial thermal expansion of the core support barrel> together with thermal expansion of the p$ nsi also produces uniform compressive stress 1n the p$ ns.
In contrast>
the relative ax)al thermal 16
expans1on of the core support bar rel causes a vert1cal force on the p1n ends 1n contact w1th 1t.
Th1s force bends the p1ns downward and produces add1t1onal compress1ve stresses.
In comb1nat1on w1th the stresses created by the prel oad~
rad1al thermal expans1oni and p1n expans1on>
FTI has calcul ated that th1s bend1ng for ce can produce the overstress referred to above.
There 1s no 1nd1cat1on 1n CE~s fa1lure analys1si Ref. (1)i that CE cons1dered the bend1ng force 1n calculat1ng str esses 1n the pos1t1on1ng p1n s.
In calculat1ng the stresses 5ust descr1bedi one must ass1gn coeff1c1ents of fr1ct1on for the threads on the pos1t1on1ng p1ni and fot the area of contact between the end of the p1n and the bear1ng pad on the core support barrel.
(F1gure 5~
page 21>
shows a threaded pos1t1on1ng p1n 1n contact w1th the bear1ng pad).
FTI has used a coeff1c1ent of fr1ct1on of 0.15 as typ1cal for these surfaces when they, are lubr1cated for 1n1t1al A
1nstallat1on.
FTI has exper1mentally der1ved a coeff1c1ent of 0.29
- for the area of contact between p1n and bear1ng pad once th1s 1n1t1 al
=
lubr1cat1on 1s lost.
However~
FTI~s calculat1ons of the stresses 1n the pos1t1on1ng p1ns show that the p1ns were overstressed regardless of whether the coeff1c1ent for lubr1cated (0.15) or unlubr1cated (0.29) surfaces 1s used.
In the former case>
the upper and lower p1ns were overstressed (relat1ve to the1r y1eld strength) by 44%
and 25%i respect1vely.
In the latter casei the calculated overstress 1ncreases to 106%
and 78%~ respect1vely.
17
The conclusion fran the stress analys)s
$ s that loss of preload and r estrade ning action of the p$ ns dur) ng zero-power operation
$ s attrfbutablei at least par teal lyi to the large ovet stress>
and that the upper pins are the most overstressed.
~ Z~Mxuf Iharmal ~ld,~~ Ruuuzf; fhuxe1 F)gure 4 Illustrates the overall geometry of the core support barrel (CSB) and thermal shovel d (TS) >
both being located fnskde the reactor pressur e vessel (not shown).
If a mor e accurate representation 3 s required refer to Figure 1> Appendix B.
18
CSB TS Figure 4.
Core Support Barrel and Thermal Shield 19
The core support barrel
$ s suspended free the reactor vessel at the top and 1s restra)ned radially at the bottan by snubbers.
The thermal shield
$ s supported at Its top by 9 lugs welded to the core support barrel at 40 degr ee interval s around its c$ rcumfer ence which mate with gr ooved cyl)ndrfcal support pins fftted and welded to the thermal shield.
Rad(al restraint
$ s provided by 9 upper and 17 lower pos1t1on$ ng p$ ns threaded from the outside through the thermal shield and bearing against the core support barrel.
The upper pins are located 15 inches below the top of the support lugsi and the lower p1ns are uniformly spaced around the
,circumference approximately 21 1/4 Inches above the bottom of the thermal shield.
All materials for these components consist of Type 304 sta)nless steel.
The core support barrel has an Inside d$ ametet of 148 Inches and a
m1nkmum wall thickness of 1 3/4 inches.
The thermal shield has an
$ ns$ de diameter of 156 3/4 inches and a wall thickness of 3
inches.
The pos$ t$ on$ ng pins cons)st of 2 Inch nominal diameter threaded. bars.
These pins bear against pads welded to the core support barrels as shown
$ n Figure 5.
The thermal shovel d support consists of gr ooved cyltndrfcal support p1ns attached to the thermal shfeldi and hung on lugs attached to the core barrel.
20
Thermal Shi'el d gore. Support
,Barr.el-Figure 5.
Sketch of Pos) t$ on) ng Pin CUBI M~ Zl The positioning pins wer e initially sub)ected to an installation torques r esul ting in a
design pin'rel oad.
This load is increased const derabl y during ful 1 power operati on because of radi al thermal expansion of the core support bar rel relative to the thermal shield and thermal expansion of the pins.
It is also deer eased somewhat because of steady-state hydraulic.pressure differences that exist between the outside and inside of the core support barrel. Additional pin loadings result fran
)nit)al weight tr ansfer and radiation-induced stress relaxation>
as 21
characterized in Chapter 7 of Ref. (1).
Table 3 gives average net positioning,pin loadings during full power operation which are summations resul ting from prel oad>
thermal load>
hydraulic pressure load>
and weight transfer and radiation-induced stress relaxation.
Also given in Table 3 are average net positioning pin loadings during zero power operation when no heat is being generated by the nuclear fuel.
all reactor internals are at the same temperature.
and all thermal loads are absent.
All calculations are mean values for the time period 1976 to 1980.
Bmxfuz I'<md lm~
Eull Zamr Pin Location:
upper lower upper 1 owel Net Pin Load (1976-1980) 45i600 39>800 1
4i900 5 i3 00 According to CE> it is the 1 oad reductions at zero-powers together with other reductions from postulated installation errorsi that probably caused complete loss of pr eload in the positioning pins during zero-power operations.
No evidence of any installation errors has been provided by CE>
and FTI~s review of the thermal shield installation documents has uncovered no such errors.
Additional loss of preload could'owevers have resulted fran overstress of the pins during full power operation.
This possibility was 22
0
not fully addressed by CE since CE considet ed compression loads but emitted bending loads in the pins.
Overstress of the pins during full
'\\
power operation.
coul d cause pl astic deformation and permanent shortening of the pins.
and lead to further loss of preload during zero-power oper ation..'n addition>
distortion of the threaded connection of the pin could also occur due to overstress during full power operation~
leading to loosening and misalignment of the pins during zero-power operation.
Detailed analysis of the stresses in the positioning pins investigating this possibility is described in the remainder of this section.
~ SMammM Based on the loads given in Table 3i the uniform compressive stresses in the pins during power operation can be calculated by dividing the net loadings given in Table 3 by the pin area (3.14 sq inches).
It must be rememberedi howevers that the positioning pins are generally subjected not only to uniform compressive stressesi but also to bending stresses.
In particular>
as the reactor is brought from zero power to full power operations the core support barrel is heated more than the thermal shield>
with greater vertical as well as radial thermal expansion of the core support barrel relative to the thermal shields thus causing bending of the pins not accounted for in CE~s analysis.
Figure 6 illustrates the bending geometry for this situation.
0
POSITIONING PIN TS CSB F
(sketch exagger ates defi ect)on)
Figure 6 - Bending of Poshtfonfng P)n Due to Relative Thermal Displacement of Core Support Barrel.
Detailed stress calculations for combined compression and bending were made by FTI using the average value of the net p$ n load)ngi as given
$ n Table 3.
These calculations were made using two fr$ct$ on coeff$ c$ ents>
24
a value of 0.15 rept esentat$ ve of Initial lubricated conditions and a
realistic value for cond1t$ ons after fn)teal lubrfcatfon
$ s lost.
0.29.
The former value
$ s typical for lubr $ cated surfacesi 'nd the latter value was determ$ ned experimental 1 y by FTI.
The str ess val ues derived are tabulated
$ n Table 4.
Table 4 Maximum and Minimum Stresses for Two Different Fr1ctkon Coefffcfents (Compressive Stress js Positive)
Thermal Shield End (Threaded End)
Frkctfon Coef.
Pfn Locat)on Upper Lower S max (ps<)
0.15 0.29 26~600 38il00 23 i200 32z900 S min (ps')
0.15 0.29 2e330 9p040 2i200
-7 z560 Core Barrel End (Bearing Pad)
Upper Lower 25e200 36e900 22w700 32~700 3 F800 2>670 The stress d$ str$ but$ on on both ends of the upper p$ ns
$ s Illustrated
$ n Figure 7 for the case where the more real fstfc coeff1cfent of friction of 0.29
$ s assumed.
25
TS CSB 9,040i I 36,900 psi 38,100 Fig. 7.
Stress Distributions on Ends of Upper Positioning Pins
~
~
During Normal Power Operation (friction coefficient = 0.29)
~ Xl ld lit @loth According to Ref. (3) the yield strength of Type 304 stainless steel at the operating temperature of 580 degrees F is S
=
18>500 psi By comparison of this value with the calculated stress values in Table 4i it can be seen that the calculated stresses in the pins exceed the yield strength on both. ends of the pins regardless of whether the initial friction coefficient (0.15 for lubricated sut faces) or the more real istic value (0.29 for non-1 ubricated surfaces) is used for the V
calculations.
For the former friction coefficient.
the upper and lower pins are overstressed (relative to the yield strength) by 44%
and 25%~
26
respectively.
For the latter friction coefficient they are overstressed by 106%
and 78$ respectively.
Considering the stress distributions of Figure 7i it is seen that the yield stress has been exceeded and significant plastic deformation can be expected on the lower part of the pins and threads at the thermal shield end and on the upper part at.the core-barrel end..
The actual stress distributions will in fact be somewhat more uniform because plastic deformation will tend to relieve the larger stresses and increase the smaller stresses at each end.
In any event.
the overstress and resulting plastic deformation can cause a permanent overall shortening. of the pins'hereby contributing to loss of pr eload during zero-power operations when the thermal stresses are absent.
Of course>
as noted earlier. the large stresses acting on the thermal shield end of the pins can also cause thread damage.
leading to further loss of preload and po'ssible misalignment.
With yielding'he threaded joints are also subject to ratcheting because the mating parts may become loose on r eturn to zero-power conditions and start the next power operation in a new r elationship with each other.
When the unit returns to full power additional distort tions may occur. After a series of cycles this ratcheting can cause the threads eventually to lose engagement.
Although the major difference between FTI and CE analyses of pin stress was the incorporation of the bending effect>
somewhat different calculated values were also obtained in the two analyses for uniform 27
compress) ve stresses fn the p$ ns under ful 1 power operate on.
CE calculations
$ nd$ cate smaller uniform compressive stresses
$ n the upper p$ ns>
and larger such stresses 1n the lower p1ns than resulted from the FTI analys)s.
However.
when the effect of bending
$ s added to the CE values>
the calculated stresses for top as well as bottan p$ ns, exceed the yield strength of the materfal.
Xzaaaa Za Because of the adver se consequences of pl ast1c deformation
$ n the thr eaded connectors~
the calculated p$ n over stresses given
$ n Table 4
are considered excessive
$ n terms of accepted eng$ neer$ ng practice.
In facti Section III of the 1974 ASME Boiler
& Pressure Vessel Codex'Ref (3)>
r equ$ res for nucl ear
. power pl ants that max[mum stresses
$ n threaded connectors of pt essure conf 1 n1 ng components never exceed the y$ el d strength of the materfal and that threaded fasteners of non-pressure conf)ning canponents never exceed the yfel d strength by more than 20%.
Thfs code requirement represents accepted eng$ neer$ ng practice and
$ s a
proper standard for evaluation of the St. Lucre Unkt No.
1 reactor design.
The pos$ t$ on$ ng p$ ns of the thermal shovel d may be regarded as threaded fasteners of a non-pr essure conf1n1ng component; the maximum allowable stress
$ s thus 20$ greater than the yield strength>
or 22>200 ps<.
Comparison of thfs value with the calculated stress values in Table 4
shows that the calculated stresses exceed th1s allowable value 28
r egardless of which value of fr$ct1on coefficient 1s used.
For the 1n)teal coefficient value (0.15 for lubricated sur faces).
the maximum calcul ated stresses fn the upper and lower p$ ns exceed the allowable stress by 2(6'nd 5$ respectively>
and for the more realistic friction coef f1c1ent value (0.29 for non-lubricated surfaces) i they exceed the allowable stress by 72$
and 48%. respectively.
~ ~E~
fa D
I a~~luuf md 2miing JsuM The calculated p$ n stresses as described
$ n Sections 4.5-4.7 tndfcate that p$ n damage can occur at the threaded end.
The bending compressive stresses on the lower part of the p$ n (at the threaded end) add to the uniform compressive stresses to produce calculated compressive stresses at this location greater than the yield stress of the mater)al.
On the upper part of the ping howevers the tensile bending stress subtracts from the uniform compressive stress to 'g1ve calculated stress values less than the yield stress (see Figure 7).
Thus'nitial pin damage at the threads fs expected on the lower s$ de of the threads only.
After failure of the St.
Lucre thermal shield support system was discovered.
two upper pos$ t$ on$ ng p1nsi Pfn X and Pfn Z from lug pos$ t$ ons 7 and 9. resp'ectfvely.
were recovered from the bottcm of the reactor.
Pin X was worn smooths presumably from motion while loose
$ n the reactor after d)slodgfng.
Pin Z also showed wear but to a lesser degree than Pin X.
Detailed photographs of these pins are contained
$ n Ref. (1).
Although Pin Z appears
$ n these photog aphs to have Intact threads~
29
CE advised that considerable wear of the tht eads had occurred.
FTI personnel visited BCL in March 1984 to assess the extent and pattern of wear on Pin Z.
Also available for examination was Pin Wi which had been sent to BCL for use as a control sample in activation analyses conducted to determine when Pins X and Z fell out.
Both Pins Z and W were examined>
and photographs wer e taken of the ends and around the circumference of each pin.
The condition of Pin W was of greater interest to FTI~ since it had remained in the thermal shield and thus was indicative of the effects of loads experienced.
Photographs 5 through 9i Appendix A>
show the condition of Pin W.
At 0
degree orientation~
severe thread damage is seen; at 90 degree orientationi little damage is seen; at 180 degree orientationi little damage is seen>
the notch about mid-length of the threads having been made after pin removal for sample material; and at 270 degree orientationi where the thread damage seen in the 0 degree orientation is again visible.
The conclusion from direct examination of Pin W is therefore that moor thread damage occurred on one side of the pin onlyi as seen in the 0 degree and 270 degree orientations.
This result is consistent with that expected from the combined uniform and bending stress calculations given in section 4.5.
30
i
FTI concurs with CE that large ampl)tude motion of the thermal shield 1s the most probable cause of cracking of the thermal sh)eld and Its support system.
The source 1s fdentff)ed
$ n the CE analysis as self-excited'igh-amplitude vibration of the thermal shield develop)ng under forces present during steady-state 4-pump operation of the reactor plant.
Howevers CE showed that s$ gn1f$ cant motion by sel,f-exc)ted vibration can occur only after loss of rest[ a$ nt of all positioning pins'nd extensive wear and degradation of the support system.
FTI conducted an independent analys)s of the damage to determine the cause of fakl ut e.
This included r eve ew of fracture patterns>
condition of the thermal sh)eld~
core support barrel and lugsi positioning pins~
and bearing pads~
as observed and documented by CE.
To explore cond) talons other than normal i steady state oper at1ons of the reactor plant>
the plant operating logs were reviewed to determine the pattern
'nd/or frequency of changes
$ n operation of the four reactor coolant pumps (RCP~s) which circulate approximately 370i000 gallons of water per minute through the reactor plant at a flow velocity on the order of 30 feet per second at the thermal shovel d.
Over 500 changes 1n number of pumps operating were found to have been made
$ n the seven years of oper at)on.
The changes
$ n pump operation produced asymmetric loading of the thermal 31
shield and Its support system.
Thfs cyclic loadfngi
" following loss of restraint by the posft1on)ng pins and subsequent weat and degradation of the support system by v$ brat$ on of the thermal shkeldi offers a plausible mechanism for developing the failure of the support system and damage to the thermal shield.
'TI reviewed the entire pattern of damage for further clues to the fa11ur e mode.
The orfgknal inspection of the suppor t system Indicated that the thermal shield had been displaced downward and was tilted about a
north-south axles.
with the lower sfde
$ n the area of Lug Nos.
6 to 8 (hei on the West).
Figure 2i Appendix B shows that damage to the thermal shovel d appeared to be concentrated at the North and South 1ocat)ons.
Howeveri damage to the lugs and core support barreli
$ n the form of lug breakouti cracking of the cor e support bat rel around the lugs> or wear of the lug support p$ n
$ nterfacei occurred around the full circumference.
The lower pos$ t$ on$ ng pins on the West s$ de (Pins Ki Mi Ni 0 and P) protruded from the outer sur face of the thermal.shovel di
$ nd$ cat$ ng they had been driven out.
The lower pos$ tfon$ ng p1ns on the opposite side were not protruding andi when
$ nspectedi were not contacting the core support barrel.
The pattern fndfcated a rocking motion of the thermal sh)eld with a
more or less North-South axis of rotation.
However i the movement appeared to be to one s$ de only>
with the West s$ de of the thermal shield moving
$ n toward the core support barrel and back to the neutral pos$ t1oni 32
without a
corresponding reverse rocking of the East side in toward the core support barr el.
A rocking action woul d produce a
twisting action adding a torque load to the support system.
Fur ther evi dence of the pattern of thermal shi el d movement was obtained by inspection of the condition of the positioning pin bear ing pads on the core support barrel.
CE also 1nspected all of the bear1ng pads.
and prepared a sketch of the condition of each pad for inclusion in Chapter 5 of Ref (1).
These sketches>
presented as Figures 5.4>>11 through
.5.4-36 inclusive in Ref (1)~
have been combined into a single drawing in Figur e 3>
Appendix B.
The circular ar rangement shows each pad in its position relative to other pads and to the support lugs.
The representation of position1ng pin imprint markings on the CSB bearing pads>
Figure 3i as a series of circles accurately rept esents the conditions noted in viewing photographs and videotapes.
The mar kings appeared
-to be a series of discrete impr essions caused by 1mpact of the positioning pi n on the pads rather than the resul t of a
gradual displacement while in contact.
All of the pads show an imprint at the center>
probably representing the initial concentr 1 c al ignment of posit1oning pins and bearing pads.
Most of the pads show additional imprints at successively lower positions on the pads>
with skew1ng to the tight or left.
These changes in imprint position on individual pads are described in more detail in Appendix C.
What 1s consider ed significant is the downwar d pr ogr essi on of 33
~
i
1mpr 1ntsp 1nd1cat1ng the poss1b111ty that wear of the support system and downward settl1ng of the thermal sh1el d was 1nterspersed w1th large-ampl1tude>
s1deway or rock1ng mot1ons of the thermal sh1eld>
wh1ch caused 1mpact of the pos1t1on1ng p1ns on the1r bear1ng pads.
Th1s 1nd1cates that 1 arge ampl 1tude mot1 on of the thermal.sh1el d was not necessar1ly the result or culm1nat1on of extens1ve wear and degradat1on of the support system~
but ratheri that such mot1ons of the thermal sh1eld occurred at 1ntervals dur1ng the wear of the support system.
The process proposed can be descr1bed s1mply as wear-1mpact-wear-1mpact-etc.
One mark1ng of part1cular 1nterest was the 3 1mpr1nts on the bear1ng pad of P1n Z~
wh1ch was one of two p1ns recovered from the bottan of the reactor vessel.
As ment1oned ear11er 1n Sect1on 4.8i CE conducted rad1oact1v1ty analyses of P1ns X and Zi along w1th P1n W as a
control samplers to determ1ne when P1ns X
and Z fell out of the thermal sh1eld.
Howevers CE concluded that there was too much var1ab111ty among the p1ns for a rel1able est1mate to be made by th1s method.
Nevertheless>
there was an 1ncrease 1n the level of Loose Parts Mon1tor (LPM) s1gnals 1n Marchi 1980 as well as later 1n September.
1982'h1ch may have been due to the second p1ni P1n Z~ fa111ng out of the thermal sh1eld.
The bear1ng pad 1mpr1nts show that the wear 1mpact process was 1n progress when P1n Z fell out; the LPM data suggests th1s may have been as early as March>
1980.
Th1s strengthens the poss1b111ty that the downward progress1on of 34
1mpr$ nts was the result of wear and impacts occurring over a several-year period.
~ Severe wear occurred at the lug-support p1n
$ nter face of several lugs.
Th)s
$ s 1ndkcated on Figure 2i Appendix Bi by a gap between the bottan of lug/shawm assembly and the bottcm of the slot 1n the thermal shield.
Also'hotograph 3> Append)x Ar shows a notch worn
$ n the top of Lug No.
9 at the right rear portion of the support p$ n contact area.
The largest gapsi 2 Inches h$ ghi were observed under Lug Nos.
7 and 8.
Gaps 1n the range of 3/4 to 1 1/4 Inch were found under Lug Nos. li 5
and 9~
whfle a gap of 3/8 inch was recorded under Lug No 4.
Lug Nos. 2i 3 and 6
dfd not show any gaps.
A possible mechanism 'for the severe wear observed at the thermal sh)eld support p$ n-lug Interfaces
$ s fretting corrosion.
The Interface
)nvolves Type 304 stainless steel bearing on Type 304 stainless steel~
a metal comb$ nat$ on known to Incur h$ gh rates of metal loss under conditions of fretting cor rosioni as well as problems of seizing and galling.
Wear by fretting corrosion occurs between two contacting surfaces under load>
sub) ected to sl fght oscar 1 1 atory or reck procat$ ng mot$ on.
Substanti al metal loss can be caused with the motion repeated over millions of cycles.
The thermal shovel d support system
$ s sub) ected to high fr equency fluctuations
$ n hydraulic pressures during 4-pump operation of the reactor plant due to pump discharge pressure pulsations associated with pump rotor frequency (15 Her tz) and pump blade frequency (75 Hertz).
Once radial
restraint of the pos$ t1on$ ng p$ ns
$ s lost and slight rad)al motion of the thermal shield
$ s possible 1n response to hydraulic pressure fluctuations>
cond) talons are conducive to frett) ng cor ros$ on at suppor t system interfaces.
The comb)nation of the obsef ved fracture patterns>
pos$ t1on$ ng pin
$mpr$ nts on the hard face pads~
and wear patterns on lugs and support p$ ns shows that the thermal shield was sub)ected to cons)derable forces and wear during operation that caused both downward and sideways (west to I
east) displacement of the thermal shield.
While th$ s rocking ot canting I
motion of the lower portion of the thermal sh)eld appears to be the predominant movement~
the extent of cracking indicates other dkrect1ons of motion occurred also'epending on the forces involved and the condition of the support system at any given t1me.
The damage to the 1 ower pos$ t$ on$ ng p$ ns and bearing pads was concentrated on the western portion of the c$ rcumferencei between the North and South positions which are the locations of the two outl et nozzles.
The posse baal sty of 1 arge-ampl etude movements of the thermal shield by hydraul)c forces due to cha'nges
$ n the reactor coolant flow was considered.
Dur)ng most of the operation of the reactor planti all four reactor cool ant pumps are run. constantly>
so that there are no moor variations
$ n flow patterns.
The CE
$ nvest$ gat$ on of 'hydraulic forces and structural response was 1 fmfted to th$ s cond$ t$ on of 4-pump operation.
36
However>
there are times when one or more of the reactor coolant pumps (RCP) ar e stopped for part) al shutdowns
$ n the course of normal operat3onsi and then restarted.
Also> startup and shutdown procedures for the reactor plant at outages
)nvolve a ser1es of changes
$ n pump operation between zero flow and full flow cond$ t$ ons.
Each RCP
$ s rated at 92>500 GPM for 4-pump operation and 120i000 GPM for single pump operation.
The output of each pump enters the r eactor vessel at one of the four 1nlet nozzle locat)ons shown on Figure 1.
Two Inlets are 60 degrees apart on the East s$ de>
and the other two inlets are 60 degrees apart on the West sider leav)ng 120 degree arcs between inlets at the North and South locations.
Because of the high flow rate and Inlet locat)ons>
each change 1n RCP operation (of a single pump or a group of pumps)
$ s cons)dered to pose a sfgnhf)cant change
$ n the hydr au11c forces acting on the thermal 'shield and core support barrel~
with a blas toward loading the East and West sides.
Th1s corresponds to the pattern of damage to the thermal shield.
Therefore~
the operating records of St.
Lucre No.
1 were reviewed for its entire period of operation from April 22>
1976>
when the r eactor first reached crktfcal condktfonsi to March 1983.
Each change
$ n pump operation was noted
$ n preparing a log of RCP operation.
Table 5 summarizes the ffndkngs by presenting the number of changes 1n pump operation that occurred each month from April 1976 to March 1983.
On thfs table one change
$ n pump operation
$ s defined as a change
$ n the 37
Table 5
Cf/ANGES IH NUt1DER OF REACTO LANT PUl<PS OPERATING BY r>ONTH 1976 1977 1978 1979 1980 1981 1982 1983 month Cum.
Juan.
Total Cum.
Juan.
Total Cum.
Juan.
Total Cum.
Juan.
Total Cum.
Juan.
Total Cum.
Juan.
Total Cum.
Juan.
Total Cum.
quan.- Total Jan Feb Mar Apr Hay Jun Jul Aug Sep Oct Nov Oec 4
4 2
6 29 35 6
41 (outage)
(outage)
(outage) 15 56 7
63 0
63 7
165 22 252 67 0
165 3
255 67 3
168 0
255 54 121 (outage) 3 258 0
138 0
199 0
269 4
142 0
199 26 295 16 158 10 209 20 315 0
158 16 225 0
315 0
158 5
230 0
315 13
. 134 31 199 11 269 4
138 0
199 0
269 0
138 0
199 0
269 0
315 0
315 3
318 (outage) 22 340
~
44 384 0
384 0
384 0
384 0
384 0
384 0
384 (outage) 0 570 5
510 0
570 11 521 0
570 0
384 0
521 0
384 0
521 0
384 0
521 16 500 0
521 0
500 41 562 0
500 0
562 0 '00 0
562 0
500 0
562 5
505 8
570 0
570 34 604 18 622
number of RCP's operating>
without regard to the number of pumps that may have been started or stopped at one time>
or the duration of the changes.
Included in the values shown in Table 5 are the starting and stopping of RCP~s in filling and venting the reactor coolant systems a
procedure followed each time that the system was opened during a
plant shutdown.
Changes in RCP operation were associated with shutdowns and startups of the reactor plant.
Not all shutdowns involved stopping and starting of RCP~si but when such changes in pump configuration wet e required.
a series of changes were generally made.
Reasons for reactor shutdowns covered a
wide variety of conditionsi such as problems in the steam generating or electrical generating and distribution systems as well as with the RCP~s and other reactor plant components.
These are considered to be a part of normal operation of a reactor power plant.
As evident in Table Si a total of 522 changes in RCP operation was logged in the approximately seven years of plant operation.
Changes in RCP operation were made in 35 of the 84 monthsi but only 19 of these months involved 10 or more such changes.
There were 43 months in which no changes were made (excluding 6 months of outages).
indicating continuous.
4-pump operation of the plant in these months.
The r ecord in Table S
shows a pattern of months of little or no change in the 4-pump operation of the planti interspersed with months in which a series of changes in pump configuration was made.
Each change in pump oper ation is considered to cause a change in the 38
hydr aulic loading of the thermal shield>
support system~
and core support barrel.
This pr ovides a plausible mechanism for generating a
limited number of 1 arge.
displ acements of the thermal shiel d once it became partially unrestrained.
These displacements could have led to the type of failure found in the suppot t pins and lugs.
In addition> large movements of the thermal shield~
interspersed in periods of continuous operation~
in which wear of the lugs and support pins is believed to have occurred>
would match the pattern of imprints on the bearing pads descr ibed in Section 5.2.
~ 3hdraulic Lmkiag~ Zramimi Elm Steady state flow conditions have been calculated by FTI based on geometric data provided by CEi basically confirming the CE steady state pressure values.
Transient asymmetric flow conditions are more difficult to analyze.
Howevers in an effort to quantify the hydraulic loading on the thermal shield associated with the start up of a single coolant pumps calculations have been made based on a 30 inch diameter inlet (cold leg) pipe and 92~500 GPM flow r ate (120~000 GPM for one pump operation) of pressurized water at 551 degr ees F and 2250 psi.
For purposes of calculating the steady state pressure 1oadingi the flow is assumed to enter the reactor~
strike the cor e support barrel and be defi ected downwar d with a 60 degree sector of the thermal shiel d circumference within the flow stream.
For steady state f1 owi calculations indicate that a
force of 2>250 pounds per pump (3.700 pounds if only one pump is 39
operating) would act axially downward on the thermal sh)eld over the 60 degree arc.
Thfs 1s a result of the flow stagnating on the top of the 3-Inch thick thermal sh)eld.
The short durations Initial startup pulse would be somewhat gt eater.
No attempt has been made to per form f$ n$ te element calculatkonsi nor to calculate detail s of motion from this loading.
Howevers once d$ stort1on of the pos$ t$ on$ ng p$ ns and'wear of the support system allow freedom of mot1onr these hydraulic calculations
$ nd$ cate that suff)cogently large.
asymmetric loads can act on the thermal shield to cause a rocking motion of the type Indicated by the failure pattern.
40
While CE concluded that loss of prel oad on the poskt1onfng p$ ns probably preceded deter
$ orat$ on of the thermal shovel d support systems reasons for the loss of preload were not specifically 1dent$ f$ed.
CE postul ated addkt)onal loss fran )nstallatkon errorsi but d$ d not provide
'I evidence of such errors.
Howevers CE~s analysks of loads and stresses on the p1nsi and of var Ious factors which affect the loads and st[ esses>
considered only uniform compr essive 1 oad1ng of
~ the pins.
The FTI analysts.
on the other hand> took
$ nto account bending as well as uniform compressive loads',
and showed that stresses exceeding the yield strength of the materfal would-occur under the combined loading.
Y)eldfng can result
$ n permanent shortening of the p$ ns as well as ratchet$ ng of the threaded )oint.
This add)tfonal mechanism pl ov$ des an explanation for the complete loss of pr eload which does not require speculation about possible involvement of un1dent1f$ ed Installation errors.
'n the sequence of events postulated by CE>
deter)oration of the support system began when a suf fhcfent number of p1ns had lost their preload to permit relative motion and impacting between the p$ ns and bearing pads on the core support barrel.
Resulting wear
$ n the threads of I
the pos1t$ on1ng p$ ns would cause a loss of restraint of radial motion of the thermal shield and allow
$t to beg1n to uncouple from the core support barr el.
Increased response to hydr aul $ c 1 oads (pressure fluctuation during 4-pump oper ation of the plant) would Increase the relative motion 41
between the thermal shield and the support lugs~
causing wear of the lugs and thermal shield.
With suff$ c$ ent wear on the lugs> the thermal shield may have r cached an unstable dynamic situation.
The resulting large amplitude motions could cause h$ gh stresses
$ n the thermal sh)eld>
causing failure 1n a
rel atfvely short period of t$ me.
The 1 arge ampl etude response
$ s thought to have occurred during the 'fifth fuel cycle>
damaging the thermal shield and extending damage to the core support bar rel.
The fracture analys)s by FTI assoc]ates large amplitude motion of the thermal shield.
and cycl)c loading of the thermal shield and its support system.
with changes
$ n primary coolant flow due to changes
$ n the number of RCP's operating during partial and complete shutdowns and startups of the reactor plant.
While deflectio of the support system under such "1 oadfng woul d be pr evented at the outseti loss of prel oad 1n the pos$ t$ on$ ng p$ ns and v$ brat$ on-induced wear on-the support lugs could permit s$ gn$ f$cant deflect)on of the thermal shield support system with asymmetric hydraul)c loading of the thermal shield.
A probable sequence of events associated wfth the failure of the thermal shield and Its support system
$ s postulated by FTI as follows:
(1)
When the r eactor '
irst reached full power oper ation the pos$ t1on$ ng p$ ns were overstep essed as a
r esult of combined axe al and bending loads>
causing local ized yfeldfng of the pins.
Changes
$ n power level would have added to d$ stort$ on of the pos1t$ on1ng p$ ns>
causing eventual loss of preload
$ n the pfns.
42
(2)
Loss of restra1nt prov1ded by the pos1t1on1ng p1ns allowed low-ampl1tude v1brat1on of the thermal sh1eld dur1ng steady-state aper at1oni resul t1ng 1n wear and deter1or at1on of the 1nter faces of the thermal sh1el d support p1ns and lugs (probably 1nvolv1ng frett1ng corros1on) and further deformat1on of the pos1t1on1ng p1ns as well.
It 1s also poss1ble that loosened pos1t1on1ng p1ns were further damaged by v1brat1ng 1n the sh1eld as a result of turbulent coolant flow around that part of the1r length protrud1ng 1nto the core support bar rel-thermal sh1el d annulus.
Th1s wear mechan1sm was d1scussed by CE extens1vely 1n the1r report to the NRC on the fa1lure of the three pos1t1on1ng p1ns at Ma1ne Yankees but was not d1scussed by them 1n the1r fa1lure analys1s for St.
Luc1e Un1t No. l.
(3)
Trans1ent hydraul1c 1 oad1ng caused by changes 1n RCP operat1on (start1ng or stopp1ng of one or more RCP~s) prov1ded cycl1c load1ng of the thermal sh1eld and support system throughout the seven year per1od from Aprili 1976 to Marchi 1983.
Ten or more such changes 1n RCP operat1on were made 1n 19 of the 84 months>
1nterspersed w1th 43 months 1n wh1ch all four RCP~ s operated cont1nuously.
(4)
The wear of the 1nterfaces produced a downward d1splacement of the thermal sh1eld as ev1denced by 1mpr1nts on bear1ng pads occurr1ng at progress1vely lower po1nts on the pads.
The 1mpr1nts are attr1buted to 1mpacts between thermal sh1el d pos1t1on1ng p1ns and core support barrel bear1ng pads dur1ng large excurs1ons of the thermal sh1eld under trans1ent load cond1t1ons.
43
(5)
Upper positioning pin X apparently dropped out of the thermal shield at an early date.
and probably was responsible for LPM alarms that occurred on February 2~
1978.
There apparently was no significant downward displacement of the thermal shield before this first pin dropped out as there was only a single imprint on its bearing pad.
Upper positioning pin Z dropped out at a later datei after 'ome downward displacement and after multiple impacts of the pin on the bearing pad had occurred.
On the basis of LPM alarm recordsi this may have been as early as Marchi 1980.
(6)
With deterioration of the support system in the final stages of the failure process'arge amplitude sel f-excited vibration of the thermal shield under steady-state operating conditions may have developed to contribute to the extensive cracking of the thermal shield and core suppot t barrel.
The starting and stopping of reactor coolant pumps dur ing startups and shutdowns of the r eactor plant are a
part of normal operation.
Reasons for shutdowns noted in FPL operating logs covered a wide variety of conditions.
Large excursions of the thermal shield and impacting of positioning pins on bearing pads in the FTI proposed sequence of events could not be expected to be detected by the Internal Vibration Monitoring (IVM) system ot the LPM system previously described.
The IVM system is useful only when all four reactor coolant pumps are oper ating>
and hence would not be 44
II
used when changes were made
$ n pump conf$ gurat$ on.
The LPM can be used with baseline background no$ se established for different numbers of.pumps oper at$ ng>
but not for the transit)on fran one pump conf 1gurat1on to another.
45
(I)
FTI concurs with CE that damage of the thermal shiel d was preceded by wear and deterioration of the support pins and lugsi and that this wear of the support pins and lugs was preceded by loss of preload of the positioning pins.
(2)
FTI does not agree with CE that 1 oss of pt el oad in the positioning pins was due solely to the combination of factors considered by CE (i.e.i pr essure differentials due to fluctuating hydraulic loads and due to steady state hydraul icsi thermal shiel d weighted irradiation stress rel axati on and postul ated install ati on error s).
Rather i FTI concludes that loss of preload was most likely due to overstress in the threaded region of pins and at the bear ing end resulting from combined unifor m and bending stresses.
These str esses were the result of differenti al thermal expansion of the core suppor t bart el and thermal shield in both a radial and an axial direction.
(3)
When a
real isti c friction coefficient for non-1ubr icated surfaces (0.29) is used and the bending stresses on the upper and lower pins are taken into accounts the calculated maximum stresses on the upper and lower pins at e approximately 38i000 and 33~000 psii respectively.
These stresses are significantly in excess of the 18i500 psi yield stress of the material.
They are also considerably greater than the allowable stress for ordinary threaded fasteners (22~200 psi) as specified by the section of the ASME Boiler and Pressure Vessel Code applicable to 46
construction of reactor components not intended for pressure loads.
(4)
FTI does not agree that the 1 ar ge ampl itude motions of the thermal shield were solely attributable to sel f-excited vibrations of the shield.
The additional mechanism proposed is that large hydraulic load cycles acting on the thermal shield and core support barrel were interspersed with a high number of low-amplitude vibrations that produced wear of the suppot t system during 4-pump operation of the reactor plant.
The intermittent hydraulic load cycles are associated with moor changes in primary coolant flow that occur when reactor coolant pumps are started and stopped.
Some 522 such pump start/stop events 'occurred during the seven years of operation and provided a cyclic loading that is considered to be the pr imary cause of the cracking that ultimately occurred.
(5)
Nothing in the sequence of events leading to damage of the thermal shield can be attributed to operating procedures by Florida Power and Light.
(6)
Indications are that the failure is generic in natu< e since at least two other similarly designed CE units have also experienced thermal shield support problems.
47
Respectful ly subm$ ttedr Thomas H.
Dawson~
Ph.D.
Senator Mechanical Engineer Robert B. Nfederberger>
P.E.
Senator Metallurgical'ng)neer ose D. Gfllerlafn> Jr.>
h.D.
Sen)or Fluid Dynamfcfst
~
~
REFERENCES
REFERENCES 1.
Combustion Engineering Power Systems Report:
Final Report on the St. Lucre Un)t 1 Post Cycle 5 Plant Recovery Program Chapters 1 - 10:
February~
1984.
Chapter 1 - Summary and Chronology of Events Chapter 2 - Descrkptfon of the Reactor Internals and Reactor Vessel Inter faces Chapter 3 - Pressurized Thermal Shock Chapter 4 - Nondestructive Examfnatfon Techniques Chapter 5 - Nondestructive Examination Inspection Results Chapter 6 - Core Support Barrel Repair Chapter 7 - Failure Mechanism Analysfs Program Chapter 8 - Core Support Barrel Structur al Integrity Chapter 9 - Safety Analys)s Chapter 10 - Monftor<ng and Inspect)on Programs 2.
"Theory of Plates and Shells" by Tfmoshenko and Wofnowsky - Krfeger.
McGraw-Hf1 1 i New York'959.
3.
ASME Bofler and Pressure Vessel Code>
1974
APPENDIX A Photographs
4
PHOTOGRAPH CAPTION SHEET Photograph 1 -
. Fracture of support pin and thermal shield at Lug No.
1 and break out of Lug No.
1 fran Core Support Barrel.
Photograph 2 - Loss of entire support p$ n and pos1t1ons of thermal sh)eld at Lug No. 4.
Photograph 3 - Fracture of support p$ n and thermal shield~ with loss of parts of each't Lug No. 9.
Extensive wear on top of lug
$ s apparent.
Photog aph 4 - Loss of Positioning Pfn Z from location below Lug No. 9.
Photograph 5 - Pin W-0 degrees Photograph Photograph 6 - Pfn W-90 degrees 7.- Pfn W-180 degrees Photog aph 8 - Pin W-270 degrees Photograph 9A/B - Pfn W-Top End and Bottom End
Photograph 1
Photograph 2
J
~ '
Photo gra ph 3, I
w Photogra ph 4
f
~
~ ~
-m?"
gt v.
wi~~>"'+: ~
Photograph 5
i/
K
~., I y
Photograph 6
Photograph 7
Photograph 8
p 3
.c
,i'd@(, jy r, '$ '
~
~h r
Photograph 9A
0
Photograph 9B
APPENDIX B Figures
0
r
- yZ.-;~/e"--
Su'C'/ Or/7 5yxrsg DEPPI/4'5'8AS/5:
CE'rgrci/iS
<~ DK S4/93C/-/GO-<>/)
gc5.55.I.IZZ.
rccc r5cccwc <<r~
ruu /creole rS)
.Srcrncrc c 5! /a/ rccC/+r
+EAC704 yQ~ PQE/SuRE VFSIZZ
/go 5"
I
.al COapF paaa/aaany ALI G NMENT PIN ro / / /r///cc
///
>ac/roc S///cw>
os z7 P5 ozZZE A
>/o 5
THERMAL SH1 ELD I I]
SNU88ER~8 CORE STOP 42" IO OUTLET NOZZLE ll
/i'O>>7
/Ora///g(
P/r/s I
CdE'F DF7-g) 5 I
~
5 I:I III ::IIII i
I I I a
5 I V
a
~
a II I. I li
! I5 I
Ill5 I II ll; I
/o/Ae' 5
i 5n-in X INLY NOZZLE (7yrg CORE SUPPGR I
8ARREL FUEL AS SEM8LY FLOW SK IRT pppsg
+>~/7ro/5/rr/G
~ p//5/
&&4Aczsp 7ijF/rc/AL S+/FM Dc/rc/ryrr
/g'E'7 p
C'OWE s'5/pwca ~
BAiK~/(Csb)
Zoezg POSr&/OI/re'G
/c pr/c/
ti//7"PXAcF
~ a PSB 4'.
SEE
~ppoz -
-7>E /4r=
CSB W&uacr!EQ (9 r sacra II
/OP rye
- 0 (Noir+)
r'5 o~/ew<~AroM egg~ pi 4o-ZZE 4 ogden.or j~~
ceca cacc supporcr
.4OG (P/5/Pt Gc= CSQ) 5ncccac Iia55" flEvA7rou'I-A. ~EPICS 7~<Rswz +eau (PGIs'/7/dezu'G P/4$$
r
~lf" g, r/re 5e:55cc<<gg taa ~.-c-rgq'rlg5 WFg ~4'a@
SpProgp u'rg/4 r.
~W OP/~/~dr'/Pg ldi
e~u, I
/~~/
~
re@ /A fEAC7OA' Oo<W'7 sPSfu/- gybe'GFPPA'7 (DASPS','.CE'4 4 N PPSg7-X/o -Zoogao/)
H~4W Veal
do p,"",A
- H ZrIS&46, R
2 rr ttttg
.r /
~:E,
ttttwI)tw'4 tt Vi'FVZlP r.
TYPrnz.
+< -~A)
(~
tw
~ SuPPc fp;. +> s r w
I.
t ZV4i (t..
9
~/~
EEb
-S iX LoW:pZy G /t (E-~P) wx.~
L
~~ ~~3 8
E.)leg;.. t-;p Fttt~~
cfaAA'rrpt pan te:fttt~&g
~ ~ i.t~
/ g P yP
.7,t Zrcr r S arnis ggppor, r rLICv/Poslr rob/~/lC P;rJ JBOr./E rr rC cot@'-'uPt= <'f.rt:t.'c (cstt.)
Tilt=-g.L1/L Si(tet.D(TS) t 9~/s ~
llld'cg eW' Wrrt/ d3 (7yP) 5 OPPFg (7ypy bi] 148 4p
- Po5>
toJt06 pgsl- '+~pl E)CPO5em 7QDS cm
~ TS-cf
~.So7 tottrt N
gp Ig, 8
~gti~
DC)cyvlA t tC Rt PRE5&Jp7/O~
~
~
~~ z~+~~ ~w Dews sm -.-.>7=:~
7 zc~gc..d re coRE <car'Pu BJA'EEL
&r,SrS -Z DeS S~ i9aaV-rbO-V5S)'
0
q fisz')
r(roy >
G3o9
.~ (re )
/// IP
/i' 1
~/A kl g~.
>7o If
/F T
~
~
(
O)
Y' viz'r. rr/'i/.-F/i".
~c~if/]//j-romp>
(7 Fg.
(/r/l://-r 4 )
/ -:--x i(27(k'ocr')
(7
~ /'>
+D.-"54= 4OCFr /drl~
(?/0>
(>4/i~ Cr<c erg B
(95 $
I (7o')~
a~a.eras Pans (red)
/
R(Fo) i A>(]if q(
(ie)
~
OP PAP
~
~
Cry/)
q'(aSD3
/-
P(s3i 3
~/r-, /i /
0/jf/Z.'-'l(23$
)
4eQC
>uPP0(7 8/'f'/:E
+><->-.=-WFW4. SgazS>: ZgpgurrS Od, tJPPgX/~@~FR,:PaS(rr odruC/
. Pul PAP5.
Qea'-CZ.
P'rC 5;g-r// ra g.(-3g), Iql
APPENDIX C Pos1t1on1ng P1n Impr1nts on Bear1ng Pads
4
Pos1t1on1ng P1n Impr 1nts on Bear1ng Pads The representat1on of pos1t1on1ng p1n 1mpr1nt mark1ngs on the core support barrel bear1ng pads>
F1gure 3i Append1x B>
as a ser1es of c1rcles accur ately represents the cond1t1on noted 1n v1ew1ng of photographs and v1deotapes.
F1gure 3 shows each pad 1n 1ts relat1ve, pos1t1on to other pads by a c1rcular arrangement.
The pads for the n1ne upper pos1t1on1ng p1ns are shown as the 1nner c1rcle 1n F1gure 3 rotated so that the bottom of each pad 1s outwar d.
All show mult1ple 1mpress1ons and a downward sh1ft except the pad fot P1n X at Lug No.
7 (on the West ax1s).
Th1s pad showed only a s1ngle contact c1rcle at the center of the bear1ng pads and s1gn1f1cantly>
only very
'1ght contact on the bottom quarter of the c1rcul ar 1mpress1on.
Pos1t1on1ng p1n X 1s bel1eved to have been the f1rst of two pos1t1on1ng p1ns to fall out of the thermal sh1eldi poss1bly early as 1977 (LPM alarm levels were f1rst exceeded 1n Feb'aryi 1978).
The other pos1t1on1ng p1n
. found at the bottom of the reactor vessel was P1n Z; th1s p1n had been 1nstalled located below Lug No.
9 (to the North).
The correspond1ng pad exh1b1ts a ser1es of 1mpress1onsi show1ng th1s p1n d1d not fall out of the thermal sh1eld unt11 after some 1mpact1ng and d1splacement of the sh1eld had occurred (3 1mpact 1mpress1ons are v1s1ble for th1s p1n).
Note that for some p1ns as many as 4-6 1mpact 1mpress1ons are v1s1ble (P1n J
shows 6
1mpact 1mpress1ons; altogether 10 p1ns show 4 or more 1mpact 1mpress1ons).
0
e The sideways displacement of contact areas on the upper pads show that Pins R (at 30 degrees) and Z (at 350 degrees) at the north location dropped to the left; pins at locations to the east and west>
S (at 70 degres)>
T (at 110 degrees)i W (at 230 degrees)>
and Y (at 310 degrees) dropped to the r ight; and Pins U and V at the south location dropped straight downward.
The pads for,the lower positioning pins'hown on the outer circle in Figure 3.
provide further substantiation of the thermal shield rocking motion described previously.
Pads for Pins A through Hi on the eastern half of the core support barrels do not exhibit any evidence of multiple impressionsi whereas those for Pins I through Q
on the western half of the core support barrel show a
significant pattern of multiple impressions.
. Pin N to the West (at 281 degrees) appears to have dropped straight downwards but with scee slight shift to the left and then to the right.
Pins Oi Pi and Qi in the NW quadrant to the left of Pin N show a
shift down and to the left in a series of impressions.
Pins Ii Ji K. Li and M.
in the SW quadrant to the right of Pin Nu show a shift downward to the righti again with a series of impressions except in the case of Pin I (on the South axis).
This pattern suggests that as the west side of the thermal shield moved in toward the core support barrel> it may also have tended to flatten or become somewhat oval in shape.
It is possible that deflection of the core support barrel was involved also.
F i 1