ML19254E728
| ML19254E728 | |
| Person / Time | |
|---|---|
| Site: | Crane  |
| Issue date: | 10/31/1979 |
| From: | Bland W PRESIDENT'S COMMISSION ON THE ACCIDENT AT THREE MILE |
| To: | |
| Shared Package | |
| ML19254E707 | List:
|
| References | |
| NUDOCS 7911020252 | |
| Download: ML19254E728 (22) | |
Text
.i s
~.
=-
~
k E
TECHNICAL STAFF ANALYSIS REPORT OS
-. m i..c u. =.=.
D Cn'. 8 D = '. ~' c '.~. =
uu e
TO PRESIDENT ' S CO?O:ISSIO: ON THE n u C.r Dr. - i
. u. o. r_ _r e.r.Lr T _e r.n. -
ADVANCE COPY NOT FOR PUBLIC RELEASE v s s,a n c o.
1 7. g 1 0. /,0 nr Jww.
u.w - w un 1' y en--
v.-_ m :
Or
- r
- e
" -.- n.
r%.s -, y s
s.
1232 348 101it opo25R
s E
5 i
s TECHNICAL STAFF ANALYSIS REPORT ON CONDENSATE POLISHER BY William M. Bland Technical Assessment Task Force OCTOBER, 1979 WASHINGTON, D.C.
- }
1232'349
g i
C s
This document is solely the work of the Cc:rrission staf f and does not nccessarily represent the views of the President's Commission or any member of the Commission.
This pre-publication copy is a final document and will be subject only to minor editorial changes in its published form.
1232 350
1 SUMMAdY On Markh 28, 1979, the Three Mile Island, Unit 2, Nuclear Power Plant experienced.ihe most swere accident in the U. S. commercial nuclear power plar2 operatNig hi:, tory.
The accident, which cccured at about four o' clock
{
in the morning wis started by a loss of normal feedwater supply to the steam generatorr which lee rapidly erough a normal shutdown sequence of events for reactor shutdcwr..
Our.ng this normal sequence the pilot-cperated-relief valve cpened and stuck (failed) in the open position in which it remained for a considerable length of time before being isolated by action of the operators. This failure felicwed by early operator action that throttled the HPt (high pressure injection) pumps initiated an abnormal secuence of events that led to this most severe accident.
The significant findings and a conclusien are listed belcw:
Findincs 1.
The Ccndensate Pelisher effluent valves closed at the beginning of the accident on March 28, 1979.
2.
These valves had cicsed unexpectedly twice before under similar surveillance procedures; and again later under a related condition.
3.
Tests at TMI-2, to ate, have not confirmed a reason for cicsure.
A possible reason ha. ceen postulated by an analysis which shows a way water accumulated in the control air line could trigger valve closure.
4.
The Ccndensate Pclisher as used at TMI-2 had essentially zero cc cational margin.
5.
The Pclisher bypass valve is not designed fcr automatic emergency opening nor even for manual opening in an emergency, though its availability is noted in the FSAR and appropriately designec could provide substantial operational margin.
6.
The Ccndensate Polisher bypass 2 failed to open by remote c:ntrol during the accident and at least once before and corrective action had not been acc0mplished.
7.
Resin remcval from polisher has been a chronic problem; has existed since early system tests, corrective action has been ineffective; and removal problems appear to contribute to unexpected closing of effluent valves.
Cesign problems were not werked with polisher designer.
8.
Available drawings at TMI are significantly different frem ccndensate polisher; no interface drawings or integrated schematics are available for use of precedure writers, maintenance people or system engineers.
9.
Condensate Polisher is not classified as safety related; did not receive '
detailed design analyses, did not receive Quality Control coverage during operation; did not receive management and management review group attentien for prcb! ems with hardware or for procedure control.
1232 351 I
2 10.
Thereis a history of operational problems and a large amount of maintehance work on the polisher.
C Conclusion $
{
The condensate polist ar, thmgh vital to the operation of the plant, did not receive appropriate attention in design and attention from assurance function, engineering, management, and management review groups, proper at ention could provide a significant increase in plant reliability.
e 1232 352
3 r-e_
5 INTRODUCTION I
On March 28,1979, Unit 2 of the Nuclear Power at Plant TF. ee Mile Island (TMl; near Middletown, Pa., experienced the most severe accident in the history of ccmmercial nuclear power plant operations in the United States.
This accident has been and continues to be the subject of much investigation in order to determine the primary cause 1 ehind the initiation and furtherance of the accident.
- number of analyses, involving studies, inspections and tests have been conducted to understand what caused this accident and why it happened.
One of these analyses, that of the Condensate Polishing System, is described in this report.
This report, which is really more investigative than analytic in nature, has been prepared to present the facts that can be gathered that relate to one part of the accicent and to examine these facts to see what lessons can be learned that can be used to possibly prevent a recurrence of the March 28, 1979, accident.
The function of the Condensate Polishing System, which is described in reference i, witn other parts of the TMI-2 nuclear pcwer plant, is to maintain water quality by removing impurities from the condensate; the cbjective being the prevention of problems in the Power Conversion System caused by scale formation, corresion carryever, and caustic embrittfement.
In addition, the system design provides for removing impurities in the condensate caused by inleakage in the steam generator of reacter ccolant liquid, and intermittent inleakage in the concenser of cooling water frcm the Circulating Water System. The design also provides a bypass of the entire condensate pelishing system. The design of the pclisher units and regenera-tion equipment is based on a 23 day pericd. The system is so designed that 7 of the polisher units can handle the full condensate flow, while the remaining one is being replenished.
The equipmer't in the Condensate Pclishing System is in accordance with ASME Sciler and Pressure Vessel Code, Sections Vlli and IX, however it is not classified as " safety grade" and thus does not receive the same care and attention, as from Quality Control during its operational life, nor did it receive such care when it was designed, fabricated, transported, stored, installed, and checked out.
Pigure I is a simplified line sketch of the Polishi6g System fcr TMi-2.
1232 353
4 The cokdensate discharged from the condensate pumps enters a header in the Condensate Polishing System and is routed through seven (7) of the eight (8) pdishe.s; the water ficws down through the polisher resin bed i
and out to sirvice.
Each polisher contains a mixed resin bed consisting of anion.catior icn exchange resins. There are nine (9) resin charges in the polishing ',ystem: one charge in each of the eight polishers and the ninth charge in either the receiving tank or the mixing and storage tank.
Under normal operating conditions tnere is a flow,f 2500 gpm of condensate through each of the seven in-service polishers; the eighth polisher is on standby.
Each pclisher remains in-service until its resin bed has been exhausted, i.e., until the polishers is no Icnger providing effluent of the desired quality.
Radicactivity will appear in the Condensate Polishing System if a leak.
cccurs in the steam generators alicwing primary coolant to leak into the secondary system.
ANALYSIS Related Activitv Just Prior tc the Accident As reported by two sources, reference 2 and 3, and as noted in various interviews, deocsitions, and hearings, the TMI-2 plant ocerating staff had been working cn the polisher for scme cime to clear resin frcm polisher tank Number 7 when the accident was initiated.
This work was reportedly being acccmplished in accordance with Operating Procedure 2106-2.2, 03/21/79, and had been in progress for about eleven hours pricr to the accident.
The work involved the use of compressed air and water, as per the procedure to force the spent resin from the tank.
At the time the turbine trip was announced an operator reported that the concensate polisher panel incicators shcwed condensate pclisher isolation, which indicated no ficw (reference 2) thru the polisher.
This condition could be caused by closed polisher effluent valves.
This state of no ficw at this time was confirmed thecugh a review of records by an NRC investigat:r (reference 2), and by this review, appendix J.
Finding The condenstate pelisher effluent valves closed at the beginning of the accide,t.
1232 354
5 Recerted Po$t Accident insoection Results A num@r of items that can be classified as ncnconformances, or thir,s f
out of the ordinary, have been reported as found on the Polisher after the event. The reported items are as follows:
Water in the instrument and service air receiver tanks (references 2 and 3)
Check valve in service air line stuck cpen (reference 2)
Failure of remote opening of condensate polisher bypass valve (reference 2)
Effluent valves :or. trol scienoids improperly wired (references 2)
Withcut discussion these items convert to findings, as felicws:
Findings Water was fcund in the air systems Polisher bypass valve failed to open Pre-Accident Related Excerience Experience at TMI-2 relative to the capacity of the instrument air system and problems with water in the air system is recounted in reference 2.
Frem this discussion, interconnection of the air systems, the instrument air systems and the service air system, to provide adequate capacity may not have been a wise decision because of subsequent concerns and modifications.
For instance, Mr. Zewe's note of 5/15/78, appendix A, reccmmended modification which contains concern as evident by the folicwing quotation; "It's time tc really do scmething on this problem" (subject of note is Water in Service Air and instrument Air) before a very sericus accident occurs.
If the polishers take themselves offline at any high level of power resultant damage could be very significant."
Zewe's note called for making the pelisher cycass valve autcmatic.
That this recuest was not implemented is confirmed by inspection and the valve problems during the TMI-2 accident.
Apparently, it met the same fate as a ehrlier request of November 3,1977, to improve the operating performance of the bypass valve, which was disposed of by the Met Ed test superintendent en November 17, 1977, as fol!cws, "no further action required by this PR (Preblem Recort).
If, when the plant is restcred the problem is bet er defined, we will rescive the problem."
1232 355
6
~
An eveFr earlier (November 14, 1977) memorandum, " Water in the instrument.
Air Lines at3he Condensate Polisher Control Panel and Regeneration Skid" discussed a 'E:ss of feedwater condition in Unit #2 on October 19,1977, (also
(
contained in appendix A) by Brummer and Ross recommended extensive modification to prevent recurrence of that event. The description of the October 19, 1977, event shows close similarity to the ever.ts believed associated with the polisher on March 28, 1979. To illustrate the similarity a descriptive paragraph from the November 1977 memorandum is reproduced.
"Ouring er shortly after the attempted transfer of resin fecm the mix bed polisher =3 to the receiving tank on the regeneration skid, the Auxiliary Operator ncted water running cut the air operated recorders en the condensate polisher control panel, No. 304.
Shortly thereafter the discharge valves on the condensate pclishers closed resulting in a total LOSS OF FEEDWATER concition.
Ucen detection, the Centrol Recm Operator immediately tried to cpen CO-V12, condensate polisner bypass valve; hcwever, he was unable to cpen this valve from the control recm. The auxiliary cperatcr was then notified to manually open Co-V12.
After abcut 60 minutes, and assisted by another Auxiliary Operator, CO-V12 was opened.
If this would have haopened while at power the unit could have been placed in a severe transient condition resulting in an Emergency Feedwater Actuation, Main Steam Relief to Atomesphere, Turbine Trip and Reacter runback with possible trip."
Still earlier prcblems in removing the spent resins in the polisher, as expressed in Field Questionnare 1577, 2-9-7-',
resulted in modification to the polishers; modifications apparently completec about 4/20/77, appendix A.
On May 12,1978, as indicated frcm a review of TMI-2 chemistry-polisher log, that is maintained by technicians to record normal activity with the equipment, and reported in appendix J, the effluent valves of the ccndensate polisher c!csed unexpectedly.
At the 'ime of c!csure the Icg reports the cperators were changing out pclishers; polishers 5, 7, and 8 were invcived with tranfer of resin frcm 7 being dcne at time of the event.
Water was found in the air lines. This event was not reported in the MOR (menthly cperating report) for May; plant was dcwn for other work at the time.
By interpretation of requirements for the MOR, significant cperaticnal events are to be included.
In this case the plant was already down so this event was not censidered to be significant.
This briefly treated history of problems with the Unit 42 condensate pelisher indicates problems that appear to have been chronic relative to removing spant resin, water in the air supply and reluctance of the pelisher bypass valve to operate.
Recorts had been made, and at least some corrective -
action had been performed; however, from what is believed kncwn about the polisher's role in the March 28 accident each of these three main concerns were present.
1232 356 1
7 f_
E incident of TMI-2 tripping because of a.oss of feedwater is i
Anothe reported in reference 2 and appendix A-1; ".
.on.'ovember 3,1978, the unit tripped from 90 percent power on high pressura due to less of feedwater.
This occurred when an instrument techician mistakenly opened the car. trol power supply breaker to the condensate polisher control panel causing all polisher outlet valves to cicse."
Findings The con densate polisher effluent valves had cicsed unexpectedly in 1977 and in 1978 when resin was being transferred frcm a polisher tank with water and air as was being done on March 28, 1979.
Design deficiency in bypass salve not corrected.
Preblems in removing spent resin remained unresolved.
Preblems were not thorcughly investigated and analyzed.
Continued problems experienced with water in the instrument and service airlines.
Recommended corrective actions were not thoroughly censidered.
Corrective actions were not verified as being able to accompiish intended action before use.
The condensate polisher effluent (outlet) valves closed in 1978, while unit was at 90 percent power wnen electrical power to condensate pelisher control panel was interrupted.
4 en-
~
1232 357
8 Desion
^ -
=~
t The Ur$.t 2 condensate pelisher was originally designated for use at the E
Oyster CreeE facility, appendix P, where it was to be required to process I
100 pe cent of the condensate flow.
On relocation the changes made in the desipa were the minimum required to accommodate the geological differences of the TMI site.
Possibly the most important deficiency is that the Unit 2 polisher did not have designed into it an automatic 'ast-acting bypass valve, that could be utilized in the event of flcw problems thru the polisher.
Aisc, from the reports of investigation into the bypass valve, appendix P and appendix J, and the photogra;.ns in appendix J-3 it can be seen that the valve was not designed for emergency manual actuation.
Its location and crientation precluce rapid access and manual actuation.
However, the content of the TM!-2 FS AR, page 10.4.8, reference 1, implies that a working bypass exists, "the design also provides a bypass of the entire condensate polisher system." As noted in appendices M and N, there have been scme differences of opinion in the design cbjectives in the TMI-I and TMI-2 Condensate Polisher by-pass capabilities. Hcwever, use of the bypass valve, either to carry scme of the flow in parallel with the conder. sate pclisher, or with a working automatic valve could provide operational margin that would make the plant less sensitive to trips caused by perturbatiens in the condensate polisher outputs.
This is substantiated by the study conducted on equipment conservation, appendix Q, which included the condensate polisher as one of three types of equipment analyzed.
It was reported that the condensate polishing system as cesigned and used at TMI-2 had essentially :ere performance margin when the plant was operating at rated power.
The analysis further indicated that acpecpriate use of a properly designed condensate bypass valve could increase the performance margin significaritly.
The appropriateness of this analysis is strengthened by a limited survey made. cf other nuclear pcwer plants use of Condensate Polisher by pass valves. The results, shown in acpendix 0, incicate there are a number of cifferent designs including, for example, the recent use by the Crystal River plant of an automatic by-pass valve that savec a plant trip.
It also utilized a deep bed demineralizer instead of the preccat type, as in Unit 1.
The type used in Unit 2 has had a number of handling problems as noted in the previous section in the discussion of the October 19, 1977, and May 12, 1978, less of feedwater event, as noted and in the problems in removing spent resins, 2/19/77, and in the problems apparently associated with removing resins from Polisher 1 en 3/22-3/23/79 as discussed in the next secticn.and again on 3/27-3/28/79, reference 2.
From a review of the pclisher log,iit has been noted that problems in tranferring resin frcm polishers ccourred frequently, about 1 cut of every 12 transfers, accendix J.
This is an indicaticn of a chronic prcblem.
1232 358
9 It is ncded that the condensate polisher is not classified as safety related equip' ment, appendix P, and appendix F.
This meant that it did not
{
receive con ~ erations in its design of requirements for analyses and tests k
as in QA Pr ram.
As noted in appendix Q, "although the condensate polisher is vital to the turbine operation and plant output, no analysis was performed nor requirements formulated to insure that sufficient active polisher legs would be in operation to adequately feed the condensate booter pumps at all operational flow conditions."
It is also noted in appendix Q that no design studies or analyses were performed to identify the varicus failure modes of the polisher and the effects of the modes en plant operation.
The supplier of the condensate pelisher, L. A. Water Treatment, has produced a number of pclisher systems, as shcwn in appendix L, including nine for nuclear clants other than TMI-2.
Reviews of the design process followed has not disc!csed any other significant problem that can be related to the accident until 1971.
Earlier problems appeared to have been coordinated with the equipment supplier and resulting changes show up in the supplier engineering drawings. Subsequent to about 1971, the coordination of problems for analysis and corrective acticn does not appear to have included the services of the equipment supplier, appendix K and appendix Q.
Six field changes applied since 1971 are listed in appendix K; two significant enes are changes related to problems of remcVing resin from polishers.
During the pcst accident review at TMl with condensate polisher operators and system engineers from GPUS, it was noted that the available drawings were significantly different from the actual condensate polisher equipment, aopendix J.
For example, valves were shown in wrong locations and had improper identification; components were shcwn in incorrect locations; and air lines locations and interconnects were not properly shcwn.
It was also noted that interface drawings and integrated system schematics were not available.
Drawings have been marked up to show examples of changes needed to make drawing the same as the polisher equipment appendix J-2.
Without accurate drawings, interface drawings, and integrated schematics cp;rators, maintenance personnel, and precedures preparers have difficult, if not impcssible, tasks to do.
Findings The condensate pclisher was not classified as safety-related equipment, thus did nct have design analysis and tests required in QA Program a p plied.
The cchdensate polisher as used at TMI-2 had essentially zero operational margin.
Use of an apcropriately designed bypass valve in parallel with pclisher could provide substantial operaticnal margin.
The condensate polisher bypass valve was not designed to autcmatically actuate upon demand, ncr was it designed for emergency manual actua-tion; though apparently censidered sc in the TMI-2 FSAR.
1232 559
10 There were problems associated with resin emoval; problems developed about one time in each 12 transfers.
y c
t Designphanges were not coordinated through the designer after 1971.
g Available system drawings were significantly di:ferent from equipment; no interface drawings or integrated system schematics were available.
Precedures From reference 2 it is believed that Condensate Polishing System Operating Procedure 2106-2.2, appcndix 8, 8/21/79 revision 9, was being used by the plant operating staff just prior to the accident on March 28, 1979.
That this revision of tne procedure was being used is supported by comments in appendix C, hcwever, latter communications, appendix D, indicate some uncertainty as to whether revision 9 or revision 8 of the procedures was being used.
The procedure is very long and provides for the positioning of many valves in precise order and alignment; althcugh the procedure is arranged to that the steps can be checked off when completed.
Statements In r Wem 2 indicate that progress has been made mest of the way through the procedure at the time the accident was initiated, this is confirmed by appendix D, which indicates that progress through the procedure stopped between steps 4.1.4E5 and 4.1.5ES, fcr either revisien of the procedures.
This step involved the use of water to trans'ar the resin.
Revision 9 of this procedure had been extensively revised en March 20, 1979.
The earlier precedure, revision 8, is included as Appendix E.
The procedure change request is included in appendix r.
i he reasons given for the change were:
"to incorporate new acid precedure, new short regen procedure, new trouble shcoting section and additicnal cperatcr guidance and instructicn."
The changes appear to affect the maicrity of the pages in the precedure, scme cnly slightly-others mere so.
The change action was initiated en 1/25/79 and the revised peccedure was approved en 3/21/79, when it was signed by the engineering supervisor and the Unit 2 superintendent, it did not require the approval cf PORC because it was not classified as a nuclear safety related precedure, or require Quality Assurance, and did 'not show verification before use.
It can not be qstablished which revision of the procedure was in use en Maren 23, _.:
thus indicating a lack of dccument centrol and work centrol.
7(P l17^Ls/
JUd
s i
.a w
4 r
a KA '
4 Mdb44[W*tM,,'eQ,,{*.f hi%[O:
y afMU M'.
la f.:@.&ep f.m f
fk's;sQ,* A?v$
- c.
2&
g.
9,T; M
b ?,:y3 i N
h y n n5 w;s'g wQ s
Wu;,ne$ y?O ;gk
,.f*
i f
'g N
e:ka,,.
W s
$ {$ i.m>n$h dw;Wu 9a w$lt$ ^;dly ne k
- 4j9 m */4;;Q.2; W
'l d
h 5'
kI M+
sp k-gp:J H'eleiV
$O 0
k g$9 n b
DE hr W :en i
hY$:+=4,dw%.
r s
W
+W
- %b. h f y
S w*ae 4
f x
& pap w 4rh N,.
h,f
'X
- 3 h?
yP u
h*$k 11 v; ye 9 4 i
L.
fR khpMAw Mw..ae hg%n ts q.
y 2a r,,
fy v
- ymg 8
h MnWq 3 -[h
$9g w:. y
- @u 3
N wgww4n n.
' > g..n : s r.d?$0 u; E
flg Af A
g f
y ith
+na&
S m
'geygegga
[g fg a
e Qie ft
$@b,s fi" D. '.
mm$gn-y q lI g u:q e g s
% %4 7
?
1
(
y 4
$@$aM a.v
~
d9 g
"%7.g.+gg,gm:
~;'
5 2
4$.
4
~y 9{$sp @mp g s[e@n a
g4 N.p w
i.
9 e
1' y
gy f';9~g;@g wgnm$
$~ y y p $$ ga g wu g$
y.
ge
+
gy a
W
.-:L h.
q g
$a
?f,3
.. ' q }p 9
& p, gs ygb$iw dih /l[k w p
p '. ;,.
+y 6 E$apyn
.(@h,5*[m,;xgh,<g@
t m
~
>gq%
kfgq p s
m.
tn i
,LM
(
- y g
pey&M W t,b. ;;., yVgf'#
g:
1 S
r
,' yu )
jf['-
e f
},'
.g c
9 p.
~ e,
A>*
/Q h
q)q,;+,t>>
s, <Ats k gg
's
,.....- T -
TEST TARGET (MT-3) 1.0 if m El a L H g4 "2.2 E 114 l,l h
bb 1.8 l.25 1.4 1.6 4
6" 4%
$+dh
'4
- $Itfi 1<v Azzzzz
t 4
$++$>,
$<#h
'+
-eE Ev <e 1,.
TEST TARGET (MT-3) 1.0 l#IBA E4 y l@ lliOE I.l i.'"lLM l.8 1.25 1.4 1.6 4%
++4 "N$'h
ht;Ah?
O/////
<v
6 TEST TARGET (MT-3) 1.0 lgmLu
','y EE I.I j, E_M l.8 1.25 i.4jg
=
s..
=
4%
$+db 4
- 5Iff*if 3
s, y
11 Excerpfs taken frcm the condensate polisher log for Unit #2, appendix G,
{
indicate tha(polisher 7 was worked on in the period 3/22-3/23 but some problem wry parently encountered and it was put back in service before I'
being comr:Gied.
The problem that was encountered is not known.
Another indicatior that there was a prcblem on 3/22-3/23 is that pclisher was again started ihrough the procedure on March 27 (the normal time between werk cn a given polisher is about 28 days, reference 1.
Findings Control of surveillance procedure was lax.
A recently revised surveillance procedure may have been used in the TMl-2 condensate polisher en March 28.
It is not clear that the revised precedure was verified befera being put into use.
There were difficulties encountered with Polisher No 7 before March 23 and en March 28.
Records Search and Ouality Contrei All documents used to record equipment pecblems at TMl-2 have been searched. The files of Werk Requests, Equipment History Cards, Nonconfor-mance Reports at TMI were searched, the NRC LER files were searched, and sampling was conducted on the files of Problem Reports, Filed Question-naires, Discrepancy Report, Unsatisfactory and inspection Report; to establish the recorded history of the TMI-2 condensate polisher for TMI-2 as summarized and reported in appendix H.
It is noted that the condensate polisher was one of the equipments at TMi-2 with the poorest maintenance history (apoendix H).
(The status reccc of a post accident insoecticn indicates a lack of preventative mainte-nance.) At the time of the March 25, 1979, evert there were thirteen open work request; against the condensate pelisher.
Most of them concerned leaks and malfunctioning instrumentation.
In addition, thirteen work requests had been worked during the three months prior to the event.
These problems were also associated with leaks and malfunctiening instru-mentation.
It is obvious frcm Tis maintenance history that the problems with this piece of equipment were excessive and that there are possiole design preolems. It was also noted that these work orders were not covered by Quality Centrol, that inspections were not performed as the work was accomplishec' or after it was completed to verify proper accomplishment cf the work and to verify that unauthorized work or disturbances were not done.
Additional details are contained in appendix R.
The enly history of problems was in the Werk Requests file.
This is discussed in general terms in the above parag aph.
The specific wcrk requests are listed in summary form in an encicsure to appendix H. The condensate polisher is not listed as safety grade or safety related equipment; l a, 7 J 01
12 Z
this may explain why reports related to its condition or performance are not
}
in the other iscrepancy files, for example, no LER's on the condensate g
polisher wers found.
No trend information has been located. That the
{
condensate polisher was not classified as safety related meant that this equipment was riot covered by inspection (Quality Control), that changes to its procedures were not reviewed by independent reviewers, management review groups, and that it did not receive attention frem management, despite the fact that its function was vital to the continued reliable cperation of the plant.
Findings The concensate polisher was not covered by Quality Control curing cperation.
Nc systematic evaluaticn of its " health" was acccmplished There is a histcry of leaks and malfunctioning instrumentation related to the condensate polisher The number of work requests to fix the condensate polisher is large when compared to history of other equipment Thirteen Work Requests were cpen at the time of TMI-2 accident, ncne related to Pclisher number 7.
No problems on the condensate polisher, which was not safely related, were repcrted on LER's; none were required.
Management and management review groups were not aware of the "healtn" of the ccndensate pclisher, despite it being vital to plant operation.
Relaticn Tc Arcther Ecuicment problem Another pcssible clue to what caused the polisher effluent valves tc cicse during the March 28, 1979, accident is noted in reference 2 where it was noted that the emergency feedwater centrcl valves appeared to not respond normally, as thcugh they had Icst instrument air supply, during the early part 1233 002
13 of the acciddiTt.
Both the polisher effluent valves and the emergency
{
feedwater vafves are reported to depend upon the instrument air supply operation. This suggests a possible '.sss of instrument air supply to both j
actuating dJices, at approximately lhe same time; perhaps a single point s
failure exists that has been un ecognized.
Findings The instrument air supply may have been Ics. to more than one activity at the beginning of the accident.
Pcs: Accident Investigation of Gondensate Polisher As summariced in appendix I, tests have been perfccmed on the Unit 2 condensate pclisher in attempt to discover the reason the effluent valves went to the c!csed pcsition en March 28, 1979. The tests have been done with only one pelisher and with partial simulation of the conditions that existed en March 28 and with water in the instrument air line.
The effluent valves fluttered but cid not close.
Subsecuent to these tests, while replacing one pclisher with another on July 5 the effluent valve of the one polisher in use was observed to flutter and then c!cse.
After removing water from the air supply line the polisher was placed back in service and performed satisfactorily; this event is in the encicsure to appendix 1.
A systems analysis has been performed and recerted in appendix P inat pcstulates a way in which water accumulating in control air line, frem the water sluicing used to remove resin, could trigger a series of events which pcssibly could result in reproducing the closure of effluent valves en Marcn 28.
Key in the analysis is c!cse reproduction of the ficw conditions, valve pcsiticns, and water in the control air line as they were on March 28.
Fincings Tests, to date, at TMI Unit 2, have not confirmed a reason for closure of the effluent valves en March 28.
Effluent valve en the ene pclisher in operation en July 5,1979 went to closed pcsition when polishers were being exchanged; water was fcund in the air line.
A systems analysis pcstulates one way in which water in the air lines cculd pt:ssibly cause effluent valves to c!cse as en March 28.
7 1233 003
14
~-
DETAll FINDINGS f
?
k The individual findings in.arder of their appearance in the text are listed below for c nvenience.
Taey have been consolidated into significant findings in the following section of this report.
The Condensate Polisher effluent valves c!csed at the beginning of the accident on March 28, 1979 Water was found in the air systems Pelisher bypass valve failed to cpen The c0ndensate peiisher effluent valves had c!csed unexpectedly in 1977 and 1978 tests under similar surveillance procedure condi-tiens Cesign deficiency in bypass valve not corrected Pr blems in removing scent resin remain unresolved Problems were not thoroughly investigated and analysed Centinued problems experienced with water in the instrument and service airlines Rec:mmenced corrective action were not thoroughly censidered Corrective actions were not verified bef:re use The condensate polisher was not c!assified as safety related equipment; thus did not recuire design analyses and tests as in QA program The condensate polisher as used at TMI-2 has essentially :ero coerational margin Use of an apor:priately designed bypass valve could provide substantial cperational margin The condensate polisher bypass valve was not designed to autcma-tically actuate; nor was it cesigned for emergency manual actuation; tricugh apparently considered so in the TMI-2 FSAR.
There were problems with resin removal from polisher tank, problems deveicped about once each 12 transfers Cesign changes were not always c:crdinated thrcugn the designer after 1971 1233 J04
15 DETAll FINDINGS (CONTINUED) -
t A%311able system drawings were significantly different from equpime-t, {
no interface drawings n_or intergrated system schematics were available.
Control of survei!!ance procedure used on pclisher was lax A recently revised surveillance precedure, not verified before being put into use, ma', have been in use on Marcn 28, 1979 There were difficulties en ountered with Pelisher number 7 during the surveillance procedures hefore March 28, 1979, and again on March 28 The concensate pclisher was not ccvered by Quality Control curing operation No systematic evaluation of its " health" was acccmplished There is a history of leaks and malfunctioning instrumentation related to the condensate polisher The number of work requests to fix.
.cndensate polisher is large whan ccmpared to history of cther equipment Thirteen work requests were open at the time of TMI-2 accident, none related to polisher 7 No pecblems with Condensate Polisher, whicn was not classified as safely related, were repcrted on LER's; ncne were recuired to be recorted.
Management and Management Review Gecues were not aware of the
" health" of the condensate pclisher, despite it being vital to plant operatien.
The instrumentation air may have been Icst tc mcre than ene activity at the beginning of the accident Tests to date, at TMI-Unit 2, have not confirmed a reascn for c!csure of the effluent valves on March 28, 1979.
Ekluent valve on the one pclisher in operaticn en July 5,1979, I
went to cicsec position when polishers were being exchanged; water was found in the air line.
A system analysis postulates one way in which water in the air line cculd pcssibly cause effluent valves to c!cse as en Marcn 28, 1979.
1233 005
16 SiGNIFICANT FINDINGS AND CONCLUSION 3
Studiesj analyses, inspections and tests of the Three Mile island, Unit 2, Consensate Polisher has a resulted in the felicwing significant findings
(
and conclusion.
Findincs 1.
The Ccndensate Pclisher effluent valves closed at the beginning of the accident on March 28, 1979.
2.
These valves had closed unexpectedly twice before under similar surveil-lance procedures; and again later under a related condition.
3.
Tests at TMI-2, to date, have not confirmed a reason for c!csure.
A possible reascn has been pcstulated by an analysis which shows a way water accumulated in the control air line could trigger valve closure.
4.
The Ccndensate Polisher as used at TMI-2 had essentially zero cpera-tional margin.
5.
The Polisher bypass valve is not designed for automatic emergency cpening nor even for manual cpening in an emergency, though its availacility is noted in the FSAR and apprcpriately designed could provide substantial operational margin.
6.
The Condensate Polisher valve failed to ccen by remote centrol during the accident and at least once before and ccerective action had not been accomplished.
7.
Resin remcval from polisner has been a chronic problem; has existed since early system tests, corrective acticn h s been ineffective; and remcval problems apcear to contribute to unexpected cicsing of effluent valves.
Design problems were not worked with pelisher cesigner.
8.
Available drawings at TM1 are significantly different fecm condensate pclisher; no interface drawings or integrated schematics are available for use of precedure writers, maintenance pecple er system engineers.
9.
Ccndesate Pel: sher is nct c:assified a, safety related; did not receive detailed desi n analyses, did not recei /e Quality Centrol coverage G
during operation; did not receive management and management review group attentien for problems with hardware er for procecure control.
1233 106'
17 10.
There is a history of operational problems and a large amount of i
mainten,ance work on the pelisher.
5 Conclus. ion -
k The condensate polisher, though vital to tha operation of the plant, did not receive appropriate attention in design and attention from assurance function, engineering, ma. agement, and management review groups; preper attention could provide a.:gnificant increase in plant reliability.
l233 J07
18 k
REFERENCES k
I.
"Finai Safety Analysis Report" (MI-2 2.
NUREG-C600 investigation into the March 28, 1979 Three Mile Island Accident by Office of Invention and Enforcement, NRC August 1979 3.
NSAC-1 Analysis of Three Mile Island - Unit 2 Accident, Nuclear, Safety Analysis Center, EPRI, July 1979 1233 008
19 Z
APPENDICES 5
t 5
I A.
Package, " History" (Assession No. 8100 45) ccitection of notes and records on TMI-2 condensate Pclisher PC9120059 A-1 Letter, lletrcpolitan Edisen, Transmitting Monthly Quarterly Report for Month of November 1978 PC9120060 B.
Condensate ~ lishing System, Operating Precedure 2106-2.2 Rev. 9, 3/2 9 PC9120061 C.
Letter, Shaw, Pittman, Pctts and Tcwbridge letter of August 7,1979.
PC9120062 D.
Letter, Shaw, Pittman, Potts and Tcwbridge letter of August 23, 1979.
PC9120063 E.
Ccndensate Polishing System Operating Procedure 2101-2.2 Rev. 8, 4/15/79 PC9120064 P.
Precedure change request, No. 2-79-020, crocedure 2106-2.2 Rev. 8, 4/15/78 PC9120065 G.
Ccndensate Polisher Icg, March 21, 1979, thrcugh early March 29, 1979. PC9120066 H.
Memerandum for Reccrd, " Discrepant Hardware investigation,"
Arthur Carr, August 10, 1979.
PC912N49 1.
Let er, status Recort Inspection of the Ccncensate Polish System, TMl Nec! ear Station No. 2, Mat
- hew A. Opeka, August 27, 1979.
PC9120067 J.
Final Report inspecticn of the Condensate Polishing System, Three Mile Island Nuclear Station, Unit 2, Matthew A. Opeka, September 12, 1979.
Dash one Text, Dash 2 drawings, Dash 3 photographs PC9120068 K.
Memorandum, polisher drawing change, D. Van Larberg, August 29, 1979. PC9120069 L.
" Partial list of L. A. Water Treatment, Electric Utility and Atcmic Energy Installaticn with L. A. Water Ecuipment in use or Precured."
L. A. Water Treatment Div. 05/79. PC9120070 1233 009
20 M.
Letter,pystem Deveicpment Coi caration, October 5,1979 PC 9290011 N.
Memoracidum for Record, "Conde'isate Colisher By-Pass, Information, s{
W. M. Sand, September 28,1979 PC 9290012 O.
Letter, S3 6em Development Corporation, Ocotber 9,1979 PC 9290013 P.
Analysis Report to President's Ccmmission on the Accident at Three Mile Island on Equipment Conservation, Appendix 1, System Deveicpment Corporation, September I,1979 PC 9290014 Q.
Analysis Report to President's Ccmmissien on the Accident at Three Mile Island on "Equicment Ccnservation," System Development Corporation, September 1,1979 PC9290015 R.
Analysis Recort to President en the Accident at Three Mile Island, investigation and Pcssible Explanation of the Condensate Polisher System Operaticn en March 23, 1979, System Develocment and Cooperation, August 31,1979 PC 9290016 1233 010
}b
$2
. --. 1 2
e
- 2
- 3..
..A>
. w=...
- b 543 2
7 g
r.....
l
,..I
,--(
u
>W=
Qt q3' EO q)
T e%
w" W
n d
y 6e at
,4,
- f. w J
J" g t) 4
.- g gC 4
4
53
- =t '.
3
=
,,S ~;.. i : ? ca 3 -.-
z -
...-o *. 2-
~
5 e
s
\\
u.
c.
.g-m W
u
[h i
3
)
T
.=
e CC C
l M l -r*
)
. C__
o i
'~
f 02 -
1 u
~
y ag-n k
a.
d
. (
/g, 2
e.
. i a :3 <
a
,- 2
\\
n d -
V j
e,
~
5g e
7 J
d 4
J n
Pa v n
-*ae w
u w
w a
Q p
m o
i T
s S
s
=
T Q
2
=
Q y
t s
t
?
e p.
~
A S
.a I
y.._.
-..s f
sa a
f, A
O y
0 l
S'
~
- e a
I" c1 1
M L
s
,4 Y
- a l
I
-..,e OV M.
)
T l
ygy-y _
1 a M q.
3
-r =
l t I e
- i' Y
['-
j.
._ s.
d.
M,g n
e.
22 s-
=.:.
- e. -
._ a. g._.
wg wo a-u 4.d y'V A
3 2
4"
--- 7*.
1 - x e., C 4 F
xes
- w. --
l u
g g _,
'y f.._
j..
W
-N ~
a Y,o ' ' ' ' ' ' ~
t o
j
.)
~
~
t c, _
..-.- g O-a h-
,J 6
'~~<~~"I
0
., o. v O
e r
t c
,A
[
~g.- -.
~>; --
?
t a
a se.
~~)
l O
d l
o T ?
-W tb b y
7
~
-o
....l...-Q-_-.
o o
4ny X-
\\
-x<-
kn-f 6
o,,._
r)
I O
O
\\
~~
v-a
,~"
Q g
e
=
h
' ~~
e a
l ;y.
q-C C I
.}.
U
. = -
P'--
I O
.4
'4
%) 1
~
w q C es A
'A i 5
}233 bl 35 8 m)
.