ML20080C767

Corrected Version of Memorandum on HPI Actuation Mystery Man Issue

ML20080C767
Person / Time
Site:	Three Mile Island
Issue date:	08/16/1983
From:	KAYE, SCHOLER, FIERMAN, HAYS & HANDLER
To:
Shared Package
ML20080C763	List: ML20080C759 ML20080C767
References
NUDOCS 8308290287
Download: ML20080C767 (18)

Text

- _ - - . _ _ - - . .

. i

- . i I

i i

+

MEMORANDUM ON THE 5:41 HPI ACTUATION 4 " MYSTERY MAN" ISSUE

.i 3

f i

)

1 1

)

i 4

1 August 16, 1983 i

e 8308290287 830825 PDR ADOCK 05000289 )

G PDR ,

INTRODUCTION The first reference to a so-called " Mystery Man" was in the opening statement in the GPU v. B&W trial by B&W's counsel, who coined the phrase in the context of his argument that (1) there was evidence in prior testimony that high pres-sure injection ("HPI") was actuated at about 5:41 a.m.- (2) if HPI had been maintained at that time, the core would not have been uncovered; and (3) since no one testified to having turned off HPI near that time, there must have been a " Mystery Man" who did so.1 Prior to this assertion by B&W's counsel, none of the studies and investigations by the NRC and other indepen-dent bodies had found or even speculated as to the presence of a so-called " mystery man" in the control room. After the ar-gument was raised at trial, GPU's trial counsel retained EDS Nuclear Inc., a firm of expert consultants, to perform an analysis of this issue. The EDS study conclusively deter-mined, on the basis of hydraulic analyses, that there had been no actuation of high pressure injection during the time period in question. Since the high pressure injection was proven by scientific analysis not to have been actuated at about 5:41 a.m., no " mystery man" turned it off.

1 Trial tr. 148-49, 159.

I

THE EDS ANALYSIS EDS Nuclear Inc., an engineering consulting firm with an expertise in thermohydraulic analysis, provides engi-neering, design and analysis services to the electric power industry.2 The EDS analysis of the alleged 5:41 full flow ac-tuation of high pressure injection was conducted by Dr. James Holderness, who is the manager of an EDS group specializing in thermohydraulic analysis.3 Dr. Holderness' qualifications to perform the analysis consisted of ten years' past experience in performing hydraulic analyses, including seven years with Combustion Engineering, where Dr. Holderness was responsible for performing analyses in support of licensing applications for nuclear power plants and for verifying the efficacy of the models used in those analyses.4 Dr. Holderness' analysis of the alleged 5:41 actua-tion of high pressure injection was reviewed within EDS both 4

by another EDS engineer and by the EDS Nuclear Quality Assur-ance Department.5 The objectives of the EDS analysis were to determine whether certain indications of operation of the emergency core cooling system, which were recorded on the day of the acci-2 Holderness, trial tr. 5590-91.

3 Id. at 5591-92.

4 Id. at 5590-91.

5 Id. at 5592.

2 l

l

dent, could be analyzed to identify an actuation of high pres- l sure injection and, if so, to determine whether an actuation occurred at or about 5:41 a.m.6'.

There are two potential' emergency cooling water sup-plies to the high pressure injection pumps: the make-up tank and the borated water storage tank.7 EDS constructed analyti-cal computer models which established that the response of the make-up tank level would indicate whether there was an actua-tion of high pressure injection.8 EDS confirmed the accuracy of its models, and of the conclusions derived from those mod-els, by comparing those conclusions with data on make-up tank level acquired during undisputed actuations of high pressure injection.9 The EDS analysis found that make-up tank level could be used as an indicator of full flow high pressure injec-tion,10 and further found that "the 5:41 response of the make-up tank level does not exhibit the characteristics of [an ac-tuation of high pressure injection]."ll The EDS analysis t

6 " Analysis of Reactor Coolant System Make-Up During the Three Mile Island Unit 2 Event," EDS Nuclear Inc., Decem-ber 29, 1982, p. 1. The analysis was marked at trial as GPU trial exhibit no. 2233.

7 Holderness, trial tr. 5594.

8 Id. at 5594; 5600.

9 Id. at 5600-01; 5608-09.

10 GPU trial exhibit no. 2223, p. 2.

I 11 Id. at p. 3.

3 l

l

flatly concluded that the make-up tank level behavior at 5:41 a.m. established the absence of an actuation of high pressure injection.12 Dr. Holderness, in; testifying at trial, demon-strated that given the behavior of make-up tank level at or i about 5:41, it was impossible for there to have been an HPI actuation at that time.13 OTHER INDEPENDENT STUDIES The conclusion of the EDS study that there had not been a 5: 41 actuation is consistent with the findings of other independent studies regarding the March 28, 1979 accident.14 Among the technical analyses of the Three Mile Is-land accident containing detailed sequences of events are the NRC Special Inquiry Group (SIG) Reportl5; NUREG-060016; and 12 Id. at p. 3.

13 Holderness, trial tr. 5637; 5700.

14 See discussion in the NRC staff's " Report of the Review of the Babcock and Wilcox - General Public Utilities Law-suit Trial Court Record," March 28, 1983, at pp. 6-9.

15 "Three Mile Island - A Report to the Commissioners and the Public," Nuclear Regulatory Commission, Special In-quiry Group (NRC-SIG), Volume II, Part 2, "The Accident

and Its Analysis" ("SIG Report").

16 " Investigation into the March 28, 1979 Three Mile Island Accident by Office of Inspection and Enforcement," Inves-tigative Report No. 50-320/79-10, NUREG-0600, July, 1979

("NUREG-0600").

l l

l 4

. NSAC-80-1.17 None of the sequences of events in these studies include a 5:41 HPI actuation, and none of these studies con-cluded that there was such an e, vent.

A GPU sequence of even'ts, TDR-044,18 refers to a 5:41 full flow manual HPI actuatibn. That entry, however, was based on statements by operators, which are reviewed below.

Significantly, the TDR limits the possibility of a 5:41 actua-tion to a maximum of five minutes, a conclusion consistent with the findings of GPU-sponsored studies of borated water storage tank levels that a 5:41 actuation could not have last-

! ed for more than five minutes.19 As we will show, this maxi-mum five-minute figure in itself casts doubt on the operators' ability to recall the timing of the full flow HPI manual actu-ation.

Moreover, Dr. Holderness of EDS Nuclear discussed the EDS study with the principal author of TDR-044, Dr. Van Witbeck of En'ergy, Inc. After reviewing the EDS report, Dr.

Van Witbeck concurred that no studies had been done in prepa-17 " Analysis of Three Mile Island - Unit 2 Accident," Nucle-ar Safety Analysis Center, NSAC-80-1 (NSAC-1 Revised),

March, 1980 ("NSAC-80 1").

18 GPU trial exhibit no. 2079, GPU Nuclear Technical Data Report, TDR-044, " Annotated Sequence of Events, March 28, 1979," February 6, 1981 ("TDR-044").

19 GPU trial exhibit 2079, TDR-044 at figure 60; Zewe dep.

821-22,_826-28, 830-37.

5

ration of the TDR which contradicted the EDS conclusion as to the absence of any 5:41 actuation.20 OPERATOR STATEMENTS IN INTERVIEWS REGARDING A 5:41 ACTUATION

In early interviews of the operators, their recol-lection of the approximately 15 hour1.736111e-4 days <br />0.00417 hours <br />2.480159e-5 weeks <br />5.7075e-6 months <br /> accident sequence gener-ally was not assisted by computer data or engineering analyses of the accident.21 Zewe testified at trial that soon after the accident he spoke to and was interviewed by "a great many people"; that he had had no opportunity to review data to con-firm the accuracy of his recollections; and that some of these early interviews were conducted in motel rooms, cars and trailers.22 Moreover, as Zewe noted during the trial, the op-erators' post-accident recollection often tended to " compress" the timing of various actions taken on the day of the acci-dent.23 Zewe explained in an early interview

". . . I was very far off in the times. Times were much longer and I felt that they were much shorter. . . . [0]n a few of the graphs and things that we have seen, we have noticed that the times

! there were a lot different than we have previously imag-ined."24 20 Holderness trial tr. 5659-60.

21 Zewe trial tr. 3031.

22

_Id.

23 Id. at 3032.

24 TMI Staff interview, 4/6/79, p. 1.

, 6

~,

i In order to delineate and understand the sequence of

! events in a nuclear plant transient, large amounts of data are

! recorded and then subjected to engineering analysis. As with i i

the " black box" used in airplane accidents, it is more accu-4 rate to rely on engineering analysis of recorded data than on human recollection of numerous, technically complex events.

No one person could be expected to recall precisely the time and sequence of the thousands of indications and actuations which occurred over a period of some fifteen hours on March 28, 1979.

Unfortunately, alarm printer data was not available ,

i for a period of time on March 28, 1979 that included 5:41 a.m.25 The alarm printer data, where available, recorded each actuation and termination of high pressure injection. The alarm printer records one full flow manual actuation of HPI at l

l 7:20. This followed the shutting off of a reactor coolant .

pump at 7:13.26

! Faust i

Craig Faust, an operator on shift on the day of the accident, stated in an NRC interview that manual HPI actuation

! occurred prior to the time of the shutting off of the reactor coolant pumps, at 5:41.27 However, Faust later appeared to

, 25 GPU trial exhibit no. 2079, TDR-044, p. 42. (Alarm i printer data not available from 5:13 to 6:48 a.m.)

26 GPU trial exhibit no. 2079, TDR-044, 27 Faust, NRC I&E interview 4/21/79, pp. 48, 51.

r 7

4- -- -- - --

I

^

l realize that his recollection of the manual actuation had been displaced in time, and thct he had confused the 7:20 full flow manual actuation with the 5:41 time period.28 Additionally, in response to what an ILE interviewer characterized as a

" monday morning quarterbacking" question about what he would have done differently, Faust stated that he had felt uneasy about turning off the reactor coolant pumps, and that, had he known at that time what he had learned after the accident, he would have tried high pressure injection rather than turning off the pumps.29 This conclusion appears to be inconsistent with Faust's statement that full flow HFI had been initiated I

at the time of the 5:41 reactor coolant pump shutdown. Howev-er, the interviewers apparently did not assist Faust in recon-ciling this divergent recollection, either by bringing it to his attention, or by providing relevant data.

Zewe Nearly two months after the accident, at an inter-view on May 25, 1979, William Zewe, shift supervisor on the day of the accident, referred to a 5:41 full flow, manual ac-28 Specifically, during an interview on May 29, 1979, Faust, after a lengthy discussion of what was unquestionably the 7:20 manual actuation of HPI, said "[t]his (the 7:20 ac-tuation) would be the manual actuation which would be right around when we were stopping the pumps. The one I could think of. So that would be earlier than what it should be." NRC I&E interview, 5/29/79, p. 10 (emphasis added).

29 NRC I&E interview 4/3/79, pp. 39-40, 48.

8 i

• - - = - . . . . . - . - - . . --, . , - - . . , . , . . . , . , --

tuation of HPI when the remaining reactor coolant pumps were secured.30 As of the May 25th interview, Zewe had attended

~

several meetings at which Faust.had mentioned a 5:41 actua-tion.31 Zewe's understanding of the time of the actuation ap-parently derived from the earlier statements of Craig Faust.

For example, during the May 25th interview, in response to a question as to whether he was sure full HPI had been on around 5:41, Zewe stated that "I was not as sure as the operator who actuated it was. He is sure, Craig Faust."32 Notably, Zewe's understanding that Craig Faust had actuated HPI was contradicted by Faust himself, who stated that he thought that high pressure injection had been initiat-ed by Edward Frederick at the time that he, Faust, recalled having secured the reactor coolant pumps.33 Frederick Edward Frederick was the operator assigned to the HPI controls throughout the day of the accident.34 He has never testified to a 5:41 actuation, either during the GPU v.

B&W trial, during his deposition or in interviews. In an ear-30 Zewe, TMI Staff interview 5/25/79, p. 31.

31 Zewe trial tr. 2760-62.

32 TMI Staff interview, 5/25/79, pp. 5-6; see Zewe dep. 824-l 26.

33 Faust, NRC I&E interview, 4/21/79, p. 39; Faust, Presi-dent's Commission deposition, p. 158.

34 Frederick trial tr. 3492-96; Zewe trial tr. 2124.

9  ;

1 i

ly interview, Frederick characterized the manual HPI actuation as having occurred during a time when Gary Miller, Station Su-perintendent, was present in th'e; control room.35 Miller ar-rived in the control room at approximately 7:05 a.m.

Frederick also placed the HPI actuation as occurring "after the relief valve had been isolated"36 which was at 6:19 a.m.,37 and as occurring "during the time when we had decided that we did not have natural circulation,"38 which was at about 6:54 a.m. when reactor coolant pump 1B was restarted in an attempt to induce forced flow circulation, marking the point at which the operators ultimately were convinced that natural circulation was not going to work.39 These statements establish Frederick's recollection that a manual actuation of full flow HPI occurred, not at 5:41 a.m., but rather after 7:00 a.m. -- conforming to the alarm printer entry at 7:20.

35 Frederick, NRC I&E interview, 4/23/79, pp. 48-57.

36 Frederick, U.S. Senate Subcommittee interview, 8/22/79, pp. 16-17.

37 GPU trial exhibit no. 2079, p. 47.

38 Frederick, U.S. Senate Subcommittee interview, 8/22/79, pp. 16-17, 39 Frederick, NRC I&E interview, 4/3/79, p. 36; Faust, NRC I&E interview, 4/21/79, p. 52; GPU exhibit no. 2079, TDR-044, p. 51.

10

P r ,

"GPU v. B&W" DEPOSITION AND TRIAL TESTIMONY 4 Faus1 Faust's deposition. testimony further confirms that he did not have'a sound basis for his earlie,r. statement re-garding a 5:41 full flow HPI actuation. He testified that. 's "before" they took the pumps off, high pressure injection was initiated; that he did not phys $cally do it, but he heard it ,

was done or was going to'be done.; that he did not know if i t ,

was at full flow; and that.he was primarily involved with the secondary side of the plant.40 Zewe During his GPU v. B&W deposition, after he had had access to relevant technical data, Zewe testified that he had recalled that "we did high pressure injection at or about the time we secured the second two pumps. But I cannot void my-self of everything else that has happened and whai is fact from other sources," referring to studies of borated water storage tank level responses to actuations of HPI.41 Thus, Zewe was unable to distinguish in his mind between what he had previously said based _on what he had' heard from Faust, and facts that he knew to be scientificall'y accurate. .

Zewe's' trial testimony further indicates that ,his recollection of a 5:41 actuation derived from Craig Faust's 1

40 Faust dep., pp. 525, 540. ~

2 41 Zewe dep.,,821-24. -

J s -

/

11 . ,

ga g T,.

% \.

5 ,_. , .-. , , . . _ . . . . ,, ,

,_ . -. ._ ._..r r _ _ - _ ..

j -

statem,ents. Thus, Zewe testified at trial that Faust had said, , referring <to a 5:41 actuation, that he, Faust, felt

+

t '

sure that's when it was" and thEt Zewe had then said:

"if you are sure7 thai's'when it was, then that's when it was. . . .

"42

y. ,

_i.

Zewe also testified regarding his own early recollection of 3

'the time of the manual HPI actuation: "I for one could not remember exactly when it happened. One of the operators, as I

recall, was pretty sure of the exact time."43 J

During the GPU v. B&W trial, GPU's counsel asked Zewe a series of-questions which he had not been asked in any

~

lof his prior testimony or interviews by the NRC or GPU person-

.nel, it. e .,: How many HPI actuations had there been? How many of these vere automatic? How many of these were manual? Zewe testified that he recalled four full flow actuations, three of wh'ich were automatic and one of which was manual.44 Specifi-cally, Zewe recalled a single manual actuation; the initial automatic actuation a few minutes into the accident; an auto-raatic actuation "af ter the declaration of the site and general emergency"; andlan automatic actuation in the early afternocn associated with the hydrogen burn.45 Zewe recalled that the

manual

actuation was the second actuation of the four.46

> 42 Zewe trial tr. 2761. .

43 Zewe trial tr. 2173, 2761.

44 Zewe trial tr. 2115-18, 2153, 2156.

i ,

45 gewe, trial ,tr. 2115-18; 2153, 2156.

s 46 Zewe trial.tr. 2117, 2121-22.

s ~

i 12 S

U 4

k e 7

n , , _ . - - , . . , . - - - . - r.~, . . . - - , _ - ,,.~---n.c e.- ,,,-., ,,- ,, , --r--., ,-,-,,n,r,-w ,- ,-n - --

This recollection is corroborated by recorded plant data'. Computer alarm data establish three automatic actua-tions of full flow HPI: at two' minutes into the event; at 7:56 a.m., or 30 minutes after the general emergency declara-

. tion; and at 1:50 p.m., the same time as the hydrogen burn.47 I The computer alarm data lists only one manual actuation of full flow HPI, at 7:20 a.m. Thus, the manual actuation was i

! the second HPI actuation, just as Zewe recalled.

+ Zewe also testified that the manually initiated full flow HPI that he recalled had been maintained "for a consider-able period of time,"48 which he stated meant from ten minutes

! to half an hour.49 This accords more closely with the record-ed 7:20 a.m. actuation, for seventeen minutes, than with the l

3 1

five-minute possibility at 5:41 a.m. in TDR-044.50 Zewe's ultimate conclusion at trial was that the closest that he could pinpoint his recollection of the single manual high pressure injection actuation at full flow was that it occurred after the reactor coolant pumps were turned off (at 5:41 a.m.) but before the second automatic actuation (at 7:56 a.m.).51 i

4 47- GPU trial exhibit no. 2084, alarm printer data.

48 Zewe trial tr. 2115.

j 49 Zewe trial tr. 2173-77.

50 GPU trial exhibit no. 2084, alarm printer data; GPU trial

exhibit no. 2079, TDR-044, figure 60.

. 51 Zewe trial tr. 2117, 2121-22.

13 i

n-~~ ,e- -,.---.w. - , - - - - , < . - , , . . _

The operators' earlier confusion may have resulted in part from the fact that some time after the reactor coolant pumps were shut off, the operators began to inject borated wa-ter into the reactor coolant system in order to regulate the boron concentration. This action was necessary to counteract symptoms that the reactor was returning to criticality.52 Furthermore, the 7:20 manual actuation was -- according to the recorded data -- in fact juxtaposed to the turning off of a reactor coolant pump,53 an event which Faust and Zewe linked to the manual actuation of HPI. A reactor coolant pump was shut off at 7:13, following an attempt to restart the pumps.54 Frederick Frederick, like Zewe, testified during the GPU v.

B&W trial that he recalled only one manual actuation of HPI.

.I He put the time of the manual actuation as "around the time of the site emergency."55 Frederick had no recollection of hav-ing actuated HPI manual.ly at full flow at around 5:41 a.m.,

and was quite certain chat no " mystery man" could have sneaked by him to actuate or turn off HPI while he was at the con-trols.56 52 Zewe trial tr. 2110-13; Frederick trial tr. 3477-79.

53 Zewe trial tr. 2153 et sec., GPU trial exhibit no. 2084, ,

alarm printer data. l l

54 GPU trial exhibit no. 2084, alarm printer data.

55 Frederick trial tr. 3493. The site emergency occurred at 6:50 a.m. GPU trial exhibit no. 2079, TDR-044, p. 50.

I6 Frederick trial tr. 4046.

14 l

l

At trial, B&W's counsel examined Frederick on the issue of a 5:41 actuation pointing out that Frederick had been

~

, present in post-accident interviews when Faust and Zewe re- ,

i l

ferred to a 5:41 actuation.57 Frederick consistently denied ever having agreed that there was a 5:41 actuation,58 and B&W produced no statements of Frederick to the contrary. Freder-ick testified:

I just never disagreed with [Zewe] that's all. Until we put all the data together, there was nothing to disagree about. . . .

I think that they think I agreed because I i

didn't disagree."5s Frederick also testified to data concerning the be-havior of make-up tank levels. On cross-examination, B&W sub-j mitted a chart. prepared by its engineers during the prior few days, reflecting the behavior of make-up tank levels at 7:20, which B&W's counsel claimed was similar to the pattern at 5:41 a.m. In order to resolve the issue in a reliable and defini-tive way, GPU's trial counsel decided to call in EDS Nuclear.

As discussed above, Dr. James Holderness testified to the EDS Nuclear study which established that it was not possible for HPI to have been actuated at full flow at or about 5:41 a.m.

in light of the analysis of make-up tank levels. In particu-lar, Holderness explained that while the 5:41 a.m. and 7:20 57 Frederick trial tr. 3876-77.

58 Frederick trial tr. 3877-79; 3883-84.

59 Id. at 3888.

15 1

- - - -,, , , , . - - . - , . - - . ---,--e -, ,. . , - - , - -~ , . , , . - -- -

.-e

a.m. make-up tank level patterns might appear similar to the l

naked eye, computer analysis plainly showed critical differ-ences and disproved any 5:41 a.m. actuation at full flow.60 CONCLUSION The alarm printer data show an undisputed manual ac-l tuation of high pressure injection at 7:20.61 Thus, the alle-gations of a 5:41 manual actuation must hypothesize an addi-l tional manual actuation. Yet, uncontradicted trial testimony establishes that both Zewe and Frederick had a clear recollec-tion of there having been only one manual actuation.62 Fred-erick recalls specific circumstances that place the manual ac-i

' tuation near to the 7:20 time recorded on the alarm printer; Zewe accurately places the one manual actuation as being the second of the total of four HPI actuations, as is recorded on the alarm printer at 7:20. The recorded 7:20 actuation fol-lowed a shut-off of a reactor coolant pump, also recorded in plant data, at 7:13. Accordingly, the trial testimony is con-sistent both with the available alarm printer data, and with the conclusion of the EDS technical analysis that there was no 5:41 actuation.

] 60 Holderness trial tr. 5637, 5700; GPU trial exhibit no.

2223, p. 3.

61 GPU trial exhibit no. 2084, computer alarm data.

62 Zewe trial tr. 2153; 2156. Frederick trial tr. 3493; 3876.

16 2

1

+ - - . - , , - - - - ,, ,,--w.--r-.,,n,-m,.,,,-,-,,-- c-e----r, , , - -w.v., e.,--- ,,,, ,,n.., ,ye-~ e e

i *

. I Earlier statements made by another operator, Craig  ;

Faust, and relied upon by his superior, William Zewe, mis-i placed the time of the manual actuation back by an hour and I forty minutes, to an earlier shutting off of the reactor cool-ant pumps, at 5:41 a.m. In the absence of data, this was an understandable confusion of recollection.

Since the EDS analysis scientifically proves that that there was no full flow manual actuation of high pressure

.1 a injection at or about 5:41 a.m., the colorful charge that a

" mystery man" turned it off was conclusively rebutted.

i 1

1 i

1 17 3

. - - _ . _ , _ _ _ . . . -_ -- . _ . ._ _ . . _ . , _ , _ _ _ - . - - -..__.m.- .-..,..m. . . - - . - . . . . . , - _ . , _ . - - _ _ . . , , , ,,