ML20147E442

Forwards Info from Contractors Re Code Selection.Provides Comments on THOR,RELAP-5 & Trac Codes by Bnl,Inel & Lasl, Respectively

ML20147E442
Person / Time
Issue date:	09/15/1978
From:	Fabic S NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES)
To:	NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES)
Shared Package
ML20147E439	List: ML20147E434 ML20147E442 ML20147E455
References
NUDOCS 7812210080
Download: ML20147E442 (45)

Text

- - - -

SEP151978 MEMORANDUM FOR: Those On The Attached List LST CY FRON: S. Fabic. Chairman Code Selection Comittee N

SUBJECT:

INFDRMATI0ft FROM CONTRACTORS Enclosed is the information sought at our code selection meetings with the contractors. The sumary is as follows:

BNL ' INEL LASL Ytem (THOR) (RELAP-5) (TRAC)

Completion of fast 17 MY 7.0 MY* 4 MY running PhA-BE LOCA (by9/79) (by12/79) code To get 1 hr.

running time on CDC 7600 Conversion of that -

code to EM 17 MY 3 MY* 3 MY (by9/80) (by6/80)

Completion of a BWR 34 MY 2 MT* 3 MY l transients code. (by9/82ff (by 12/80) without verification insequence)

• EG&G estimate does not include developmental verification of code (s) nor documentation

• • EG&G assumes availability of an existing and acceptable 1-D kinetics routine. that can be converted / adapted Original Signed By S. Fabic S. Fabic Chairman -

Code Selection Comittee

Enclosures:

as stated '

M BiaQ10 D 80

s 8

• e APPENDIX II . .

REVIEW COMMENTS AND ASSESSMENT OF THOR, TRAC AND RELAP-5 CODES FOR -

t POTENTIAL FUTURE USE IN LICENSING AUDIT CALCULATIONS. ,

V CONTENTS: INDIVIDUAL COMMENTS A) by Code Selection Committee Members - - -

S. Fabic, dated 9/17/78 L. M. Shotkin, dated 9/18/78 N. Zuber, dated 9/19/78 W. C. Lyon, dated 9/20/78 -

F. Odar, dated 9/21/78 B) by Consultants:

Professors T. Theofanous, dated 9/13/78 P. Griffith, dated 9/14/78

~

A. F.. Henry, dated 9/15/78 . -

l$

'; ' ' . . 4

~.

., s b , 1 '

1

'[

t

/.d.

h' .,p A;,_'

. .] ,

.u

.o .

....- . g. ...

I _

.. . . . . . . . ~ . . ... .

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __________b'- , .

S. Fabic 9/17/78 ASOSSMENT OF SIMPLIFIED, FAST RUNNING ADVANCED SYSTEMS CODES (THOR, TRAC, RELAP-5)

Contents: I. General Observations II. Items Considered During Selection of Simplified Advanced Codes (Table of ratings)

III. Justification of the rating shown in II IV. Listing of available options V. Items Considered in rating options VI. Conclusions / Recommendations

)

h

w I. Gen,eral Observations

1. Calculational speed should not take precedence over reliability of calculated results.

2. While initial drive in formulating simplified advanced systems codes emphasized calculation efficiency as well as accuracy, the emphasis on speed reaches a point of diminishing return after a certain number

.of code development years. For example, the current yearly development -

costs for THOR' code are around $1,300K. This amount of funds would pay for 1625 runs on a CDC computer, 'each lasting 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br />; this translates to 5 1/2 years of performing one such (2 hour2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br />) run each working day

~

(by or for NRC). Current estimates show that completion of THOR would require a minimum of two years just for a Best Estimate LWR LOCA analysis.

3. From what was observed so far, it is anticipated that an integral LOCA analysis with TRAC (collapsed noding) will take about 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br /> on CDC 7600, on the average. Development of detailed (3-D) TRAC will proceed regard-less of the choice of the simplified advanced code, candidate for EM analysis. It is also clear that this planned development of TRAC will allow for running the code with collapsed noding.

4. Develupers of RELAP-5 stated that, to obtain predictive capabilities, it may be necessary to extend the code to more than one dimension _in treating certain vessel regions. While their claims that this could easily be accomplished may be valid, it appears that, in that case, both RELAP-5 and TRAC are likely to have ccaparable running times.

.....+...4  ; . : ' 4. . . . . . . . . . e +#n . .... * -. e-.i-eMa'*wo-..'. n w oe *,-ad%* -e-

1 s - l

5. The current running times of TRAC, for sample problers chosen by the code selection committee were equal to those of RELAP-5 for two sample problems. For the third (and most important) sample problem dealing with an integral LOCA calculation, TRAC ran twice as fast'as RELAP-5.

Both LASL and INEL claim that further running time improvements are foreseen.

6. Developers of RELAP-5 have achieved remarkable progress in the last two years with a relatively small group of dedicated personnel'(6 engineers).

While they owe a good part of their success to the breakthrough achieved at LASL in development of numerical solution techniques for transient two phase flow, their ability to render excellent service must be made use of in the NRC research program.

7. Current limitations regarding computer accessibility can be easily over-come by renting a CDC 7600 computer and installing it at one national labora tory. Estimates show that a combined cost of renting plus service overhead, would actually result in savings of about $500K/ year, in comparison with the current computing expenditm es at a national laboratory (basedon$425/ hour). Computer accessibility should, therefore, play ,

no role in making a code selection.

l i

l l

i

-u... ~ ~ . . - . ....6. . -. . - -

\\ 8 b4 b / (*.

In II. ITEMS TO BE CONSIDERED DURING SELECTION k y 0 y

0F SIMPLIFIED ADVANCED CODE j[ (k GROUP A: CURRENT CAPABILITIES MODELING OF THERMAL-HYDRAULICS T 8 #

GE0 METRIC DESCRIPTION; ADAPTABILITY 6 4 8 NEUTRONICS -

3 5- r USER CONVENIENCE 4 4 7 COMPUTATION EFFICIENCY g z e VERIFICATION (PERFORMED THUS FAR) .

4 / 3 3o 14 41

'/' o 52 40 vs GROUP B: ADAPTABILITY TO FUTURE.NRC NEEDS AND PROJECTIONS OF DEVELOPMENTAL EFFi.",

STEAM GENERATOR TUBE BRELS P t y STEAM LINE BREAK 7 , v ALTERNATE ECCS (E.G., UHI) 7 r f PWR LOCA - EM p. e f BWR LCCA - BE ^ND EM Ir 5- 7 ATWS (PWR AND BWR) p 4 g RIA (PWR AND BWR) P 4 e OPERATIONAL TRANSIENTS (PWR AND BWR) # d' f sg r1 tr 71 57 77

/

$*f4 h/Y ee II. ITEMS TO BE CONSIDERED DURING SELECTION  % 4  %

OF SIMPLIFIED ADVANCED CODE (cont.) $

GROUP C: ASSESSMENT OF_ CODE DEVELOPMENT' TEAM AND FACILITY .

RECORD OF PAST PERFORMANCE 7 3 8 KNOWLEDGE OF INFORMATION OBTAINED IN

1. # 7 REACTOR SAFETY RESEARCH PERSONNELPOOLACCESS(WITHKNOWLEIiGE '

- IN THERMAL HYDRAULICS, REACTOR SAFETY-INFORMATION, LICENSING REQUIREMENTS) 7 4 /

COMPUTER ACCESS (ABILITY TO PERFORM f g 7, ON-CALL ASSISTANCE)

52. .

20 37 l

l 4

.- l

. - a-:~n . - . .. .. .

l

~

III. Justification for Rating in Section II

1. Group A: Current Capabilities (a) Modeling of thermal hydraulics in TH0R is superior. Disadvantages: 1 Need to develop property profiles for each system component, for

• each type of transient. A new situation, if and when encountered, would require code developer's intervention. Ability to analyze system transients has not yet been demonstrated; however, that is ~only a question of time. This, however, offers a dilemma:

Is the sizeable future development cost worthwhile if other, currently available techniques can give very acceptable answers?

To this reviewer modeling in TRAC is superior to that of RELAP-5.

, l Further improvements are planned.

(b) Geometric adaptability of TRAC is superior primarily because consequences of multidimensional behavior can be logically ex-plored and accounted for in determining the collapsed noding.

Inflexibility of TH0R became apparent in attempting to calculate the Marviken experiment.

(c) In the area of neutronics TH0R group has done most development (1-D kinetics will be needed for ATWS; it may be needed for small break). However, it was not yet operational. Point kinetics in TRAC was. operational. RELAP-5 did not consider it , yet.

e.e.wwwseS usum s-met me.<=<e te +e e w +-Mw- 'AI

=* w e 1* ,

e, g..A p o t e as .g-s e

~

6-(d) User convenience. The planned routines for self-initialization not yet incorporated in RELAP-5 and THOR. Existing in TRAC.

(e) Computation efficiency of TRAC is superior to that of RELAP-5.

TH0R could execute only the simplest test problem and even then LSe choked flow (the main purpose of this test problem) was not calculated per method described in documentation.

(f) Verification of TRAC, so far, has considered many more cases than that of RELAP-5, with acceptable accuracy.

2. Group B: Adaptability to Future NRC needs TRAC and RELAP-5 score the same. While TH0R could also (theoretically) perform all the anticipated analyses, it is clear that its adaptation would be more time consuming, primarily because of the need to develop specific profiles.

l

3. Group C: Assessment of Code Development Team (a) Record of past performance is better at LASL considering their sustained success and progress over the last 31/2 years. That record greatly overshadows the record of BNL team in spite of BNL's headstart.

(b) Knowledge of RSR information appears somewhat better at INEL, mainly because of their long term involvement in reactor safety l

work. LASL is catching up fast.

(c) & (d) No explanation needed.

l t

. ' . , . , _ b.m . .-w .,+ - t .- .,w,... ../......?.n.A.....'..4.,.<- . - . - * - ..

' ~

L

.e IV. Op_tions (Not Prioritized)

I,A., TRAC: Development per current scheoules include; development of detailed BE analyses of LWRs. Also ability to collapse

,s w noding for faster,4 detailed calculations. Reasonable running times with collapsed noding; could be greater thcn 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br />, Possibility of conversion to CRAY computer to obtain very fast running times, after FY 80.

RELAP-5: Discontinued.

INEL code development effort on RELAP-4/N00 7 remains unchanged.

Code independent assessment effort INEL switched to assessment of TRAC.

Fersonnel currently working on RELAP-5 re-directed to perform TRAC sensitivity studies, including testing of own models for constitutive relations and break flow.

During FY 80 all INEL code developers switched to assessment of TRAC (LASL version only).

In FY 81 Simplified TRAC converted to EM (PWR and BWR LOCA)

THOR: Discontinued In FY 79 all effort given to NRR specified mddifications in foreign non-LOCA codes (ALMOD, ALMOS, RAMONA).

In FY 80, 50% of work on item as above, 50% on assessment of TRAC concerning basic and separate effects tests.

After FY 80, all work on TRAC assessment, emphasizing basic and separate effects tests.

-.w..m<~ u se wee i+ - se e*<w=- d* ** * ' ' * ' * * """

~

-..L .. _. l.- ..--.. -

^ ~

- . .-- . --. z..

e

- Development of detailed BE analyses of LWR's per current (B.) TRAC:

plans

- Work initiated on fast running version of TRAC, suitable for EM analyses, with running times less or equal to one hour (CPU) on CDC 7600.

- EM conversion for PWR & BWR LOCA performed at LASL, during FY 1981.

RELAP-5: Discontinued Work at INEL same as for option (A) except that work on conversion of TRAC to EM, during FY 81, not performed but, rather, TRAC assessment and sensitivity studies continued.

Possible help with intermediate uncertainty study with TRAC.

THOR: Discontinued Work at BNL same as for option (A).

- Same as for option (A)

(C) TRAC:

RELAP-5: On one year probation INEL continues work on RELAP-4/ MOD 7 at current level (and completes it at end of FY 79).

- Work on RELAP-5 continues at same level. During Sept. 79 RELAP-5 assessed again against TRAC.

Work during FY 80, on code development, to be defined by NRC, pending Sept. 79 assessment.

Code assessment work during FY 79: All on development checkout of RELAP-5. Work after FY 79 to be determined by NRC.

P b - n.e e 4

~...e 4 i~-..,.... , .. . . . .. E .

a .

THOR: Discontinued Work at BNL same as per option (A).

[D.' TRAC: As per option (A)

RELAP-5: Discontinued Work at INEL as per option (B) except, during FY 81.

THOR: On one year probation .

Work at BNL proceeds per current level.

TH0R re-assessed during September 1979 (compared with TRAC).

Future work pending results of September 1979 assessment.

[E., TRAC: Continue work per current plans and funding, as per option (A).

x. -

RELAP-5: On one year probation. Work at INEL same as per option (C).

THOR: Also on one year probation. Work at BNL same as per option (D).

(F) TRAC: per option (A)

RELAP-5: Accelerated development.

Work on RELAP-4/ MOD 7 discontinued. All code development rsersonnel working on RELAP-5, including developmental checkout.

During FY 79, independent assessment group at INEL performs TRAC assessment and sensitivity studies per NRC requests. ,

o, After FY 79 s,ame, to be determined, fraction of their work j may involve independent assessment of RELAP-5. It is.not yet clear that a future EM code needs independent assessment.

l THOR: Discontinued Work at BNL per option (A).

l L.

W - . . . , . . . . , , ,, ,a

l

.. l

'.' l

\

RATING OF OPTIONS i 0PTIONS Item Weight A B C D E F l l

(1) 3 3x4 3x4 3x3 3x2 3x3 3x3 (2) 2 2x7 2x7 2x7 2x5 2x7 2x7 (3) 2 2x7 2x7 2x8 2x5 2x8 2x8 (4) 1 3 3 2 2 1 1

~

(5) 2 2x3 2x2 2x2 2x2 2x1 2x1 (6) 2 2x3 2x2 2x2 2x2 2x1 2x1 (7) -1 0 -3 -1 -1 -2 -2 (8) 2 2x1 2x3 2x2 2x2 2x3 2x3 (9) 1 0 0 1 1 2 3 (10) 1 0 0 1 3 2 1 (11) 1 3 2 1 1 1 0 (12) 1 -3 -2 -1 -1 0 -1 (13) 1 1 1 2 2 3 2 (14) 2 3 3 2 2 0 1 (15) -3 0 0 -3x1 -3x3 -3x3 -3x2 TOTAL SCORE 61 58 55 41 47 48 Total Score, 21 18 16 15 8 9 Omitting first Three Items

.. . . . . . . . . . , , . . . . . . .. .. . ,, . . . . ., ,,.4.

VI. Conclusions / Recommendations

1. On technical and fiscal grounds the optimum option is (A); providing the INEL and BNL personnel would be willing to help NRC in its effort to get a thoroughly qualified best estimate code and if their assessment of TRAC could be constructively formulated, resulting in the needed improvements. Ideas and means for such improvements should be explored at both INEL and BNL and finally implemented by LASL if acceptable to NRC. TRAC assessment would thereby be greatly ac elerated.

Many comparisons with test data need to be made, and many sensitivity runs performed, to reach a desired goal. Accomplishment of that goal requires a multi-laboratory effort and cooperation.

2. The second choice is the option (C). I do not think that it would be wise to abandon RELAP-4/M00 7 at this stage, as stipulated in the option (F). While TRAC exists today and can calculate PWR LOCA,it is a very young code. A mature code such as RELAP-4 sho- rot be dropped before a desirable level of confidence in TRAC ability has been reached by all its users. It may even turn out that RELAP-4/ MOD 7 is faster than RELAP-5 and equally accurate. Both RELAP-5 and RELAP-4/

MOD 7 should have the ability to ca.lculate PWR integral LOCA transients at the end of FY 79, if work were to proceed at current levels.

3. If NRR insists that running time for simplified analysis must be less or equal to 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> on CDC 7600 then my choice would be the option (C).

1

. .. .J ..a . . - . . ..w.. , . . , , . . . . . . .

&u c -

A disadvantage implied by this choice is the fact that INEL expertise would not be directed towards improvements and assessment of the best estimate code, TRAC. It is this code, rather than a simplified (fast running) code which will ultimately be used for the purpose of evaluating the margin of safety .

3.. .

'W',,'

O-' "*

,. L. M. Shotkin 9/18/78 The results of the code selection meetings of 9/6 + 9/8/78 show that the choice is between simiplified TRAC and RELAP 5; TH0R is not a serious contender. My relative ranking of the 3 codes is given 'in Table I.

TABLE I: RELATIVE RANKING OF CODES Perfect Code Score RELAP-5 THOR TRAC Present 10 5 1 8 Capability Future 10 7 3 8 Prospects Code 5 3 1 5 Team Facility 5 4 2 3 Expertise i

Total 30 19 7 l

24 A comparison of the 3 problem results presented at the meeting is given in TABLE II while TABLE III details the justification for the ranking in TABLE I. Finally, TABLE IV compares pros and cons of continuing with TRAC or RELAP-5; TH0R is so far behind the other two codes, it was not included. .

TABLE II: PROBLEF COMPARISON RUNNING TIME (MIN., DATA COMPARISON CODE RELAP-5 TH0R TRAC RELAP-5 TH0R TRAC Marviken 4 2-4 " Poor Good 4 g{

S. P.16 100 Cou n' t 100 Fair None Fair Run Code To 150 360 Couldn't 135 Good None Good Run sec) l l

i

_._.._u. _._.,.:..._.. _ , . . . _ . . . .. .. . . . , _ . . _ .

l

__ .__. -- _ _ _ _ _ . _ _ _ . . _ _ - _ _ _ - _ _ _ - _ _ _ _ _l

2 TABLE III: DETAILED PROS AND CONS: RELAP-5 Pros Cons Present Capability Thermal-Hydraulics Break-Flow Model. No Reflood Heat Transfer.

Much Improved over RELAP-4 Overall, not much different than TRAC. Used different multipliers (0.85 and 0.95) to fit different data.

Geometric Description More Nodes Than RELAP-4 Limited to 1-D (Downcomer, Upper Plenum and Core Cross-Flow Modeling too Simplified)

Neutronics Not in Yet flunerics Reasonably fast running Unimaginative treatment of time, until ECC water flows water packing thru system Verification Limi ted: No refill or reflood, just blowdown. Will require separate effort from TRAC.

Future Prospects 7 g LOCA-BE [Givesindependent, 5 (Limited to 1-D, LOCA-EM if similar, capability will probably require ATWS,RIA Operational &

Non-LOCA to TRAC

) ore effort than TRAC.j l

6 l

TABLE III Continued 4

, Pros Cons a

l Code Team i

8 Past Performance Small Team Worked Well SLOOP Experience j Safety Research INEL Facility Most Team Experience seems ,

Knowledge Experienced Limited to Blowdown i

Facility Personnel Pool Can draw from RELAP-4

.i Computer Access OK for present limited use Will be saturated in 1980

-i t

J i

}e-m er war p r., ,

R Ai,

. . . . ~ . . _ . .L._. , ;

i

_ _ . . _ _ . . _ _ l

, a .,.

\

TABLE III: DETAILED PROS AND CONS: TH0R l

Pros Cons Present Capability Thermal-Hydraulics Level Tracking in some Limited to 4 Eq'n.fDF Break Components Model not working, Code system Moving flow and Heat-Transfer not working. Requires profiles Boundaries for accuracy.

Geometric Description Limi ted to 1-0 Not as ficxible as standard finite-difference, uniess profi.les accurate.

Neutronics BNL has 1-D package developed Not in yet Numerics Technique potentially very System not working. Technique rapid not yet demonstrated.

Personnel have little successful experience future Prospects I

(

LOCA-BE Limited to 1-0, Requires know-l LOCA-EM ledge of profiles for each case ATWS, RIA Based on past performance, l

Operational & questionable if can meet goals ,

n n-LOCA

<i

\n any reasonable time.

l l

l 4Ih iTN $

' %tra b _e e e - - ~~w- ,

=

TABLE III Continued (THOR)

Pros Cons Code Team ,

Past Performance After 4 years of effort, Goals not met.

Has trouble working together as a team.

Safety Research Other BNL Groups active in Not demonstrated Knowledge licensing issues Facili ty Personnel Pool Individuals with seemingly Very little with successful good knowledge of basic code development experience Two-phase flow, but have trouble applying to large engineering problems Computer Access OK for limited usage Computer Down-time excessive l

l ,

_ . _ _ _ _ . _ _ _ . _ . , . _ _ _1 _._ _~~ _ . . _ _ . _ _ _ _ _

l

. ,1 * ..' i l

l TABLE III: DETAILED PROS AND CONS: TRAC 1 Pros '

Cons

Present Capability

. Thennal-Hydraulic Can handle entire LOCA, with Collapsed noding capability first models not yet demonstrated Geometric Description Flexible, with Multi-D New models may be needed for where needed collapsed noding Neutronics _

" Point" Capability 1-D not in yet, Feedback models -

not verified Numerics Competitively fast Verification Most extensive of 3 codes Must redo for collapsed noding i Future Prospects LOCA-BE Complex Version will be Will have to hire additional LOCA-EM designed to handle these people to accomplish these ATWS, RIA anyway tasks, for simplified vertion Operational & Non-LOCA of TRAC Code Team Past Performance Excellent team; work well together. Try to respond to NRC schedules Safety Research Knowledge Increasing familarity with Still need to hire people with experimental facilities thru more experience comparison of TRAC with data Faci li ty Personnel Pool Would have to hire additional

people for simplified TRAC tasks.

Computer Access May be resolved in FY 79 Probably won't be resolved until FY 80.

TABLE IV: PROS AND CONS OF CONTINUING WITH SIMPLIFIED TRAC OR RELAP-5 TRAC RELAP-5 (simplified) i Pros Must develop simplified version, anyway, 1. INEL experience with 1-D modeling.

for sensitivity and uncertainty studies.

2. Independent approach to full-scale extrapolation Cons May require major remodeling (i.e. , 1-D) Separate verification effort to get 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> running time with " accurate" required, answers i

,-. ..,m.4 ..

N. Zuber 9/19/78 CODE SELECTION Recommendations

1. The TRAC code should be selected for advanced licensing audit analyses pertaining to LWR LOCA and non-LOCA transients.

2. The TH0R code activity should be reduced to a two or three man developmental effort to demonstrate the capabilities of the approach and assess the potential for attaining high computational speed coupled with accuracy and level training abili ty.

3. With' the exception of being more modular, RELAP 5 in its present form, does not offer any technical advantage when compared to TRAC. Furthermore, its development lags TRAC by more than a year. Consequently, I do not see any rationale, based on technical argumen'ts, fo? continuing the develop-ment of RELAP-5.

4. Ccda development resourcas at li!EL and 0::L should be maintained and used by NRC to test and assess the capabilities of TRAC and convert it to a code suited to licensing needs.

Discussion

1. The overall rating given to the three codes was as follows:

Group A: RELAP 5: 40/60 THOR: 30/60 TRAC: 52/60 Group B: .RELAP 5: 60/80 THOR: 50/80 TRAC: 70/80 I

a - ww + - b . + - . +

.i l

Group C: RELAP 5: 34/40 THOR: 24/40 TRAC: 32/40

2. The main advantages of TRAC compared to the two other cocies are:

a) Conversion of TRAC to a licensing code will require a minimum of additional code assessment effort b) It provides a natural framework for reducing 3-D to 1-D computations.

c) It offers the fastest method for providing NRR with an advanced EM code.

3. Some of the disadva'.t c e? TRAC are:

a) TRAC in its prs " an, is not a truly modular code. For example, the entire vessel is' treatdd as'one component. ,

b) TRAC does not differentiate flow regimes and flow regime correlations according to components and processes. At present, the same flow regine criteria and correlations are used throughout the system. This approach, as evidenced during LASL presentation, can lead to erroneous results.

c) TRAC uses the staggered mesh approach.

4. The only advantage that RELAP-5 has presently vis-a-vis TRAC, is 'that it -

is more modular. However, in any other respect. TRAC is superior to RELAP-5.

5. TH0R has not demonstrated yet that is can:

a) Run system computations b) Calculate cold ECC injection c) Achieve high computational speeds ,

6. TH0R has demonstrated that it can track mixture levels. Since this capability is important to small break analyses, it seems desirable to continue this

prog ram. However, this effort should be at a greatly reduced scale with the objectives of:

a) Demonstrating the capabilities of the TH0R approach b) Asaessing the potential of attaining a truly fast running code.

Thus, it these objectives are attained, NRC will have a computational tool available to use if the need arises (or continues) for fast running cal-culations coupled with accuracy and level tracking capability.

, 7. Both INEL and BNL have code development teams with experience in thermo-hydraulic modeling and reactor licensing needs. These resources should be maintained and used by NRC. The most efficient and rational method of utilizing them would be to ask both laboratories to participate in a testing and assessment program of TRAC and in converting it into a code suitable to licensing requirements.

i

l

. W. C. Lyon 9/20/78 l ADVANCED CODE COMPARISONS

1. Summa ry The following are my conclusions regarding the three advanced codes:

1. TRAC clearly is further developed than the other codes, and is capable of handling the proposed investigations.

2. RELAP exists as a blowdown code, but can be made available in the amplbte version on a schedule and at a cost comparable to that for TRAC in a 1-D version. It can handle most of the proposed versions, i

and may be able to handle all of them.

3. TH0R doc- exist as a working code and' the effort required to bring

, it to that .,utus is prohibitive. I see little potential benefit in continuing the THOR development.

4. RELAP, on a comparable basis, appears to be more accurate than -

in the test problems investigated.

5. On a comparable basis, RELAP is slightly faster than TRAC,'for the three test 1 (exclusive of refill - reflood on S-06-3). The di fference significant.

l 6. RELAP is easier to use, with better input processing, easier to use input format, and better output capability.

7. RELAP modeling of nonequilibrium phenomena, choking, and area changes is superfor.

8. The RELAP solution technique is accurate, stable and requires no water pack "fix". These benefits are accomplished at the expense of running time which could be a problem during periods of ECC injection at low pressure. Mass conservation is excellent, whereas there is an un-explained 20% error in TRAC for S-02-6.

W.K h

. _ . . . , , . . - . . , , . .~. . w , . .. - . . : -. - ..: . : .

- ~

9. RELAP's basic formulations readily- take the form of two fluid, dri. .

flux, or homogeneous descriptions, either equilibrium or nonequilibrium, which allows easy incorporation of EM modeling and comparisons with other codes.

10. Na energy partition information is needed for RELAP.

II. Recommendation l Terminate TH0R and retain TRAC and RELAP as complimentary codes. I consider 1

the ability to independently assess PWRs to be very dmportant. TRAC's 3-D L capability is required and the ability to reduce to I-D is very attractive.

i INEL's 1-D'and engineering experience will provide valuable input, and will provide a code familiar to the RELAP users that is designed as an engineering i

l tool.

III. Code Assessment L THOR does not exist as a working code and requires considerable effort to '

be developed to this 'atus. I do not believe it competitive, and will not discuss it further.

TRAC is the most completely developed of the codes, but does not exist in i a true 1-0 version. It clearly is capable of a blowdown - refill - reflood 1

calculation and has sufficient flexibility to be applied to experimental systems and to PWRs. Run time is acceptable in the test problems, with considerable improvement possible. I am concerned with the poor mass conservation ( 20% error for S-02-6) and the potential for the water pack phenomenon determining the course of ECC flow (as has been seen in RELAP-4). Treatment of break flow requires detailed nodalization and I s *w o . .e-. g

question that the numerics converge to the correct solution when taken in the limit. This means break flow modeling is an " art", an undesirable feature. - Improvement is also indicated in modeling internal structures such as core support plates, where water holdup may occur. The code is somewhat cumbersome to use (which can be corrected with a relatively small ef fort) .

RELAP exists as a blowdown code only, although it was run through refill and reflood for S-06-3. The heat transfer package was added within roughly the last month, and checkout is incomplete. Calculational error is likely.

RELAP is easier to use than TRAC, has several features that are not in TRAC, and is more of an engineering tool designed for full utilization of our user and 1-D experience with acceptable approximations.

Code capabilities are compared in Table 1.

I consider the test problems to be an important assessment in code evaluation, l

i and have spent considerable effort in comparing results. The calculations were not performed on comparable bases and one must ,iudge what would have happened if a comparable calculation were performed. For example, the TRAC Semiscale calculations were performed with a more coarse mesh in most of the primary system than used in RELAP. The selected TRAC noding introduced calculational error. On a comparable basis, TRAC will run slower and will provide better accuracy than demonstrated by LASL.

l l

l

. - . . . n,.. - _ -

at .. )

y

_4 .' 2

( TABLE 1. CURRENT CAI' ABILITIES .

i j

ITEMS RELAP-5 THOR TRAC i i

T-H Modeling H excellent on 1-D modeling Theory appears good, H good, mistes nonequib. effects.

limited to BD, CHF deficiencies, little demonstration of T improvements indicated. H not calcs. lig holdup on plates, etc workabili ty. Not as as flexible as RELAP. May require due to modeling approach. Treats flexible as RELAP and more constitutive info than RELAP break flow and tees simply. TRAC. (energy partition for example).

Geometric & Highly flexible and can be I'm not convinced code Good flexibility. Can be applied Adaptability applied to almost any test can be applied to range to 3-D and 2-D applications where geometry or component. (No of test geometries and others may use 1-D approximations, BWR yet, see also TRAC) facilities without ex- the adequacy of which hasn't been tensive additional work. demonstrated.

(See also TRAC.)

Neutronics Not avail. Point kinetics and 1-D can be easily added.

Overall, I see little advantage of one code over another here.

User convenience Excellent, but self-init. Not assessed Input by number fields required.

not available yet. All in- Output somewhat complicated, but put processed for errors not bad. One error in input can prior to termination. Input terminate processing with no independent of number fields. analysis of remaining information.

Excellent plot pkg. in exis-tance and planned. Card order is not required for

, ' aut.

t

^

~ '

l TABLE 1. Continued .-

ITEMS RELAP-5

] TH0R TRAC

. Computation Acceptable for BD. Not demon- Not demonstrated Acceptable. Can be improved Efficiency strated for reflood. Can be significantly.

. improved significantly. Run times on a comparable basis are - -

faster than TRAC.

Verification ' Very limited. Has done an *None with systems code Limited. Good results have been excellent job generally on the obtained generally. Promising.

problems which have been run.

Promising. Better fit to nonequib.

than TRAC and slightly better on three test problems.

l l

-___.-___~

I have considered these effects in comparing the calculations on a side-by-side item-by-item basis. The results are summarized in the following sub-sections and are detailed in the appendix. I've also provided my assessment on a 10 is perfect basis.

A. Marviken RELAP-5 9.5, TRAC 8.5, TH0R <j TRAC and RELAP are comparable on running time (a RELAP run'with the two nodes downstream of the choke point removed ran in 1.04 min vs a TRAC best of 1.2 min). Both can be improved readily by a factor of 2 or 3.

RELAP has the edge on calculation accuracy, with the ability to calculate the early nonequilibrium behavior, a better fit to data for AP216 (where TRAC calculated the anticipated AP reversal that was not seen in the data), and better calculation toward the end of the run where TRAC pre-dicted a sudden pressure drop. RELAP mass flux is closer to the data.

TH0R had to be fed pressure upstream of the break in order to run, and essentially missed all phenomena. THOR run time shows no advantage, and I have little info to predict an improvement.

B. S-02-6 RELAP-5 9, TRAC 8.5, TH0R 0

=

TH0R could not do the first time step. TRAC reduced their nodalization too far to attain the run time of 74 min. (85 min. with self-initialization)

Proper nodalization would add 50% to this run time. RELAP initially took 90 minutes, which was reduced to 55 minutes when a junction error was removed which did not affect the results. On a comparable basis, my guess is RELAP is roughly twice as fast in the existing configuration.

7-On pressure behavior, there is little advant 'ge overall, with each code exhibiting minor weakne:ses. RELAP generally follow' es w rate phenomena better than TRAC, which in some calculation; :. . . room the coarse nodalization. Both show regions of excellent agreement s some regions of significant disagreement with the data. Fluid temperatu*e l-appears somewhat better calculated by RELAP a -. k:*.3P generally does a I much better job of calculating rod teinperat - ;s, s . .;gh it calculates thermal spike and uncovering after 350 to 400 see that are not reflected in the data. TRAC results are strongly affected by the coarse heat slab nodalization, c

Both codes do good job on density, with RELAP following discontinuties better, but sometimes missing the event times. I would give RELAP a slight edge after taking the TRAC nodalization into account.

C. S-06-3 RELAP-5 9, TRAC 8 '. 5 , THOR 0 THOR could not do the first time step. TRAC used 112 nodes to 100 in

RELAP, but on a "RELAP basis" TRAC really used 50 to 55 nodes. (TRAC used 18 nodes for each nozzle (2) where RELAP requires none,15 in the

! steam generator to RELAPs 5, and 24 in the vessel, of which half are i

duplicates due to the requirer;ent for a mir um of two nodes in the angular direction.) TRAC's description of ' s of the primary loop is not as detailed as RELAP's. Run times are comprable early in the blowdown, and TRAC becomes faster during periods of ECC delivery at low pressure by about a factor of two. On a comparable node basis, RELAP will be faster early and of about the same speed later on. As gp ser) at 46 9 . htmeng$e 86s

8-in S-02-6, both codes can be speeded up by roughly a factor of three.

However, we really don't know what RELAP can do during the later stages of refill and during reflood since only the blowdown version of the code was available.

Neither code has.an advantage in calculation of pressure and differential pressures, with both exhibiting similar differences with respect to the da ta. Accuracy is good. RELAP is better at calculation of density, probably in part due to the coarse TRAC nodalization. However, RELAP provided some excellent comparisons to 150 see whereas TRAC only presented the first 40 sec and only a few comparisons. Some of the oscillatory phenomena associated with ECC caused condensation was followed very well by RELAP. This comparison could not be made for TRAC. The codes are generally comparable on mass flow rate behavior except during the first second or two, where RELAP is better.

Neither code is as good as I would like in computation of rod temperature.

TRAC is a little closer to the data at mid-core for the blowdown. I do

'not consider RELAP applicable for reflood due to use of the blowdown correlations.

IV Adaptability to Futur g NeedsChis is summarized in Table II. I believe both RELAP and TRAC can

/

be applied to the requir ments as outlined. RELAP has an advantage for 1-D oriented problems eser TRAC when multidimensional investigations are to be

.)

I

+ -

9 performed. (Note that RELAP's structure is compatible with multidimensional calculations when required. INEL indicated this could be added with little di fficul ty. My Teeling is it would be a significant effort unless of a very limited nature.).

TABLE II Adaptability to Future Needs Item Comments S. G. Tube Break N2 needed, comparable (reasonable) effort for both RELAP and TRAC. A better S. G. model may be necessary for RELAP.

Steam Line Break Potentially requires 2-D or 3-0 analysis in reactor vessel, which TRAC should be capable of treating now. Need has been neither established or disproved. Even if 3xis ts , I believe 1-D code is applicable with suit __ . ossumptions and models. RELAP will need a better S. G. model.

Alternate ECCS RELAP has demonstrated better ability to follow nonequili-brium phenomena and ability to hold up liquid at area restrictions with no special modeling considerations.

Multidimensional aspects would have to be approximated in l

RELAP where TRAC may treat now. (Downcomer adequacy has notbeenestablished.)

t

-. ~

TABLE II Continued Item Comments PWR LOCA-EM RELAP hydrodynamic formulation is compatible and structure is designed to accommodate. I suspect TRAC will be more difficul t.

BWR LOCA-BE and EM No BWR work has bee'i accomplished in RELAP.

ATWS, RIA and RELAP has been structured to accommodate these events.

Transients However, multidimensionality requirements and RELAP suit-ability remain to be established, where TRAC hydrodynamic capability exists (RELAP can be extended to more than 1-0 if required - I would consider only on a few special cases).

4

%1h ::

L . .

V. Code Development Team and Facility Assessment Team and facility background is summarized in Table III. On a value received basis, RELAP is a bar&9f n for the - 10 MY effort. expended. INEL also jumped ahead in their schedule by several months to perform the test problems on a short term basis, an impressive performance. In the same time frame, they provided good documentation and most of the data comparisons we requested. LASL provided the test problems on schedule, but some of the requested data comparisons have not been provided. As usual, they appear fully qualified for the work in question. Brookhaven provided good documentation, but failed in the major purpose of the investigations, as opposed to INEL and LASL, who clearly demonstrated a response capability.

P 1

.s,.,.,.. ,,

W"*~~

• i ~ :

TABLE III ASSESSMENT OF CODE DEVELOPMENT TEAM AND FACILITY -

i

• ITEM RELAP-5 TH0R . TRAC

' j- Past Performance Generally outstanding. Occasion- Generally poor. Misses Generally good to excellent.

, ally misses schedules, sometimes schedules. Cannot assess Cooperation with NRC appears early. RELAP-4 team generally cooperation with NRC. In- good to excellent. Internal

good. Cooperation excellent when ternal attitude poor. attitude and leadership appear ,

NRC identifies a priority need. Leadership appears lacking good to excellent.

Internal attitude excellent .

(cooperation within organization).

Leadership good to excellent.

Reactor Safety Excellent. Fully aware of most Can't fully assess. Appear Can't fully assess but appear

{

Research Knowledge U.S. research. Weak on some weak on systems knowledge, well versed in basic areas and -  ;

foreign programs. good on basic hydrodynamics. improving in systems tests,  !

I although still weak. +1 Personnel Pool Outstanding knowledge of thermal- Can't fully assess. Entire Can't fully assess. I've seen Access hydraulics, reactor safety in- THOR team has demonstrated no demonstration of licensing fonnation, licensing requirements, they have little knowledge related knowledge and suspect 1 systems tests. Depth in basic of licensing requirements it has not been developed.

f hydrodynamics limited to a few and needs. Hydrodynamics- Hydrodynamics depth appears to j '

t personnel, but adequate so far, depth appear.e excellent. be adequate.

Computer Access Excellent - computer is not in  ?  ?

full time use.  !

l

VI Cost and Schedule THOR is too expensive, and I will not comment furt.her. On TRAC, I doubt they can get a checked out EM for 3MY, double that is more likely.

(I cannot conceive the job is easier than for RELAP.which was designed with EM in mind following the PWR-BE, and which is basically compatible with most EM assumptions.) The 6/80 estimated date appears reasonable.

I do not feel qualified to comment on the BE and BWR transient estimates beyond stating the dates appear reasonable and the estimates appear low.

.RELAP is estimated too low and they have not allowed enough time. The INEL detailed breakdown is given on the next page. Values in par!Intheses are my guesses where I disagree with the INEL estimates. (I have not discussed these with INEL - they are simply my guesses based on experience.)-

I have not looked far enough into the BWR transient that I feel confident of the estimates.

_ , _ _ . , . __ . . ., , . u a _ m ~ . 4 ;

,'.  ;//' /::.'/. - COST ESTIMATE FOR REl.APS PWR BE, EM Arid BWR TRANSIENT PROGRAM A

^ '

• TASKS PWR DE PWR F.M RWR TRANSirNT MAN MONillS MAft mNillS Matt MONTils Raflood Ifeat Transfer P4 [S6) 15 Non-condensible Gas 8 Reactor Kinetics 1 [.2) 4"[gj Steady-State Initiali7ation 6 _

Hydrodynamics Maintenance 8 3 g j[@),- em /r

1. [

Numerical Improvements 6 Output Options 4

[/)) h) --

Fue1 Model 6 (/?) 3 (b)

Accumulator 2 General Angle Tee 2 [3) llorizontal Phase Separation 2 [6)

BE Test Calculations (FLECili, Semiscale, LOFT) 12[/?)'

EM End of Dypass Critorion 1 [il)

EM Test Calculations 12['f)

/

Servo Control loop 4 Jet Pump (RELAPS branching and Area Change) 4 Steam Water Separator 1 (S)

BWR Test Calculations 12 0f r - f/))'

TOTAI.S MM R1 fl/ 3) 3A [J/5 25 ['fl)

MAN YEARS 6.75(f f) 2.08[7f)

CllMULATlVE MY 6./S((1') 2.83 (S '/))

9. 5Fi [/7, / 11.66[N-ARLIEST AVAlt.ARLE DATE 8/79[jp/77)10/79[f8c') / 2/00b#Iof/

OMPUTER $

50K [/do/c) SOK (/co /<-)

These estimates are for code /roi< development o not and (Am) include user paration and developmental verification required for release of the respective code versions.

Assumes availabili ty of an existing and acceptable 1-D kinetics toul.ine that can be converted / adapted

APPENDIX TEST PROBLEM COMPARISONS Marviken Comparison Item TRAC RELAP-5 Run time (1) 3.9 min (1) 10.6 min with two nodes downstream of choke plane (2) 2.2 "

" (2) 4.2 min. with reduced (3) 2*6 backpressure (4) 1.2 "

(3) 1.04

-3 sec (1) 27858 accepted No. of time steps (1)17x10 (2) 37 (3)19 (4)49 Mass conservation  ? 0.08% (1)

No. of Nodes (1)60 (1,2) 20 vessel, 6 pipe 2 downstream of choke plane (2) 60 (3) remove 2 downstream of (3) 40 choke (4) 40 Early mass flux good good, possible overshost?

Pressure, vessel misses dip ~ 5% low at almost perfect, including mid BD, drops quickly t "" "

Bottom (M106) { p{,a aybe 1 er at 43 sec. mid-BD Overall massflux Misses initial valley, initial fit excellent,

- 15% low at 12 sec, - 6% low at 12 sec drops sharply at 42 sec Pressure 101 Misses initial valley, good good to 20 sec, above after(datalow) u.

. er * %

1. 'e"- * *

• g - - y a 4 g p s. , , gg ,

APPENDIX Continued Marviken Comparison Item TRAC RELAP-5 T 401 Misses dip, good to 42 sec, 2 high after dip drops early (1 high)

T 505 Misses dip, almost perfect almost perfect to 42 see T 510 misses dip, good to 42 sec good, - 2 high at end

(~ 2 high towardend)

T 515 Misses spike at 6 sec good Same as TRAC except OK for to 42 see when it drops t > 42 sec before data. Initiel rise slower than data.

T 419 Same as above except spike in data is at 12 sec.

T 402 TRAC & RELAP identical except t > 42 sec.

T 531 T 404 (RELAP~l higher)

T 532 P 216 Predicts reverse AP early with Follows data but always respect to data, then follows lower AP data but with lower AP than RELAP aP 220, 230 poor -

6P 202 -

Good aP 244- 246, RELAP shows large swings prior to passage of 114 level. Why?

4 I

i -

S-02-6 TRAC - RELAP COMPARISON Item Comments PV 10 T (TRAC) better near 100 sec, R (RELAP) toward end. R shows odd behavior near 100 see that isn't seen by data Mass Flow 23 T and R both start low, T drops slowly crossing data at - 70 sec and passing thru data again at 350 sec. T shows little correspon-dence to data. R drops sharply as does the data, but about 40 sec late. Shows excellent agreement to - 350 see where it jumps at the same time on the data, but goes significantly above the data (T missed most of this).

Mass Flow 13 .T better early (almost perfect) with R poor, RELAP good from 150 to 350 where TRAC slowly approaches data, both miss jump at - 350 sec.

Temperature TRAC RELAP D4-29 good to 250 sec, drops good to 400 sec, rises after.

rapidly after Spike at 200 sec not in data hrsund Cold Leg Fair to slightly high to Fair to good to 300 sec, spike at fM# 300 sec, then high 330 sec, then good, Orops drasti-cally at 500 see Broken Vessel Side Good to 60 sec, then high Good to 100, high to 250, to 200, then drops below in data to 500 l.

UP Poor to terrible -

Run Time 500 sec in 85 min, include 55 min (w/o junction error) steady state. for 600 se; 75 nodes, 74 min w/o ss

.N wy

18 -

S-02-6 Continued Temperature TRAC RELAP' 04-50 Starts high, crosses data Exhibits uncovering not seen D5-39 around 300 sec, in dat 8 Good at 380 sec D5-29 becomes too low. Only Some uncovering, but not as much D4-14 faic (small spike at 350 sec)

Cold leg p near Excellent to 300 sec. Drops Excellent following of behavior, vessel too slowly at 80 sec. Misses late on drop (120 vs@ )

chugging for t > 300.

Broken leg a although some of trend for t large is calculated.

Ho t l e,g y Excellent but misses noise Excellent

.e owl Near./;thF around 500 sec p - bottom of Good to excellent Excellent, a little noisy for DC, top of LP t > 300 sec Core mass flow - Generally excellent, rate occasional spike To get comparable results - TRAC nodalization must be increased with - 50% increase in run time. RELAP is about twice as fast on this basis. (Nodalization is not optimized.)

• • initial behavior generally under predicted

• • . gwgg og, ,4, ,g.g , g, gg ,, ,gg39 ,p

('*4dirhe W mh e 44 *'sw * ** = im w m 2 deqe ge oa ,

-_ . _ . _ _ _ . . - - ~ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ____

l ~ '

l i

! TRAC - FROM OUTPUT T VAPOR FRACTION t Cold leg Broken hot leg Bottom of DC/

D5-29 D4-14 D4-50 D5-39 leg i near vessel near vessel Lower Plenum Steady state reached in 3550 time steps, - 400 sec CPU time 0 820 920 739 0 0 0 0 20 733 799 693 0 0/.9(3) .3 0 90 584 590 580 .7 .6/.9(3) .7 0 120 576 583 573 .8 .8/.98(3) .8 0 60 588 596 585 .3 .~07 9(3) .6 0 200 560 566 566 .9 .9/.99(3) .98 0 280 536 542 533 1 1 1 .15/0 300 529 534 526 1 1 1 .19/0 400 480 485 477 1 .92/.99(3) 1 .24/0 490 455 460 453 1 .97/.87(3) 1 .15-0 A

l 26833 ats CPU = 4448 for run d

e k

k

S-06-3 TRAC - RELAP Comparisons Item TRAC RELAP-5 Pressurizer P Both codes 10% low at 20 sec, good otherwise Upper Plenum P 20% 16 sec Bk Hot Leg P " "

slightly low at 20 sec, good otherwise k Fluid T. cold leg 110 low at 20 sec, low 15 low at 20 sec, good to 80 sec, afterward, misses before reproduces data Noise during except t = 0, Not presented refill for t > 40 sec Lower core Not presented Excellent to 75 sec density -

Pump inlet Excellent Good density Intact cold g Ex e gnt to 20 sec, poor Excellent to 75 sec, good to leg density tog40 sec excellent during oscillations to 150 sec Bk p (coldleg) Not presented Excellent to good. Follows data oscillations very well for 90<t<l50 sec Core inlet Both fair to good. RELAP better t < 2 sec where TRAC is a factor mass flow of two off, RELAP on.

Intact cold Both Codes Good leg mass flow AP across Both codes significantly below data for first 40 sec.

simulated pump Hot leg break Both codes fair to good to 15 sec, good afterward ,

flow -ate Cold leg break Both codes good, RELAP initial flow better (30% error TRAC, flow rate 10% RELAP)

Item TRAC RELAP

&c. r Temperatures Ave. perfect to .10 c c, TRAC perfect to 12 sec, RELAP drops at Rod, Center continues upward to 925K, 20 sec and data do not drop signifi-stays there to 50 sec, & cantly. RELAP T max is 850K BD, decreases. Quench predicted lower reflood. Data -900K BD, 125 sec, data at 150 sec 950K RF.

Hot Rod Center 80 peak : 1100K, data : 1150 RELAP BD peak : 1200K, (D5-29 not for 05-29, 1050 for D4-29 not shown), good to 65 sec, then (not shown by-TRAC), TRAC over data to 150 sec. _

good for remainder, slightly -

late on quench Run time (min) to 150 sec 135.4 367.2 30 sec 24.5 49.5 60 sec 61.2 107.9 1

l 100 sec 108.1 224.8 10 sec 11.5 13.2 20 sec 18.2 28.1 No of nodes 112 100 (18 each nozzle, 15 SG 24 vessel, 54 on RELAPbasis.)

h e

_ . _ _ . _ _ _ _ . _ _ . . _ _ _ _ _ _ _ _ _ . . _ . _ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __. _