ML20206D425

Rev 0 to Ltr Rept, Technical Review of Gpu Nuclear Three Mile Island Unit 1 Cable Functionality Assessments

ML20206D425
Person / Time
Site:	Three Mile Island
Issue date:	02/26/1998
From:	Nowlen S SANDIA NATIONAL LABORATORIES
To:	Ronaldo Jenkins NRC (Affiliation Not Assigned)
Shared Package
ML20206D405	List: ML20206D400 ML20206D414 ML20206D420 ML20206D425
References
NUDOCS 9905040106
Download: ML20206D425 (33)

Text

-

A Technical Review of the GPU Nuclear Three Mile Island Unit 1 Cable Functionality Assessments A Letter Report to the USNRC February 26,1998 l Revision 0 l

l Prepared by:

Steven P. Nowlen Accident and Consequence Analysis Dept.

Sandia National Laboratories Albuquerque, New Mexico 87185-0748

)

Prepared for:

Ronaldo Jenkins Electrical Engineering Branch Office of Nuclear Reactor Regulation U. S. Nuclear Regulatory Commission Washington,DC 20555 USh7C JCN J-2503

~~

9905040106 990420 PDR ADOCK 05000289 ATTACHMENT 1 F PM s

e 1

4 kT .

FORWARD I

The United States Nuclear Regulatory Commission (USNRC) has solicited the suppon of Sandia l

National Laboratories (SNL) in the review oflicensee submittals associated with fire protection i

and electrical engineering. This letter repon represents the second repon in a series ofreview repons associated with cable functionality assessments from the GPU Nuclear Corporation for the -

l Three Mile Island Nuclear Generating Station Unit 1 (TMI-1). This review effon was initiated in response to a licensee submittal ofDecember 1996 that documented analyses of a number of specific fire barrier elements for compliance with the fire endurance performance criteria of Supplement I to Generic Letter 86-10 based on analytical assessments of cable functionality j

under extrapolated fire exposure conditions. SNLs initial review of that submittal was competed in June of 1997. Based in part on this SNL review, in August 1997, a RAI was forwarded to the l licensee by the USNRC. The current repon documents SNL's findings and recommendations l resulting from a review of the licensee's response to the USNRC RAI as documented in a licensee l

submittal ofDecember 1997. This work was performed under Task Order 4 of USNRC JCN J-2503. ,

l l  !

l1 l

I

)

iii

TABLE OF CONTENTS:

section P. age FORWARD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii 1.0 l INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I 1.1 Obj ectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I 1.2 Limitations to the Current Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Organization of Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.0 B ACKGROUND DISCUS SIONS . . . . .$ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1 Overview of the Licensee Assessment Approach . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 An Underlying Issue . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . .. . . .5 3.0 THE RAI AND THE LICENSEE RESPONSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3.1 RAI Item 1: The Data Extrapolation Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3.2 RAI Item 2: Burn-Through . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 l 3.3 RAI Item 3a: Cable Damage Thresholds . . . . . . . . . . . . . . . . . . . . . ... . 15 3.4  !

RAI Item 3b: Circuit Performance Requirements . . . . . . . . . . . . . . . . . . . . . 16 j i 3.5 RAI Item 3c: Power Cable Self-Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 '

l 3.6 RAI Item 3d: Adequacy of the 57 Minute Barrier Rating . . . . . . . . . . . . . . . 17 4.0

SUMMARY

OF FINDINGS AND RECOMMENDATIONS . . . . . . . . . . . . . . . . . . 19 i

5.0 REFERENCES

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 Appendix A: An Assessment of the Licensee Extrapolation Method . . . . . . . . . . . . . . . . . . . 23 i

l IV *

1.0 INTRODUCTION

1.1 Objectives As a part of the overall resolution of fire barrier concerns usociated with the fi Thermo-Lag @ 330-1 at Three Mile Island Nuclear Generating Station, Unit 1 (T December 31,1996 GPU Nuclear submitted a request to the USNRC for exempt nominal fire barrier fire endurance performance criteria established in Supplement I t Letter 86-10. The exemption request covers a specific subset of the installed fire barr TMI- 1, namely, certain 1-hour conduit and armored cable barriers. It is based on us functionality method of test evaluation for cenain test items from the Nuclear E (NEI) generic fire barrier testing program that failed the nominal temperature rise cri ASTM testing standards. Sandia National Laboratories (SNL) was asked by the USN review and assess the cable functionality related ponions of this licensee submittal an j

recommendations regarding the technical acceptability of those analyses.

{

SNL completed an initial review of the 12/96 submittal in June 1997.8 Based in pan on this SNL review, by letter dated August 19,1997, a RAI was forwarded to the licensee by the The current repon documents SNL's findings and recommendations resulting from ajl the licensee's response to the USNRC RAI as documented in a licensee submittal ofDec 30,1997. J This work was performed under Task Order 4 of USNRC JCN J2503. The documents re under the current effon are:

l Letter, James W. Langenbach, GPU Nuclear /IMI-1, to the USNRC Document Control Desk, Item 6710-97-2489, Dated December 30,1997 (with four attache Note that to suppon this review, SNL requested and was provided access to a non I version of the following Nuclear Energy Institute (NEI) document:

NEl Application guidefor Emluation of Thermo-Lag 330 Fire Barrier Systems, Document number 0784-00001-TR-02, Revision 1, Volumes 1 and 2.

This NEI document (hereafter referred to as the "NEI Application Guide") was cited e by the licensee in its assessments. It should be noted that the SNL effons were not in and have not, provided for a comprehensive review of this NEI document. SNL's object obtaining this document was simply to allow us to verifv the licensee references and inference cited as deriving from the NEI Application Guide.

2See SNL letter repon to the USNRC "An Initial Review of the GPU Nuclear Three Mile Island Unit 1 Cable Functionality Assessments," dated June 25,1997.

I

1.2 Lin~tations to the Current Review This SNL review is limited to those ponions of the licensee submittal directly assoc those cable functionality assessments that are being forwarded by the licens justifying an exemption from the nominal fire endurance test acceptance criteria es Supplement I to Generic Letter 86-10. These functionality assessments are relatively small subset of the licensee installations, and the related analyses c ponion of the original licensee submittal. These assessments did, however, r focus ofthe 8/97 USNRC RAI.

In particular, the licensee submittal includes extensive discussions of the fire hazards with specific fire areas / zones that are not directly Felevart to the functionality asses themselves. SNL has not attempted to assess the adequacy nor technical merits fire hazards assessments.

As a final note, SNL has not attempted to assess the acceptability of those fire barrier i that the licensee has chosen to upgrade so as to meet the direct nominal NRC per acceptance criteria. For these cases, the licensee's own assessments of the base installation demonstrated an inadequate performance margin, and the licensee has chosen barriers consistent with the NEI Application Guide. An assessment of the acce NEI Application Guide for this purpose is beyond the scope of these effons, and rema purview of the USNRC.

1.3 Organization ofRepon Section 2 of this repon provides a brief background discussion including an overview o licensee's approach to its cable functionality assessments. Section 3 provides review of the licensee RAI responses. Section 4 provides a summary of the SNL recommendations. Section 5 cites the referenced documents. Appendix A provides a more detailed review of the licensee calculation in which the test data from encompass a full 60 minute exposure period.

I 2 .

- l

2.0 BACKGROUND

DISCUSSIONS 2.1 Overview of the Licensee Assessment Approach The licensee has performed a review of all ofits installed fire barrier systems to address the concerns raised by the USNRC in Generic Letter 92-08. For each fire banier, one of three approaches to resolution has been pursued. This was noted by SNL in our review of 6/97, and -

remains unchanged for the licensee's most recent submittal. The three paths are summarized as follows:

For many nominal 1-hour installations the licensee has chosen to upgrade the base installation to achieve a minimum 1-hour fire endurance rating fully consistent with the USNRC and ASTM /NFPA test acceptance criteria. This apparently includes all of the 1-hour cable tray installations. For these installations, no exemptions are being requested because the installations will be upgraded to comply with the NRC acceptance criteria based on the guidelines of the NEI Application Guide.

For at least some of the installed nominal 3-hour barriers the licensee has not been able to demonstrate a 3-hour fire endurance rating but has demonstrated s ininimum 1-hour ASTM /NFPA fire endurance rating for the base installations. The licensee proposes not to upgrade this performance to achieve the full 3-hour rating, but instead, is requ-sting an exemption from the Appendix R requirements for protection of the fire area by automatic '

suppression in lieu of 3-hour barrier separation.

For a limited subset of fire barriers the licensee has not been able to establish an ASTM /NFPA fire endurance rating of at least one hour for the base as-installed system, 4 and proposes not to upgrade the barriers to achieve the desired fire endurance rating.

Rather, for these cases the licensee has invoked an altemate cable functionality based performance assessment approach to demonstrate the equivalent of a one-hour " cable qualification rating." For these applications the licensee is requesting an exemption from the nominal ASTM /NFPA fire performance evaluation criteria in favor of the modified barrier ratings derived from the functionality assessments.

Of these " paths" only the third and last is ofinterest to the current review; namely, the path in which the " cable qualification rating" is assessed based on functionality arguments. The licensee has followed the " functionality path" for fire barriers in ten specific fire zones. Some of these zenes contain numerous individual barrier systems. All of the cable functionality evaluations are associated either with cables in conduits (including straight sections of conduits, radial bends, and condulets) or with armored cables. No functionality assessments for cable trays have been put forth by the licensee. One additional commonality is that all of the functionality assessments have been based on the extrapolation of test results from a single fire endurance test; namely, NEI Test 2-1.

It is important to note at the outset that the licensee is drawing a clear distinction between the

" actual rating" or " fire endurance rating" and the " cable qualification rating" (CQR) of the fire barriers. The " actual rating" or " fire endurance rating" is that rating that derives from a direct compliance with the USNRC and ASTM /NFPA test acceptance criteria. In contrast, the CQR is 3

a licensee construct based on an assessment of cable functionality during an expos beyond that at which the nominal test acceptance criteria were exceeded. SNL will also us same terminology in this review.

One point to note is that for those barriers considered in the functionality assessments, the conresponding " actual rating" is based only on the nominal single point temperature rise criteria established in the ASTM E-119 and NFPA 251 fire test standards (325'F No consideration has been ;given to the average temperature rise criteria (.250'F above a this assessment. The licensee argues that the single point criteria is most relevant because this will be most characteristic of anticipated cable failures. In the specific context of the function assessments SNL agrees with this reasoning. That is, single point cable failures are most likely; hence, consideration of the single point high tem;ierature is an appropriate basis for the functionality assessment. However, some care should be taken to avoid confusing this " actu rating" with the true ASTM /NFPA fire endurance rating. ASTM /NFPA clearly requires that b

• the single point and average temperature criteria be met. The licensee's assessments bas on the single point criteria may be more or less conservative, but is not fully equivalent in all

- cases. This observation has no significant impact on the technical validity of the functionality assessments and is noted only in the interest ofclarity.

The licensee's overall functionality assessment process has not changed significantly as a resu the USNRC RAI although some details of the assessments have been adjusted to account for the concerns raised ir the RAI. As was noted in SNL's 6/97 review, this overall process can be summarized as follows:

The licensee has estimated the " Cable Functionality Temperature" (CFT) for its cables.

This is the temperature at or below which a cable is assumed to be capable of perfonning its intended function. In the original submittal it appeared that a value of 698'F (370*C) was used in all cases. The most recent submittal indicates that one of four values has used. For power cables one of three CFT values has been applied, ranging from 299'C to 396*C, depending on the cable material. For instrument cables, a damage threshold of 150*C was uniformly assumed. The values are based on the results of SNL testing of cables as reported in NUREG/CR-5546 and a relatedjournal article [1,2).

The licensee has compared its installed, non-upgraded fire barrier systems to those described in the NEI Application Guide. The licensee has concluded that the installations used in NEI test 2-1 are representative of all of the conduit and armored cable base installations as originally implemented by TMI-I and for which functionality-based exemptions are being requested. This one NEI test that is then used as the basis for all of the subsequent cable functionality evaluations.

The licensee has taken the NEI test 2-1 data and made a direct assessment of the barrie

" actual rating" as described above. In all cases relevant to the functionality-based exemption requests, these " actual" endurance times are less than the nominal 1 hour1.157407e-5 days <br />2.777778e-4 hours <br />1.653439e-6 weeks <br />3.805e-7 months <br /> rating of the TMI-1 installations. In particular, in NEI test 2-1 the conduits exceeded the nominal temperature criteria in 27-50 minutes depending on which of the individual conduits is considered (larger conduits performed better).

4

1 In ceder to assess the cable functionality at 60 minutes, the licensee has performed a full j ualytical simulation of the anticipated cable thermal response for the fire barriers tested in NEI Test 2-1. This includes an estimation of the performance beyond the point at which the test was, in reality, stopped. That is, NEI test 2 1 was terminated after a 50 minute exposure and no temperature data beyond this time is available. For the purposes of functionality analysis the licensee has analyzed the fire barrier thermal perfonnance for conduits from both NEI Test 2-1 and NEI Test 1-6, and has used the " average" behavior from these two tests to perform thermal simulation of Test 2-1 for a full 60 minute exposure. The result is an estimate ofwhat the temperature response might have been had the test been continued for a full hour.

The resulting temperature response is then~ compared to the CFT. The comparison includes the consideration of both plant ambient and the initial cable operating temperature (an area ofimprovement in the current submittal). Based on this comparison, the licensee determines a duration time that it defines as the " Cable Quali6 cation Rating"(CQR). That is, the CQR is the time at which the extrapolated thermal response exceeds the CFT. This represents the licensee's assessment of when installed cables would have actually reached a potential failure thredold dudng the fire test had such measurements been made and had the test been extended in time to one hour or beyond.

For those cases where a CQR of 57 minutes or greater is determined, the licensee requests an exemption for acceptance of the cable functionality or CQR calculation as an appropriate basis for compliance with the Appendix R requirements.

In each case, the assessments are also supported by a discussion of the local fire hazard conditions (these hazards assessments have not been specifically reviewed by SNL).

As can be seen, the licensee assessment is relatively simplistic. In essence, the method " boils down" to a comparison of a damage threshold temperature to the extrapolated thermal response of the NEI Test Article 2-1. Effectively, the licensee is assuming that so long as the thermal response of test anicle does not exceed this threshold temperature, then the performance of the circuits would be assured.

The specifics of these assessments were provided as Attachments 1 and 2 of Enclosure B in the licemee's original 12/96 submittal. They appear in the form of an extensive set of 6re barrier

" element" evaluations. An " element" in this context would, for example, be a straight section of cable tray, a straight section of conduit, or a conduit radial bend section (there are 34 three-hour elements and 30 one-hour elements identified in these assessments). In these " element" evaluations the licensee has summarized its assessments of the " actual rating" ofits installed fire barriers as well as the functionality assessment results. SNL has not attempted to review all of these individual element evaluations, but rather, only those " elements" which are actually cited in the specific functionality-based exemption request ponion of the submittal.

2.2 An Underlying Issue In its 6/97 review, SNL recommended that the USNRC determine "as a matter ofpolicy" whether or not the extrapolation of a fire test beyond the period of performance for that test was an 5

acceptable practice in principal, particularly in the context of the fire banier material Thermo-Lag @ and the known vulnerabilities of that material. The USNRC RAIitem I included the '

following statements by the USNRC:

"Further, although Supplement 1 of GL 86-10 states that cable functionality assessments can be made if they are ' based on a comparison of the fire barrier internal temperature profile measured during the fire endurance test to existing cable specific performance data', altemative methodologies can be used provided that the technical basis can be established for the assessment of the thermal exposure conditions."

SNL interprets this passage as answering this " policy" question in the affirmative. That is, i

extrapolat an offire test data is acceptable in prin'cipal, provided that an adequate basis for analysis is c;*ablished. SNL has proceeded with the current review on this basis.

i i

6

3.0 THE RAI AND THE LICENSEE RESPONSES 3.1' RAI Item 1: The Data Extrapolation Basis 3.1.1 Summary of SNL's Initial Review Findings In our initial review of 6/97, SNL found that the licensee had not established an adequate basis for" the extrapolations of fire endurance test data beyond the actual test period of performance. These extrapolations were cited by SNL as the " lynch pin" of the licensee functionality assessments, and the acceptability of these extrapolations was critical to the overall acceptability of the licensee -

assessments. It was recommended that the licensee be asked to provide and explain the supporting calculations, to discuss the validation of the method, and to address the issues of '

material bum-through.

3.1.2 Synopsis ofthe USNRC RAI The USNRC RAI reiterated SNL's findings, and requested the licensee to provide a copy of the supponing calculation and "a detailed explanation of both the uncertainties and conservatisms in these extrapolations. Specifically, the licensee should address the following points:

How has the analysis treated the potential for material burn-through in extrapolating the NEI test data assessment?  !

I On what basis has the licensee tnermal extrapolation model been validated?

Documentation of the extrapolation model should be in detail so as to allow independent implementation and verification. The licensee is requested to provide l the conesponding validation results and studies for staff review." i 3.1.3 Synopsis of the Livnsee Response i

The licensee has provided the subject calculation for review. In addition, the licensee has provided significant supplemental discussion of the model, its basis, its validation, ud the issue of material burn through.  !

With respect to the issue of material burn-through, the licensee cites two arguments for dismissing this potential. First, the licensee cites that:

"there is no evidence from the NEI 2-1 Test TC temperature data, post-exposure test observations, or from calculated Thermo-Lag barrier thermal performance data that leads i to suggest that eitner of the 2" and 4" aluminum conduit specimens in this test were about i

to experience a barrier ' burn-through' during the remaining 10 minutes of the test."

Second, the licensee cites that in NEI Test 1-6, a similar but upgraded barrier was tested. The licensee calculation included consideration of this test as well, and concluded that the fire barrier performance in the two tests was "very similar." Test 1-6 did proceed for the full 60 minute exposure, and no burn-through was observed. By implication, the licensee appears to be citing 7

i

b this as evidence that burn-through would not have occurred in Test 2-1 had the test continued for 60 minutes as well.

For validation, the licensee refers to the information provided in the calculation itself. The licensee also cites an internal process of model verification through independent implemen of the same model using an alternate coding and comparison of the results.

2.1.4 SNL Assessment of the Licensee Response SNL has reviewed the licensee supporting calculation as detailed in Appendix A below. Those portions of the calculation related to the direct analysis of the NEI test data were found to be of high quality, were well thought out, and were well executed. This portion of the analysis was, i fact, rather sophisticated in its treatment of the dominant heat transfer mechanisms. In parti it included extensive consideration of radiation heat transfer behaviors; a very dominant factor in an ASTM Sre endurance test. The model was exercised by the licensee on a total of seven conduits (three from NEI Test 2-1 and four from NEI Test 1-6).

However, SNL has identified three specific concerns related to the manner in which the licensee has utilized the results of the data analysis and has simulated the performance of Test 2-1. In considering the following findings, note that the licensee is not extrapolating the existing test data for NEI Test 2-1, but rather, is actually using an " averaged" thermal response from the combined performance of both of the NEI Tests 1-6 and 2-1 as the basis for a full simulation of the NEI Test 2-1 thermal response. The following speci6c concerns were identified regarding the calculation process itself:

The licensee extrapolations for the 2" conduit in NEI Test 2-1 was based on use of the average net thermal conductance for the 2" conduit in Test 2-1 and the two 3" conduits in Test 1-6. Similarly, the 4" conduit in Test 2-1 was simulated on the basis of the average thermal performance of that conduit and the 5" conduit from Test 1-6. It is well known that for a given barrier system, larger conduits perform better than smaller conduits.

Indeed, the licensee data analysis clearly illustrates that there is a sharp " time <

compression" effect apparent in the performance of smaller conduits in comparison to the performance oflarger conduits. This was identified by examining a number of characteristic behaviors in the conductance curves, and comparing the time at which these ,

behaviors were observed (see discussions in Appendix A below). This is also reinforced  !

by test experience from TUE and TVA.

In the specific case of the 2" conduit, the licensee data analysis shows a clearly

. inconsistent behavior as compared to the 3" conduits that were included in the averaged thermal response behavior estimates. Indeed, the averaged thermal conductance values fail to reflect the steadily rising trend in the thermal conductance of the 2" conduit observed during the last 11 minutes of Test 2-1.

In the specific case of the 4" conduit, the licensee did demonstrate some consistency in the long time behavior of the 5" conduit. However, the early time >

' behavior of the 4" conduit showed significant inconsistencies in comparison to the 5" conduit that are masked by the averaging process.

8

Based on the known impact of conduit size on barrier performance, and on the results of the licensee data analysis, SNL finds that it is inappropdate to predict the behavior of a smaller conduit based on the measured performance of a larger conduit.

The licensee extrapolations for the two conduits ofinterest in Test 2-1 are based on thermal conductance values averaged with conduits from Test 1-6. Indeed, since Test 2-1 ended at 50 minutes, during the actual extrapolation period the analysis is based entirely on the performance ofthe conduits in 1-6. NEI Test 1-6 was a test ofan upgraded barder system and included upgrading with both " stress skin" and trowel grade matedal at all joints between abutting sections o'the fire barrier material. The licensee cites that "the test No.1-6 upgrades are of a structural nature and do not aff'e ct the thermal performance of the barrier." SNL disagrees with this cbnclusion. That is, while the upgrades might be characterized as " structural" in nature, they do profoundly impact the thermal behavior.

This is danonstrated, for example, by the very fact that NEI Test 1-6 passed the exposure test successfully while the non-upgraded barriers of Test 2-1 all failed in less than 50 minutes. There is a simple explanation for the importance of the " structural" upgrades to thermal performance:

One of the two primary mechanisms of failure for Thermo-Lag @ fire barriers is related to the behavior and integrity of materialjoints.2 These failures can be attributed to two separate mechanisms. First is the simple structural integrity of thejoint itself. That is, testing reveals that thejoints between butted panels will tend to open unless protected. This can, in an ofitself, lead to a failure by allowing the fire exposure environment direct access into the protected envelope.

However, a second critical behavior is also at work. That is, as thejoint begins to open, the butt-end of the material panel will be directly exposed to the heating of the furnace. Hence, the material is " thermally attacked" not just from the exposed outer surface inward, but also along the exposed butt-end. This leads to a "2-sided attack" that can degrade the barrier adjacent to the butt joints much more rapidly than the degradation observed in locations remote from the joints. Both of these behaviors will have a profound impact on the thermal behavior of the fire barrier system. In the context of the licensee calculations, either process would be i reflected by a sharp and sudden increase in the thermal conductance.

Hence, SNL finds that the " structural" upgrades to NEI Test 1-6 clearly had a profound impact on the thermal performance of that test article in comparison to the thermal performance of Test 2-1. Hence, SNL finds that the licensee practice of simulating the performance of a non-upgraded barrier system, Test 2-1, based on the performance of an upgraded barrier system, Test 1-6, is inappropriate.

.- The thermal model used to estimate the conduit-to-cable heat transfer process was ,

optimistic and unnecessarily crude. In particular, the licensee has assumed that the cable '

and the conduit are fully separated by an annular air gap. In reality, the cable will rest on the inside bottom of the conduit, and will make significant direct contact which would ,

enhance the heat transfer rates significantly. This is also especially true in bend areas that !

are known points of performance weakness. Alternate and more reasonable methods of The second primary mechanism is the lack of sufficient thickness away from the joints to survive the full exposure, which is not impacted directly by the Test 1-6 upgrades.

9 -

i 1 1

t i I analysis are readily available to the licensee (see for example the works of Neher, et. al.

[4,5]). The licensee treatment underestimates the rates of heat transfer from the conduit to the cables; hence, it is optimistic. Ultimately, the difference appeared to be an underestimation of the thermal response by between 10*F and 25'F (the exact impact cannot be clearly discerned because it seemed to vary from case to case and was also mixed with other interacting effects).

In addition, the licensee was asked to specifically address the issue of material burn-through.

SNL cited in its 6/97 review that joint failures and fire barrier burn-through are not amenable to predictive analysis. SNL reiterates this earlier finding. Funher, SNL finds that the questions of material burn-through and joint failure are inherently outside the scope of the licensee calculation.

Hence, these issues must be addressed separately: The licensee has provided an extensive discussion of the topic. The initial discussion focuses on semantics and the definition of what actually constitutes a burn-through:

For the purposes of this discussion, SNL accepts the licensee definition ofmaterial burn-through; namely, burn-through is the condition where both the virgin material and char layer has beca breached exposing the test article. SNL acknowledges that the role of the char layer is important to the material performance and does provide some insulating protection. Its role in the overall process of the fire barrier's protection should, indeed, not be neglected.

The broader question remains; namely, what assurance is provided that burn-through would not have been observed had Test 2-1 been continued for the full 60 minutes. This questic:: remains critical because if a burn-through is realized, then the licensee thermal calculation would no long apply. This condition lies inherently outside the scope of the calculation as noted above. As discussed in 3.1.3 above, the licensee conclusion of dismissal is based on~ (1) the lack of a bum-through in NEI Test 1-6, and (2) a lack of evidence ofimpending bum-through in the data for NEI Test 2-1. With regard to these two arguments:

SNL finds that the lack of burn-through orjoint failure in Tests 1-6 is not relevant to the evaluation of that potential for Test 2-1. In particular, the " structural" upgrades implemented in Test 1-6 would have a clear and profound impset on the potential for burn-through to be observed. As discussed above, there is a clear and well known relationship between joint integrity and the potential for either a direct joint breach or a burn-through adjacent to ajoint. The upgrades applied in NEI Test 1-6 were aimed directly at these known joint vulnerabilities. Hence, SNL finds that the lack of a burn -

through in Test 1-6 has little or no relevance to the question of bum-through for Test 2-1.8 Further, the licensee assessments of the thermal conductance for Tests 2-1 and 1-6 show clear inconsistencies between the two tests. SNL finds that these inconsistencies render the comparisons of bum-through potential for the two tests highly suspect. On this basis, SNL finds that the lack of burn-through during NEI Test 1-6 is not sufficient to i

2At most it might be argued that NEI Test 1-6 provides some indication that burn through away from thejoints was unlikely in NEI Test 2-1 extrapolated to 60 minutes, but the question of j

behavior near thejoints is not resolved by this test. i 10

r ..

1 conclude that burn-through would not have occurred during the NEI Test 2-1 had the test i

been extended for an additional 10 minutes.

i SNL finds that the lack of and actual burn through during the 50 minutes of Test 2-1 l provides no clear assurance that a burn-through would not have been realized during th l

50-60 minute period. It is SNL's assumption that the " burden ofproof" lies with the l

licensee rather than with the USNRC. Hence, it is incumbent upon the licensee to prov l

positive evidence that burn-through (including the consideration of thejoint effect on l

burn-through discussed immediately above) or an actualjoint failure would not occur. If, l

as recommended immediately above, the comparison to Test 1-6 is not credited, then the licensee is left with only a citation of a lack of evidence that burn-through was immanent.

l SNL does not find this assenion to be equivalent to a positive assenion that the undesired behavior would not have occurred had the test continued.

• In an attempt to more fully explore this concem, SNL has examined the available test repor might shed light on this behavior. SNL cites three sources of added information that lead us to

. conclude that burn through cannot be dismissed for the 2" and 4" conduits in NEI Test 2-1:

One such source is the NEI Test descriptions themselves as was cited by SNL in its review of 6/97. The NEI Test 2-1 report cited that several locations had no remaining uncharred L

' barrier material present. While this does not constitute an actual burn-through, this does indicate that the only protection for these regions was provided by the char layer. SNL cited this as evidence that burn-through was all but inevitable given enough time. The unanswered question, of course, is how much time would be requ* red to actually realize a burn through. SNL once again cites this result as indicating, at the least, a very marginal performance for the tested fire barriers in Test 2-1.

A second source ofrelevant insight is realized by a comparison ofthe thermal conductance profile for the 3/4" conduit and 2" conduit in NEI Test 2-1 (see Appendix A for a more I complete discussion). That is, the licensee analysis included an examination of the 3/4" conduit from Test 2-1, and this conduit di.! expedence an actual burn-through as reported I

by the testing laboratory. The behavior of the thermal conductance for this case clearly reflects the burn-through. That is, for a 12 minute period between 29 and 41 minutes the l

' estimated thermal conductance was rising steadily from its " post-plateau" minimum value.

SNL sttributes this behavior to breakdown of the char layer. In the case of the 2" conduit, a very similar behavior is noted during the last 11 minutes of the test. That is, between 39 and 50 minutes, the estimated conductance for the 2" conduit was observed to be increasing steadily. For the 3/4" conduit ber;ca 41 and 43 minutes the conductance value rises sharply and suddenly to a very high value before the conduit temperature data i was lost. This is a clear indication of the onset of burn-through. For the 2" conduit there!

was no corresponding indication of an actual burn-through. However, given the similarity l between the end of test behavior of the 2" conduit and the pre-burn-through behavior of l

the 3/4" conduit, it would appear that such a burn-through was indeed immanent.  !

A third source ofinsight can be derived from an examination of the Texas Utilities Electne j

(TUE) tests, in particular, TUE Test Scheme 9-3 [3). This test involved testing of 0.75",  ;

1.5" and 2" steel conduits 'vith installed fire barrier systems that appear nominally identical l

11 '

to those tested by NEI in Test 2-1. Note in particular the following banier installation descriptions from each repon. From the TUE report:

"Each rigid conduit raceway was covered first prior to installing material on the conduit support members using 1/2 in. nominal thickness Thermo-Lag @ 330-1 Pre-Shaped Conduit Material except the 3/4 in. e conduit system which used 3/4 in. nominal thickness pre-shaped material as described below. Alljoint, seams and built-up areas were pre-caulked with 330-1 Trowel Grade Material and secured in place with stainless steel tie wire and metal banding matedal.

The UniStrut trapeze type support member was covered with Thermo-Lag @ Flat Panel material for a 9 in. distance extending from the closest Thermo-Lag @ Pre-Shaped section leaving the remaining UniStrut support steel surface unprotected 1 from the fire source.

Each raceway LBD fitting was covered with a flat panel material in a manner similar to an L-shaped box configuration. Alljoints were pre-caulked with 330-1 Trowel Grade Material and secured in place with stainless steel banding matedal.

The LBD " box" configurations were then upgraded as described below."

Note that the TUE upgrades only impacted the LBD (or condulet elbow) box enclosures and the abutting conduit sections within 2" of the box. Otherwise the barrier system was a nominal installation consistent with the original manufacturer recommended practices.

The upgrades did not impact the general sections of the conduit barrier in any way. Now consider the following description of the NEI test 2-1 installation:

"All portions of the conduits were covered with Thermo-Lag @ 330-1 1/2 in.

nominal thickness Pre-Shaped sections. Joints between sections were pre-caulked w',th Thermo Lag @ 330-1 Trowel Grade material. To allow for curvature, the sections installed on the radial bends were miter-cut into individual wedge shaped pieces and fit to the conduit radial bend. Stainless steel bands were installed on the radial bends such that one band was located in each miter-cut wedge section.

Stainless steel bands were installed around the remaining sections ofconduits within 2 in. of a top or bottom panel butt joint and at 12 in. intervals thereafter.

Once the conduit assemblies had been covered, the support members and conduit LBs were completely covered using Thermo 1.4g@ 330-1 V-Ribbed baseline panel material ... The panels were installed such that the side panels were sandwiched -

between the top and bottom panels, and thereby, placed into compression when the external banding was tightened. Alljoints and seams between panels were pre-caulked with Thermo-Lag @ 330-1 Trowel Grade materials. The panel covering the outside of the 6 in. LB (over the cover) was scored across its surface to accommodate the curvature of the LB. The support members and the LB box barrier envelopes were secured with stainless steel bands placed within 2 in. of a panel butt joint or the end of a panel and at 12 in. intervals thereafter."

Given these descriptions, it would appear that the barriers in these two tests were virtually identical. The major difference appears to be that (1) the TUE tests used steel conduits 12

p ,

and NEI used aluminum, (2) TUE upgraded the barrier covering the LBs or LBDs with

' added stress-skin and trowel grade whereas the NEI test used no upgrades, and (3) the TUE test used two condulet (or LBD) elbows whereas the NEI test used one LB) and one radial bend elbow section instead. Each of these differences should have to a better performance for the TUE test than that of the NEI test.' None-the-less, the TUE test failed to achieve the thermal performance objectives, and the USNRC has not to date accepted cable functionality arguments as an appropriate basis for acceptance of the TUE results. Of panicular note to the current discussion is the fact that all three of the TUE tested conduits, including the 2" conduit, did suffer burn-through as reponed in the TUE submittals. Given the similarities in both the test anicles and the installed fire barriers, SNL cites the observed bum-through for all conduits in TUE Scheme 9-3 as positive evidence of the same potential in the nominally identical barriers tested in NEI Test 2-1. At the very least this potential is clearly demonstrated for all conduits 2" and r, aller, but this clearly casts doubt on the likely performance of the larger conduits as v P. Unfonunately, the TUE tests did not involve conduits larger than 2".

To SNL's knowledge there have been no other tests of a similar nature submitted for USNRC consideration. TVA did test 3" and larger conduits using a nominal S/8" base material installation without upgrade, but did not test the non-upgraded 1/2" nominal banier material. Also Florida Power and Light did perform and submit ampacity tests of a base conduit banier system, and may have performed some baseline installation fire endurance tests as well. However, if the tests were performed, repons have not been submitted to the USNRC.

3.1.5 Findings and Recommendations SNL has reviewed the licensee calculation, information not available at the time ofour original l

' review in 6/97, and has iden'.ified three specific weaknesses in the licensee simulations / predictions of the thermal performance of NEI Test 2-1. Each of these items is found to compromise the integrity and reliability of the cited temperature response results. These weaknesses are:

inappropriate application of estimated thermal conductance factors from larger conduits to the simulation of smaller conduits, inappropriate application of estimated thermal conductance factors for an upgraded barrier system to a non-upgraded barrier system, and L an optimistic model for the conduit to cable heat transfer behavior.

Funher, SNL finds that the licensee has not provided an adequate basis for concluding that fire barrier burn-through can be dismissed for NEI Test 2-1. In contrast, SNL has cited passages -

from the NEI test repon, results of the licensee's own thermal analyses, and results from testing i

l

' Note in panicular that the licensee has stated that testing of aluminum conduits would conservatively bound the performance of a steel conduit due to the higher mass density of steel and its ability to absorb more heat with less temperature rise than aluminum. SNL has concurred with this conclusion. Also, the radial bend sections in the NEI tests were the focus of subsequent .

upgrade effons in later tests due to their relatively poorer thermal performance  !

13 -

by TUE that indicate that burn-through cannot be easily dismissed and may indeed have been l immanent, at the very least for the 2" conduit. On this basis, SNL recommends that the licensee l

response to this RAI item is inadequate, and that the licensee estimations of the thermal performance of the fire barriers in NE! Test 2-1 not be credited by the USNRC.

Note that the concerns identified here will be difficult, ifnot impossible, to address through a I refinement of the calculation. There is very little data available to support these assessments, and .

the licensee has already made extensive, and inappropriate, use of the available NEI data. The only other known test of a similar non-upgraded 1/2" nominal conduit barder system was TUE Scheme 9-3, [3). The conduits in that test also failed to pass the temperature rise acceptance criteria, all of the conduits experienced actual burn-through as reported by the testing laboratory, and to date the USNRC has not accepted TUE's cable functionality assessments for that test as an alternative acceptance criteria. Hence, direct consideration of the data from this test can likely only hamper the neensee position. Further, SNL reiterates our earlier findings that the onset of material burn-through orjoint failure is currently not amenable to predictive analysis. Hence, SNL does not expect that further licensee interactions would be helpful in resching the identified concerns.

SNL does note that at least two altematives are readily available to the licensee. That is, the I licensee could either upgrade the barriers to achieve compliance based on thejoint upgrades of NEI Test 1-6, or perform a confirmatory test to resolve the uncertainty. Given the nature of the applications, testing could be limited to a single test using : sufficient number of conduits to bound the licensee installations. Note that given the existing test results, none of the conduits l

would be expected to pass on temperature rise alone; rather, the ultimate pass / fail assessment will  !

likely depend on cable functionality arguments. Hence, SNL recommends that in the event that tests are proposed, the USNRC request that the licensee use representative plant cables and perform cable functionality (insulation resistance) measurements during the fire exposure to

)

eliminate any uncertainty regarding cable performance. SNL also considers it unlikely that

{

conduits of 2" or smaller would survive the exposure without burn-through. The performance of larger conduits remains indeterminate. l 3.2 RAI Item 2: Burn-Through 3.2.1 Summary of SNL's Initial Review Findings SNL had cited burn-through as a specific problem with Thermo-Lag @ not amenable to analysis.

Passages from the NEI test report were cited as supporting SNLs concern for the potential adverse impact of burn-through on the licensee assessments.

3.2.2 Synopsis of the USNRC RAI The USNRC reiterated SNL's concerns in this regard and requested that the licensee " discuss the post-test physical inspection results for the relevant test items from the NEI test report and specifically to address the implications of the statements .... with regards to the application of the NEI test results for extrapolation purposes such that the subject cables will be able to continue to  ;

function during the full one hour period."

14

[ ..

3.2.3 Synopsis of the Licensee Response The licensee cites it response to RAI item 1 (see 3.1 above), and reiterates the arguments put fonh in response to that question. The licensee agam states that observations such as "O inches of unburned material remaining" are not indicative of a burn through and that the char layer is an imponant factor in barrier performance. The comparison to NEI Test 1-6 is also re-stated.

3.2.4 SNL Assessment of the Licensee Response l

As noted in SNL's assessment of the licensee response to RAI item I in 3.1.4 above, the licensee is correct that the lack of unburned material in and ofitselfis not equivalent to a burn-through.

SNL, however, stands by its earlier finding that th'e lack of any remaining unbumed material is at the very least indicative of a very marginal barrier performance even given the 50 minute test time period. With no un-burned material remaining, burn-through is inevitable, and only the time of the burn-through remains uncenain. SNL also has noted that the timing of a material burn-through is currently not amenable to analysis. Funher discussion on this issue is provided in 3.1.4 l above and in Appendix A below.

3.2.5 Findings and Recommendations As discussed in 3.1.5 above, SNL finds the licensee discussion of burn-through to be inadequate l

to resolve SNL's concerns in this regard. Given the critical impact of this issue on the reliability

) of the licensee's calculation, SNL recommends that the licensee thermal performance simulation l results not Le accepted as an adequate basis for assessment of barrier performance in the 50-60 minute time peiiod.

3.3 RAIItem 3a: Cable Damage Thresholds l 3.3.1 Summary of SNL's Initial Review Findings SNL had cited that the licensee had not established an adequate basis for the cable thermal damage threshold used in the calculations. In particular, the supporting memorandum was not provided in the submittal.

l 3.3.2 Synopsis of the USNRC RAI ,

The licensee was requested to provide the cited memorandum.

3.3.3 Synopsis of the Licensee Response 1

The licensee has provided the cited memorandum.

3.3.4 SNL Assessment of the Licensee Response The licensee memorandum makes clear how the damage thresholds were established. The values cited in the table are appropriate to this purpose. In panicular, SNL notes that the values have 15 -

included consideration of the cable insulation material, and that a more stringent performance requirement was established for instrument circuits.

3.3.5 Findings and Recommendations l

I SNL finds that the licensee has given adequate and appropriate consideration to the cable material in establishing the cable thermal damage thresholds. The licensee response was fully adequate to resolve the identified concerns. No further actions on this RAI item are recommended. '

3.4 RAIItem 3b: Circuit Performance 'quirements

)

3.4.1 Summary of SNL's Initial Review Finding's SNL had cited that the licensee did not establish an adequate basis for the criteria used to assess the functionality requirements of a given circuit.

3.4.2 Synopsis of the USNRC RAI The licensee was requested to characterize the vohage, current and function of each circuit analyzed, and to provide an assessment of the circuit performance requirements for both a limiting power and limiting instrumentation circuit.

3.4.3 Synopsis of the Licensee Response The licensee cites that there are no instrument circuits impacted by the exemption request. The cited damage thresholds are applied to " low to medium voltage" power and control cables ranging from 600-5000V. The liccnsee cites that these voltage levels are consistent with the SNL test findings.

3.4.4 SNL Assessment of the Licensee Response >

The licensee response does address the identified concern. The cited nominal damage criteria are appropriate for use in low-to-medium voltage power and control applications, and the licensee cited limit of 5kV is a reasonable upper bound on this classification. SNL also notes that the licensee response to RAI item 3a also demonstrated a clear understanding of the concerns raised by SNL. In particular, the licensee memorandum requested in RAI 3a did include a discussion of a far more restrictive performance criteria for instrument circuits, even though ultimately no such circuits were included in the exemption request.

-3.4.5 Findings and Recommendations SNL finds that the licensee has given adequate and appropriate consideration to circuit performance requirements in establishing the cable thermal damage thresholds. The licensee response was fully adequate to resolve the identified concerns. No further actions on this RAI item are recommended.

.i 16 -

l i

i

3.5 RAI Item 3c: Power Cable Self-Heating 3.5.1 Summary of SNL's Initial Review Findings SNL had cited that the licensee had not proivided adequate treatment of the power cable self-heating effects, and that the licensee assumptions in this regard were inconsistent with the licensee's own ampacity assessments. -

3.5.2 Synopsis of the USNRC RAI i

The licensee was requested to provide additional information on the analyzed circuits and to more

~

fully address cable self-heating concerns.

3.5.3 Synopsis of the Licensee Response The licensee has provided updated calculations that explicitly include consideration of ampacity limits, actual cable ampacity loads, and ambient temperature.

3.5.4 SNL Assessment of the Licensee Response The licensee has revised the calculations as requested to include an appropriate consideration of cable self-heating effects. SNL has " spot-checked" the modified calculations and identified no errors in this regard.

3.5.5 Findings and Recommendations SNL finds that the licensee has given adequate and appropriate consideration to the cable self-heating effect in the modified calculations. The licensee response was fully adequate to resolve the identified concerns. No funher actions on this RAIitem are recommended.

3.6 RAI Item 3d: Adequacy of the 57 Minute Barrier Rating 3.6.1 Summary of SNL's Initial Review Findings SNL had cited that the licensee assessments often cited a Sre barrier qualification rating of 57 minutes where a nominal 60 minute rating was required. No specificjustification was provided for this assumption.

3.6.2 Synopsis of the USNRC RAI The licensee was requested to providejustification for this assumption.

3.6.3 Synopsis of the Licensee Response The licensee cites that the adequacy of a 57 minute rating is based on the fire hazard discussions provided for each application.  ;

17

(

3.6.4 SNL Assessment of the Licensee Response i

The licensee has provide no substantial new information in its RAI response. As was noted by SNL in our 6/97 review, the licensee fire hazard assessments are outside the scope of the SNL efforts and have not been reviewed in any detail by SNL. Hence, we are unable to assess the I adequacy of the licensee assessments in this regard.

3.6.5 Findings and Recommendations SNL makes no specific recommendations regarding de acceptability of a 57 minute barrier rating in lieu of a 60 minute rating as nominally required by Appendix R. SNL recommends that the USNRC evaluate the licensee fire hazards assessnients in order to make a determination of the acceptability of a 57 minute banier rating.

l l

l l

18 .

F 4.0

SUMMARY

OF FINDINGS AND RECOhedENDATIONS SNL finds tic . the licensee has adequately responded to certain of the issues raised in SNL's previous revis and in the USNRC RAI of 8/19/97. In particular, the licensee's actual cable functionality assessments have been significantly improved, and have adequately addressed the fouritems raised in RAIItem 3a-3d. However, the functionality assessments are dependent on the thermal environment simulation results for the extension of NEI Test 2-1, and SNL finds these -

calculations to be deficient.

Three specific technical weaknesses in the licensee calculations were identified; namely, (1) the inappropriate application of thermal conductance values for larger conduits to the simulation of smaller conduit thermal response, (2) the inappropriate application of thermal conductance values from an upgraded barrier system to the simulation of a non-upgraded barrier, and (3) an optimistic treatment of the conduit to cable heat transfer. While the third point is a relatively minor factor that could be easily resolved, the first two weaknesses relate to factors deeply ingrained in the analysis apprcach. Each results from a fundamental lack of appropriate data upon which to base the assessments. The licensee has already made extensive, ifinappropriate, use of the most favorable data sets currently available; hence, there is no clear path to resolution of thew concems other than to perform a confirmatory test of a non-upgraded 1/2" nominal conduit fire barrier system to provide the appropriate data. This would, of course, render the calculation moot.

Ultimately, the first two of these concerns cannot be readily addressed through licensee interactions or modifications of the calculation.-

In addition, SNL finds that the licensee has not adequately addressed the potential that a material burn-through might have been realized during the 50-60 minute data extension period. The issue of material burn-through is inherently outside the scope of the thermal simulation method. As an alternative, the licensee offered two main arguments to resolve the bum-through concerns. SNL's assessment of these arguments is summarized as follows:

The licensee offers the fact that no burn-through was noted in NEI Test 1-6 and implies that this provides evidence that burn-through was not likely in NEI Test 2-1 even had the test progressed to the full one-hour exposure. SNL finds this argument to be i inappropriate because NEI Test 1-6 involved significant upgrades to all of the barrier i

joints that were not present in NEI Test 2-1. Contrary to the licensee conclusions, SNL i finds that these upgrades would profoundly impact the thermal behavior of the envelope and were directly aimed at known material weaknesses that have often contributed to hot-shot and burn-through behavior.  ;

The licensee cites that there was no evidence that the onset of burn-through was immanent at the time NEI Test 2-1 was terminated. SNL disagrees with this conclusion (see discussion immediately below), and in any case, does not find this argument to be adequate to resolve the identified concern. It is SNL's assumption that the " burden of proof' lies with the licensee and that some positive evidence that burn-through would not have occurred is needed to address this concern. Rather, this argument simply cites a lack of evidence that a burn-through would occur. This is not equivalent.

I 19 -

SNL funher finds that there is evidence of a likely burn-through had NEI Test 2-1 continued for the full 1-hour exposure, and that this evidence is sufficient to raise significant doubt regarding the one-hour performance of the tested barrier system. In panicular, SNL cites the following three observations:

As noted p cviously, the NEI test reports include observa. ions for all of the Test 2-1 conduits of"O inches of unburned material remaining" in several locations on each test ccnduit, and an actual burn-through on the smallest 3/4" conduit. While SNL concurs with the licensee interpretation that the lack ofvirgin material does nct constitute an actual bum-through, SNL reiterates its previous 6ndings that this is indicative of, at best, marginal barrier performance at the 50 minute test duration, and that bum through was all

~

but inevitable.

SNL has compared the performance of the 2" conduit from NEI Test 2-1 to that of the 3/4" conduit from that same test based on the licensee thermal conductance estimates for each item. SNL sees a clear parallel between the pre-burn-through behavior of the 3/4" conduit and the end of test behavior of the 2" conduit (a steadily rising thermal conductance factor preceding the onset of bum-through). SNL finds this behavior to be a clear indication that burn through was, indeed, immanert for the 2" conduit at the time the

, test was stopped. It is considered unlikely that this test item would have survived a full 1-hour exposure without experiencing a burn-through. While the same conductance behavior was not noted for the 4" conduit, the result does cast further doubt on the reliability of the licensee method in general, and on the simulations of the 4" conduit performance in panicular.

SNL has also considered the results of TUE Test Scheme 9-3. This test involved three conduits ranging in size from 3/4" to 2" and a fire barrier that appears nominally identical to the barriers in NEI Test 2-1; however, the test was carried out for the full one-hour exposure. Three diff'erences in barrier construction were identified, and each appears to favor the TUE test articles; that is, the TUE 9 3 test specimens should have performed better than the NEI 2-1 test specimens. All three of the TUE conduits experienced burn-through. Further, to date the USNRC has not accepted any cable functionality arguments for acceptance this test article. This test provides clear added evidence that burn-through in the nominally identical NEI Test 2-1 was likely had the test continued for the fidi exposure period, at the very least, for the 2" conduit. Again, the performance of the 4" conduit remains indeterminate.

On this basis, SNL finds that the licensee has not adequately addressed the potential that burn-through might have been observed had the NEI Test 2-1 been subjected to the full 60 minute exposure. Funher, SNL finds that the licensee calculation is inherently incapable of addressing the bum-through issue. The onset of bum-through is currently not amenable to predictive analysis.

Given (1) the fact thu the licensee calculation does not address bum-through, (2) the lack of an adequate alternate resolution basis for the burn-through concern, (3) direct evidence that burn-through was indeed likely at the very least for the 2" conduit, and (4) the identified technical weaknesses, SNT must recommend that the licensee extrapolations of the test data not be l

\

20 -

accepted by the USNRC as an adequate basis for the fire barrier perfonnance assessments.

Because the functionality assessments are, in turn, dependent on the thermal performance simulation results, SNL also recommends that those assessments should not be accepted as the basis for acceptance of the fire barrier performance.

SNL notes that the licensee has at least two clear alternatives to reliance on the cited calculations.

First, the licensee could choose to upgrade the barriers based, for example, on the as-tested joint upgrades used in the successful NEI Test 1-6. Second, the licensee can perform a confirmatory test to assess the performance of a non-upgraded barrier for the full 1-hour quali6 cation period.

A single fire endurance test with a sufficient number of conduits to bound the licensee applications would be adequate. In this regard, it is further recommended that, because cable functionality arguments are likely to form the basis for the test acceptance, should testing be proposed, the USNRC request the licensee (1) to include cable samples representative of those installed in the plant, (2) to use a lower bound cable fill in the test article construction, and (3) to make direct measurements of the cable insulation resistance during the furnace exposure.

21 1

a

5.0 REFERENCES

1. Nowlen, S.P., An Investigation of the Effects of Thermal Aging on the Fire Damageability of Electric Cables, SAND 90-0696, NUREG/CR-5546, SNL, May 1991,

2. Nowlen, S.P. and Jacobus, M.J., "The Estimation of Electrical Cable Fire-Induced Damage Lhnits," SAND 92-1404C, Fire andMaterials,1st Intemational Conference and Exhibition, Sept.

24-25,1992, Washington DC.

3. TUEC Comanche Peak Stenm Electric Station: Fire endurance test ofa 7hermo-Lag 330-1 fireprotective envelope (3M in.,1-1/2 in. & 2 in. Conduits), Test Scheme 9 Assembly 3, Omega Point Laboratories, December 28,1992.

4. F. H. Buller and 3. H. Neher, "The Thermal Resistance Between Cables and a Surrounding Pipe or Duct Wall," AIEE Transactions, Volume 69, pgs 342-349,1950.

5. J. H. Neher and M. H. McGrath, "The Calculation of the Temperature rise and Load Capability of Cable Systems," AIEE Transactions, pgs 752-772, Oct.1957.

l l

1 22 -

J

a .

Appendix A: {

An Assessment of the Licensee Extrapolation Method

}

A.1 Introduction

~

In response to the USNRC RAI item 1, the licensee has provided for review a copy of the I calculation in which the temperature response for NEI Test 2-1 is extrapolated for the 50-60 minute time period after termination of the actual test. This is documented in licensee Calculation C-9000-814-5310, Revision 0. The purpose of this Appendix is to provide a brief summary of SNL's insights gained from a review of the calculation. In this review SNL has focused only on the conduit calculations. The calculation includes an analysis of a cable tray case as well, but I since the exemption request only deals with conduits, SNL has not reviewed the cable tray analyses.

It is important to note that the licensee calculation is cited as proprietary infonnation. Hence, in order to prevent the public disclosure of the calculation details, this review has purposely avoided any discussion of the details of the calculation, and instead, focuses only on the broad approach and insights of the calculation.

A.2 Summary of the Licensee Approach The ultimate objective of the licensee calculation is to estimate the thermal performance of two i conduits, the 2" and 4" conduits from NEI Test 2-1. Test 2-1 was a test of a basic non upgraded '

barrier system. This test was stopped after 50 minutes, and the licensee is attempting to estimate the thermal response of the test items out to a full 60 minute exposure in order to meet the Appendix R requirements for a 1-hour barrier. The calculation is also heavily dependent on information on the thermal response of the conduits from NEI Test 1-6 which involved an upgraded fire barrier system.

The licensee calculation begins by analyzing the test dats for several individual conduits from NEI Tests 2-1 and 1-6 in order to estimate the net thermal conductance between each conduit and the exposure furnace. The conductance is estimated at each 1-minute measurement interval; hence, the values are time-dependent. The assessment is quite sophisticated in certain regards, especially including the treatment ofradiation heat transfer between the fumace flames anu the banier system. Radiation is a critical aspect of a standard fire exposure test. The licensee treatment has included quite advanced treatment of the effects, for example, of various gasses on the transmission of radiant heat energy to the test article.

Using the fumace temperature and the conduit hot-spot temperature, the licensee calculates the net thermal conductance as a function of time. The details of this process are unique to the licensee calculation, and are presumably the reason for the proprietary nature of the calculation. l This treatment avoids the need for a detailed model of the barrier behavior. Recall that Thermo- j Lag @ behavior is rather complex. The material starts as a relatively homogeneous fibrous solid,

{

undergoes a combustion / sublimation / expansion process in which significant off-gassing occurs, and as it is consumed leaves behind a char layer that also provides some insulating protection.

Ultimately, the char layer breaks down and given sufficient time the test article is ultimately 23 -

I i

l

subject -;o direct furnace exposure. The ultimate breakdown of the char layer is referred to as a bum-through, and for a test to be successful burn-through must not be realized within the desired rating period. Any bum-through would result in sharp rises in the target temperature that would very quickly (typically within less than a minute) compromise cable performance. This complex behavior would be very difficult to model. The licensee side-steps this complexity by simply modeling the process as a transient net heat transfer behavior.

Through this point in the calculation we are dealing with a data analysis and interpretation effort.

Conceptually, the licensee approach is fully valid up to this point. That is, this approach is certainly a valid means oflooking at the test data and attempting to understand the thermal behavior taking place. However, it is in the application and extrapolation of the results that SNL

~

ultimately will find fault.

Recall that the ultimate objective of the calculation is to estimate the thermal response of a cable

, inside the conduit and to extrapolate the behavior observed in a 50 minute test out to a full 60 minute exposure period. At this stage, the licensee has assembled a " family" of thermal conductance versus time curves that characterize the heat transfer between the fumace and the conduit for each of the individual conduits considered from two separate tests.

Because the conductance calculation was based on the conduit response, a model is needed to predict the heat transfer from the conduit to the cables. This is done using a very simple and unnecessarily crude model of the cable / conduit heat transfer. SNL notes that this aspect of the model has been seen commonly in the treatment of ampacity issues; hence, SNL assumes that this aspect of the mod:1 is not proprietary infonnation. The licensee treats the cable and conduit as an annular system of the conduit, an air gap, and the cable itself. That is, the cable is fully isolated from contact with the conduit by an annular air gap. This simplistic treatment ignores the actual contact that occurs between the cables and the conduit. That is, in reality the cables would rest on the inside bottom of the conduit, and would come into direct contact with the conduit resulting in enhanced heat transfer rates. This nominally implies that the conduit-to-cable model is non-conservative because in reality heat would flow more quickly to the cables due to the intermittent contact. SNL notes that altemate methods of analysis are readily available to characterize the heat transfer behavior between a cable bundle and surrounding conduit. See, for example, the work of Neher, et. al., [4,5). Ultimately, this appears to have had only a modest imp'act on the calculation as will be discussed further below. The external heat transfer mechanisms appear to dominate this problem. However, the model is clearly optimistic; hence, this concern has been identified as a weakness in the approach.

In order to perform the extrapolations, the licensee has split the conduits into groups by size, and has averaged the conductance values for each group. That is, the extrapolation for the 2" conduit is based on the average behavior of the 2" conduit from 2-1 and the two 3" conduits from 1-6.

Similarly, the 2-1,4" conduit extrapolation is based on the 'average behavior of the 4" conduit from 2-1 and the 5" conduit from 1-6. Using the average thermal conductance, the actual conduit physical dimensions, and ASTM fumace temperature profile, a prediction of the full period performance of the 2" and 4" conduits is made. The profiles are compared for the 0-50 minute period.

24

L ..

Note that in a true meaning of the term, the licensee is not actually extrapolating the test data, but rather, is performing a full simulation of the test including an extended test performance period.

That is, in extrapolation, one generally takes the available data and uses a model to extend the scope of the data beyond the measured bounds. In this case, the licensee has not simply extended the otiginal data for an extra 10 minutes, but rather, has developed an entirely new pro 61e based on the average conductance behavior of selected conduit groups taken from two tests. As noted below, the simulated response actually under-predicts the measured response in each case by a modest amount (on the order of10-20*F).

Finally, once the simulated response profiles have been developed, the resulting temperatures are used in an assessment of the cable functionality. This step is not included in the proprietary calculation, but rather as a pan of the overall licensee submittal. The process as modified in the current submittal includes specific consideration of both the actual cable current load and the ampacity limit to estimate the cable operating temperatures. The assessment also includes consideration of the ambient temperature in the area that would be expected at the outset of the fire. The simulated thermal response temperature rise is then imposed on the initial cable temperature and compared to the anticipated failure threshold. The cable qualification rating ,

(CQR) is based on the time at which the temperature rise exceeds the failure threshold.

A.2 General Initial Observations In general, SNL found the initial phases of the licensee calculations to be of very high quality, to have been well thought out, and well executed. In particular, those portions of the calculation in which the test data is analyzed and the conductance factors are developed provide a number of important and valuable insights into the behavior of Thermo-Lag @ in a fire test. SNL found no significant fault with this aspect of the licensee analyses. Ultimately, it is in the application of the analysis results that SNL will find fault. In particular, as will be discussed further below, SNL finds that the licensee has used questionable and inappropriate practices in simulating the thermal response for NEI Test 2-1.

A.3 Validation The licensee calculation does include significant validation results for the model. One aspect of this process is that the licensee cites that the model was implemented using two independent computer codes. The results compared favorably. This does provide an indication that errors of coding are unlikely, but says nothing about the general validity of the model.

The licensee has also provided a number of figures to support the model validation. In one approach for each case, the licensee has initially calculated the net heat transfer factors using the test conduit and furnace temperatures, and has tFen re-calculated the test cable temperatures using the net heat transfer coefficients (see for example page 24 of the calculation). Ideally, these values should match to within the machine precision plus the accuracy of the conduit-to-cable heat transfer model. While, as noted above, the conduit to cable heat transfer model is both optimistic and unnecessarily crude, the results uhimately show a reasonable agreement. This does provide some indication that the conduit-to-cable model has a relatively minor impact in comparison to the external heat transfer behavior. The close match also indicates, again, that the model was properly coded and implemented.

25

The next level of validation is based on using the heat transfer coefficients derived from one test item (NEI Test 1-6,3" steel conduit) to predict the behavior of a second test item (NEI Test 1-6, 5" aluminum conduit). Again the results compare reasonably well (see page 34 of the calculation). In this case, the simulation is somewhat conservative. This is consistent with our expectation given that larger conduits perform consistently better for a given barrier system than do smaller conduits. Hence, use of a 3" conduit response to predict a 5" conduit response should, and does, yield a conservative result. Of course, if the process were to be reversed (i.e., the heat transfer coefficients for the 5" conduit were used to predict the behavior of the 3" conduit) then the results would be modestly non-conservative. This case provides some interesting results indicat.ng at the least some consistency in the behavior of the various test items in this one test.

This is also as expected given that the barrier systems are nominally identical.

The next level of validation is based on a comparison of the various calculated heat transfer coefficients from the three test items ofNEI Test 1-6 analyzed by the licensee. This is shown on page 47 of the calculation. Once again, all of these cases are tr. ken from a single test of an upgraded barrier that is nominally identical for all three items. Given this similarity one should anticipate consistencies in the behavior as well. Nonetheless, some interesting trends can be i observed.

The figure does show some very pronounced differences in behavior during the first 35 minutes of the test, and much more consistency in the later stages of the exposure.

For the first 15 minutes the calculated thermal conductance goes through wide swings in value over a range of about one order of magnitude. This presumably reflects the initial stages of the swelling and ignition of the fire barrier system surface. Indeed much of the volatility in this stage may be due to intermittent ignition at the surface of the barrier.

Between 15 and 20 minutes the thermal conductance stabilizes at a relatively high value.

This may well reflect the sustained ignition of the banier surface. The fact that the thermal conductance is somewhat lower than the peak values observed during the first 15 minutes of the test can be attributed to the fact that by this stage some thickness of an insulating char layer would have formed.

Between 20 and 25 minutes the thermal conductance declines sharply to the minimum value observed throughout the test (seen at about 25 minutes). When compared to the test temperatures, one notes that this roughly corresponds to the conduit reaching a  !

temperature around 200-220"F. At this point it is quite common to observe a " plateau" in the temperature response for some pedod of time. That is, for some time the barrier internal temperature will stabilize before again beginning to rise. Given the temperature at which this consistently occurs, this stabilization is likely an Mifact of the remaining water in the barrier system reaching the boiling point and the resulting off-gassing ofwater vapor overwhelming any other behavior.

Immediately after the plateau stage (at about 35 minutes into the test) the conductance value rebounds to a relatively high value. However, the values also tend to " settle down" and no longer display the pronounced volatility observed in the early test.

26 I

Between about 35 and 50 minutes, the heat transfer coefficient declines slowly. This presumably reflects the final stages of virgin barrier material sublimation and buming and the final formation and thickening of the char layer.

At about the 50 minute mark, the conductance appears to reach a post-plateau minimum, and shows some signs ofincreasing again. As will be discussed below, this behavior is more pronounced in the analysis of the conduits from Test 2-1. The development of the -

. increasing conductance trend likely reflects the early breakdown of the char layer and loss of the charlayerinsulating power.

The most important result of this comparison is to illustrate the relatively consistent behavior in the later stages of the test. Again, given the consistency in the construction of the three Sre barriers, there should be a high level of consistency in these results. The early volatility indicates that simulating the early stages of the fire test will be problematic at best. However, it is the later

, stages of the test that we are especially interested in.

The licensee then moves on to its analysis ofNEI Test 2-1. The first case assessed is the 4" conduit from that test. The behavior of the heat transfer coefficient for that case is compared to the three cases from Test 1-6 in a figure on page 55 of the calculation. Note that the behavior is markedly diff'erent. The early stages of the test show far less volatility than the previous cases.

The low again occurs near the 25 minute mark (the beginning of the plateau stage), but recovers far more slowly than in the previous cases. The reasons for this remain unclear. Ifthis stage does reflect the loss of water, then the longer recovery might be indicative of a higher moisture content. However, the cure time for Test 2-1 was approximately 30 days and that of 1-6 about 34 days. Hence, the moisture content should have been comparable. AAer the plateau stage, the heat transfer coefficient again recovered, and the value was significantly greater than the post-plateau maximums observed in the analysis oftest 1-6. By 50 minutes, the end of the test, the conductance had dropped to a value consistent with those observed in 1-6.

The next step was to use the calculated conductance values to simulate cable response for the 4" conduit. In this case (see page 57 of the calculation) the prediction lags the measured data somewhat. This is likely indicative of the crude conduit-to-cable treatment and the lack of consideration of the intermittent contact between the cables and conduit. The diff'erences are not, however, pronounced (on the order of 10'F or less). Once again, this indicates that the external heat transpon dominates the problem, and the optimistic treatment ofinternal conduit to cable heat transfer plays a more minor role.

The next step in the validation process involves analysis of the Test 2-1,2" conduit. Here again the behavior is markedly different for this conduit as compared to the previously analyzed conduits. In particular, the onset of the plateau stage occurs earlier in the test, about 19 minutes versus 25 minutes for the earlier cases. The minimum also is not sustained for any significant time period (indicating a very shon plateau stage), and the heat transfer coefficient has recovered to its post-plateau maximum by 20 minutes. In contrast, the other tests did not see the end of the

" plateau stage" until 27-35 minutes into the test. The post-plateau minimum is noted at 39 minutes as compared to 50 minutes in the other cases considered. This case also shows a clear and pronounced trend ofincreasing thermal conductance for the last 11 minutes of the test; again, a likely reflection of breakdown in the char layer. i 27 i

The results for the 2" conduit appear to reflect a " time compression"in the behavior as compared to the larger conduits. That is, given SNL's interpretations, the plateau stage is reached more quickly, and lasts for only a very short time. In the post plateau stages, the char layer completes its formation more quickly, and displays signs of breaking down earlier in the test. Overall these results do indicate that the time response of the smaller 2" conduit is markedly compressed in comparison to the other test items. This is consistent with general experience that larger conduits perform better for a given barrier system than do the smaller conduits. It does, however, also raise uncertainty regarding the application of heat transfer behaviors derived from larger conduits to the analysis of a smaller conduit.

The next step in the validation process is one of the most interesting aspects of the licensee calculation. This involves the analysis of the 3/4"' conduit from Test 2-1. This conduit is of particular interest because it is the only conduit considered in the analysis that suffered an actual burn through a noted in the test report. The conductance behavior for this conduit is shown on page 93 of the calculation. Again, the general behavior of the conductance plot is quite similar in the general trends displayed to those of the previous conduits up to about 40 minutes into the test, although, once again, as in the case of the 2" conduit, we see a clear " time compression" effect in the response. For example, the post-plateau stage begins at about 18 minute compared to 20 minutes for the 2" conduit and 27-35 minutes for the 3" and 5" conduits. The post-plateau minimum is reached at about 29 minutes compared to 39 minutes for the 2" conduits and 50 minutes for the 3" and 5" conduits. In the period from 30 minutes through 41 minutes we observe a pronounced trend ofincreasing thermal conductance quite similar to that noted for the 2" conduit from 2-1 between 39 and 50 minutes. Once again, this likely reflects the progressive breakdown of the char layer. The most inter: sting behavior of the plot is realized in the period beyond about 40 minutes. At this point the conductance begins to rise rather sharply, and at about 41 minutes a very abrupt rise in the conductance is realized. This very likely corresponds to the onset of barn-through for this test item.

The final case analyzed is the 3/4" conduit from Test 1-6. This test displays a markedly different behavior than any of the previously analyzed cases. In this test, the fire banier was much thicker, apparently involving an upgraded 3/4" thickness barrier. The " plateau" stage is entered at about 35 minutes into the test, and lasts virtually to the end of the test. This test bears no resemblance to any of the other tests; hence, it is oflittle value in the attempt to understand Test 2-1.

A.4 The Net Results and Case Extrapolations The net results of the case examples as ultimately applied by the licensee are illustrated on pages 70 and 71 of the licensee calculation. In these two figures, the licensee averages the heat transfer coefficients for the 4" and 5" cases, and for the 2" and 3" cases respectively. It is these values that are apparently applied in the simulation of the test data for the 4" and 2" conduits from Test 2-1. Again, as noted above, the licensee is actually performing a full simulation of the test response based on the averaged conductance factors rather than performing a direct extrapolation of the measured test data. These simulation results are shown on page 74 and 75 of the calculation.

The most significant potential concern in this regard is that the marked upward trend of the heat transfer coefficient from the 2" conduit in Test 2-1 is not reflected in the averaged small conduit 28

, response values. This illustrates (1) the fallacy of predicting the behavior of a smaller conduit based on the behavior oflarger conduits, and (2) the fallacy of using the thermal response of an upgraded barrier to simulate the behavior of a non-upgraded barrier.8 SNL finds that the small conduit group average conductance profile fails to capture, in panicular, the actual behavior of the 2" conduit adequately. This is taken up funher in A.5 below.

Nominally, the results of the predictions track the test data reasonably, even surprisingly, well for the 50 minute test duration. One will note that in both cases the predicted response modestly lags the actual test response (by about 15-25'F). This is likely a result of a combination of three factors; namely, the inclusion of thermal performance factors for an upgraded barrier in the analysis of a non-upgraded barrier, the size effects, and the non-conservative conduit-to-cable

~

heat transfer model.

Ultimately, the most imponant issue which remains unaddressed is, again, the question of bum-through. The licensee has assessed just one case involving an actual bum-through, the 3/4" conduit from Test 2-1. Hence, there is only limited information available on the conductance behavior leading up to the onset of burn through. However, SNL notes some rather troubling

+

similarities in the behavior of the 3/4" and 2" conduits from Test 2-1 that may indicate that the onset of burn-through was closer for the 2" conduit than assumed by the licensee. The truly unique behavior observed in the Test 2-1,3/4" conduit is the precipitous rise in the thermal conductance at 41 minutes into the test. However, we also note that the sharp rise occurred about 12 minutes after the post-plateau conductance minimum had been observed (at about 29 minutes). For the 12 minutes between the 29 and 41 minute marks, the conductance was distinctly rising at a relatively steady rate. As noted above, SNL atti;butes this rise to the breakdown of the protective char layer. The most ',roubling aspect of this is that the 2" conduit also displays a similar rate ofincrease in conductr.nce for the 11 minute period from about 39 minutes through the end of the test at 50 minute: 7.- contrast, whet we observe the other larger conduits in both 2-1 and 1-6, by the end of each psure the .onductance had apparently reached a minimum and wasjust beginning to show signw! e wbsequent period ofincreasing conductance.

In SNL's view, the 2" conduit from Test 2-1 in panicular has more in common with the 3/4" conduit from that same test than it does any of the others conduits considered by the licensee.

For the 2" conduit in panicular, SNL interprets the steadily increasing conductance observed during the last 11 minutes of the test as a clear indication of a breakdown in the char layer. Based on the comparison to the 3/4" conduit behavior, this also appears indicative ofimmanent burn-through. Based in part on this observation, SNL considers it very unlikely that the 2" conduit would have survived the full 60 :ninute exposure without suffering a burn-through.

The behavior of the 4" conduit in Test 2-1 is more difficult to interpret. Indeed, this test bears little resemblance to any of the other tests, having displayed a far more prolonged plateau period than any of the other tests. There is no clear explanation for this rather unique behavior. Hence, it is difficult to draw any parallels to the other test profiles. At the end of Test 2-1 at 50 minutes, SAs discussed in the body of this report, SNL disagrees with the licensee assertion that the barrier upgrades were of a structural nature only and would not impact the thermal performance.

Rather, it is quite clear that the upgrades did significantly impact the thermal performance.

29 -

I

-. , ]

the conductance had not yet displayed a rising behavior. Clearly this conduit was performing better than its " siblings", the 3/4" and 2" conduits. However, given the other uncensinties and concerns associated with licensee performance simulations, SNL considers the results suspect. A lot can happen in a fire test during 10 minutes, especially if a seam between two panels of the material begins to open. There is little or no assurance that such an event would not have been realized had the test progressed for the full 1-hour period.

A.5 Summary ofInsights and Final Observations The insights derived from this exercise include the following:

The early time behavior of the test'anicles appears quite chaotic and volatile.

The upgraded barriers in Test 1-6 did display a marked self-consistency as one should expect given nearly identical baniers on each item.

The baniers in Test 2-1 displayed a marked inconsistency in behavior as compared to the baniers in Test 1-6. This does raise questions regarding the validity of applying the heat transfer coefficients from Test 1-6 to the simulation oftest 2-1.

Each of the test items displayed at least some evidence of the " plateau" behavior observed as the inside of the banier reaches a temperature near the boiling point of water. During the plateau stage, the temperature profile " flattens out" for some period of time. This is clearly reflected in t!.at the conductance values go though a sharply defined minimum corresponding to this plateau before rebounding to a relatively high post-plateau maximum value. The time ofonset and the duration of the plateau was affected by both the size of the conduit and the banier construction.

l The banier conductance values did display more consistency when the long-term, I " post plateau" behavior was considered. In particular:

! At the end of the plateau period, each barrier realized a post-plateau peak l in conductar.cc.

l -

Following this peak, each barrier went through a period of declining thermal conductance until a post-plateau minimum was reached. SNL attributed this behavior to the progressive consumpion of the virgin barrier material and the development and thickening of tha char layer.

This was generally followed by a period ofincreasing conductance, although in some cases the end of the test was reached at about this time so the behavior is not fully illustrated. This behavior was, however, especially evident for the two smaller conduits in Test 2-1. SNL attributes this behavior to the break down of the char layer.

Only one test was analyzed that actually involved a material burn-through as noien in the test reports. In this case, following about 12 minutes ofincreasing conductance, a precipitous rise in conductance was noted. This apparently reflected the first onset ora burn-through.

30 l 1

The results clearly reflect a " time compression" of the thermal response behavior for smaller conduits as compared to larger conduits. This was especially reflected when the calculations for the 3/4" and 2" conduits in Test 2-1 were considere This is also consistent with experience that shows that zor a given barrier system

' larger conduits generally perform better than smaller conduits. This does raise same concern over the practice of predicting small conduit performance on the -

basis oflarger conduit performance

' A.6 Summary of Findings and Recommendations '

In general, those portions of the licensee calculation dealing with the analysis of the test data and the calculation of thermal conductance values were found to be of a high quality. The calculations are well documented and clear throughout. The case studies provide many insights into the behavior of the fire barriers. In this regard SNL does not find any significant fault with the data analysis.

However, SNL did note three specific points of technical concern regarding the actual simulations of the test performance for the 2" and 4" conduits from NEI Test 2-1. These are summarized as follows: 1 The licemee thermal model for the conduit-to-cable heat transfer is unnecessarily crude and underestimates the actual rate ofheat transfer to the cables. Ultimately, this appeared to have a relatively minor impact on the simulation results. Hence, this point is noted for completeness, but is not considered a major point of concern. This is, however, one area of weakness for which a clear and preferable alternative does exist and is readily accessible to the licensee (sem& mees 4 and 5 of the text).

In performing th . wanse simulations, the licensee has averaged the conductance factors for 2" and 3" conoints to predict the 2" conduit response, and those of the 4" and 5" conduits to predict the 4" conduit response. In general it is well known that for a given fire barrier system larger conduits perform better than smaller conduits. Hence, SNL finds that this practice is generally inappropriate. This is ofparticular concern in the case of the smaller 2" conduit simulation given the pronounced differences in behavior noted for this particular conduit in comparison to the others. In effect, the smaller conduits appear to experience a " time compression" in the thermal behavior, and the licensee approach does not reflect this. This weakness will be quite difficult to nddress given the current licensee approach, again especially in the context of the 2" conduit. This is because there are very few tests of a non upgraded fire barrier available upon which to base the simulations, and the licensee has already made extensive, ifinappropriate, use of the most favorable of the available data sets. SNL considers it unlikely that the licensee can fully address this concern without performing additional confinnatory tests of the non-upgraded barrier system.

The simulations for Test 2-1 are based in large part on the application of heat transfer conductance factors derived from Test 1-6. However, Test 2-1 involved a non-upgraded barrier system while Test 1-6 involved significant upgrades. SNL disagrees with the 31 -

licensee's contention that the upgrades were limited to structural enhancements and that they would not impact the thermal behavior. Rather, the construction upgrades in Test I-6 were specifically aimed at a known weakness in the barrierjoints cor.anonly identi6ed as a contributing factor in hot-spot and bum-through behavior. Hence, SNL finds that the upgrades had a clear impact on the thermal behavior and that significantly different thermal behavior would be anticipated due to the upgrades. SNL finds that application of thermal conductance values derived from an upgraded barrier to the analysis of a non-upgraded barrier system is inappropriate. Again, this is a concern that the licensee will be unable to address in the absence of additional confirmatory results for a non-upgraded fire barrier system.

Finally, the licensee has consider .1just one casein which an actual burn-through of the fire l barrier was noted. Hence, there is only very limited information available on the behavior associated with the onset of burn-through. Two points were noted in this regard:

The one case of burn-through considered in the calculation was not included in the i development ofconductance values used in the final simulations. Hence, SNL finds that the question of material bum-through lies outside the scope of the licensee calculation.

SNL finds that there is a marked similarity between the pre-bum-through behavior of the 3/4" conduit from Test 2-1 and the end of test behavior of the 2" conduit in Test 2-1.

SNL finds this to be an indication that burn-through on the 2" conduit was likely and immanent. Hence, for this case in particular, SNL sharply questions the validity of the ,

licensee extrapolations. While similar behavior was not noted for the 4" conduit, this j observation does cast doubt on the acceptability of the licensee approach for that conduit i as well. (Further related discussion are provided in the text.)

l Based on these observed weaknesses, SNL recommends that the licensee calculations not be accepted as the basis for estimsting the full 1-hour fire banier perfonnance of the subject conduits from NEI Test 2-1.

As noted above, two of the three specific technical concems are considered to be of a fundamental nature and SNL does not expect that further licensee interactions would be helpful to resolve the concems. These two concerns derive from the fundamental lack of appropriate fire endurance data sets, are deeply rooted in the licensee method, and will not be readily addressed through minor modifications of the method. Rather, to address these concems will require the availability of appropriate test data on representative conduits and non-upgraded fire barriers. Of course, given such data the calculation becomes moot. Ultimately, SNL must recommend that the current approach is not an acceptable basis of analysis, and that the approach is fundamentally i

flawed in at least two regards. Funher, the method is inherently incapable of addressing i analytically the issues of material bum-through especially as related to joint effects; hence, in the  :

absence of direct experimental evidence this will remain a major point of uncertainty.

32 -

t