A seismic margin assessment procedure

Nuclear Engineering and Design 107 (1988) 61-75 North-Holland, Amsterdam

A SEISMIC MARGIN ASSESSMENT R.P. K E N N E D Y

1, R.D.

61

PROCEDURE

C A M P B E L L 2 a n d R.P. K A S S A W A R A 3

1 RPK/Structural Mechanics Consulting, 18971 Villa Terrace, Yorba Linda, CA 9"2696, USA 2 NTS Engineering, 6695 East Pacific Coast Highway, Long Beach, CA 90803, USA 3 Electric Power Research Institute, 3412 Hillview Avenue, Palo Alto, CA 94303, USA

Received 1 June 1987

This paper presents recommendations for a seismic margin assessment program for operating nuclear power plants. The goals of such a program are stated, and an overview of two alternate approaches is presented. Brief guidance for performing a seismic capability walkdown (an integral part of such a margin program) is provided. One of the prime purposes of such a walkdown is to screen out from review those components which will not be significant contributors to seismic risk. As guidance, the dominant contributors to seismic risk as found from existing seismic probabilistic risk assessments are presented. Seismic capacity screening guidelines are also provided. Lastly, a conservative deterministic failure margin (CDFM) for estimating a high-confidence-low-probability-of-failure seismic margin level is summarized.

1. Introduction

Nuclear power plant structures and safety-related systems have been generally designed conservatively for a safe shutdown earthquake (SSE) and more conservatively for a smaller operating basis earthquake (OBE). Depending upon the relative conservatism of the design criteria, either the SSE or the OBE will control the design. For plants with SSE levels less than 0.2g, often non-seismic loadings control the design. In recent years, increasing knowledge in the geoscience field has led to a better understanding that, although highly unlikely, it is possible for the nuclear power plant to be subjected to earthquake ground motion greater than that for which the plant was designed. For this reason, interest has developed in demonstrating that nuclear plant structures and safety-related systems can safely withstand earthquake ground motion larger than their design earthquake ground motions (SSE and OBE). Within this paper, this larger-than-design earthquake ground motion will be called the seismic margin earthquake (SME) to distinguish it from the design earthquakes. The plant has already been designed. Therefore, for the SME, the goal is not to design the plant, but to determine the performance of already-designed structures, components, and systems when subjected to the SME.

The primary impetus for having seismic margin reviews performed has come from the Advisory Committee on Reactor Safeguards (ACRS), which has expressed interest in the capability of nuclear power plants to withstand earthquakes greater than the SSE for a number of years, and from seismic probabilistic risk assessment studies (SPRAs) which have shown that, although the seismic risk is small, it is generally not negligible when compared to other sources of risk. For U.S. plants in lower seismic zones for which SSE levels are commonly set between 0.12g and 0.25g, these SPRA studies have indicated that the dominant seismic risk comes from earthquake ground motion that is 2 to 5 times greater than the design SSE level. This statement implies large margin over the SSE level. However, SPRA studies have large uncertainties, both in the seismic hazard and seismic fragility (capability) aspects. For this reason, interest has been expressed in establishing more direct, simpler, and less controversial methods to evaluate the seismic margins of nuclear power plants that have been indicated by SPRAs. E P R I / N R C held a workshop on the subject of seismic margin for nuclear power plants in October 1984 [1]. The consensus was that large seismic margin generally exists over the SSE. It was concluded that seismic margin reviews should be treated as safety reevaluations and not as design evaluations. New ap-

0 0 2 9 - 5 4 9 3 / 8 8 / $ 0 3 . 5 0 © Elsevier Science Publishers B.V. ( N o r t h - H o l l a n d Physics Publishing Division)

62

R.P. Kennedy et al. / A seismic margin assessment procedure

propriate criteria and approaches should be established for conducting practical and economical seismic margin reviews. The purpose of margin review should be to demonstrate adequate margin over the SSE for plant safety and to find any "weak links" which might potentially have less than adequate margin to safely withstand some SME bigger than the SSE.

2. Overview of margin assessment approaches In mid-1984, the NRC formed an "Expert Panel on the Quantification of Seismic Margins" to make recommendations of an approach for seismic margin review. This panel has produced two reports [2,3] on an approach for seismic margin reviews. This panel recommended an approach which they considered to be very cost-effective for demonstrating high confidence of a low probability of seismic-induced core damage for earthquake ground motion levels up to about 0.3g and for determining "weaker links" which might have highconfidence-of-a-low-probability-of-failure ( H C L P F ) capacity less than this level. The recommended approach relies heavily on earthquake experience data, generic qualification and fragility test data, extensive use of expert judgment and experience to reduce the level of reevaluation work required, and substantial levels of seismic margin plant walkdowns to discover "weaker links" and to determine where more detailed evaluations needed to be performed. The "Expert Panel" recognized that this approach was most cost-effective for SME levels less than about 0.3g, where an extensive data base exists and where expert judgment and experience is most easily applied. As the SME level is increased above 0.3g, the approach becomes significantly more difficult as less data exist in the "data base" at higher ground motion levels and expert judgment becomes less well-founded. At levels above about 0.5g, the approach breaks down into essentially a complete reevaluation of all required safety systems which might be vulnerable to seismic loading and no longer could be considered cost-effective. At the same time, the panel recognized that it is highly unlikely that the vast majority of nuclear power plants in the U.S. would be asked to demonstrate high confidence of low probability of seismic-induced core damage for ground motion levels above 0.25g because of the extremely low annual probability of exceedance of such ground motion levels for these plants. For this reason, the panel concentrated its efforts on SME levels of 0.3g and less, but also provided a lesser level of information for SME levels of 0.5g and less.

The Electric Power Research Institute (EPRI) initiated a program in late 1985 to further simplify and expand upon the N R C "Expert Panel" recommendations for utility implementation of seismic margin reviews. This paper highlights the EPRI program recommendations and compares and contracts these with the recommendations of the "Expert Panel". The primary difference is that the EPRI program recommends a more deterministic (less probabilistic) approach. It is believed that such an approach will be more amenable to use by experienced design engineers (including utility staff design engineers) who may not have any probability background or who may not be comfortable with probabilistic approaches. The goals of a seismic margin assessment program are: (1) To determine a ground motion level for which one has high confidence of a low probability of seismicinduced core damage. (2) To identify any "weaker-link" components which tend to reduce the seismic margin capability of the plant for withstanding earthquakes larger than the SSE. These goals can be achieved by performing a seismic probabilistic risk assessment (SPRA). However, unless the SPRA already exists, performing an SPRA is not a cost-effective or practical approach to achieving these goals. In order to establish a seismic margin assessment (SMA) methodology, one must decide how much of the SPRA methodology should be retained versus how much can be "thrown away" and still address these goals. One of the primary benefits of SPRA is the coupling of system evaluations with seismic capability assessments, and this coupling needs to be retained. However, simplifications can be made in both system evaluations and seismic capability assessments. In the system aspects, the NRC Expert Panel reports recommend retention of the need for developing nearly complete sets of fault and event trees, but for a reduced number of functions (called Group A functions). From these trees, simplified cut sets (Boolean expressions) of components should be developed for the end-point of core damage. However, in the EPRI approach, it is recommended that the process can be further simplified. In lieu of developing detailed fault and event trees, it is only necessary for the system engineers to define those components required for an operational sequence of plant systems that will bring the plant to a stable condition (either hot or cold shutdown) and maintain that condition for at least 72 hours. Herein, this set of components is called a "success path". Several possible success paths exist. The idea is to select the success path

R.P. Kennedy et al. / A seismic margin assessment procedure

for which it will be easiest to demonstrate an adequate seismic margin. Only those components within this success path need to be reviewed in the seismic margin assessment (SMA). Selection of the success path which will be studied in the SMA is a joint responsibility of the plant operators, system engineers, and seismic capability engineers involved in the SMa with the system engineers taking the lead role. Sometimes, it might not be obvious at the outset which success path will most easily demonstrate an adequate seismic margin. In this case, a primary path and one or more alternate paths might initially be selected. The more alternate paths which are considered, the more likely that a success path with high seismic margin will be found. However, the more paths which are considered, the greater the number of components which must be investigated. From a practical standpoint, the choices should be quickly narrowed down to no more than one primary and one alternate success path for which the SMA is to be performed. The seismic margin capability (expressed in terms of HCLPF) for any success path is then equal to the seismic margin capability of the weakest seismic capability component in that success path. The key benefit of this success path approach is to reduce the amount of system modeling required down to only one or two success paths and to reduce the number of components investigated by the seismic capability engineers. Furthermore, this approach is purely deterministic. These benefits are gained at the expense of possibly not finding the path with the higher seismic capability. Thus, a lower level SME might be reported for which a high confidence of a low probability of seismic-induced core damage is demonstrated than would be reported if the N R C Expert Panel-recommended procedure is followed. Whichever route is taken for the system investigations, the seismic capability engineers must provide a lower-bound estimate of the ground motion level for which they have high confidence of a low probability of failure (HCLPF) for each component not screened out of the margin review by the system engineers. This statement does not mean that they must compute a specific capacity number for each component. If the SME review level is set sufficiently low (such as 0.25g), then it will be possible to use earthquake experience data, generic qualification and fragility test data, and combined experience and judgment of the seismic capability review team to eliminate many components from review (subject to a walkdown verification, which is always necessary), on the basis that they have high

63

confidence that the capacities of these components exceed the SME level. For these screened-out components, it can only be stated that the HCLPF ground motion level exceeds the SME level. Thus, only a lower-bound estimate is provided. For components which are not screened out, a specific HCLPF capacity estimate is provided. In this regard, both the methodology recommended in this report and that recommended by the N R C Expert Panel are identical. For those components for which HCLPF capacity estimates are required, there are multiple ways to make such estimates, but these ways fall into two main approaches. One approach is to estimate the median seismic capability (fragility) of the component and then to account for random variabilities and uncertainties to back off from this median capability to one in which one has a HCLPF. This approach is well described in the N R C Expert Panel reports. An alternate approach is to use a set of pre-established conservative deterministic failure margin (CDFM) criteria and procedures to directly compute a seismic margin capability for which a HCLPF of the component is demonstrated. This alternate approach, briefly described in the N R C Expert Panel Reports, is the only approach recommended in the EPRI program. Its advantages are that it is prescriptive, deterministic, and repeatable by different groups of experienced engineers. Its disadvantages are that a definitive quantitative statement about the level of confidence or failure probability levels that result from its use cannot be made. All that can be stated is that because of the inherent conservatism of the approach, a HCLPF exists at the computed component capability (i.e., only a qualitative statement). However, even though confidence level and failure probability numbers are produced by the probabilistic fragility approach, these quantitative probability statements are suspect because of inherent uncertainties, limits to the data base, and the subjectiveness of many of the assigned variabilities. Whichever approach is chosen, the SMA has certain minimum requirements. First, it should be an integrated effort between plant operations and systems engineers and seismic capability evaluation engineers. Second, the SMA cannot just be limited to the minimum front-line and support system components required for a safe shutdown. Structures that house the components and connecting electrical, fluid, and gas systems must also be considered. Any non-essential structure or components whose failure is likely to render the essential components inoperable should be included.

64

R.P. Kennedy et al. / A seismic margin assessment procedure

3. Steps of seismic margin assessment A Seismic Margin Assessment (SMA) consists of the following steps: 1. Selection of Seismic Margin Earthquake (SME) level 2. Selection of Assessment Team 3. Preparatory Work Prior to Walkdowns 4. Systems and Elements Selection ("Success Paths") Walkdown 5. Seismic Capability Walkdown 6. Subsequent Walkdowns (as needed) 7. Seismic Margin Assessment Work 8. Documentation Because of space limitations, it will not be possible to discuss each of these steps. However, several of the more important seismic capability aspects will be highlighted.

4. Seismic capability walkdown In order to be cost-efficient, a seismic margin review should incorporate a step where elements are quickly "screened out" from further review, or a detailed SMA, based upon experience and judgment concerning their seismic ruggedness to withstand the specified SME level. Such "screening out" of rugged elements enables the margin review to quickly concentrate on those elements for which one might have some legitimate concern about seismic ruggedness. Thus, the effort is con° centrated to where it might have some value rather than being spread more thinly over documenting seismic capacity of all dements. Essential ingredients for this quick "screening out" of elements are: (1) A plant walkdown by experienced people capable of exercising the necessary judgment to "screen out" from further review, seismically rugged dements within a specific plant. (2) A general industry and regulatory consensus that, in fact, there are wide classes of components (elements) in nuclear power plants which have demonstrated a substantial seismic ruggedness either because of their performance in past earthquakes, available generic ruggedness or fragility data, or because generally accepted seismic margin capacity evaluations have been previously performed on like components in previous seismic margin or seismic PRA studies. The first ingredient is covered by the performance of a plant-specific walkdown, while the second ingredient requires the establishment of some guidelines which are

generally accepted as conservative for screening. Reconunendations for such guidelines are summarized in Section 6. Seismic margin analysis work should concentrate on failure modes which have actually occurred in heavy industrial process plants and fossil fuel power plants in past earthquakes. Cost-effective SMA requires elimination from consideration those elements which, based upon historical performance, fragility or high-level qualification testing, and judgment can be quickly assessed to have high-confidence-low-failure-probability (HCLFP) capacities greater than the SME level. However, when real failures have occurred, a significant percentage of them have been due to seismic spatial system interactions (SI) due to heavy masses impacting critical elements, or seismic-induced flooding of critical elements. A cost-effective SMA cannot ignore the most credible SI failure modes, but it should not become bogged down into a typical nuclear industry SI study. It requires experience to quickly spot the most critical SI failure modes without becoming overwhelmed in an SI failure mode walkdown study. Quick screening-out from the SMA those rugged elements which are clearly safe at the SME level, and screening-in of the most critical seismic-spatial-systems-interactions (SI) by judgment requires considerable experience in earthquake engineering for critical process facilities. One method to accomplish the plant-specific screening is through the use of a three- to five-member Seismic Review Team (SRT) who between them should possess the following qualifications: (1) Knowledge of the failure modes and performance of components and structures during strong earthquakes in heavy industrial process plants and fossil fuel power plants including structures, tankage, piping, process and control equipment, and active electrical components. (2) Knowledge of nuclear design standards and seismic design practices for nuclear power plants including structures, tankage, piping, process and control equipment, and active electrical components. (3) Ability to perform fragility/margins-type capability evaluations including structural/mechanical analyses of the above mentioned elements. (4) Some general understanding of seismic probabilistic risk assessment conclusions and systems analysis. It is not necessary that each member of the team individually have strong capability in all of these areas or strong seismic experience for all of the dements identified in the success paths being considered. However, between them at least one member and preferably two members of the SRT should have strong experience

R.P. Kennedy et al. / A seismic margin assessment procedure

in each of the above-mentioned areas so that in the composite, the SRT is strong in all of these areas. The SRT then takes primary responsibility for the seismic capability walkdown. The purposes of the SRT walkdown are to: (1) Aggressively screen out from the margin review all elements for which they are prepared to state high-confidence-low-failure-probability at the specified SME level based upon their combined experience and judgment, using earthquake experience data as appropriate. (2) Clearly define the failure modes (such as anchorage failure or relay chatter) and elements (such as water storage tank or a lightly reinforced block wall) which are not screened out and the types of review which should be conducted. (3) Add to the SMA any SI items which the SRT judges to be potentially serious problems (such as air lines to air-operated equipment which will not fail safe or has an accumulator, or heavy, questionably secured space heaters or lights over critical batteries, etc.). Because of the benefits of intercommunication, the SRT should not break down into groups which are too small. At least two (2) members and preferably the entire SRT should visit each element together. As much as possible, decisions should be made on the spot, and the walkdown should proceed at a pace that makes this possible. Decisions to screen out should be unanimous. Otherwise, the concerns should be documented and left in for further review.

5. Dominant contributors to seismic risk

The seismic margin review should concentrate on those elements which have been judged to be the dominant contributors to seismic risk. The results of 20 seismic probabilistic risk assessment (SPRA) studies have been reviewed. Each SPRA was performed on a plant with an SSE level between 0.1g and 0.25g located in the eastern half of the U.S. From these past SPRA studies, the dominant contributors to seismic risk have consistently been: 1. Loss of power 2. Electrical equipment 3. Water storage tanks 4. Civil structures 5. Soil failure (liquefraction) 6. Non-load bearing walls 7. Major NSSS component supports 8. Reactor scram devices 9. Systems-interactions

65

In most of the more recent and more detailed seismic PRAs, loss of power (station blackout) has been either the overwhelming dominant contributor to seismic-induced core damage or at least a major contributor. A seismic-induced anticipated transient without scram (ATWS) resulting from failure to insert the control rods plus either the loss of the diesel generators or loss of adequate cooling water has also generally been an important contributor to core damage. Some form of large LOCA with containment bypass has often dominated the seismic risk of early fatalities. The most dominant cause of this seismic-induced LOCA has been the seismic failure of major NSSS component (reactor vessel, steam generator, pressurizer, etc.) supports. Such failures can be postulated to lead to sufficiently high overpressure to cause breach of containment. It should be noted that, with one possible exception, seismic-induced structural failure of the containment has never appeared to be a dominant contributor to seismic risk. The seismic margin of the containment is sufficiently large relative to the margin of other components (Primarily NSSS component supports) to prevent a seismic-induced containment failure from having any dominance. Inertial (dynamic shaking) induced piping failures have also never been reported to be a significant contributor to seismic risk. On the other hand, in some cases where piping interconnects between two buildings and has inadequate flexibility between supports in different buildings, excessive building deflections have been considered to lead to piping failures which result in seismic risk. This observation from seismic PRAs that inadequate piping flexibility and excessive relative support deflections are the more likely contributor to seismic-induced piping failures rather than dynamic shaking effects is totally consistent with observed behavior of industrial facility and fossil-fuel power-plant piping during past earthquakes with ground motion up to 0.5g. Each of these nine dominant contributors to seismic risk will be briefly discussed. 5.1. Loss of power

Station blackout requires the loss of both offsite and onsite power. Both sources of power have vulnerable features. For offsite power, the station switchyard, with its large yard-mounted switchgear and ceramic insulators, plus the above-ground transmission lines have proven to be vulnerable in past earthquakes. A study of fossil fuel plants in California indicates that the seismic fragility of the plant connection to the power grid is highly variable even when one only considers ground motion records with broad frequency content, and rea-

66

R.P. Kennedy et al. / A seismic margin assessment procedure

sonably long duration from about magnitude 6.0 and greater earthquakes. With such ground motion, at least one power plant lost connection to the power grid due to switchyard and transmissions line damage at a ground motion of about 0.13g. On the other hand, in a number of cases, plants have stayed on line with ground motions in excess of 0.5g. A study of this limited data seems to indicate a median fragility (50% probability of failure) of about 0.30g to 0.35g, with high confidence (95%) of a low probability (5%) of failure at 0.10g and certainly less than 805[ probability of failure at 0.50g. Analytical studies tend to support these fragility estimates drawn from past earthquake experience. In other words, loss of offsite power is certainly possible in a moderate (0.2g to 0.3g) ground motion, but should not be taken as 100% probable as was done in some of the earlier seismic PRA studies. Once offsite power is lost, the emergency diesel generator must start and operate. In the more recent seismic PRAs where random failures and seismic failures have been coupled, it has been consistently calculated that for ground motion less than about 0.3g, the dominant seismic-induced cause of core damage is seismic-induced loss of offsite power coupled with random failure of the diesels. In other words, if one wishes to reduce the seismic risk from ground motion less than 0.3g, one should concentrate on increasing the reliability of the diesels rather than performing seismic margin studies or upgrades on other components. At ground motions in excess of about 0.3g, seismic-induced failures of the diesels become significant contributors to the risk. The diesel generators themselves have not been the low seismic capacity components. In fact, the authors are not aware of a single case where a diesel generator has ever been damaged in any earthquake of any size. The postulated failure have all been with the peripherals such as the fuel oil storage tank, the day tank, the oil cooler tank, and the heating and venting ducting. The postulated critical failure modes for these peripherals have all been structural modes (i.e., attachment bolt and support failures). Fragilities have primarily been estimated using analytical approaches [4,5] and have varied from plant to plant. However, even for the lowest capacity components, the median capacities have generally exceeded 0.9g and the high confidence of a low probability of failure (HCLPF) estimate has generally exceeded 0.3g. These estimates are consistent with the past earthquake experience data base in which emergency diesels and their peripherals have both demonstrated substantial seismic ruggedness. Even so, if one must reduce the seismic risk of nuclear plants for ground motions in

excess of 0.3g, seismic PRA results indicate that the peripherals of the emergency diesel generators are strong candidates for further review. During the Systematic Evaluation Program (SEP) in the United States (1977-1980), batteries and their racks were identified as major seismic-induced risk contributors because of the lack of seismic bracing in the racks and lack of lateral and longitudinal support of the batteries. Fragility testing of batteries has demonstrated that standard 125V DC batteries have substantial seismic ruggedness (HCLPF ground motion estimate of about 0.7g) when properly supported. The problems with battery racks identified in the SEP now appear to have been corrected at most U.S. nuclear power plants. Batteries and their racks have never been identified in a seismic PRA to be a significant contributor to risk. Unless a seismic walkdown of a nuclear plant identifies an obviously deficient battery rack or longitudinal support of batteries in such racks, batteries and their supports do not appear to be a seismic margin or seismic risk issue in low-to-moderate seismic regions. 5.2. Electrical equipment

In the majority of seismic PRAs, it has been assumed that operators could recover from seismic-induced relay chatter and circuit breaker trip. Whenever this assumption has been made, electrical equipment such as motor control centers and in-plant low-voltage and metal-clad switchgear have never been reported to be significant contributors to risk. This finding is consistent with experience in past earthquakes and the conclusions of the Senior Seismic Review and Advisory Panel (SSRAP) [6], which reported that such equipment had an inherent seismic ruggedness so long as it was properly anchored. There does not appear to be a single case of a reported seismic-induced structural failure of this equipment or its cabinet or failure to operate after the earthquake has ended within the U.S. earthquake experience data base, so long as this equipment remains achored. Furthermore, proper achorage is easy to achieve. Seldom has anchorage of this equipment failed in industrial and fossil fuel plants subjected to ground motions up to about 0.5g. Within the experience of the authors, anchorage of such equipment at nuclear plants in the U.S. is generally better and at least as good as that in the data base plants subjected to past occurrences of strong ground motion. The HCLPF capacities calculated in seismic PRAs are generally larger than 0.4g. Clearly, unless a seismic walkdown indicates poor anchorage practice at a nuclear plant, structural failures of such equipment do not appear to be an important

R.P. Kennedy et al. / A seismic margin assessment procedure

seismic margin or seismic risk concern in low-to-moderate seismic regions. However, relay chatter and circuit breaker trip are another matter. Seismic PRA sensitivity studies which assume that there is a relatively significant probability that the operators cannot recover from relay chatter induced circuit breaker trip have shown that this electrical equipment is a major contributor to seismic risk of core damage. Qualification and fragility test data as well as earthquake experience has demonstrated that some relays do chatter at low input shaking levels. Thus, one cannot dismiss the possibility of relay chatter at ground motion levels much above the SSE for which the equipment was qualified. Much further work is needed to properly treat the issue of relay chatter in seismic PRAs. Detailed circuit design, detailed information on the types of relays present, the probability that relay chatter leads to breaker trip, and the probability that the operator cannot recover from breaker trip must all be considered. 5.3. Water storage tanks

Past earthquake experience has demonstrated that steel flat-bottom, skirt-mounted, and leg-mounted vertical water storage tanks have highly variable seismic capacity. Some thanks have failed at ground motions less than 0.2g, while many tanks have survived 0.5g motions. Seismic design practices have been highly varied even in the nuclear power industry. Prior to about 1975, most tank designs were based upon the improper assumption that for impulsive effects the tank could be treated as rigid and be designed for the support acceleration rather than an amplified spectral acceleration. In most cases, about a factor of 2 unconservatism was introduced by this approach. Such unconservatism removes most of the conservatism introduced by other parts of the design process. As a result, tank failures at ground motions only slightly greater than the SSE level must be considered possible. Seismic fragilities can be estimated for tanks by analytical methods [4,5]. Published seismic PRAs have shown high confidence, low probability estimates as low as 0.2g with median (50% probability of failure) estimates as low as 0.5g. Despite such low estimated capacities for some water storage tanks, these tanks have never been reported in a seismic PRA to be a significant contributor to risk. There is redundancy of water storage at a plant, and tank design has been highly non-standard even at a single nuclear plant. Thus, while one tank may have low seismic capacity, some other thank has high seismic capacity. Here is a case where non-standardization has

67

proven to be very beneficial, If all tanks had been designed the same, the benefits of redundancy in a seismic event would have been lost. 5.4. Civil structures

Category I civil structures which have been designed for seismic events have substantial seismic capacity unless a design or construction deficiency occurs. Earthquake experience has shown that one would seldom expect severe structural distress for structures designed similar to nuclear plant structures at ground motions less than 0.5g. Seismic PRA studies on low-to-moderate SSE design plants have reported HCLPF capacities for most civil structures of about 0.5g with median capacities around 1.5g. Even so, at least one weak-line civil structure with an HCLPF capacity around 0.3g and median capacity around 0.9g is found in most published PRAs. Weak-links have been generally due to low ductility in the shear wall or braced frame design, weak floor diaphragms due to large cutouts, foundation failure, or building sliding. Despite the rather high capacity of even the weak-link civil structures, such structures have generally been reported to be significant contributors to the seismic risk. This situation arises because of the conservative assumption made in seismic PRAs that severe structural distress to the civil structure results in failure of all components within the structure. This conservative assumption must be reevaluated before one can actually assess whether civil structures are significant contributors to risk. Building-to-building impact due to an inadequate seismic gap between buildings has also been identified as a major contributor to seismic risk in some PRAs. However, these PRAs may have overemphasized the damage that might result from such impacts. Experience shows that building impacts tend to only lead to local damage. 5.5. Soil liquefaction

To date, no published seismic PRA has identified soil liquefaction as a dominant contributor to risk because their site conditions have not been particularly vulnerable to liquefaction. However, several nuclear plants in the U.S. are founded on soils for which one would expect significant probability of severe soil liquefaction effects at effective ground accelerations of 0.3g to 0.5g from magnitude 6 and greater earthquakes. Such liquefaction is likely to result in tilting of surfacefounded buildings (diesel generator buildings is one

68

R.P. Kennedy et al. / A seismic margin assessment procedure

example) and severing of interconnected buried piping between buildings. For these sites, liquefaction would be expected to be the dominant contributor to seismic risk from effective ground accelerations greater than about 0.3g. Due to the limited data and potential for serious damaging effects, soil liquefaction should be fully investigated (at susceptible sites) for seismic margin earthquakes larger than the plant design basis. 5. 6. Non-load bearing walls

risk of early fatalities only begins to become significant at effective ground accelerations in excess of 0.6g. Such severe ground motion is extremely unlikely in low-tomoderate seismic regions and significantly exceeds the ground motion levels for which experience data exists for fossil fuel and heavy industrial plants in California. Thus, the fragility estimate of these component supports must be based purely on analytical studies, since no experience data exists at these high ground motion levels.

Eastern U.S. nuclear plants often have used either unreinforced or lightly reinforced non-load bearing masonry walls as room dividers and fire barriers. Such walls have demonstrated very poor performance in past earthquakes. Both earthquake experience and seismic analyses indicate that one can expect median capacities of such walls to generally lie in the 0.3g to 0.4g range even after these walls have been strengthened to satisfy the SSE ground motion. Often, these walls have been placed around the battery racks and as fire barriers separating the individual diesel generators where their failure might damage either the batteries or diesels. These walls have seismic capacities similar to that of offsite power. Thus, these walls have sometimes been significant contributors to core damage from station blackout. Here is an example where walls built to reduce fire risk without adequate consideration of seismic risk have increased the seismic risk and possibly the overall risk (fire plus seismic). Clearly, such walls are a primary candidate for seismic margin reviews for earthquakes greater than the SSE level.

In seismic PRAs, the seismic-induced failure to fully insert the control rods results in a postulated ATWS, which in turn has often been a moderate contributor to the seismic risk of core damage. The predominant seismic causes of failure to fully insert the control rods have been reported to be seismic failure of the control rod drive housing and seismic-induced severe distortion of the core geometry. Fragility estimates are generally quite high with HCLPF ground motion estimates between 0.3g to 0.4g and median fragilities around 1.0g. The estimates for distortion of the core geometry must generally be based on NSSS-supplied proprietary studies and tests so that independent verification is nearly impossible. No earthquake experience data exists beyond about 0.15g for these components since they do not exist on fossil fuel or heavy industrial plants. Based on these results, the capacity of control rod drive systems should be reviewed for seismic margin earthquakes exceeding the plant design basis.

5. 7. Major N S S S component supports

5.9. Systems interactions

Published seismic PRAs have indicated that the major NSSS components (reactor vessel, steam generator, pressurizer, etc.) have very substantial seismic capacity. Typically, the weak link has been reported to be the component supports where the fragility estimates are based upon analytical evaluations [4,5]. Such studies have indicated HCLPF capacities generally exceeding 0.5g to 0.6g, with median capacities generally exceeding 1.5g (excluding some BWR recirculation pumps). In other words, very large seismic margin will result beyond the SSE levels (0.12g to 0.25g) because of large conservatism in the design practice for such supports. Even with such high capacities, these component supports have consistently appeared to be the dominant contributor to seismic-induced early fatalities because of the severe consequences of such support failures. In each of the published seismic PRAs, seismic-induced

No systematic or exhaustive study of system interaction (seismic failure of a non-seismic designed non-safety component leading to failure of a safety component) has been performed in any of the published seismic PRAs. However, plant walkthroughs have been conducted as part of each of these PRAs, and the most apparent systems interactions have been incorporated. Several such interactions have been found to be significant contributors to seismic risk and have been eliminated as a result of the seismic PRA. Some examples are presented. Hung ceilings in control rooms can fall. Such failures have occurred in fossil fuel plants with ground motions in the 0.3g to 0.5g range where operators have been injured and access to control boards has been severely impaired due to hung ceiling debris. No damage resulted to the control board, but operator performance was impaired. Cases where non-seismic

5.8. Reactor scram systems

R.P. Kennedy et al. / A seismic margin assessment procedure

designed hung lighting, space heaters, and block walls might fall on batteries and exposed tubing and controls of diesel generators have also been observed in seismic PRA walkthroughs. Non-seismic designed fill lines into safety-related water storage tanks have not been valved close to the tank and have been attached to the tank in a way such that their failure would drain the tank. Each of these potential failures could have a profound effect on the seismic risk in the 0.2g to 0.4g ground motion range where seismic PRAs would otherwise show the seismic risk to be very small. For such ground motion, a systematic review of system interactions would probably result in greater reduction in seismic risk than would a seismic upgrade or reevaluation of safety equipment for earthquakes greater than the design SSE level.

69

that spans between buildings, however, should be reviewed for potential failure due to relative displacement effects for margin earthquakes greater than 0.3g. Fans and cooler units in HVAC systems have generally been shown to have HCLPF capacities in excess of 0.3g. For moderate earthquakes larger than 0.3g, the supports of these systems should be reviewed. Additional attention may be required if shock mounting is used for the HVAC equipment. Cable trays have also been found from past PRA analysis, testing, and earthquake experience to possess capacity in excess of 0.5g. For moderate earthquakes larger than 0.3g, the review should ensure that cables have sufficient slack to accommodate displacements.

6. Seismic capacity screening guidelines 5.10. Other components

There are a few other components which are not found to be consistent risk contributors but may be important in some cases and should be mentioned briefly. Heat exchanger supports have, in some instances, been found to have low high-confidence-of-alow-probability-of-failure (HCLPF) values, although experience from past earthquakes has not indicated that such failures occur. One concern with the failure of a heat exchanger support is that the tank may leak or rupture, or piping may fall. Some plants depend on dams, levees, a n d / o r dikes to provide cooling water. Failure of such structures during an earthquake can also potentially lead to flooding of the plant. In one instance, failure of an upstream dam was found to be a contributor to seismic risk. Such scenarios are highly plant-specific; and since dams are often not designed to have the same high structural capacity as required for nuclear plant components, each structure should be reviewed on a case-by-case basis. Pumps, valves, and piping, in general, have been shown by analysis, earthquake experience, and testing to be quite rugged. For moderate earthquakes larger than 0.3g, there is some potential for possible failure modes related to displacement effects (inertia-induced stress need not be reviewed unless the margin earthquake is significantly larger), particularly piping runs between buildings; motor-operated valves on small lines and in cases where the operator is anchored to the structure but the immediately adjacent piping is not, could present problems; and service water pumps with long cantilevered casings should be reviewed. As for piping, ducting has been shown in PRAs to have a HCLPF capacity exceeding 0.5g, considering inertia-induced stresses. Ducting

Based partially on this information from SPRA studies, the NRC-sponsored Expert Panel on the Quantification of Seismic Margins has provided panel-wide consensus seismic capacity screening criteria [2]. The EPRI Seismic Margin Program has reviewed additional data and has refined the N R C Expert Panel recommendations. The EPRI Seismic Margin Program screening guidance is given in tables 1 and 2. Basically, tables 1 and 2 are intended to provide generic conservative estimates of the ground motion below which it is generally not necessary to perform a seismic margin review for particular elements because, in general, they have demonstrated HCLPF capacities above this ground motion level. Thus, for a given ground motion level, these tables list the elements which should, in general, be "screened out" from margin review because of their generically good performance at this review level. In this way, one concentrates on those components for which a legitimate margin issue might exist. Neither table should be used except in conjunction with a walkdown of plant-specific elements by experienced seismic capacity engineers. Both tables should simply be used as guidance to assist the SRT in "screening out" components during their walkdown. The SRT should exercise their own experience and judgment in the use of this material for any specific component. Tables 1 and 2 provide screening guidance in terms of mean peak ground acceleration (average of two orthogonal horizontal components). Mean peak ground acceleration (pga) is only a very approximate damage descriptor, even for stiff components mounted in stiff buildings. Damage is heavily influenced by the duration and breadth of frequency content of the ground motion [7] and where the component is mounted in a building

70

R.P. Kennedy et al. / A seismic margin assessment procedure

subjected to this ground motion. Tables 1 and 2 are primarily geared to earthquakes with Richter magnitudes between 6.0 and 6.5, producing ground motion with broad frequency content input spectra and total durations of 10 to 15 seconds. These recommendations are keyed to median spectra shapes given in N U R E G / CR-0098 [8]. These recommended screening ground motion levels might become slightly unconservative for a Richter magnitude 7.0 earthquake and could be seriously unconservative for a Richter magnitude 8.0 earthquake. On the other hand, they are excessively conservative for a low-magnitude earthquake which produces high frequency and narrow-banded input spectra with short durations and negligible spectral displace-

Table 1 Summary of civil structures screening criteria for seismic margin evaluation (adapted but modified from [2]) Type of structure

Peak free-field ground acceleration < 0.3g

0.3 to0.5g

> 0.5g

Concrete containment (posttensioned and reinforced)

no

(1)

(2)

Freestanding steel containment

(3) (4)

(3) (4)

yes

Containment internal structures

(5)

(6)

yes

Shear walls, footings and containment shield walls

(5)

(6)

yes

Diaphragms

(5)

(7)

yes

Category I concrete frame structures

(5)

(6)

yes

Category I steel frame structures

(5)

(8)

yes

Masonry walls

(9)

yes

yes

Control room ceilings

(10)

(10)

yes

Impact between structures

(11) (12) (12)

yes

Category II structures with safety-related equipment or with potential to fail Category I structures

(13)

yes

yes

Dams, levees, dikes

yes

yes

yes

Soil failure modes, soft liquefaction, and slope instability

yes

yes

yes

Notes to table 1:

(1) Major penetrations should be evaluated. (2) Major and minor penetrations should be evaluated. The concrete containment structure only needs to be evaluated for peak ground acceleration exceeding 0.8g. (3) No evaluation required if base mat is integral part of pressure boundary or steel pressure boundary is keyed to base mat to prevent slipping. (4) Mark I tori require evaluation for earthquakes exceeding the design basis. (5) Evaluation not required for Category I structures if design was for an SSE of 0.1g or greater or if structure essentially meets 1985 UBC. Zone 4 requirements. (6) Evaluation not required for Category I structures if design was by dynamic analysis for an SSE of 0.1g or greater, and if the structure complies with ACI 318-71 or ACI 349-76 ductility detailing requirements. (7) Evaluation not required for Category I structures if design was by dynamic analysis for an SSE of 0.1g or greater, and if the diaphragm complies with ACI 318-71 or ACI 349-76 ductility detailing requirements, provided the diaphragm seismic loads were explicitly calculated. (8) Evaluation not required if structures were designed using dynamic analysis and meet the requirements of AISC, 7th Edition, 1970 or later. (9) Evaluation not required for walls which have been externally reinforced to withstand an SSE of 0.1g or greater. (10) Inspect for adequacy of bracing or safety wiring. (11) Investigation not required if structures designed by dynamic analysis and displacements are not dominated by soils flexibility. (12) Investigation can be limited to potential for electrical malfunction (relay or contactor chatter) and loss of equipment anchorage in immediate vicinity of impact. (13) Evaluation not required provided the structure is capable of meeting the 1985 UBC Zone 4 requirements. ment. In other words, these recommendations should primarily be coupled to a N U R E G / C R - 0 0 9 8 - t y p e median spectral shape and moderately long durations. Furthermore, tables 1 and 2 screening guidance is primarily intended for components mounted fairly low (less than about 40 feet above grade) in stiff, nuclearpower-plant-type structures. Care should be exercised in using this guidance for components mounted significantly more than 40 feed above grade or in other spots where large resonant buildup of the input motion might occur.

7. Conservative deterministic failure margin approach The E P R I Seismic Margin Program recommends the use of a conservative deterministic failure margin ( C D F M ) approach to estimate the H C L P F capacity of

R.P. Kennedy et al. / A seismic margin assessment procedure

components which have not been screened out of the seismic margin review. In the C D F M approach, the seismic margin earthquake (SME) loadings are only combined with normal operating loads (NOL) which would be expected to occur concurrently with the SME and to use load factors of unity for all loadings. In other words, no conservatism is added in the load combination equations when determining margin for the SME. One exception is that for the containment, which represents the last line of defense, it is probably prudent to combine an accident pressure (Pa) with NOL + SME. This accident pressure should be based upon a postulated small or medium LOCA and might be substantially less than the design pressure. When computing the SME Demand (structure response, or in-structure response spectra), conservatism should be introduced only at the location of greatest uncertainty, which is the specification of the ground input response spectrum, i.e., in the definition of the SME. Preferably, the SME input should be defined in terms of an 84% non-exceedance probability (NEP) site-specific response spectrum. With the SME conservatively defined at about the 84% NEP, there is no need to add additional conservatism in the lesser uncertain response parameters in order to assure about 84% NEP response values at all locations throughout the structure and for input to all components. Ideally, damping, structure modelling parameters, and soil-structure-interaction (SSI) evaluations should all be median-centered. Considerable controversy exists over the damping to use in dynamic models. One should select a conservative estimate of median damping to cover this controversy. The important point is that one should aim to use median damping and not to intentionally introduce conservatism at this point. For structures, N U R E G / CR-0098 [8] damping value ranges are considered appropriate, so long as the structure is significantly stressed by the SME (i.e., damping values in the range of 7 to 10% are appropriate for concrete structures when linear analysis is performed). Other generally appropriate damping values are: Cabinets bolted or welded to floor: 5% Piping: 5% Cable trays: 15% Median-centered SSI evaluations mean that one should take full credit for such things as: vertical spatial variation of the ground motion, kinematic interaction, and scattering of energy (radiation damping) from the structure back into the ground. All of these effects should be median estimated with no intentional con-

71

seroative bias. T h e SME should be defined at a control

point represented by either the free-ground-surface or at the top of a hypothetical stiff material outcropping. For embedded structures on soil sites, these recommendations are expected to reduce computed responses throughout the structure by as much as a factor of 2 below conventional nuclear power plant design practice. Conventional SSI analysis practice in the nuclear industry represents one of the most substantial sources of seismic margin for embedded structures which can rationally be reduced. Considerable uncertainty exists in structural frequency estimates, vertical spatial variation of the ground motion, and SSI frequency shifting effects. Each of these has substantial influence on the SME Demand to components mounted on the structure (i.e., in-structure response spectra). It is unrealistic and potentially very unconservative to ignore this uncertainty and only use median centered structural frequency modeling or median-centered SSI analysis. The effects of structural modeling frequency shifting, SSI frequency shifting, and vertical spatial variation of the ground motion frequency shifting on in-structure spectra should be included by appropriate parameter variation. One should attempt to encompass the effects of a moderate plus or minus one standard deviation parameter shifting variation on instructure spectra. In-structure spectra generated from analyses which incorporate these parameter variations should be treated as a representative input (Demand) to components. Primarily these parameter variations result in frequency shifting of the in-structure spectra. Whenever possible, one should not encompass this frequency shifting effect by enveloping the full range of possible in-structure spectra. Such enveloping creates a broad frequency content envelope in-structure spectra which contains more power than could possibly be produced by any single SME. Broadening, rather than frequency shifting, introduces considerable conservatism which might not be desirable for an SMA. Responses computed in the above manner certainly have less than a 16% probability of being exceeded if the SME were to occur. In order to achieve a HCLPF, one must conservatively estimate the structure or component seismic capacity for comparison with the estimated demand. One should have high confidence that there is less than a 5% probability of failure, even if this demand is reached. Since it is unlikely that the computed demand will be reached, the net effect of this approach is a HCLPF. For failure modes evaluated by analysis, a seismic capacity estimate requires an estimate of: 1. material strength,

72

R.P. Kennedy et al. / A seismic margin assessment procedure

Table 2 Summary of equipment and subsystems screening criteria for seismic margin evaluation (adapted but modified from [2]) Type of structure NSSS primary coolant system (piping and vessels) NSSS supports Reactor internals Control rod drive housings and mechanisms Category I piping Active valves Passive valves Heat exchangers Atmospheric storage tanks Pressure vessels Buried tanks Batteries and racks Diesel generators (includes engine and skid-mounted equipment) Horizontal pumps Vertical pumps Fans Air handlers Chillers Air compressors HVAC ducting and dampers Cable trays Electrical conduit Active electrical power distribution panels, cabinets, switchgear, motor control centers Passive electrical power distribution panels, cabinets Transformers Battery chargers Inverters Instrumentation and control panels and racks Temperature sensors Pressure and level sensors

Peak free-field ground acceleration < 0.3g

0.3 to 0.5g

> 0.5g

no (1) (2) (26) (4) (5) no no (8) yes (8) (10) (11)

no (1) (2) (3) yes yes (5) (6) (25) no (9) yes (9) (10) (11) (25)

yes yes yes yes yes yes (7) yes yes yes yes yes

(12) no no (14) (14) (14) (14) (5) no (18) (19) (20)

(12) (25) (13) (25) (15) (15) (15) (15) (5) (16) (17) (25) (18) (19) (20) (25) (19) (21) (22) (25) (23) (23) (19) (20) (24) (24)

yes yes yes yes yes yes yes yes yes yes yes

(19) (21) (22) (23) (23) (19) (20) (24) (24)

yes yes yes yes yes (24) (24)

Notes: see next page.

2. static capacity or failure equation, 3. inelastic energy absorption capability. Each of these parameters should be conservatively estimated to achieve the above-recommended level of capacity conservatism. Material strengths used in the C D F M approach should be sufficiently conservative so that there is very little likelihood that actual strengths are less than those used in the margin review. When test data are available, one should use approximately 95% exceedance probability strengths to achieve this goal. Otherwise, one should use code or design specified m i n i m u m strengths. In most cases, the use of 95% exceedance actual test data

strengths will result in a 10 to 20% increase over code or design-specified m i n i m u m strengths. In most cases, static capacity estimates should be based on code-specified ultimate capacity approaches, since there is very little possibility of failure at capacities less than given by these code equations. For concrete, the ACI ultimate strength approach with the appropriate capacity reduction factor, q~, included should be used. For structural steel, the AISC-Part 2 maximum strength approach with a load factor of unity is considered appropriate. For ASME components, the Service Level D approach is adequately conservative for structural failure modes. Functional failure modes may

R.P. Kennedy et al. / A seismic margin assessment procedure

73

Notes to table 2:

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

(12) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26)

BWR piping with suspected intergranular stress corrosion cracking may require evaluation. Evaluation not required if supports are designed for combined loading determined by dynamic SSE and pipe break analysis. Evaluation recommended for PWR pressurizer supports and BWR reactor vessel and recirculation pump supports. Evaluation not required if CRD housing has lateral seismic support. Walkdown of representative piping and ducting systems should be conducted. Evaluation recommended for MOVs in piping lines of 2 inches diameter or less. Walkdown to assure that valves do not impact adjacent structures or equipment. Margin evaluation only needs to consider anchorage and supports. For vessels designed by dynamic analysis or equivalent static analysis enveloping vessel inertial and piping loading, only the anchorage and supports require evaluation. For vessels not meeting these criteria, all potential failure modes require evaluation. Evaluation of piping connections is required. Other failure modes do not require evaluation. Batteries mounted in braced racks designed for seismic loads or qualified by dynamic testing do not require evaluation. Rigid spacers between batteries and end restraints are required. Batteries should be tightly supported by side rails. Supports and anchorage should be reviewed. Margin review should be conducted for anchorage and attachment of peripheral equipment. Can be done by visual inspection for an SME of 0.3g or less. Margin evaluation required for vertical pumps with unsupported lengths of casing below the flange exceeding 20 feet or pumps with shafts unsupported at their lower end. All units supported on vibration insulators require evaluation of anchorage. Evaluation should focus on anchorage and supports. Evaluation required only for potentially large relative displacements between structures or equipment and structures. Guidance beyond scope of paper. No evaluation required if supports meet the National Electrical Code. Anchorage evaluation required. Walkdown required to verify device attachment to cabinets or panels. Relays, contactors, switches, and breakers must be evaluated for chatter and trip if functionality during strong shaking is required. Anchorage evaluation required. Liquid-filled transformers require evaluation of overpressure safety switches. Solid state units require anchorage checks. Others require evaluation. Insufficient data are available for screening guidelines. Emphasis should be on attachments. Units mounted on structures at elevations exceeding 40 feet above grade should be reviewed if 5% damped horizontal floor spectra exceeds 2g. Insufficient data to enable recommendations to be made.

require lesser limits. In some cases, it is known that code equations for capacity are excessively conservative. In these cases, failure capacity equations based upon actual failure test data or more rigorous evaluation may be used. However, these failure equations should not be median-centered. They should be sufficiently conservatively biased to encompass nearly all of the failure test data. Because of the scarcity of failure test data, it is often difficult to extrapolate to the 95% exceedance static capacity equation. Secondly, because of conservatism introduced in the material strength and inelastic energy absorption factors, it is not necessary to introduce such large conservatism in the static capacity equation for a CDFM review. Moderate conservatism (about 84% exceedance) is sufficient at this step. Nearly all structures and components exhibit at least some ductility (i.e., ability to strain beyond the elastic limit) before failure. Because of the limited energy content and oscillatory nature of earthquake ground mo-

tion, this ductility is highly beneficial in increasing the seismic margin against failure for structures and components. The inelastic energy absorption factor, Fv, represents the ratio of the SME at which a certain system ductility/~ is reached to the earthquake level for which failure would be predicted by linear elastic analysis. The additional seismic margin due to this inelastic energy absorption factor F~ should be considered in any failure margin review. Ignoring this effect will lead to unrealistically low estimates of the failure margin. It is impossible to correlate performance of structures and equipment in past experience with capacities predicted by elastic analyses without considering the F~ factor. For a CDFM, this inelastic energy absorption should be conservatively estimated. In most cases, one will want to use linear analysis techniques. With linear analysis, the easiest way to account for the inelastic energy absorption capability is to multiply computed seismic stresses by a reduction

74

R.P. Kennedy et aL / A seismic margin assessment procedure

factor, k, to obtain effective stresses to add to nonseismic stresses:

(1)

Oeffective seismic = k ° s e i s m i c ,

where: k =

I/F..

For all but the most brittle failure modes, one can very conservatively choose k = 0.8. Alternately, one could perform some form of nonlinear evaluation to justify a much lower k value in many cases. Often, functional failure modes will have to be assessed using test data. Component-specific qualification or fragility test data may be used for this purpose. Alternatively, one might use a generic equipment ruggedness spectrum (GERS). Such GERS are currently being developed by EPRI for many specific types of equipment [9]. The GERS are intended to represent the highest 5% damped test response spectrum (TRS) available for a specific type of equipment for which no failures were observed. The GERS are always set lower than the lowest TRS for which failures were observed, if any such failure test data exists. The GERS are sufficiently conservatively biased to cover as broad a generic equipment class as possible. If one were to assume that either component specific qualification test data or the applicable GERS represents an upper bound on seismic input TRS for a piece

of equipment, then one would clearly need a margin factor between the computed in-structure required response spectrum (RRS) and the TRS in order to achieve a HCLPF capacity. However, one should recognize that either GERS or component specific qualification tests are often very conservatively biased estimates of the maximum seismic capability (fragility) of a piece of equipment. In most cases, the actual factor of conservatism is highly uncertain. This issue needs further investigation. The following recommendations may be too conservative. Even so, at this time, it is suggested that the following criteria be used in an SMA review: (1) Within the frequency range of concern, over any 20% frequency shift bandwidth (for example, 10 to 12 Hz), the average ratio of TRS to any individual frequency shifted but not broadened RRS should exceed a factor of 1.3. (2) Over frequency bandwidths corresponding to less than 20% frequency shift, any condition where an individual frequency shifted but not broadened RRS exceeds the TRS should be individually assessed as to its consequences. It is highly unlikely that a small exceedance (up to a factor of 1.5) over a narrow frequency band could lead to functional failure. These recommendations are illustrated in fig. 1. The most common failure mode for equipment is anchorage failures [6]. A margin review should heavily concentrate on the review of anchorage failure modes unless the SRT can dismiss the possibility of such

INDIVIDUAL FREQUENCYSHIFTED, BUT UNBROADENED RRS's ASSESSCONSEQUENCES OF THESEEXCEEDANCES

_J ¢_1 c_)

TRS/~ L

20% FREQUENCYSHIFTBANDWIDTH, TRS/RRS>I.3 IS ACCEPTABLE

FREQUENCY (HZ) Fig. 1. Comparison of individual frequency shifted RRS and TRS.

R.P. Kennedy et al. / A seismic margin assessment procedure

failures during their walkdown. Such failure modes can be evaluated analytically. EPRI has previously funded an extensive effort to develop seismic anchorage guidelines for nuclear plant equipment [10]. It is recommended that these guidelines be followed for seismic margin reviews.

[4]

[5]

[6]

Acknowledgement The work summarized in this paper has been funded by the Electric Power Research Institute, R P 2722-1. [7]

References [1] Proceedings: EPRI/NRC Workshop on Nuclear Power Plant Reevaluation to Quantify Seismic Margins, EPRI NP-4101-SR, Electric Power Research Institute (August 1985). [2] R.J. Budnitz, et al., An approach to the quantification of seismic Margins in Nuclear Power Plants, NUREG/CR4334, Lawrence Livermore National laboratory, prepared for U.S. Nuclear Regulatory Commission (August 1985). [3] P.G. Prassinos et al., Recommendations to the Nuclear regulatory Commission on Trial Guidelines for Seismic Margin Reviews of Nuclear Power Plants, NUREG/CR-

[8]

[9]

[10]

75

4482, Lawrence Livermore National Laboratory, prepared for U.S. Nuclear Regulatory Commission (March 1986). R.P. Kennedy et al., Probabilistic seismic safety study of an existing nuclear plant, Nucl. Engrg. Des. 59 (1980) 315-338. R.P. Kennedy and M.K. Ravindra, Seismic fragilities for nuclear power plant risk studies, Nucl. Engrg. Des. 79 (1984) 47-68. R.P. Kennedy W.A. yon Rieseman, P. Ibanez, A.J. Schiff and L.A. Wyllie, Use of Past Earthquake Experience Data to Show Seismic Ruggedness of Certain Classes of EquiPment in Nuclear Power Plants, Seismic Qualification utility Group (August 1984). R.P. Kennedy et al., Engineering Characterization of Ground Motion-Task 1, Effects of Characteristics of Free-Field Motion on Structural Response, N U R E G / CR-3805, Nuclear Regulatory Commission (May 1984). N.M. Newmark and W.J. Wall, Development of Criteria for Seismic Review of Selected Nuclear Power Plants, NUREG/CR-0098, Nuclear Regulatory Commission (May 1978). Seismic Equipment Qualification Using Existing Test Data, EPRI NP-4297, Prepared by ANCO Engineers for the Electric Power Research Institute, Interim Report (OCtober 1985). Development of Anchorage Guidelines for equipment in Nuclear Power Plants, Prepared by URS/John A. Blume & Associates for Electric Power Research Institute, Draft (November 1985).