ML14058A060
| ML14058A060 | |
| Person / Time | |
|---|---|
| Site: | Oconee  |
| Issue date: | 03/16/2010 |
| From: | Jeffrey Mitman NRC/NRR/DRA |
| To: | Cunningham M NRC/NRR/DRA |
| Shared Package | |
| ML14055A421 | List:
|
| References | |
| FOIA/PA-2012-0325 | |
| Download: ML14058A060 (17) | |
Text
March 16, 2010 MEMORANDUM TO: Mark A. Cunningham, Director Division of Risk Assessment Office of Nuclear Reactor Regulation THRU: Lois M. James, Chief Probabilistic Risk Assessment Operational Support Branch Division of Risk Assessment Office of Nuclear Reactor Regulation FROM: Jeffrey T. Mitman, Senior Reliability and Risk Analyst Probabilistic Risk Assessment Operational Support Branch Division of Risk Assessment Office of Nuclear Reactor Regulation
SUBJECT:
GENERIC FAILURE RATE EVALUATION FOR THE JOCASSEE DAM The Probabilistic Risk Assessment Operational Support Branch (APOB) evaluated data on large dams in the U.S. to determine a generic failure rate applicable to the Jocassee Dam.
This evaluation was initially performed in the middle of 2008. The attached evaluation updates and documents that analysis.
The staff estimated generic dam failure rates for large rockfill dams, which it considers applicable to the Jocassee Dam, as 2.8E-4/dam-year. Given the nature of the data and the assumptions involved in narrowing the applicable failure events and subset of the U.S. dam population comparable to this specific dam, the staff performed a Bayesian analysis. Using available data on the domestic inventory of dams and dam failures, the range obtained varies between 1.3E-4/dam-year and 4.8E-4/dam-year (5 th - 9 5 th percentile) around a mean of 2.8E-4/dam-year.
Enclosure:
Generic Failure Rate Evaluation for Jocassee Dam
MLXXXX rV'L- y )(UVI Hr I\
OFFICE NRR/DRA/APOB NRR/DRA/APOB NRR/DRNAPOB NRR/DRA NRR/DRA/APOB NRR/DRA NAME JMitman FFerrante JVaiN SLaur LJames MCunningham DATE
C US.NRC UNITED STATES NUCLEAR REGULATORY COMMISSION ProtectingPeople and the Environment Generic Failure Rate Evaluation for Jocassee Dam March 15, 2010 Probabilistic Risk Assessment (PRA) Analyst: James Vail, Reliability and Risk Analyst, NRR/DRAIAPOB Probabilistic Risk Assessment (PRA) Analyst: Fernando Ferrante, Reliability and Risk Analyst, NRR/DRA/APOB Probabilistic Risk Assessment (PRA) Analyst: Jeff Mitman, Senior Reliability and Risk Analyst, NRR/DRAIAPOB Peer Reviewer: Steven A. Laur, Senior Technical Advisor NRR/DRA SESITIV_ INFORMATION -OFFICAL USE O`LY
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM BY DIVISION OF RISK ASSESSMENT'S PRA OPERATIONAL SUPPORT BRANCH The following documents a generic dam failure rate analysis applicable to the Jocassee Dam performed by the PRA Operational Support Branch (APOB) of the Division of Risk Assessment (DRA) in the Office of Nuclear Reactor Regulation (NRR). The analysis, technical justifications, and databases used in support of the calculations for the derived value are briefly discussed.
Portions of this evaluation were initially performed in 2007 but not formally documented at that time.
Approach The approach used in deriving a generic failure rate value applicable to the Jocassee Dam included: (i) an evaluation of the physical characteristics and description of the dam, (ii) an assessment of the overall U.S. dam population for those with similar features to the Jocassee Dam, (iii) a study of U.S. dam performance information for failure events that may be applicable to this subset of the overall population, and (iv) a calculation of a point estimate, as well as consideration of the uncertainty involved, for the failure rate given the observed failure events and the observed time period (in dam-years).
Jocassee Dam Description The Jocassee Dam is located in northwest South Carolina, forming a reservoir (Lake Jocassee) with a 7565-acre surface area, a water volume of 1,160,298 acre-feet, and a total drainage area of 147 sq-miles at full pond (1,110 feet elevation above mean sea level). The reservoir was created in 1973 with the construction of the dam. The Jocassee Dam is an embankment dam with an earthen core and rockfilled and random rockfilled zones (see Figure 1).
(b)(7)(F)
S NLY
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc The dam is 385 feet in height (1,125 crest elevation above mean sea level) and 1,825 feet in length and, along with two homogeneous earthfill dikes and a reinforced concrete spillway, is part of a hydroelectric station and pumped storage project. The underground powerhouse generating units receive water from two cylindrical intake towers through eight openings. The water is channeled from the intake towers to four hydro turbines by two bifurcated power tunnels which are constructed through the bedrock of the east abutment. Two gates 33 feet in height and 38 feet in width control the outflow of the spillway.
Databases The staff used two databases to obtain information about the population of dams in the US: the National Inventory of Dams (NID), maintained by the US Army Corps of Engineers, and the National Performance of Dams Program (NPDP), developed by the Department of Civil and Environmental Engineering at Stanford University. The NID database contains data describing multiple attributes such as dimensions, type, impoundment characteristics, etc. The NPDP database contains a collection of dam incident reports searchable by various parameters including dam type, incident type, and consequences.
Failure Events Table 1 lists the applicable dam failures initially derived from the NPDP database. To choose these 13 failures, the analysts used criteria based on the previously discussed dam characteristics (i.e., dam type and height). However, due to the ambiguity in the classification of the dam type (i.e., based on material composition) between and within the NID and NPDP databases, as well as the lack of information to establish an exact link with the Jocassee Dam characteristics for every data point, the staff considered both rockfill dams and mixed-rockfill dams (i.e., those classified exclusively as rockfill dams as well as mixed dam types that include rockfill in their categorization). It should be noted that the NPDP database does not list any failures post-2006 and at least two well-known large dam failures in the U.S. are not included:
the Big Bay Dam in Mississippi (March 2004) and the Taum Sauk Reservoir (December 2005) in Missouri. While the Big Bay Dam was an earthen dam (i.e., excluded based on dam type),
the Taum Sauk Reservoir consisted of a concrete-faced rockfill dam approximately 100 feet in height and was, therefore, included in the current analysis.
Additionally, the list was screened to take into consideration (i) failure events observed between 1900 and 2005, and (ii)failure events observed between 1940 and 2005; under the assumption that events prior to these construction periods could produce different results representative of distinct design practices. In part, this choice was due to the lack of information on the exact construction date of several dams in the database. The staff expended an extensive effort to determine the construction completion date for several dams for which the information was missing in the NPDP database (this information is included in Table 1).
Several failures listed in Table 1 have (or are assumed to have) occurred within a few years of either the start or completion of construction (e.g., the Lower Hell Hole Dam and the Frenchman Dam failures). Based on the information available and the estimated completion dates, the staff screened out such failures since the occurrence of the events was assumed to be related to the construction phase and, therefore, not applicable to a mature dam such as Jocassee.
Finally, the analysts chose to include the Dresser No. 4 Dam failure, because they deemed this dam to be similar to the Jocassee Dam in composition (i.e., a large mixed earthfill-rockfill dam),
-SENSITIVE INFOFFII 2
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc despite the fact that it is listed as a tailings dam (i.e., a dam theoretically built under lower standards of quality and maintenance).
Therefore, the final list of failures of dams similar to, and therefore applicable to, the Jocassee Dam includes 6 failures occurring between 1900 and 2005. These six failures are highlighted in Table 1. The staff included these failures based on the following criteria: (i) rockfill or mixed-rockfill dam type, (ii) dam height above 50 feet, (iii) failure occurring after 1900, and (iv) no failures during or within a few years of completion of construction. Note that if failures occurring prior to 1940 are screened, then only 4 events remain: (1) Taum Sauk, (2) Dresser No.4 Dam, (3) Skagway, and (4) Kern Brothers Reservoir. It should be noted that there are 1 to 3 failures of dams built between 1940 and 2005 depending on whether the entries with unknown construction dates are excluded or not, respectively (in similar fashion, there are 3 to 5 failures for dams constructed between 1900-2005 excluding or not entries with unknown construction dates, respectively).
Total Dam-years Calculation To calculate the dam failure rate, the staff needed to obtain the total number of dam-years of both failed and non-failed dams. The analysts extracted a subset of dams from the NID database based on a set of parameters to narrow the US population of dams to those reflecting the characteristics of the Jocassee Dam discussed above, i.e., large rockfill dams. They assumed that dams above 50 feet in height appropriately reflect design practices and structural characteristics of larger dams such as Jocassee. This height criterion was consistent with the large dam definition (WCD, 2000) established by the International Commission on Large Dams (ICOLD) which "defines a large dam as a dam with a height of 15m or more from the foundation." If dams are between 5-15 meters high and have a reservoir volume of more than 3 million cubic meters, ICOLD also classified such dams as large. Hence, the staff used this definition as a screening criterion. The dams considered for calculation of the total dam-years were those in the NID database that were categorized exclusively as 'Rockfill' dams (i.e., those listed under the 'ER' abbreviation, intended to correspond to rockfill dams for NID cataloguing purposes).
The staff included the dam-year contributions from Skagway and the replacement for the failed Frenchman Dam, while those from Kern Brothers Reservoir, Dresser No. 4 Dam, Penn Forest, and the failed Frenchman Dam were not included. This was because the staff judges that including the dam-year contribution from these specific dams would not significantly impact the resulting dam-year total. The staff calculated the final result using the difference between the last year in the available data (2005) and either 1900 or 1940. For the 1900-2005 period, the staff obtained a total of 21,490 dam-years; while for 1940-2005 the result was 13,889 dam-years. See Appendix A for a tabulation of the dams and the associated dam-years.
"-ENSITIVE INFORATIO, OFF-',CIAL UE3_. ."
3
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc TahbA I1: Initial Liqt nf dam faiirp Avpnh-Taum Sauk Overtopped due to over-pumping of reservoir. Independent analysis.
2005 1963 Ove0cping RocM 94 Indicated several root causes (eqg., lack of mofitorng, spillway).
Iwer No.4 Earth Rocilt Dam 1975 Unknown Piping /Tailin 105 Catastrophic failure that created a breach 300 feet wide inthe levee.
Skagay 165 125 Inflow Fioed-Skagway 1965 1925 Hydrologic Event Rockfill 79 The dam failed duinng aflood in1965, Hell Hole 1964 1964 Not Known Rockfill 410 Dam failed during construction. Overtopped by 100 feet -washing out most of the fill.
Penn Forest 1960 1960 Piping Concrete Eat 151 Partial failure, Sinkhole occurred inupstream slope of dam.
Rocklill Frenchman 1952 1951 Inflow Flood- Rockfil6 Runoff from melting snow. Adike section was overtopped early Dam Hydrologic Event morning Apni 15,1952. Later that day, dam breached, Kern Brothers 1949 Unknown Settlement Earth RoIll 54 Failure due to excessive settlement of fill.
Reservoir Blowout failure under concrete spillway weir structure during period Lake Francis 1899 1899 Piping Earth Rockfill 79 of heavy spillway flow. Spillway failure thought to be due to piping in soft saturated foundation.
Lafayette 1928 1928 Embankment Slide Earth Rockfill 132 Foundation slide during construction (at 120 feet). Height raised to 170 feet in1932. Not sure ifthis isconsidered afailure.
Manitou 1924 1917 Seepage Earth Rockfill 123 Partial failure was disintegrating and converted into gravel fill.
Failure by piping through abutment; undermined by passage of water Lyman 1915 1912 Piping Earth Rockfill 76.4 under cap of lava rock which flanked dam and extended beneath spillway. Main part of dam uninjured.
Lower Otay 1916 1897 Spillway Earth Rockll 154 Foundation slide during construction (at 120 feet). Height raised to 170 feet in1932, Not sure Ifthi isconsidered afailure.
Failure by piping through abutment; undermined by passage of water Black Rock 1909 1908 Piping Earth Rockfill 70 under cap of lava rock which flanked dam and extended beneath spillway. Portion of spillway dropped 7feet; some fill at south end washed out. Main part of dam uninjured.
4
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc Generic Point Estimate of the Dam Failure Rate The staff calculated the point estimate by dividing the number of applicable dam failures (see Table 1 above) by the total applicable dam-years (derived as described previously). Assuming a 1900-2005 range for the year of occurrence of the failure events and the dam-year estimation (based on completion year), the analysts obtained a failure rate of 2.8E-4 per dam-year. When considering a 1940-2005 range, the staff obtained a result of 2.9E-4 per dam-year.
Because the NID database does not give information regarding the quality of design, construction and/or maintenance, and the NPDP database does not consistently supply information on the dam health (i.e., is it well maintained?) at time of failure, the staff could not derive failure rates for above or below average built and maintained dams. This lack of information precluded the staff from making any judgment as to whether Jocassee is or is not an above average designed, constructed and maintained dam deserving of a failure frequency different than an average failure frequency.
Additionally, the staff recognizes that ambiguity and lack of complete information with respect to dam type, construction completion data, and dam incident reporting, may result in variations in the failure rate estimation. Therefore, the staff performed a simple sensitivity study in order to evaluate the changes due to screening failure events and cut-off year criteria. The results are shown in Table 2 for an assumed number of failures and clearly indicated that the results exhibit small variations for the period cut-off selected (1900-2005 and 1940-2005) and the number of failures considered (6 and 4, respectively). Additionally, the extent of the variation in the point estimate is shown for other number of failures and cut-off years based on the subset of dams selected. The table illustrates that the order-of-magnitude failure frequency estimate does not change significantly ifthe number of failures is increased or decreased slightly.
Table 2: Failure Rate Sensitivity Analysis ASSUMED NUMBER OF FAILURES CUT- DAM-OFF YEARS # DAMS 1 2 3 4 5 6 7 ALL 25137 484 4.OE-05 8.OE-05 1.2E-04 1.6E-04 2.0E-04 2.4E-04 2.8E-04 1900 21490 466 4.7E-05 9.3E-05 1.4E-04 1.9E-04 2.3E-04 2.8E-04 3.3E-04 1910 19778 449 5.1E-05 1.OE-04 1.5E&04 2.OE-04 2.5E-04 3,0E-04 .3.5E-04 1920 18389 434 5.4E-05 .1.1E-04 1.6E-04 2.2E-04 2.7E-04 3.3E-04 .3.8E-04 1930 16475 410 6.1 E-05 1.2E-04 1.8E-04' 2.4E-04 3.OE-04. 3.6E-04 4.2E-04 1940 13889 373 7.2E-05 1.4E-04 2.2E-04 2.9E-04 , 3.6E-04 4.3E-04 5.0E-04 1950 12269 346 1960 8453 270 1970 3242 143 1980 1339 82 ml 1990 381 36 RATE GIVEN # NUMBER OF FAILURES AND CUTOFF YEAR I CENIT-1,TVE IhFORMATIMON - OF FIkCI;AL U3E ONLYr 5
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc Bayesian Estimate of the Dam Failure Rate To evaluate the dam failure rate uncertainty, the staff conducted a Bayesian analysis of the failure rate for the 1900-2005 period via a Bayesian analysis approach (Atwood et al, 2003). In this approach, a prior distribution was assumed from the number of failures and dam-years for all large dams (according to the ICOLD definition) identified in the NID and NPDP databases.
Failures identified as 'infantile failures' in NPDP were excluded and only dams built since 1900 according to NID were used for total dam-year calculation. Under these assumptions, the total number of failures for all large dams for 1900-2005 was 84 with a total of 260,960 dam-years.
This corresponds to a point estimate of the failure rate equivalent to 3.2E-4/dam-year. A distribution was fitted around this mean. The number of dam failure events was modeled as a Poisson distribution for which its conjugate prior was assumed to follow a Gamma distribution (i.e., the conjugate prior in a Gamma-Poisson model). The staff, based on judgment, chose a Gamma distribution with the point estimate obtained from the large dam failure rate above and a 5th percentile corresponding to 1E-5/dam-year. With these assumptions, the staff obtained a prior Gamma distribution with parameters a = 0.8333 and P = 2589, which has a 5th percentile equivalent to 1E-5/dam-year and a 9 5 th percentile corresponding to 1 E-3/dam-year. The staff updated this prior distribution with the data used to obtain the large rockfill dam point estimate (e.g., 6 failures in 21,490 dam-years) to calculate the posterior distribution. The resulting posterior has a mean of 2.8E-4/dam-year, a 5th percentile of 1.3E-4/dam-years, and a 95th percentile of 4.8E-4/dam-years (with parameters a = 6.8333 and P3 = 24,079). Figure 2 shows both the generic large dam prior and the posterior specific to rockfill dams.
Conclusions The staff estimated generic dam failure rates for large rockfill dams, which it considers applicable to the Jocassee Dam, as 2.8E-4/dam-year. Given the nature of the data and the assumptions involved in narrowing the applicable failure events and subset of the U.S. dam population comparable to this specific dam, the staff performed a Bayesian analysis. Using available data on the domestic inventory of dams and dam failures, the range obtained varies between 1.3E-4/dam-year and 4.8E-4/dam-year (5th - 9 5 th percentile) around a mean of 2.8E-4/dam-year.
A literature review performed by the authors for statistical studies of dam failures appears to corroborate this conclusion. Such studies were found in Baecher et al (1980), Martz and Bryson (1982), Donnelly (1994), ICOLD (1995), Foster (2000a), and Foster et al (2000b).
6
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc Figure 2: Failure Rate Probability Distributions Used in Bayesian Updating 5000 4500-
-or 4000 - Posterio
,-,3500-~
C a) 3000-1, Z- 2500 %
p2000 %
o a- 1500 -
1000" 500 0-----------------------------
2 4 6 8 10 Failure Rate (per dam-years) x 104 References Baecher, G. B., M. E. Pat6, and R. De Neufville (1980), "Risk of Dam Failure in Benefit-Cost Analysis," Water Resource Research, 16(3), 449-456.
Martz, H.F., and M.C. Bryson (1982), "Predicting Low-Probability/High-Consequence Events,"
Proceedings of the Workshop on Low-Probability/High-Consequence Risk Analysis, June 15-17, 1982, Arlington, Virginia.
Donnely, R. (1994), "Issues in Dam Safety, ACRES International Innovations Autumn Edition":
http://www.hatch.com.cn/Hatchenerqv/Innovations/autumn2004/feature.html ICOLD (1995), "Dam Failures Statistical Analysis," Bulletin 99, International Commission on Large Dams.
WCD (2000), "Dams and Development: A New Framework for Decision-Making - overview," The Report of the World Commission on Dams.
Foster M, Fell R, Spannagle M (2000a), "The statistics of embankment dam failures and accidents," CanadianGeotechnicalJournal,37, 1000-1024.
Foster M, Fell R, Spannagle M (2000b) "A method for assessing the relative likelihood of failure of embankment dams by piping," CanadianGeotechnicalJournal,37, 1025-1061 7
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc C.L. Atwood, J.L. LaChance, H.F. Martz, D.J. Anderson, M. Englehardt, D. Whitehead, and T.
Wheeler (2003), "Handbook of Parameter Estimation for Probabilistic Risk Assessment,"
NUREG/CR-6823, US NRC.
-&SENT.E INFORMT',, - 3r 8
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc I
3S DAM IMPOUNDMENT MPOUNOMENT
)RAGE IMPOUNDMEPCT DAM SENSITIV, IFR MATION ,FFICAL UC- ,.L' 9
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc (ON RINGSLAKE PLANTDAM YELL IECT I
DDAM WagiOka voir xLLOWTAiLINGS DAM b--,CIT IFDRM^',IO, 1 OFFICIAL. UZ- ONL 10
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc NS"E"8I IiV' INFORMATIO1 - OFFICIAL L 11
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc 9SERVOIR 11 (NP&LFERC)
SENSITIV.-E INF.R.ATI^N "FGFAL" l- USE-.,NL 12
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc 74 27S 87 76 690 67 68 239 67 90 2100 67 264 182,600 67 441839 320 4,190 67 RTH DAM 57 68 61 250 6o 500 69 63 721 69 65 23.000 69 IR'n-I DAM 69 110 2,000 140 23066 66 63 395 70 80 70 90 93850 70 0EK 130 3ý8OO 70 3,480 70 143 568 70 190 38,505 71 87 196 72 138 25900 72 54 1,D90 73 55 32 74 86 107 74 332 141.900 74 554 75 65 987 75 67 905 75 85 704 75 90 32 76 132 4,250 77 3.237 50 77 50 280 77 52 5,20 77 66 543 77 103.60 122 77 147 ý1342 76 125 2O0 76 175 79 85 79 107 5r38 79 57 3.570 80 1,010 90 2305500 80 44,100 61 63 168 9,769 61 188 10,372 61 56 5oo 83 so 91 472 82 1143 105 92 62 21,ao0 63 71 50 150 65 1970 43.193 65 13,750 86 56 511 69 143 16i313 69 63 91 145 02 950 93 Ur500 93 181 93 0 127 64 90 5,500 54 50 95 55 3-04 05 61 12,B3o 95 145 15,122 85 54 145 45,410 95 60 340 97 90 170 87 77 81376 87 89 67 54 253 86 90 612 90 I5 2*210 190 90 45 190 64 so 190 280 103 53 77 102 93 103 103 103 310 104 50 1.285 61 105 77 7.400 105 63 6,B18 105 55 1,140 107 61 672 109 66.5 3195 106 63 1:075 111 63 112 60 427 113 61 117 114 13
GENERIC FAILURE RATE EVALUATION FOR JOCASSEE DAM.doc SENSITIVE iNFORMATiON - QFF~iOAL-' O"-L1 14