Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Available online at www.sciencedirect.

com
annals of

NUCLEAR ENERGY
Annals of Nuclear Energy 35 (2008) 323334 www.elsevier.com/locate/anucene

Uncertainty analysis of LBLOCA for Advanced Heavy Water Reactor


A. Srivastava *, H.G. Lele, A.K. Ghosh, H.S. Kushwaha
Health Safety and Environment Group, Bhabha Atomic Research Centre, Mumbai 400085, India Received 27 September 2006; accepted 24 February 2007

Abstract The main objective of safety analysis is to demonstrate in a robust way that all safety requirements are met, i.e. sucient margins exist between real values of important parameters and their threshold values at which damage of the barriers against release of radioactivity would occur. As stated in the IAEA Safety Requirements for Design of NPPs a safety analysis of the plant design shall be conducted in which methods of both deterministic and probabilistic analysis shall be applied. It is required that the computer programs, analytical methods and plant models used in the safety analysis shall be veried and validated, and adequate consideration shall be given to uncertainties. Uncertainties are present in calculations due to the computer codes, initial and boundary conditions, plant state, fuel parameters, scaling and numerical solution algorithm. All conservative approaches, still widely used, were introduced to cover uncertainties due to limited capability for modelling and understanding of physical phenomena at the early stages of safety analysis. The results obtained by this approach are quite unrealistic and the level of conservatism is not fully known. Another approach is the use of Best Estimate (BE) codes with realistic initial and boundary conditions. If this approach is selected, it should be based on statistically combined uncertainties for plant initial and boundary conditions, assumptions and code models. The current trends are going into direction of the best estimate code with some conservative assumptions of the system with realistic input data with uncertainty analysis. The BE analysis with evaluation of uncertainties oers, in addition, a way to quantify the existing plant safety margins. Its broader use in the future is therefore envisaged, even though it is not always feasible because of the diculty of quantifying code uncertainties with suciently narrow range for every phenomenon and for each accident sequence. In this paper, uncertainty analysis for the Large Break LOCA (200% Inlet Header Break) of Advanced Heavy Water Reactor (AHWR) has been carried out. The uncertainty analysis was carried out for the peak cladding temperature (PCT), based on the two dierent methods i.e., Wilks method and the response surface technique. Their ndings have also been compared. 2007 Elsevier Ltd. All rights reserved.

1. Introduction to AHWR Recent inclination in nuclear industry is towards providing as many passive features as possible, including those during operating conditions. These enhance safety substantially. One of such conceptual design is natural circulation pressure tube boiling water reactor. The proposed Advanced Heavy Water Reactor (Sinha and Kakodkar, 2006) is a 920 MWth vertical pressure tube type boiling light water cooled and heavy water moderated reactor. One of the important passive design features of this reactor
*

Corresponding author. Tel.: +91 22 25593776; fax: +91 22 25505151. E-mail address: abhisri@barc.gov.in (A. Srivastava).

is that the heat removal is achieved through natural circulation of primary coolant at all power levels with no primary coolant pumps. The Main Heat Transport System of the proposed reactor consists of reactor core, core inlet, core outlet, bottom extensions, inlet feeders, tailpipes, steam drums, downcomer and inlet header as shown in Fig. 1. The reactor core consists of 452 Pressure Tubes. In each Pressure Tube, a 54-element single fuel bundle is being housed. A displacer rod is placed at the centre of the bundle and the fuel elements are arranged in concentric circles. In the rst circle, second circle and third circle 12, 18 and 24 fuel elements are arranged respectively. The coolant leaves the Pressure Tubes with an average exit steam quality of 18.7% and

0306-4549/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.anucene.2007.02.017

324

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

Fig. 1. Schematic of AHWR plant.

reaches four steam drums through 452 core top extensions and 452 tail pipes. The steam drum provides a large interface area, which provides the natural separation without using mechanical separators. At the central location (between two plates) of the steam drum bottom, feed water enters which leads to a subcooled condition after mixing with the saturated water. This subcooled water leaves the steam drums through 16 downcomers to a single ring Inlet Header. From the Inlet Header the coolant enters the 452 Pressure Tubes through 452 inlet feeders and 452 core bottom extensions. Emergency Core Cooling System (ECCS) is provided to limit the fuel temperature rise within acceptable limits in the event of Loss Of Coolant Accident (LOCA). The ECCS consists of 1. High pressure injection from advanced accumulator, 2. Low pressure injection from Gravity Driven Water Pool (GDWP), 3. Long term core cooling by recirculation & cooling of reactor cavity water. The advanced accumulators inject large amount of cold water by passive means, at high pressure, directly into the core for 2 min and then a relatively small amount of cold water for next 13 min to quench the core. This objective is achieved with the incorporation of Fluidic Flow Control Device (FFCD) at the bottom of accumulator tank, which reduces the ow after 2 min, by passive means. When the ECC Header pressure reaches below the GDWP head at 2 min after LOCA, then low-pressure injection starts from GDWP. The GDWP inventory of 6000 m3 ensures low pressure injection, by passive means, for 3 days

without any operator action. The supply of water from GDWP to the core continues even after the accumulators are exhausted, keeping the temperature well within the limit. The inventory of MHT, advanced accumulators and GDWP, after passing through the core, spills out through the ruptured location, into the reactor cavity. After 3 days, this inventory is pumped to the GDWP, after cooling it in the heat exchangers to ensure long-term cooling of the core. 2. Methodologies in safety analysis 2.1. Conservative approach A conservative approach has been widely used in the safety analyses of a nuclear installation. In this approach, a code based on conservative models is used. The initial plant parameters, availability of safety control components and systems and operator action are selected over the nominal values in order to exaggerate the course of unexpected events. 2.2. Best estimate approach with uncertainty evaluation (BE) In contrast to the conservative approach, BE methodology is characterized by applying the best estimate codes and the realistic input data along with the nominal (or best estimate) initial and boundary conditions, which means the initial facility parameters are dened at their nominal values. The introduction of nominal values makes it imperative to perform the uncertainty analysis. The

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

325

major sources of uncertainty in the area of safety analysis are represented by the uncertainty of the code (associated with the code models, and correlations, solution scheme, model options, data libraries or deciencies of the code), representation uncertainties (accuracy of the complex facility geometry, 3D eects, scaling, control and system simplications) and plant data uncertainties (unavailability of some plant parameters, instrument errors and uncertainty in instrument response). 2.3. Conservative v/s BE approach The basic advantage of the BE methodology over the conservative approach is the quantication of the safety margin between the realistic value of the calculated parameters and the dened safety limits. A conservative method can successfully show that the facilitys parameters are well below the safety limits, but it fails to predict exactly how far. 3. Methods in uncertainty analysis

neously with the BE calculation, thus avoiding the uncertainty-methodology-user eect and the need of resources for uncertainty prediction. However, a suitable error-database must be made available to pursue this approach. In light of above discussion, two popular methods are widely used for the uncertainty analyses. The methods are discussed below. 3.1. Wilks method The determination of uncertainty relies on statistical sampling technique. It is possible to determine the tolerance limit from unknown distribution by randomly sampling the character in question (i.e. PCT). Wilks (1941) study showed that the proportion of the population between two-order statistics from a random sample is independent of the population sampled; it is only a function of the particular order statistics chosen. Using the well-known Wilks formula, one can determine sample size for desired proportion at a given tolerance interval. NP log1 b loga 1

Nuclear Regulatory Commission (NRC) has developed Best Estimate methodology based on Code Scaling, Applicability and Uncertainty (CSAU) method (Boyack et al., 1990). The CSAU is a methodology for quantifying the overall calculational uncertainty for LOCAs. CSAU takes an expanded top-down approach in that a process identication and ranking table (PIRT) is constructed, not only for the entire system but also for its component. This method also addresses the question concerned with extrapolating the results from the reduced scale integral eects test. This method uses response surface functions. NRC licensed the Westinghouse methodology based on CSAU method for 3- and 4-loop PWRs in 1996. This method uses response surface functions developed from many WCOBRA/TRAC single parametric sensitivity runs. More recently, Westinghouse further developed Automated statistical Treatment of Uncertainty Method (ASTRUM) (Frepoli et al., 2004) which is a new realistic LOCA methodology utilizing a non-parametric statistical sampling technique. In this technique, the uncertainty contributors are sampled for each run without additional step to combine. The number of computer runs is independent of the number of uncertainty parameters and only determined by the tolerance limits desired. GRS, Germany developed uncertainty analysis code SUSA (Software System for Uncertainty and Sensitivity Analysis). University of Pisa, Italy has developed a code to carryout Best-estimate analysis with the capability of Internal Assessment of Uncertainty (CIAU) (D Auria and Giannotti, 2000). The idea of the CIAU is the identication and the characterization of standard plant status and the association of uncertainty to each status. One hypercube and one time interval identify the plant status. Quantity and time uncertainties are combined for each plant status. The CIAU allows the achievement of continuous uncertainty bands simulta-

where N is minimum sample size, a is probability contents (percentile), b is the condence level. Therefore, minimum number of computer runs N, required to get a 95% fractile (percentile) with 95% condence NP log1 0:95 log0:95 ( N P 58:404 ) ( N P 59 ) In the present study, 6 parameters have been chosen and a random value for each set of 6 parameters has been generated by Monte-Carlo simulation for 59 cases. 3.2. Response surface The Response Surface Method (RSM) is an application of statistical and mathematical techniques useful for developing, improving and optimizing response variable. One major factor that inuences the quality of response models is the selection of the data points, or sampling technique, that is used to create the model. Traditionally, the data needed for the RSM is generated eciently via a Design of Experiment (DoE) technique. There are dierent ways to arrive at these designs using DoEs like full-factorial design, BoxBehnken design and central composite design. The selection of an appropriate sampling technique is essential to the creation of accurate response models. Let Y (PCT) represent the response variable and Xi the continuous independent variables that inuence Y. If one wishes to nd the levels of each variables at which Y is maximized. One can write the observed response as a function of Xi as shown;

326

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

Y f X 1 ; X 2 ; . . . ; X n e

For the present case six important parameters were treated as independent variables. These are, (a) (b) (c) (d) (e) (f) Reactor power, Subcooled discharge coecient, Two-phase discharge coecient, Fuel conductivity, Gap conductance between clad and fuel, Decay power.

The expected response E(Y) = f(X1, X2, . . . , Xn) is called the response surface could be drawn on two areas, leading to contours of constant value of E(Y). In most RSM problems, the form of the relationship between the response and the independent variable is unknown. Thus, the rst step in RSM is to nd a suitable approximation for the true functional relation between Y and Xi. Usually, a low order polynomial can be used to describe this relation, particularly in the region described by the levels of the independent variable studied (Montgomery, 1984). This obtained response surface is further used for Monte-Carlo analysis to obtain the required percentile value. 4. Uncertainty analyses methodology The Large Break-LOCA (LBLOCA) analysis was done for postulated 200% break at inlet header, which is the largest pipe (2400 NB) in the entire reactor system. The aim was to calculate the Peak Clad Temperature (PCT). The thermal hydraulic analysis, for LBLOCA (200% in present case), involves several parameters, which aect the PCT. Many of these parameters are uncertain. To account for the uncertainties, Monte-Carlo technique is popularly adopted. It involves a series of analyses using best estimate code. Such analyses are computationally expensive and time consuming. The present uncertainty analysis has been performed using two methods Wilks method and Response surface method. Wilks methodology is based on Wilks theorem. In this method, 59 sets of dierent independent parameters are decided based on their probability distribution. It is worth noting here that there is no restriction on choosing the number of independent parameters. The thermal hydraulic analysis will be performed for these 59 sets using best estimate thermal hydraulic code i.e. RELAP5. The maximum of the PCT obtained by the 59 runs gives the 95 percentile value for the PCT. In other words, it can be stated with 95% probability and 95% condence that the PCT for any random set of input parameters will not exceed this number. The response surface method, in which response surface of PCT was determined using small number of thermal hydraulic analyses. Application of Response Surface Method (RSM) needs only a few thermal hydraulic code runs to establish an approximate model to describe the relationship between the PCT results and its dominating independent parameters. This PCT response surface can be viewed as an approximate analytical equation in terms of independent parameters. The above mentioned response surface will be further used for Monte-Carlo analysis to get the 95 percentile PCT value as well as PCT Probability Distribution Function (PDF) and Cummulative Distribution Function (CDF) also.

This selection is based on literature survey and estimates on their contribution to sensitivity of PCT. Initial rigorous best estimate analyses, for dierent break sizes, revealed that clad temperature does not exceed 800 C with large margin. It was also found that critical heat ux (CHF) does not occur. Hence phenomena pertinent to post-CHF and high temperature regime need not be considered. After the present analysis also this assumption holds true. The analysis also revealed that PCT occurs during the initial blowdown phase itself and hence strongly inuenced by the fuel-stored energy and critical ow parameters. This further justies the selection being limited to above six parameters. The statistical properties of these six parameters are widely discussed in literature (Boyack et al., 1990; Prosek, 2000). The analyses, for LBLOCA, were done using RELAP5/ MOD 3.2 (Fletcher and Schultz, 1995) thermal hydraulic computer code. After the independent variables have been selected, a few RELAP5 runs are performed to obtain the information necessary to build the response surface, the input variables having dierent values in each run. The technique to predict the number of code runs and the value of each independent variable is called the experimental design. In other words, an experimental design is used to determine a set of sample runs to be performed. All the parameters were varied either within the range of 3 times the standard deviation (3r) or within minimum and maximum admissible values. The resulting PCT values were determined and these in turn were used to generate response surface in terms of all the six independent parameters. The non-monotonic behaviour of response surface, with respect to a given parameter, was taken into account by taking thermal hydraulic runs at ner intervals. After generation of response surface the 95 percentile value of PCT was determined using Monte-Carlo method. 4.1. Response surface generation procedure In order to carry out probabilistic analysis response surface was generated, to obtain the relation between PCT and the independent parameters. For the present initial assessment, the independent parameters are reactor power, subcooled discharge coecient, 2-phase discharge coecient, fuel conductivity, gap conductance between clad and fuel and decay power. The statistical properties of these 6 parameters (designated as X1, X2, X3, X4, X5 and X6) are given in Table 1.

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334 Table 1 Statistical properties of independent parameters in thermal hydraulic analysis Name of parameter [Designation]a Reactor power [X1] Subcooled discharge coecient [X2] 2-Phase discharge coecient [X3] Fuel conductivity (as fraction of mean value which is function of temperature) [X4] Gap conductance (as fraction of nominal value) [X5] Decay power (as fraction of nominal value) [X6]
a

327

Distribution Uniform Normal Normal Normal Uniform Uniform

Mean value (l) NAa 1.0 1.0 1.0 NA NA

Nominal value 100% 1.0 1.0 1.0 1.0 1.0

Standard deviation (r) NA 0.042 0.062 0.1 NA NA

Range covered in analysis 97% to 113% 0.8741.126 (3r) 0.8141.186 (3r) 0.7 to 1.3 (3r) 0.21.8 (80% about nominal value) 0.91.1 (10% about nominal value)

The parameters are designated as X1, X2, . . . , X6 and NA denotes not applicable.

The range of variation, of each of these 6 parameters, is also listed in Table 1. The range of variation of most of the parameters was based on suggestions made in literature, (Boyack et al., 1990; Prosek, 2000). However, the range of reactor power variation is based on this reactors design philosophy. It is assumed that 200% break occurs when reactor is at full power. In this reactor the trip is set based on neutron power of 110%. Considering 3% instrumentation error the upper bound turns out to be 113% and lower bound is 97%. The PCT was determined for all the cases, which correspond to dierent combination of independent input parameters. The response surface is usually tted based on output corresponding to predetermined combinations of uncertain parameters. A procedure was adopted in which initial array of PCT was generated based on 2n + 1 design (Bucher and Bourgund, 1990). n is number of uncertain parameter. Since the relationship between the PCT and input independent variables is not known exactly, hence it was expressed in the form of algebraic polynomial, Eq. (3). This was done by multiple regression analysis of PCT results. PCT a0
i6 X i1

Although the procedure discussed above do not conrm to conventional experimental design but leads to minimum number of computationally expensive best estimate code runs. The procedure of considering only selective cross terms was suggested by Lee and Kwak (2006). The response surface was validated, by taking additional runs, for dierent combination of independent parameters. These combinations corresponded to worst combination in terms of higher PCT, best combination in terms of lower PCT. It was observed that for all such combinations, the PCT value predicted by response surface and as evaluated using thermal hydraulic analyses are reasonably close, in the region of interest. 5. Modelling of AHWR main heat transport system 5.1. Main heat transport (MHT) system The MHT system Nodalisation is shown in Fig. 2. The nodalisation scheme models 452 channels and considers 4 parallel paths in the core. Each parallel path i.e. one-fourth core path comprises of one hot channel and 112 lumped channels. There are 4 steam drums and one parallel path of the core is connected to single drum. Sixteen downcomers coming from 4 steam drums are represented by 4 parallel paths and are connected to inlet header. Description of nodalisation is as follows. Descretisation scheme used for the simulating the reactor coolant loop consists of RELAP5 specic models for the reactor core, core top and bottom extensions, inlet feeders, tail pipes, steam drums, downcomers and inlet header. Component number 100, 101, 102 and 103 represents average channels and 121,122,123 and 124 represents hot channel. Eight parallel ow paths are used for simulating core inlet extensions, inlet feeders, coolant channels, core top extensions and tail pipes. Volume 4 to 8 of pipe components 100, 101, 102 and 103 represent 112 parallel average channels and components 121, 122,123 and 124 represents single hot channel. Forty heat (20 average and 20 hot channel) slabs are connected to these core volumes with appropriate axial power prole. The AHWR fuel cluster consists of three rings of fuel rods. The outer rings fuel rods have

bi X i

i6 X i1

bi X 2 d 2 X 3 e 2 X 4 i 2 2 3

c14 X 1 X 4 c15 X 1 X 5 c45 X 4 X 5

In the above equation a0, bis, cis, d2, e2, c14, c15 and c45 are polynomial constants. The constant a0 correspond to combination in which all the uncertain parameters are at nominal values, that is, the best estimate case. The constants bis and cis correspond to the combinations in which all the parameters, except one, are at nominal values. This represent a set of runs in which one of the parameter was chosen equal to lower or upper value of the range given in Table 1. The PCT variation with respect to X2 was found to be mildly non-monotonic. Therefore, additional combinations were analysed with intermediate values of X2. This results in constants d2 and e2 corresponding to third and fourth order terms of X2. The constants c14, c15 and c45 correspond to cross-combinations among uncertain parameters X1, X4, and X5.

328

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334


120

112

671
511 314 357 665 353 663 351 661 505 SD 10 Vol 206 181 220 148 DC 119 202

Governer valve
512 313

116

Safety Valve 672


513 Bypass Valve 413 670 358

Governer valve
514

Junction Relief Valve Volume


Feed Header

414 Feed water pump 669 355 667 677 150 SD 10 Vol 203 222 215 221 678 676 677 675 214 104

360 664 352

359

361 668 354

356

Steam Drum
678 222 506

662 105 SD 10 Vol 204 161 DC 108 208 676 215 106

666 405 SD 10 Vol 205 171 146 201 DC 110 223

6 6 0

659 658

657 656

654 655

6 5 3

6 0 0

221 406

Downcomer
151 DC 107 207

145 200

219

147

Inlet Header, 109 12 11 10 9 5 8 4 7 3 6 2 5 1 4 3 2 102 1 12 11 10 9 8 7 6 5 4 3 2 1 124 12 11 10 9 5 8 4 7 3 6 2 5 1 4 3 2 103 1 12 11 10 9 8 7 6 5 4 3 2 1 123 12 11 10 9 5 8 4 7 3 6 2 5 1 4 3 2 101 1 12 11 10 9 8 7 6 5 4 3 2 1 122 12 11 10 9 8 7 6 5 4 3 2 100 1 12 11 10 9 8 7 6 141 5 4 3 2 1 121

5 4 3 2 1

5 4 3 2 1

5 4 3 2 1

5 4 3 2 1

5 4 3 2 1

209

142

210

143

212

144

211

Fig. 2. Nodalisation scheme of MHT for AHWR.

highest power in a channel. A hot pin has been modelled to take into account the radial power peaking in the fuel. Volume 1 and 2 of theses parallel paths represent vertical and horizontal portion of inlet feeders, while core bottom extension is represented by volume 3 of these pipe components. Volume 9 of these components simulates core top extension, while volumes, 10 and 11 represent tail pipes. The two-phase mixture from the reactor core with dierent parallel channels enters the rst volume at the bottom of corresponding steam drum via tail pipes. All the 4 drums (components 505, 105, 405 and 150) are modeled separately using 10 control volumes in each drum. Out of these, nine volumes are taken as saturated volumes where natural separation of steam and water takes place as the drum has a large cross-sectional area and steam exits from volume 10 of these components. The bottom part of the drum (subcooled portion) in between two baes is represented by components 406, 106, 506 and 104. The feed water along with recirculation ow enters into this part of the drums. Flow from this portion of steam drums enters the downcomer. There are four components (119, 108, 110, and 107) for downcomer. One component represents four downcomer. Steam outlets from drums 405 and 150 are connected to pipe component 356 and drum 505 and 105 are connected to other pipe component 353. Two governor valves, which is connected in both steam pipe for controlling the pressure of MHT. Relief valve (414), safety valve (413) and bypass valve have also simulated. Feed water conditions are simulated using time dependent volume (600) and two feed water pump is used to store feed water

in feed header (660). Flow coming from feed header is divided into four paths in such that every path is connected to subcooled region of steam drum. All the downcomers are attached to inlet header. Total number of inlet feeder is 452, which is attached to corresponding reactor channel. This nodalization has 112 inlet feeder clubbed into one component, which is connected to average channel of reactor. So four average channel component and four hot channel component. Flow coming from inlet header is divided into 8-ow path. Out of eight-ow path four-ow path are connected to average channel and rest connected to hot channel. 5.2. Emergency core cooling system (ECCS) This system simulates four advanced accumulators (component no. 900, 902, 904 and 906), GDWP, ECCS header (700) and 452 feeder and water tubes (see Fig. 3). ECCS feeder tubes and water tubes are represented by eight pipe components. Out of these eight, four represent for average channel and four for hot channel. These pipe components are divided into 5 internal volumes each, which inject water into 5 core volumes in each parallel path, when logic for ECCS injection is satised. Advanced accumulators are connected to ECCS header through valves, which changes their resistance depending on level in the accumulator and simulate resistance in ow path and isolation of accumulator at appropriate level (1 m) during transient. GDWP is modeled with ve axial stacked volumes. The pipe coming from GDWP is divided into a 15

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334


733 731 729 727 725 726 710 GDWP Header 286 715
900 Accumulator 902 Accumulator

329

709

720

285 714 288


904 Accumulator

284 713 287


906 Accumulator

282 712 283

289

901 ECCS HEADER 245 704 251

903 246 705 252

905 247 706 253

907 248 707 249

718
254 244 1 2 3 4 5 6 7 8 12 11 1 10 2 9 3 8 4 260 261 7 5 6 6 262 5 263 7 4 8 264 3 701 2 1 124, 102 12 11 10 9 8 7 6 5 4 3 2 1 257

717

716
256 243 242 1 2 3 4 5 6 7 8

700
255 241 12 11 1 10 2 9 3 8 125 4 7 126 5 6 127 6 5 128 7 4 129 8 3 751 2 1 100, 121

755 756 757 758 759

1 2 3 4 5 6 7 8

275 276 277 278 279

135 136 137 138 139

1 2 3 4 5 6 7 8

130 131 132 133 134

754

708

753 752 103, 123

12 11 1 1 10 2 2 9 3 3 270 8 4 4 271 7 5 5 272 6 6 6 273 5 7 7 274 4 8 8 3 703 702 2 1 122, 101

265 266 267 268 269

Fig. 3. Nodalisation scheme of ECC System for AHWR.

Table 2 Description of the hydrodynamic components Vol. no. 100-01, 100-02,101-01,101-02,102-01,102-02, 103-01,103-02 100-03, 101-03, 102-03, 103-03 100-04 to 100-08,101-04 to 101-08,102-04 to102-08, 103-04 to103-08 100-09, 101-09, 102-09, 103-09 100-10, 100-11,101-10,101-11,102-10,102-11, 103-10,103-11 109-01 104-01, 106-01, 406-01, 506-01 150-01 to 150-10, 105-01 to 105-010, 405-01 to 405-10, 505-01 to 505-10 351 to 356 107, 108, 119 & 110 111, 113, 114, 115 112 and 120 511 and 513 512 and 514 350 900, 902, 904 and 906 700, 716, 717 and 718 704, 705, 706, 707 701-01 to 701-05, 702-01 to 702-05, 703-01 to 703-05, 708-01 to 708-05 709-01 to 709-05 720724 710 Description of the volume Inlet feeders in 4 dierent ow paths 100, 101, 102, 103 Core bottom volume in 4 dierent ow paths 5 active core volumes in 4 dierent ow paths Core top extension in 4 dierent ow paths 2 volumes in tail pipes in 4 dierent ow paths Inlet header volume Subcooled volumes in steam drums Steam drum volumes Steam piping 4 downcomers pipes clubbed for 4 steam drums Time dependent volumes representing constant feed Time dependent volumes for turbine steam ow Time dependent volumes for safety valve Time dependent volumes for relief valve Time dependent volumes for containment Accumulators ECCS header Piping between accumulator to ECCS header ECCS feeders and water tubes GDWP GDWP piping GDWP header

volume. GDWP volume is also divided into a fteen volume so each volume is separately connected to the GDWP

pipe. ECCS header has compartmentalized into four parts and its volume number is 700, 716, 717 and 718. Each

330 Table 3 Description of the sequence of events Events Break initiation Reactor trips (1st signal) Reactor trips (2nd signal) set point Reactor trips (3rd signal) set point Reactor trips (4th signal) set point Reactor trips (5th signal) set point CIES valve closure initiation CIES valve fully closed initiation Reactor trip actuation Accumulator injection Stand pipe cut o GDWP injection Feed water cut o Accumulator injection stops End of transient

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

Postulated t = 0.0 s Containment high pressure (1.14 bar) at 1 s (SDS-1)(ignored) Containment high pressure (1.14 bar) at 1 s. (SDS-2)(taken) MSIV closure by 10% (observed) Low drum level (0.8 m) (observed) Low MHT pressure (59.6 bar) (observed) At reactor trip set point At reactor trip set point Set point with 0.5 s delay At ECC header pressure less than 50 bar Water level below stand pipe ECC header pressure less than 2 bar Deaerator low level On low accumulator level

Observed t = 0.0 s t = 1.0 s t = 1.0 s t = 1.3 s t = 6.3 s t = 13.5 s t = 1.0 s t = 1.2 s t = 1.5 s t = 21.06 s t = 55.89 s t = 119.06 s t = 309.0 s t = 372.8 s t = 2563 s

compartment is connected to GDWP and accumulator. The details of the dierent component numbers are given in Table 2. 6. Result and discussions The nominal case i.e., 200% inlet header break analysis with all the independent parameters are at their nominal value has been postulated. The details of the case are discussed in Table 3. 6.1. Pressure transients

rium between the two-phases. The pressure continues to decrease smoothly afterwards as shown in Fig. 4. Fig. 5 depicts the drum Pressure behaviour. The rate of pressure reduction is initially less as compared to the inlet header owing to the large compressibility in drum. The reactor trips on high containment pressure signal. As the pressure goes down below 50 bar, ECCS injection starts and cold water start coming into core. This led to the further void collapse in the core and pressure reduction. As the pressure goes below 2 bar, GDWP injection is initiated in the three ECC path. GDWP injection continues till end of the transient. 6.2. Flow transient

Pressure at inlet header is characterized by initial sharp drop due to large break ow owing to subcooled condition in the header. This sudden depressurisation, which leads to two-phase condition in the inlet header, reduces the critical ow. The sharp pressure drop is due to initial non-equilib-

Fig. 6 indicates the behaviour of break ow. It can be seen from Fig. 6 that the break ow accelerates initially because of subcooled condition in the inlet header. Further ow is dictated by upstream pressure and enthalpy condi-

8000000 7000000 6000000 5000000 4000000 3000000 2000000

7000000 6000000 5000000

Pressure (Pa)

Pressure (Pa)

4000000 3000000 2000000 1000000

1000000

0
0

-1000000
-20 0 20 40 60 80 100 120 140

20

40

60

80

100

120

140

Time (sec)

Time (s)

Fig. 4. Inlet header pressure transient.

Fig. 5. Steam drum pressure transient.

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334


30000 250

331

25000

200

20000

Flow (Kg/sec)

15000

Flow (Kg/sec)
0 50 100 150 200

150

100

10000

50 5000

-50

50

100

150

200

250

300

350

Time (s)

Time (s)

Fig. 6. Break ow transient.

Fig. 8. Feed water ow transient.

10 8 6 4 2 0 -2 -4 0 -6 0 50 100 150 200


0 50 100 150 200 250 300 350 400

400

Flow (Kg/sec)

300

Flow (Kg/sec) Time (sec)

200

100

Time (sec)

Fig. 7. Core exit ow transient. Fig. 9. Accumulator ow transient.

tion. It can be observed that break ow goes to maximum of 26354.8 kg/s in the initial few seconds. Core ow behaviour is depicted in Fig. 7. The core ow largely depends upon the results of the two balancing forces i.e., driving force and two phase pressure drop. As soon as the transient is initiated, due to break, ow starts reducing in the core and after some time it reverses. In later part of the transient, as ECC water gets injected in the core and further GDWP water injection leads to large uctuation in the core quality as shown in Fig. 10. This quality uctuation leads to ow as shown in the Fig. 7. The ow in the core with initiation of ECC injection is sucient to remove the decay heat for the rest of transient (see Fig. 8). The accumulator ow is depicted in Fig. 9. As soon as the pressure in the core become less than 50 bar, the Accumulator injection starts. The accumulator ow is as high as 400.893 kg/s around 40 s. At around 62 s, due to stand pipe cut o, the ow reduces to 149.9 kg/s. The ow keep on reducing based on pressure condition in the accumulator

1.4 1.2 1.0 0.8

Quality

0.6 0.4 0.2 0.0 -0.2 0 50 100 150 200

Time (s)

Fig. 10. Core exit quality transient.

332

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

and ECC header. The accumulator cut-o takes place at 372 s. 6.3. Inventory transient The MHT inventory is 365 tonnes at the beginning. Due to break in the inlet header, the inventory starts coming out through the break. This leads to the sudden fall in MHT inventory. The inventory comes down to 114 tonnes. As accumulator injection is established, the total inventory addition due to cold water coming from the accumulator, GDWP and the feed water becomes more than the inventory going out through the break, resulting in an increase in the MHT inventory. The inventory starts reducing gradually again due to accumulator cut-o and feed water cuto based on the low deaerator level. The inventory at the end of transient is around 152.4 tonnes (Fig. 11).

6.4. Clad temperature transient Fig. 12 presents the clad surface temperature transient for maximum axial power node. This gure shows the clad temperature transient for the average channel, hot channel and hot pin together. Even after reactor trip on SDS-2 signal, clad surface temperature shows increasing trend for some time. The maximum temperature achieved is 858.34 K in the hot channel after 21.36 s from the initiation of transient. This is attributed mainly due to ow to power mismatch as the stored heat is coming out of the fuel and ow is nearly zero during that time. The rise in the average channel is insignicant. The heat transfer from fuel to coolant is very less, leading to temperature rise. Further with initiation of ECC injection rst from the accumulator ow and later on ow from the GDWP, keeps the temperature well with in acceptable limit. The hot pin temperature variation shown in Fig. 13 is similar to hot channel behaviour. The maximum temperature achieved is 975.6 K after 21.34 s from the initiation of transient. 6.5. Results of uncertainty analyses The results obtained from Wilks method are discussed followed by results of response surface technique. 6.5.1. Wilks result The six independent variables along with their probabilistic distributions are used to generate the 59 sets randomly. These 59 sets of six independent parameters are required to obtain 95th percentile PCT with 95% condence as discussed in earlier section. These 59 sets of independent parameters along with other dependent parameters are used to run transient cases using RELAP5 model. The PCT obtained during the dierent runs from best estimate code is recorded along with other thermal hydraulic output. The PCT obtained from the all 59 code

400

350

PHT Inventory (Tonnes)

300

250

200

150

100 0 500 1000 1500 2000 2500 3000

Time (s)

Fig. 11. MHT inventory transient.

700

700

Hot Pin Clad Temperature (C)

600

Hot pin hot channel Average channel

600

Clad temperature (C)

500

500

400

400

300

300

200

200 100

100

-10

10

20

30

40

50

60

70

80

90

100

500

1000

1500

2000

2500

3000

Time (s)

Time (s)

Fig. 12. Clad temperature transient.

Fig. 13. Hot pin temperature transient.

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334 Table 4 Details of lower bound, upper bound and nominal cases S. no. Lower Nominal Upper Power 0.97354 1.0 1.022 Sub. Dis. Coe. 0.92784 1.0 1.031 Sat. Dis. Coe. 0.98162 1.0 0.9651 Cond. 1.087 1.0 1.193 Gap cond. 1.5415 1.0 0.2227 Decay power 0.9865 1.0 1.062 PCT (C) (hot pin) 643.61 702.6 1119.61

333

PCT (C) (hot chnl) 544.12 585.37 863.7

runs from RELAP5 analysis are ranked from the higher to lower number. The top number in the table is the 95th percentile PCT value with 95% condence. The above nding are reported in Table 4.

1100 1000 900

Upper bound Lower bound

Fig. 14 shows the clad temperature variation in hot pin with time for all the 59 runs taken using best estimate code RELAP5 for the rst 50 sec of the transient time. It can be seen from the gure that PCT variation for most of the cases are in the lower bound region. Some of them are in the upper region. Fig. 15 shows the clad temperature transient for hot pin for the 3 cases i.e., upper bound, nominal and lower bound of 59 cases analysed. 6.5.2. Response surface result The generated response surface is validated with results of additional RELAP5 runs. These validation runs have been performed taking such combinations of independent variables that will result in the maximum and minimum PCT. The combination of independent parameters that results in the minimum PCT is known as favourable case the one resulting in the highest PCT as the conservative case. The PCT predicted by the response surface for the favourable case is found to be in agreement with the RELAP5 output. But for the conservative case, the response surface prediction showed a deviation of about 15% with RELAP5 output. Further, it was realized that out of the six chosen parameters reactor power, fuel conductivity and gap conductance play the most dominant role in deciding the PCT. Hence, to improve the response surface in the higher temperature range, a cross combination of the above mentioned parameters was chosen with two parameters varying at a time (as given in Eq. (3)). A new response surface was generated with these additional cases included. The new response surface predictions deviate by 13.16% for the favorable case but for the conservative case, it deviates by 5.7% only. The details of the dierent cases have been presented in Table 5 for the initial response surface as well as the improved response surface. It can be seen that for the dominant parameters combination for the higher PCT, the improved response surface has predicted value that is very close to RELAP5 calculation, the error found to be 0.38% in the conservative side. Monte Carlo technique has been applied to the obtained response surface using 1,000,000 trials. In each trial, each parameter used in the response surface was sampled from

Clad Temperature ( C)

800 700 600 500 400 300 200

10

20

30

40

50

Time (sec)

Fig. 14. Hot pin temperature with Wilks method.

Clad Temperature Transient : Wilks Method


1200 1100 1000 900

Temperature (C)

800 700 600 500 400 300 200 100 -10 0 10 20 30 40 50

Nominal case Upper Bound Lower Bound

Time (s)

Fig. 15. Bounding hot pin temperature using Wilks method.

Table 5 Results of response surface improvement CASE Favourable Conservative RELAP5 RUN PCT (C) 587.827 1660.68 RSM PCT (C) 588.34 1369.95 % Error 0.06 15.03 Improved RSM PCT (C) 701.14 1549.498 % Error 13.16 5.7

334
0.007

A. Srivastava et al. / Annals of Nuclear Energy 35 (2008) 323334

7. Conclusion
PDF

0.006 0.005 0.004

0.003 0.002 0.001 0 500

600

700

800

900

1000

1100 1200 1300 1400 1500

PCT (deg. C)

Fig. 16. Overall PDF for PCT.

CDF
1.2
CDF

The uncertainty analyses for the PCT in LBLOCA for AHWR have been performed. The analyses have been carried out using two dierent methods i.e., Wilks method and response surface followed by Monte-Carlo method. The 95th percentile value obtained from the Wilks method is 1119.6 C and the 95th percentile PCT value obtained from the response surface method is 1182.9 C. The analyses highlights that results obtained from the response surface method envelope the result obtained from the Wilks method. The 95th percentile PCT value obtained from the response surface method has very less margin w.r.t. 1200 C i.e. fuel coolability criteria. Gap conductance plays a dominant role in the PCT. This margin can be improved further based on the experimental data obtained from the actual fuel. References
Boyack, B.E., Caton, I., Duey, R.B., Grith, P., Katsma, K.R., Lellouche, G.S., Levy, S., Rohatgi, U.S., Wilson, G.E., Wul, W., Zuber, N., 1990. Quantifying reactor safety margins, Part 1: an overview of the code scaling, applicability, and uncertainty evaluation methodology. Nuclear Engineering and Design 119, 115. Bucher, C.G., Bourgund, U., 1990. A fast and ecient response surface approach for structural reliability problems. Structural Safety 7 (1), 5766. D Auria, F., Giannotti, W., 2000. Development of code with capability of internal assessment of uncertainty. Journal of Nuclear Technology 131 (1), 159196. August. Fletcher, G.D., Schultz, R.R., 1995. RELAP5/MOD3.2 Code Manual, Idaho National Engineering Laboratory Idaho. Frepoli, C., Ohkawa, K., Kemper, R.M., 2004. Realistic large Break LOCA Analysis of AP1000 with ASTRUM, NUTHOS 6, Japan, October 48. Lee, S.H., Kwak, B.M., 2006. Response surface augmented moment method for ecient reliability analysis. Structural Safety 28 (3), 261 272. Montgomery, D.C., 1984. Design and Analysis of Experiments, 2nd ed. John Wiley & Sons, NY, USA. Prosek, A., 2000. Optimal statistical estimator for ecient generation of the response surface, International Meeting on Best Estimate Methods in Nuclear Installation Safety Analysis, Washington DC, November. Sinha, R.K., Kakodkar, A., 2006. Design and development of the AHWR the Indian thorium fuelled innovative nuclear reactor. Nuclear Engineering and Design 236, 683700. Wilks, S.S., 1941. Determination of sample sizes for setting tolerance limits. The Annals of Mathematical Statistics 12, 9196.

pdf CDF

1 0.8 0.6 0.4 0.2 0 500

600

700

800

900

1000 1100 1200 1300 1400 1500

PCT (deg. C)

Fig. 17. Overall CDF for PCT.

its respective distribution as explained in Table 1. The six independent parameters obtained in each such trial are used to get PCT value from the response surface. The result provides a range and distribution of PCT in the case of a LBLOCA. The 95th percentile value obtained from the Monte-Carlo analysis is 1182.9 C. The shape of overall probability density function (PDF) and Cumulative Distribution Function (CDF), of PCT, are shown in Figs. 16 and 17, respectively. From Fig. 17, it is observed that temperature corresponding to 50% probability is about 710.38 C. This value is very close to RELAP5/MOD3.2 result of best estimate case (involving combination of all nominal values), which is 702.6 C. This further enhances the condence in accuracy of response surface t.

You might also like