CATHENA Void Fraction Accuracy and Uncertainty Using RD-14M LOCA Tests

Transcription

1 CATHENA Void Fraction Accuracy and Uncertainty Using RD-14M LOCA Tests Geoff Waddington, Songyu Liu AECL Nuclear Laboratories Chalk River, Ontario, Canada 2012 May 3 UNRESTRICTED / ILLIMITÉ

2 Outline Introduction Background, Objectives, Definitions Code Accuracy Assessment Methodology, Data, Model, Results Code Uncertainty Estimation Methodology, Parameters, Results Summary UNRESTRICTED / ILLIMITÉ 2

3 Introduction Background CATHENA: Canadian Algorithm for Thermalhydraulic Network Analysis CATHENA used primarily for analysis of system thermalhydraulics in postulated upset conditions in CANDU reactors, and research reactors Extensive validation program to quantify code accuracy and demonstrate its fitness-for-purpose UNRESTRICTED / ILLIMITÉ 3

4 Introduction Objectives The objectives of this work are to: estimate the CATHENA code accuracy and uncertainty in predicting the channel void fraction during the initial rapid voiding phase of large LOCA demonstrate applications of the industry-accepted code accuracy assessment and uncertainty analysis methodologies illustrate the adequacy of the considered code modeling uncertainties through comparison of the accuracy assessment and uncertainty analysis results UNRESTRICTED / ILLIMITÉ 4

5 Introduction Definitions Code accuracy is the degree of closeness of a calculated quantity to its true value true value is the actual value of a quantity that would be obtained with perfect measuring instruments and without introducing any errors arising from processing or correction Code uncertainty is the uncertainty of a computer code prediction that arises from the uncertainty in code models prediction uncertainty is the overall uncertainty in computer code prediction that arises from the combined effect of the uncertainties in code models, uncertainties in plant parameters and any relevant representation uncertainty UNRESTRICTED / ILLIMITÉ 5

6 Outline Introduction Background, Objectives, Definitions Code Accuracy Assessment Methodology, Data, Model, Results Code Uncertainty Estimation Methodology, Parameters, Results Summary UNRESTRICTED / ILLIMITÉ 6

7 Code Accuracy Assessment Methodology Computer code accuracy in predicting a key parameter is assessed through comparing code prediction to the corresponding experimental measurement measured value, after corrections for any known biases, is the best estimate of the true value Code accuracy in the key parameter is represented using two parameters: code bias = mean of residuals variability in the bias = standard deviation on residuals Key parameter is the channel void fraction, as measured by the neutron scatterometer, from selected RD-14M blowdown tests UNRESTRICTED / ILLIMITÉ 7

8 Code Accuracy RD-14M Facility RD-14M schematic full-height representation of CANDU PHWR heat transport systems ten 6-metre long horizontal channels, in 2 passes 7-element fuel element simulator (FES) Breaks can be represented at various locations inlet header break tests used in present work PROTECTED - SENSITIVE / PROTÉGÉ - DÉLICAT 8

9 Code Accuracy Experiment Data Selected 23 tests for accuracy assessment, from B01xx and B03xx series: 9 multi-channel tests, 14 single-channel tests break sizes between 15 and 48 mm (8 large, 15 small ) 10 of 14 single-channel tests with power pulse, others with full/decay power transient Neutron scatterometer measured void in channel 14 Assessment period is time of rapid voiding after break Code bias and variability calculated for: each test, large and small break tests, all tests UNRESTRICTED / ILLIMITÉ 9

10 Code Accuracy CATHENA Model Representation of below header piping HS10 to HS14 shown similar for other pass (HS5 to HS9) Also idealized: primary HTS above headers secondary side heat transfer in wall models UNRESTRICTED / ILLIMITÉ 10

11 Accuracy Assessment Example Results Test B0105: 25 mm break, multi-channel test code bias = 0.014, variability = test well predicted, smallest bias Symbols denote points in accuracy assessment period UNRESTRICTED / ILLIMITÉ 11

12 Accuracy Assessment Example Results Test B0108: 48 mm break, multi-channel test. code bias = , variability = initial voiding predicted well, over-prediction in 2 nd half of assessment period UNRESTRICTED / ILLIMITÉ 12

13 Accuracy Assessment Example Results Test B0113: 18 mm break, single-channel test with power pulse code bias = 0.014, variability = predictions generally within measurement uncertainty UNRESTRICTED / ILLIMITÉ 13

14 Accuracy Assessment Overall Results Accuracy assessment results for each test: Accuracy Statistic Range Code bias ( ) ~ Variability in bias ( ) ~ Accuracy assessment results: Tests Data Points Code Bias Variability Large Break Small Break All Selected UNRESTRICTED / ILLIMITÉ 14

15 Accuracy Assessment Analysis of Results Code bias versus break size (all tests) Bias is related to break size for small break tests In general, CATHENA over-predicts void in large break tests, and under-predicts void in small break tests UNRESTRICTED / ILLIMITÉ 15

16 Outline Introduction Background, Objectives, Definitions Code Accuracy Assessment Methodology, Data, Model, Results Code Uncertainty Estimation Methodology, Parameters, Results Summary UNRESTRICTED / ILLIMITÉ 16

17 Uncertainty Analysis Methodology Code uncertainty was assessed by conducting an integrated uncertainty analysis (IUA), to obtain 95%/95%, 2-sided tolerance limits for void fraction using GRS method of order statistics Wilks formula gives minimum of 153 uncertainty simulations for 95%/95% 2 nd -order 2-sided tolerance limits The uncertainty analysis was conducted for code model parameters, excludes uncertainty from facility/test parameters 155 uncertainty cases randomly generated, with 27 uncertain code model parameters, determined by a PIRT process IUA performed for six selected tests, with objective to demonstrate that the considered code model uncertainties can account for the observed code bias and variability UNRESTRICTED / ILLIMITÉ 17

18 Uncertainty Analysis Example Results B0310: 30-mm break, multi-channel test IUA simulations and 2-sided 95%/95% tolerance limits uncertainty range is large because this is a critical break channel void is sensitive to the flow split, which is sensitive to discharge coefficient UNRESTRICTED / ILLIMITÉ 18

19 Uncertainty Analysis Example Results 1-ph discharge coefficient Colebrook-White friction factor flow form loss factors Chen HT correlation B0310: correlations between void and sample code model parameters uncertainty in break discharge coefficient is the dominant contributor to the predicted uncertainty in channel void fraction UNRESTRICTED / ILLIMITÉ 19

20 Uncertainty Analysis Example Results B0307: 48-mm break, single-channel test IUA simulations and 2-sided 95%/95% tolerance limits uncertainty range is narrow because this is such a large, fast break that the break discharge uncertainty is not significant UNRESTRICTED / ILLIMITÉ 20

21 Uncertainty Analysis Overall Results For the small break tests, the tolerance limits encompass the measured void fractions and their ±2σ measurement uncertainty For the large break tests, the two-sided tolerance limits do not in all cases cover the measured void throughout the assessment period The liquid discharge coefficient is the most significant source of uncertainty; however, its importance is less for the large break tests for all tests, the predicted flow through the break is singlephase liquid throughout the assessment period; therefore the two-phase break discharge model and coefficient play no role UNRESTRICTED / ILLIMITÉ 21

22 Outline Introduction Background, Objectives, Definitions Code Accuracy Assessment Methodology, Data, Model, Results Code Uncertainty Estimation Methodology, Parameters, Results Summary UNRESTRICTED / ILLIMITÉ 22

23 Summary Accuracy Uncertainty Comparison Accuracy and uncertainty results can be compared to assess the adequacy of the identified input parameters and their uncertainty ranges For large break tests, the identified code model parameters and their uncertainty ranges are adequate this supports the use of the same code model parameters and uncertainty ranges in CANDU plant LOCA simulations For small break tests (representative of critical break LOCAs), the identified code model parameters and their uncertainty ranges are not sufficient to cover the under-prediction range from the accuracy assessment therefore need to do more work, or resolve problem differently UNRESTRICTED / ILLIMITÉ 23

24 Summary Conclusions Main results of accuracy assessment are: Tests Code Bias Variability in Bias all LOCA large break small break Main results of the model parameter IUA are: code uncertainty range from model parameters is smaller for the larger breaks; break discharge coefficient is the most significant contributor to the uncertainty in predicted void. Estimated code uncertainty generally covers the onesided 95% under-prediction limit for large break tests UNRESTRICTED / ILLIMITÉ 24

25 UNRESTRICTED / ILLIMITÉ 25