correlations between lct integral experiments and their · pdf filecorrelations between lct...
Post on 24-Mar-2018
218 Views
Preview:
TRANSCRIPT
© 2017 Organisation for Economic Co-operation and Development
Correlations between LCT Integral
Experiments and Their Impact on
Adjustments
Ian Hill Nuclear Science Division
SG39
May 2017
© 2017 Organisation for Economic Co-operation and Development © 2017 Organisation for Economic Co-operation and Development 2
Overview
The main point is to attempt to estimate correlations between
existing ICSBEP experiments and show what the impact of
these estimated correlations would be on nuclear data
adjustments.
Not a toy exercise; meant to be best estimate of real data
• Status of Existing Correlations
• Tools to Generate Correlations
• New Correlation Data, tests? [focus of this presentation]
• Impact of Integral Correlation Data on Adjustment [focus of this
presentation]
© 2017 Organisation for Economic Co-operation and Development © 2017 Organisation for Economic Co-operation and Development 3
Status of Existing Correlations (DICE Uncertainties button)
Correspond to the correlations of benchmark model uncertainties
– Level 1 correlations show that evaluations are correlated
– Level 2 correlations give the quantitative information about the
correlations between cases
– Currently 94 cases have correlation data [level2] in DICE
(or ~2%). Level 2 required for analysis.
Also Compute Cosine Similarity
of Sensitivity Vectors!
© 2017 Organisation for Economic Co-operation and Development
NEA: Tools to Simplify Covariance Generation
LCT021-1 LCT021-2 LCT021-3 LCT021-4 LCT021-5 LCT021-6 LCT021-1 720 0 0 0 0 0 LCT021-2 0 720 0 0 0 0 LCT021-3 0 0 720 0 0 0 LCT021-4 0 0 0 500 0 0 LCT021-5 0 0 0 0 500 0 LCT021-6 0 0 0 0 0 500
Shared uncertainty in benchmark models
Uncertainties(pcm) Case1 Case2 Case3 Case4 Case5 Case6
Clad Thickness 400 400 400 72 72 72
Boron Concentration 384 384 384 130 130 130
Enrichment 338 338 338 363 363 363 Experimental Uncertalnty 300 300 300 300 300 300
Pitch 5 5 5 270 270 270
LCT021-1 LCT021-2 LCT021-3 LCT021-4 LCT021-5 LCT021-6
LCT021-1 1.000 0.784 0.784 0.408 0.408 0.408
LCT021-2 0.784 1.000 0.784 0.408 0.408 0.408
LCT021-3 0.784 0.784 1.000 0.408 0.408 0.408
LCT021-4 0.408 0.408 0.408 1.000 0.694 0.694
LCT021-5 0.408 0.408 0.408 0.694 1.000 0.694
LCT021-6 0.408 0.408 0.408 0.694 0.694 1.000
Sum shared
components
Current Situation Generate Correlation Matrix
When rho is between 0 and 1? Do we shamelessly make it up?
Made Spreadsheet Tool
Used Tool to Generate Correlations for ~600 LCT Cases FYI: Similar Tools exists for Nuclear data
V.Zerkin, “Web Tool for Constructing a Covariance Matrix from EXFOR Uncertainties”, EPJ Web of Conferences 27, 00009 (2012)
© 2017 Organisation for Economic Co-operation and Development
After Assignment: Impact of Integral
Benchmark Correlations
Some issues explored……
• How does the goodness of fit change?
• Limitations to chi squared + other tests
• How does the GLS fit change?
• Difficultly decoupling Nuclear data and Benchmark uncertainty
• What is the impact on data adjustment?
• Newly available data (4200 sensitivity coefficients)
All analysis will use SCALE, with ENDF/B-VII.0 CE, and 44 Group
Covariance Library, no filtering applied
Intent was to explore the reasonability and impact of the previously
assigned integral benchmark correlations………
© 2017 Organisation for Economic Co-operation and Development
LCT Data Set
-2.50E+00
-2.00E+00
-1.50E+00
-1.00E+00
-5.00E-01
0.00E+00
5.00E-01
1.00E+00
1.50E+00
2.00E+00
2.50E+00
3.00E+00
LCT
00
1-0
01
LCT
00
1-0
07
LCT
00
3-0
05
LCT
00
3-0
11
LCT
00
3-0
17
LCT
00
4-0
01
LCT
00
4-0
07
LCT
00
4-0
13
LCT
00
4-0
19
LCT
00
9-0
05
LCT
00
9-0
11
LCT
00
9-0
17
LCT
00
9-0
23
LCT
01
0-0
02
LCT
01
0-0
08
LCT
01
0-0
14
LCT
01
0-0
20
LCT
01
0-0
26
LCT
01
6-0
02
LCT
01
6-0
08
LCT
01
6-0
14
LCT
01
6-0
20
LCT
01
6-0
26
LCT
01
6-0
32
LCT
01
7-0
06
LCT
01
7-0
12
LCT
01
7-0
18
LCT
01
7-0
24
LCT
02
1-0
01
LCT
02
2-0
01
LCT
02
2-0
07
LCT
02
6-0
06
LCT
02
8-0
02
LCT
02
8-0
08
LCT
02
8-0
14
LCT
02
8-0
20
LCT
02
9-0
06
LCT
02
9-0
12
LCT
03
2-0
06
LCT
03
3-0
03
LCT
03
3-0
09
LCT
03
3-0
15
LCT
03
3-0
21
LCT
03
3-0
27
LCT
03
3-0
33
LCT
03
3-0
39
LCT
03
3-0
45
LCT
03
3-0
51
LCT
03
4-0
05
LCT
03
4-0
12
LCT
03
4-0
19
LCT
03
4-0
25
LCT
03
7-0
02
LCT
03
7-0
08
LCT
03
8-0
03
LCT
03
8-0
09
LCT
03
9-0
01
LCT
03
9-0
07
LCT
03
9-0
13
LCT
04
2-0
02
LCT
04
3-0
01
LCT
04
3-0
07
LCT
04
4-0
04
LCT
04
5-0
02
LCT
04
5-0
08
LCT
04
5-0
14
LCT
04
5-0
20
LCT
04
6-0
05
LCT
04
6-0
11
LCT
04
6-0
17
LCT
04
7-0
01
LCT
05
4-0
04
LCT
05
7-0
02
LCT
05
7-0
08
LCT
05
7-0
14
LCT
05
7-0
20
LCT
05
7-0
26
LCT
05
7-0
32
LCT
05
8-0
02
LCT
05
8-0
08
LCT
06
1-0
05
LCT
07
0-0
01
LCT
07
0-0
07
LCT
07
1-0
01
LCT
07
2-0
03
LCT
07
2-0
09
LCT
07
5-0
02
%(C
-E)/
C
Case
This plot can be deceiving as no ND correlations
information is shown (and no Integral benchmark
correlations)
1 Sigma ND
Benchmark Model Unc
© 2017 Organisation for Economic Co-operation and Development
Initial Chi Squared Within An Evaluation – No Integral Experiment Correlations
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75Init
ial C
hi S
qu
are
d p
er
DO
F
Evaluation ID
NOCORR
8 22 20 27 30 32 29 6 7 6 4 20 12 9 52 26 3 11 14 17 7 9 10 21 22 8 3 36 10 9 12 4 9 4 6
8 > chi/dof=1
26 < chi/dof=1
Total Uncertainty Over Estimated (Either ND, INT), Or the degree of correlation is mis-estimated (under?)
# Cases
d=c-e
total = 2.4
ND+Int.BM
© 2017 Organisation for Economic Co-operation and Development
Chi Squared Within An Evaluation – No Integral Experiment Correlations
8 22 20 27 30 32 29 6 7 6 4 20 12 9 52 26 3 11 14 17 7 9 10 21 22 8 3 36 10 9 12 4 9 4 6
28 > chi/dof=1
6 < chi/dof=1
Total Uncertainty Under Estimated (Either ND, INT or mis-estimated Covariance (over?))
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Init
ial C
hi S
qu
are
d p
er
DO
F
Evaluation ID
CORR
# Cases
d=c-e total = 4.7
ND+Int.BM
Since this isn’t a real adjustment exercise next step of chi squared filtering wasn’t done
© 2017 Organisation for Economic Co-operation and Development
Components of Chi squared
Contributing terms are:
1) Shared Errors in Nuclear Data
2) Unshared Errors in Nuclear Data
3) Shared Uncertainties in Benchmark Model
4) Unshared Errors In Benchmark Model
Contributing terms are:
1) Shared Errors in Nuclear Data [Often Similar Between CASES AND
EVALUATIONS].
2) Unshared Errors in Nuclear Data [SMALL]
3) Shared Uncertainties in Experimental Benchmark Model [Similar
between CASES, but not between EVALUATIONS (exp series)].
4) Unshared Errors In Experimental Benchmark Model ‘[All cases].
© 2017 Organisation for Economic Co-operation and Development
How Similar is the Response to ND?
9.00E-01
9.10E-01
9.20E-01
9.30E-01
9.40E-01
9.50E-01
9.60E-01
9.70E-01
9.80E-01
9.90E-01
1.00E+00
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75A
vera
ge D
ot
Pro
du
ct o
f 3
G C
ase
Le
vel S
en
siti
vity
Ve
cto
rs
Evaluation ID
Many of these have a very similar nuclear data response
The uncertainty from ND isn’t so much different
© 2017 Organisation for Economic Co-operation and Development
And Some ck’s (same as before but covariance weighted)
LCT001
1 2 3 4 5 6 7 8
1 1.0000E+00 9.9690E-01 9.9780E-01 9.9860E-01 9.9760E-01 9.9770E-01 9.9810E-01 9.9620E-01
2 9.9690E-01 1.0000E+00 9.9990E-01 9.9960E-01 9.9990E-01 9.9990E-01 9.9970E-01 9.9990E-01
3 9.9780E-01 9.9990E-01 1.0000E+00 9.9990E-01 9.9990E-01 9.9990E-01 9.9990E-01 9.9970E-01
4 9.9860E-01 9.9960E-01 9.9990E-01 1.0000E+00 9.9980E-01 9.9980E-01 9.9980E-01 9.9930E-01
5 9.9760E-01 9.9990E-01 9.9990E-01 9.9980E-01 1.0000E+00 1.0000E+00 9.9990E-01 9.9980E-01
6 9.9770E-01 9.9990E-01 9.9990E-01 9.9980E-01 1.0000E+00 1.0000E+00 9.9990E-01 9.9980E-01
7 9.9810E-01 9.9970E-01 9.9990E-01 9.9980E-01 9.9990E-01 9.9990E-01 1.0000E+00 9.9960E-01
8 9.9620E-01 9.9990E-01 9.9970E-01 9.9930E-01 9.9980E-01 9.9980E-01 9.9960E-01 1.0000E+00
LCT022
LCT054
1 2 3 4 5 6 7
1 1.000 0.984 0.936 0.901 0.880 0.843 0.841
2 0.984 1.000 0.982 0.961 0.945 0.907 0.904
3 0.936 0.982 1.000 0.995 0.987 0.953 0.949
4 0.901 0.961 0.995 1.000 0.998 0.973 0.970
5 0.880 0.945 0.987 0.998 1.000 0.984 0.981
6 0.843 0.907 0.953 0.973 0.984 1.000 1.000
7 0.841 0.904 0.949 0.970 0.981 1.000 1.000
1 2 3 4 5 6 7 8
1 1.0000 1.0000 0.9999 0.9999 0.9999 1.0000 1.0000 1.0000
2 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
3 0.9999 1.0000 1.0000 1.0000 1.0000 1.0000 0.9999 1.0000
4 0.9999 1.0000 1.0000 1.0000 1.0000 0.9999 0.9999 0.9999
5 0.9999 1.0000 1.0000 1.0000 1.0000 1.0000 0.9999 1.0000
6 1.0000 1.0000 1.0000 0.9999 1.0000 1.0000 1.0000 1.0000
7 1.0000 1.0000 0.9999 0.9999 0.9999 1.0000 1.0000 1.0000
8 1.0000 1.0000 1.0000 0.9999 1.0000 1.0000 1.0000 1.0000
© 2017 Organisation for Economic Co-operation and Development
And Some Sensitivity Profiles: LCT001, 8 Cases
U235 Total, and H1 Total. You are looking a all 8 cases
Moral is: the metrics aren’t perfect…but cases have very similar
uncertainty to nuclear data and response to changes in nuclear data.
© 2017 Organisation for Economic Co-operation and Development
LCT Data Set
-2.50E+00
-2.00E+00
-1.50E+00
-1.00E+00
-5.00E-01
0.00E+00
5.00E-01
1.00E+00
1.50E+00
2.00E+00
2.50E+00
3.00E+00
LCT
00
1-0
01
LCT
00
1-0
07
LCT
00
3-0
05
LCT
00
3-0
11
LCT
00
3-0
17
LCT
00
4-0
01
LCT
00
4-0
07
LCT
00
4-0
13
LCT
00
4-0
19
LCT
00
9-0
05
LCT
00
9-0
11
LCT
00
9-0
17
LCT
00
9-0
23
LCT
01
0-0
02
LCT
01
0-0
08
LCT
01
0-0
14
LCT
01
0-0
20
LCT
01
0-0
26
LCT
01
6-0
02
LCT
01
6-0
08
LCT
01
6-0
14
LCT
01
6-0
20
LCT
01
6-0
26
LCT
01
6-0
32
LCT
01
7-0
06
LCT
01
7-0
12
LCT
01
7-0
18
LCT
01
7-0
24
LCT
02
1-0
01
LCT
02
2-0
01
LCT
02
2-0
07
LCT
02
6-0
06
LCT
02
8-0
02
LCT
02
8-0
08
LCT
02
8-0
14
LCT
02
8-0
20
LCT
02
9-0
06
LCT
02
9-0
12
LCT
03
2-0
06
LCT
03
3-0
03
LCT
03
3-0
09
LCT
03
3-0
15
LCT
03
3-0
21
LCT
03
3-0
27
LCT
03
3-0
33
LCT
03
3-0
39
LCT
03
3-0
45
LCT
03
3-0
51
LCT
03
4-0
05
LCT
03
4-0
12
LCT
03
4-0
19
LCT
03
4-0
25
LCT
03
7-0
02
LCT
03
7-0
08
LCT
03
8-0
03
LCT
03
8-0
09
LCT
03
9-0
01
LCT
03
9-0
07
LCT
03
9-0
13
LCT
04
2-0
02
LCT
04
3-0
01
LCT
04
3-0
07
LCT
04
4-0
04
LCT
04
5-0
02
LCT
04
5-0
08
LCT
04
5-0
14
LCT
04
5-0
20
LCT
04
6-0
05
LCT
04
6-0
11
LCT
04
6-0
17
LCT
04
7-0
01
LCT
05
4-0
04
LCT
05
7-0
02
LCT
05
7-0
08
LCT
05
7-0
14
LCT
05
7-0
20
LCT
05
7-0
26
LCT
05
7-0
32
LCT
05
8-0
02
LCT
05
8-0
08
LCT
06
1-0
05
LCT
07
0-0
01
LCT
07
0-0
07
LCT
07
1-0
01
LCT
07
2-0
03
LCT
07
2-0
09
LCT
07
5-0
02
%(C
-E)/
C
Case
© 2017 Organisation for Economic Co-operation and Development
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Init
ial C
hi S
qu
are
d p
er
DO
F
Evaluation ID
NOCORRCORR
• Some series are impacted by integral benchmark correlations, some aren’t.
• At an assigned average correlation of 50% chi doesn’t move, above 80% it moves
a lot…..caution advised above 90%! [possible limits? Check MVB? PPP Limits?]
• Looking at chi squared, we can lose some appreciation of spread vs. bias
• ND should be mostly bias.
Properties: ND vs INT Correlations
95 47 54 83 80 92 81 62 84 88 64 91 67 78 35 53 80 61 30 94 59 93 95 44 92 95 93 48 93 90 97 85 68 95 94 Av corr
assigned%
© 2017 Organisation for Economic Co-operation and Development
More Tests…..
What is the spread of results compared to the anticipated
spread? (Attempt to Check Random Component)
d=(c-e)/c
Con: Very sensitivity
to bias
Pro: Good for
identifying odd series
• Benchmark uncertainty is currently treated as uncorrelated • Nuclear data uncertainty within a case is HIGHLY correlated • Random spread of C-E/C results, would be close to the
benchmark uncertainty if results are uncorrelated.
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 20 40 60 80 100 120(C-E
)/C
Data below ~same chi squared…
© 2017 Organisation for Economic Co-operation and Development
-1.80E-02
-1.60E-02
-1.40E-02
-1.20E-02
-1.00E-02
-8.00E-03
-6.00E-03
-4.00E-03
-2.00E-03
0.00E+00
2.00E-03
0 5 10 15 20 25
C-E
CASE
Mapping Data Sets…..
© 2017 Organisation for Economic Co-operation and Development
Standard Deviation of (C-E)/C Values Before GLS Fit a) STDSpread is less than Experimental Benchmark Uncertainty. Good! b) Often, spread should mostly come from the random component of Integral Benchmark Uncertainty.
Thankfully the quoted component (systematic+random) is less than the actual spread (random) mostly. You would expect that where the spread is low vs EXP, there may be high Integral Correlations!
c) The red bar is mostly bias and is ND+Integral. Since the total is fixed from the evaluation, later when I add correlations, I increase shared uncertainty/bias, and decrease expected std (blue).
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Stan
dar
d D
evia
tio
n a
nd
Su
m o
f A
BS
(Bia
s)
Evaluation ID
Exp.std (C-E)ABS |(C-E)|
© 2017 Organisation for Economic Co-operation and Development
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Stan
dar
d D
evi
atio
n a
nd
Su
m o
f A
BS
(Bia
s)
Exp.
std (C-E)
ABS |(C-E)|
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Init
ial C
hi S
qu
are
d p
er
DO
F With a high bias, putting experimental correlations doesn’t really change goodness of fit……
© 2017 Organisation for Economic Co-operation and Development
After GLS Fit Data
-2.00E+00
-1.50E+00
-1.00E+00
-5.00E-01
0.00E+00
5.00E-01
1.00E+00
1.50E+00
2.00E+00
LCT0
01-
001
LCT0
03-
001
LCT0
03-
009
LCT0
03-
017
LCT0
04-
003
LCT0
04-
011
LCT0
04-
019
LCT0
09-
007
LCT0
09-
015
LCT0
09-
023
LCT0
10-
004
LCT0
10-
012
LCT0
10-
020
LCT0
10-
028
LCT0
16-
006
LCT0
16-
014
LCT0
16-
022
LCT0
16-
030
LCT0
17-
006
LCT0
17-
014
LCT0
17-
022
LCT0
21-
001
LCT0
22-
003
LCT0
26-
004
LCT0
28-
002
LCT0
28-
010
LCT0
28-
018
LCT0
29-
006
LCT0
32-
002
LCT0
33-
001
LCT0
33-
009
LCT0
33-
017
LCT0
33-
025
LCT0
33-
033
LCT0
33-
041
LCT0
33-
049
LCT0
34-
005
LCT0
34-
014
LCT0
34-
023
LCT0
37-
004
LCT0
38-
001
LCT0
38-
009
LCT0
39-
003
LCT0
39-
011
LCT0
42-
002
LCT0
43-
003
LCT0
44-
002
LCT0
44-
010
LCT0
45-
008
LCT0
45-
016
LCT0
46-
003
LCT0
46-
011
LCT0
46-
019
LCT0
54-
002
LCT0
57-
002
LCT0
57-
010
LCT0
57-
018
LCT0
57-
026
LCT0
57-
034
LCT0
58-
006
LCT0
61-
005
LCT0
70-
003
LCT0
70-
011
LCT0
72-
003
LCT0
74-
002
LCT0
75-
006
%(A
DJ-
EXP
)/EX
P
NOCORR_%(adj- exp)/exp
CORR_%(adj- exp)/exp
These are ‘in series’ adjustments. Within LCT001 for example.
I’ll show the ‘big case’ adjustment later!
© 2017 Organisation for Economic Co-operation and Development
Standard Deviation of (C-E)/C Values after GLS Fit
• Correlated have a higher spread • Spread is less than the experimental benchmark uncertainty (uncorrelated spread) • Nuclear data is highly correlated, often higher than 0.95, so most of the spread
should be from experimental uncertainty • Experimental uncertainty either too high or correlation is underestimated
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Stan
dar
d D
evia
tio
n a
nd
Su
m o
f A
BS
(Bia
s)
Evaluation ID
Exp.NOCORR STDNOCORR ABS[C-E]CORR STDCORR ABS[C-E]
© 2017 Organisation for Economic Co-operation and Development
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02LC
T001
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Init
ial C
hi S
qu
are
d p
er
DO
F
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Stan
dar
d D
evia
tio
n a
nd
Su
m o
f A
BS
(Bia
s)
Evaluation ID
Exp.
NOCORR STD
NOCORR ABS[C-E]
CORR STD
CORR ABS[C-E]
© 2017 Organisation for Economic Co-operation and Development
Impact on Adjustment
Many XS adjusted. Chose to look at U238
inelastic and elastic in the fast energy range
MT2:MT2
MT2:MT4
MT4:MT4
U238, Fast, Inelastic Sensitivity ~ 25 pcm/%XS
U238, Fast, elastic Sensitivity ~ 7 pcm/%XS
© 2017 Organisation for Economic Co-operation and Development
-60.00
-40.00
-20.00
0.00
20.00
40.00
60.00
80.00
100.00
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
LCT_
com
.
Re
lati
ve C
han
ge in
Cro
ss S
ect
ion
(%)
Evaluation ID
u-238 n,n'(F,NC)
u-238 elastic(F,NC)
u-238 n,n'(F,C)
u-238 elastic(F,C)
© 2017 Organisation for Economic Co-operation and Development
-2.50E+00
-2.00E+00
-1.50E+00
-1.00E+00
-5.00E-01
0.00E+00
5.00E-01
1.00E+00
1.50E+00
2.00E+00
2.50E+00LC
T00
1-0
01LC
T00
3-0
01LC
T00
3-0
09LC
T00
3-0
17LC
T00
4-0
03LC
T00
4-0
11LC
T00
4-0
19LC
T00
9-0
07LC
T00
9-0
15LC
T00
9-0
23LC
T01
0-0
04LC
T01
0-0
12LC
T01
0-0
20LC
T01
0-0
28LC
T01
6-0
06LC
T01
6-0
14LC
T01
6-0
22LC
T01
6-0
30LC
T01
7-0
06LC
T01
7-0
14LC
T01
7-0
22LC
T02
1-0
01LC
T02
2-0
03LC
T02
6-0
04LC
T02
8-0
02LC
T02
8-0
10LC
T02
8-0
18LC
T02
9-0
06LC
T03
2-0
02LC
T03
3-0
01LC
T03
3-0
09LC
T03
3-0
17LC
T03
3-0
25LC
T03
3-0
33LC
T03
3-0
41LC
T03
3-0
49LC
T03
4-0
05LC
T03
4-0
14LC
T03
4-0
23LC
T03
7-0
04LC
T03
8-0
01LC
T03
8-0
09LC
T03
9-0
03LC
T03
9-0
11LC
T04
2-0
02LC
T04
3-0
03LC
T04
4-0
02LC
T04
4-0
10LC
T04
5-0
08LC
T04
5-0
16LC
T04
6-0
03LC
T04
6-0
11LC
T04
6-0
19LC
T05
4-0
02LC
T05
7-0
02LC
T05
7-0
10LC
T05
7-0
18LC
T05
7-0
26LC
T05
7-0
34LC
T05
8-0
06LC
T06
1-0
05LC
T07
0-0
03LC
T07
0-0
11LC
T07
2-0
03LC
T07
4-0
02LC
T07
5-0
06
%(A
DJ-
EXP
)/EX
P
NOCORR_%(adj- exp)/exp
CORR_%(adj- exp)/exp
Total_NOCORR_%(adj- exp)/exp
Total_CORR_%(adj- exp)/exp
Total is the big GLS with all LCT cases in one run
© 2017 Organisation for Economic Co-operation and Development
-60.00-40.00-20.00
0.0020.0040.0060.0080.00
100.00
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
LCT_
com
.
Rel
ativ
e C
han
ge in
Cro
ss S
ecti
on
(%)
Evaluation ID
u-238 n,n'(F,NC)u-238 elastic(F,NC)u-238 n,n'(F,C)u-238 elastic(F,C)
1.00E-02
1.00E-01
1.00E+00
1.00E+01
1.00E+02
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Init
ial C
hi S
qu
are
d p
er
DO
F
00.0010.0020.0030.0040.0050.0060.0070.008
LCT0
01
LCT0
03
LCT0
04
LCT0
09
LCT0
10
LCT0
16
LCT0
17
LCT0
21
LCT0
22
LCT0
26
LCT0
27
LCT0
28
LCT0
29
LCT0
32
LCT0
33
LCT0
34
LCT0
35
LCT0
37
LCT0
38
LCT0
39
LCT0
42
LCT0
43
LCT0
44
LCT0
45
LCT0
46
LCT0
47
LCT0
54
LCT0
57
LCT0
58
LCT0
61
LCT0
70
LCT0
71
LCT0
72
LCT0
74
LCT0
75
Stan
dar
d D
evia
tio
n a
nd
Su
m o
f A
BS
(Bia
s)
Exp.
NOCORR STD
NOCORR ABS[C-E]
© 2017 Organisation for Economic Co-operation and Development
Summary
• Little correlation data available for integral experiments
• When doing integral testing, it’s tough to judge the nuclear data covariance without integral covariance
• Generated integral benchmark correlations for LCT
• Took a look at their effect on various adjustment quantities
• Chi squared is good for finding outliers with odd systematic error….but its not great for checking constancy of random errors
• Deviations are good for random errors, nice to compare against integral benchmark correlation matrix
• Suggestions on checking benchmark correlations are appreciated!
• Adjustment is a statistical process….need lots of data
• Integral correlations have a strong impact on adjustments and it’s a difficult task to generate them in aggregate
• Would be interesting to expand scale…….test PIA for series etc.
top related