Project

General

Profile

Meeting 2015 09 04 » History » Version 7

Nicolas Ponieman, 09/02/2015 08:30 PM

1 1 Nicolas Ponieman
{{toc}}
2 1 Nicolas Ponieman
3 1 Nicolas Ponieman
h1. Meeting 2015 09 04
4 1 Nicolas Ponieman
5 1 Nicolas Ponieman
h2. Generic settings
6 1 Nicolas Ponieman
7 1 Nicolas Ponieman
/N02/det/trackMaxSteps 1000000000
8 1 Nicolas Ponieman
/N02/det/trackMaxLen   10 km
9 1 Nicolas Ponieman
/gun/energy 10.9181415106 MeV
10 1 Nicolas Ponieman
/N02/prec/setDeltaOne   1.0e-2 mm  
11 1 Nicolas Ponieman
/N02/prec/setDeltaInt   1.0e-5 mm
12 1 Nicolas Ponieman
/N02/prec/setEpsilonMin 1.0e-3
13 1 Nicolas Ponieman
/N02/prec/setEpsilonMax 1.0e-3
14 1 Nicolas Ponieman
/N02/prec/setDeltaChord 0.25 mm
15 1 Nicolas Ponieman
16 1 Nicolas Ponieman
MSE is the Mean Squared Error
17 1 Nicolas Ponieman
MSE = 1/N * sum_{i=1}^{n} (G4_i - AN_i)^2
18 1 Nicolas Ponieman
N = Number of output points
19 2 Nicolas Ponieman
r_err
20 1 Nicolas Ponieman
21 1 Nicolas Ponieman
h2. Experiment 1 - Vary epsilon in G4
22 1 Nicolas Ponieman
23 1 Nicolas Ponieman
(see [[Meeting_2015_08_21|August 21 meeting]])
24 1 Nicolas Ponieman
25 1 Nicolas Ponieman
/N02/det/stepMax 20. mm   (i.e., stepMax = 20mm)
26 1 Nicolas Ponieman
epsilonMax = epsilonMin = epsilon
27 1 Nicolas Ponieman
The mean square error doesn't seem to vary.
28 1 Nicolas Ponieman
The error, however, keeps growing with distance.
29 1 Nicolas Ponieman
30 6 Nicolas Ponieman
|_. epsilon |_. Simulation Time |_. RHS evaluation steps |_. Mean Square Error [mm^2] |_. sqrt(MSE) [mm] |_. Max(X_Error) [mm] |
31 6 Nicolas Ponieman
| 1.0e-3    |        0m13.725s | 16500010        | 291.21 | 17.06 | 30.54 |
32 6 Nicolas Ponieman
| 1.0e-4    |        0m13.725s | 16500010        | 291.21 | 17.06 | 30.54 |
33 6 Nicolas Ponieman
| 1.0e-5    |        0m13.725s | 16500010        | 291.21 | 17.06 | 30.54 |
34 6 Nicolas Ponieman
| 1.0e-6    |        0m21.815s | 38500010        | 294.09 | 17.14 | 30.68 |
35 6 Nicolas Ponieman
| 1.0e-7    |        0m27.980s | 60500010        | 294.22 | 17.15 | 30.69 |
36 6 Nicolas Ponieman
| 1.0e-8    |        0m41.054s | 104499999       | 294.22 | 17.15 | 30.69 |
37 6 Nicolas Ponieman
| 1.0e-9    |        0m55.830s | 159500010        | 294.22 | 17.15 | 30.69 |
38 6 Nicolas Ponieman
| 1.0e-10    |        1m25.948s | 269499999       | 294.22 | 17.15 | 30.69 |
39 6 Nicolas Ponieman
| 1.0e-11    |        2m17.996s | 456500010       | 294.22 | 17.15 | 30.69 |
40 6 Nicolas Ponieman
| 1.0e-12    |        3m50.224s | 786518324        | 294.22 | 17.15 | 30.69 |
41 1 Nicolas Ponieman
42 1 Nicolas Ponieman
A few plots for different values of epsilon:
43 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28621/G4_vs_AN_error_eps1e-3.png!
44 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28620/G4_vs_AN_error_eps1e-6.png!
45 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28619/G4_vs_AN_error_eps1e-9.png!
46 1 Nicolas Ponieman
47 1 Nicolas Ponieman
The 3 plots look exactly the same!
48 1 Nicolas Ponieman
The output for those epsilon values is different, though.
49 1 Nicolas Ponieman
50 4 Nicolas Ponieman
h2. Experiment 2 - Vary stepMax in G4
51 1 Nicolas Ponieman
52 1 Nicolas Ponieman
/N02/prec/setEpsilonMin 1.0e-3  (i.e, epsilon = 1.0e-3)
53 1 Nicolas Ponieman
/N02/prec/setEpsilonMax 1.0e-3
54 3 Nicolas Ponieman
Errors were measured for first 2 km (the file size for 10 km in the case of stepMax=0.2mm was too big)
55 1 Nicolas Ponieman
56 3 Nicolas Ponieman
|_. stepMax [mm] |_. Simulation Time |_. RHS evaluation steps |_. Mean Square Error [mm^2] |_.  sqrt(MSE)[mm] |_. Max(X_Error) [mm] |
57 3 Nicolas Ponieman
| 0.2    |        13m34.816s | 550000011        | 12.12 | 3.48| 6.16 |
58 3 Nicolas Ponieman
| 1     |        2m51.026s | 110000000        | 12.11 | 3.48 |  6.15 |
59 3 Nicolas Ponieman
| 2     |        1m28.192s | 55000000        | 12.11 | 3.48 | 6.15 |
60 3 Nicolas Ponieman
| 5     |        0m33.121s | 22000000        | 12.11 | 3.48 | 6.15 |
61 3 Nicolas Ponieman
| 10    |        0m21.907s | 22000010       | 12.08 | 3.47 | 6.14 |
62 3 Nicolas Ponieman
| 20    |        0m13.797s | 16500010       | 11.98 | 3.46 | 6.12 |
63 3 Nicolas Ponieman
| 50    |        0m8.319s | 13200010        | 11.71 | 3.42 | 6.03 |
64 3 Nicolas Ponieman
| 100    |        0m7.179s | 13200020       | 11.25 | 3.35 | 5.89 |
65 3 Nicolas Ponieman
| 150    |        0m6.848s | 13200020       |  10.79 | 3.28 | 5.76 |
66 3 Nicolas Ponieman
| 200    |        0m6.382s | 12650020        | 10.35 | 3.21 | 5.65 |
67 3 Nicolas Ponieman
68 1 Nicolas Ponieman
Although all the errors look similar, there is a clear and counterintuitive tendency for the error to become smaller as stepMax is increased. 
69 4 Nicolas Ponieman
A few plots for different values of stepMax:
70 5 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28708/G4_vs_AN_error_stepMax0.2mm.png!
71 5 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28707/G4_vs_AN_error_stepMax2mm.png!
72 5 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28706/G4_vs_AN_error_stepMax20mm.png!
73 5 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28705/G4_vs_AN_error_stepMax200mm.png!
74 1 Nicolas Ponieman
75 1 Nicolas Ponieman
h2. Experiment 3 - Vary DeltaQ in PD
76 1 Nicolas Ponieman
77 1 Nicolas Ponieman
Simulated with QSS3 method
78 1 Nicolas Ponieman
DeltaQmin = DeltaQ * 1e-3
79 1 Nicolas Ponieman
*Several values of the table were left blank deliberately as they weren't calculated yet*
80 1 Nicolas Ponieman
81 6 Nicolas Ponieman
|_. DeltaQ |_. Simulation Time |_. Equivalente in PD to RHS eval steps |_. Mean Square Error [mm^2] |_. sqrt(MSE)[mm] |_. Max(X_Error) [mm] |
82 1 Nicolas Ponieman
| 1.0e-1    | 0.41s | 5.60e5 | | | |
83 1 Nicolas Ponieman
| 1.0e-2    | 0.42s | 6.40e5 | | | |
84 7 Nicolas Ponieman
| 1.0e-3    | 1.33s | 2.40e6 | 704.43 | 26.54 | |38.08 
85 7 Nicolas Ponieman
| 1.0e-4    | 8.89s | 1.42e7 | 568.32 | 23.83 | 37.89 |
86 7 Nicolas Ponieman
| 1.0e-5    | 16.3s | 3.08e7 | 86.89 | 9.32 | 18.41 |
87 7 Nicolas Ponieman
| 1.0e-6    | 35.6s | 6.64e7 | 56.79 | 7.54 | 14.80 |
88 7 Nicolas Ponieman
| 1.0e-7    | 77.6s | 1.43e8 | 57.45 | 7.58 | 15.03 |
89 1 Nicolas Ponieman
| 1.0e-8    | 165s  | 3.09e8 |  | | |
90 1 Nicolas Ponieman
| 1.0e-9    | 354s  | 6.64e8 | | | |
91 1 Nicolas Ponieman
| 1.0e-10   | 740s  | 1.43e9 |  | | |
92 1 Nicolas Ponieman
| 1.0e-11   | 1630s | 3.09e9 |  | | |
93 1 Nicolas Ponieman
| 1.0e-12   |       |         |  | | |
94 1 Nicolas Ponieman
95 1 Nicolas Ponieman
A few plots for different values of DeltaQ:
96 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28625/PD_vs_AN_error_DeltaQ1e-3.png!
97 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28624/PD_vs_AN_error_DeltaQ1e-4.png!
98 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28623/PD_vs_AN_error_DeltaQ1e-5.png!
99 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28622/PD_vs_AN_error_DeltaQ1e-6.png!
100 1 Nicolas Ponieman
!https://cdcvs.fnal.gov/redmine/attachments/download/28626/PD_vs_AN_error_DeltaQ1e-7.png!
101 1 Nicolas Ponieman
*The last two plots look exactly the same*