10 recent dfn developments and comparison of dfn and fcm models viswanathan lanl
1. Spent Fuel and Waste Science and Technology
Recent DFN developments and Comparison of
DFN and FCM Models
Hari Viswanathan, Nataliia Makedonska, Satish Karra, Jeffrey Hyman
Los Alamos National Laboratory
Hadgu Teklu, Elena Kalinina, Yifeng Wang
Sandia National Laboratories
2017 SFWST Annual Working Group Meeting
May 24, 2017 Las Vegas, NV
2. Spent Fuel and
Waste Science and
Technology
SFWST Working Group Meeting 2
Summary of FY16-FY17 Work
Overview of dfnWorks
– 2017 R&D 100 nomination
Reduced order model developments
– Using graph-based machine learning to mimic DFNs efficiently
– Robust uncertainty quantification
Benchmark Testing of fracture models (DFN and FCM) using
relevant test case (part 1, Teklu part 2)
– Used Fuel Disposition Crystalline Rocks: FY16 Progress report
– Article submitted to Journal of Hydrology (currently under revision)
3. Spent Fuel and
Waste Science and
Technology
dfnWorks Worflow
3SFWST Working Group Meeting
4. Spent Fuel and
Waste Science and
Technology
R&D 100 Submission
SFWST Working Group Meeting 4
– Three minute movie made as
part of the entry:
http://www.lanl.gov/newsroo
m/video/index.php
5. Spent Fuel and
Waste Science and
Technology
Recent dfnWorks developments
Characterize when it is possible to replace computationally intensive
DFNs with efficient graph-based flow and transport models
– Use machine learning to mimic high fidelity DFN on graph
– Use for robust uncertainty quantification when 1000s of forward runs are
needed to bound system behavior
SFWST Working Group Meeting 5
Breakthrough Curve
6. Spent Fuel and
Waste Science and
Technology
Reduced Order Models
– Quantities of interest such as the cumulative breakthrough curve can be
reproduced directly on the graph leading to 3 orders of magnitude
computational savings in many cases
SFWST Working Group Meeting 6
Topology based pruning Graph Laplacian solver Computational Savings
Dimensionless Time
Dimensionless Time
CumulativeBreakthrough
CumulativeBreakthrough
7. Spent Fuel and
Waste Science and
Technology
DFN & FCM Comparison
SFWST Working Group Meeting 7
Discrete Fracture Network Fractured Continuum Model
Discrete Fracture Network Model generates fracture networks similar to those observed on
natural sites using site fracture characteristics.
DFN undertakes fracture percolation between boundary faces of the simulation domain.
DFN provides all necessary information for FCM for the transport and flow comparison.
8. Spent Fuel and
Waste Science and
Technology
Mapping DFN for use in FCM
SFWST Working Group Meeting 8
9. Spent Fuel and
Waste Science and
Technology To be continued …
SFWST Working Group Meeting 9
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+01 1.E+03 1.E+05
NormalizedBreakthrough
Time (year)
DFN ADE
ECM
DFN PT
Realization0
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+01 1.E+03 1.E+05
NormalizedBreakthrough
Time (year)
DFN ADE
ECM
DFN PT
Realization1
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+01 1.E+03 1.E+05
NormalaziedBreakthrough
Time (year)
DFN ADE
ECM
DFN PT
Realization2
0
0.2
0.4
0.6
0.8
1
1.E-01 1.E+01 1.E+03 1.E+05
NormalizedBreakthrough
Time (year)
DFN ADE
ECM
DFN PT
Realization3