runningqilin commited on
Commit
7e8d0ac
·
verified ·
1 Parent(s): 14acf28

HF card, cases.csv manifest, decks

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. README.md +146 -0
  2. card.json +73 -0
  3. cases.csv +111 -0
  4. decks/NB-I-320-Bullet-a-120.k +0 -0
  5. decks/NB-I-320-Bullet-a-160.k +0 -0
  6. decks/NB-I-320-Bullet-a-40.k +0 -0
  7. decks/NB-I-320-Bullet-a-80.k +0 -0
  8. decks/NB-I-320-Bullet-b-120.k +0 -0
  9. decks/NB-I-320-Bullet-b-160.k +0 -0
  10. decks/NB-I-320-Bullet-b-40.k +0 -0
  11. decks/NB-I-320-Bullet-b-80.k +0 -0
  12. decks/NB-I-320-Bullet-c-120.k +0 -0
  13. decks/NB-I-320-Bullet-c-160.k +0 -0
  14. decks/NB-I-320-Bullet-c-40.k +0 -0
  15. decks/NB-I-320-Bullet-c-80.k +0 -0
  16. decks/NB-I-320-Rectangular-a-120.k +0 -0
  17. decks/NB-I-320-Rectangular-a-160.k +0 -0
  18. decks/NB-I-320-Rectangular-a-40.k +0 -0
  19. decks/NB-I-320-Rectangular-a-80.k +0 -0
  20. decks/NB-I-320-Rectangular-b-120.k +0 -0
  21. decks/NB-I-320-Rectangular-b-160.k +0 -0
  22. decks/NB-I-320-Rectangular-b-40.k +0 -0
  23. decks/NB-I-320-Rectangular-b-80.k +0 -0
  24. decks/NB-I-320-Rectangular-c-120.k +0 -0
  25. decks/NB-I-320-Rectangular-c-160.k +0 -0
  26. decks/NB-I-320-Rectangular-c-40.k +0 -0
  27. decks/NB-I-320-Rectangular-c-80.k +0 -0
  28. decks/NB-I-320-Sphere-a-120.k +0 -0
  29. decks/NB-I-320-Sphere-a-160.k +0 -0
  30. decks/NB-I-320-Sphere-a-40.k +0 -0
  31. decks/NB-I-320-Sphere-a-80.k +0 -0
  32. decks/NB-I-320-Sphere-b-120.k +0 -0
  33. decks/NB-I-320-Sphere-b-160.k +0 -0
  34. decks/NB-I-320-Sphere-b-40.k +0 -0
  35. decks/NB-I-320-Sphere-b-80.k +0 -0
  36. decks/NB-I-320-Sphere-c-120.k +0 -0
  37. decks/NB-I-320-Sphere-c-160.k +0 -0
  38. decks/NB-I-320-Sphere-c-40.k +0 -0
  39. decks/NB-I-320-Sphere-c-80.k +0 -0
  40. decks/NB-I-480-Bullet-a-120.k +0 -0
  41. decks/NB-I-480-Bullet-a-160.k +0 -0
  42. decks/NB-I-480-Bullet-a-40.k +0 -0
  43. decks/NB-I-480-Bullet-a-80.k +0 -0
  44. decks/NB-I-480-Bullet-b-120.k +0 -0
  45. decks/NB-I-480-Bullet-b-160.k +0 -0
  46. decks/NB-I-480-Bullet-b-40.k +0 -0
  47. decks/NB-I-480-Bullet-b-80.k +0 -0
  48. decks/NB-I-480-Bullet-c-120.k +0 -0
  49. decks/NB-I-480-Bullet-c-160.k +0 -0
  50. decks/NB-I-480-Bullet-c-40.k +0 -0
README.md ADDED
@@ -0,0 +1,146 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ pretty_name: NotchBeam2D-Impact (StructBench)
4
+ tags:
5
+ - structural-engineering
6
+ - physics
7
+ - simulation
8
+ - lsdyna
9
+ - sph
10
+ - benchmark
11
+ size_categories:
12
+ - n<1K
13
+ configs:
14
+ - config_name: manifest
15
+ data_files:
16
+ - split: cases
17
+ path: cases.csv
18
+ ---
19
+ # NotchBeam2D-Impact — StructBench canonical dataset
20
+
21
+ ## Download
22
+
23
+ One case, one file — fetch exactly what you need (`pip install huggingface_hub`):
24
+
25
+ ```python
26
+ from huggingface_hub import hf_hub_download, snapshot_download
27
+
28
+ # one case
29
+ path = hf_hub_download("StructBench/notch-beam-2d-impact",
30
+ filename="<case_id>.h5", repo_type="dataset")
31
+
32
+ # the full archive (resumable; cached under HF_HOME)
33
+ root = snapshot_download("StructBench/notch-beam-2d-impact", repo_type="dataset")
34
+ ```
35
+
36
+ `cases.csv` lists every case with its split and loading/geometry
37
+ parameters plus a SHA-256 manifest; pin the dataset repo's `v0.1.0`
38
+ tag (`revision="v0.1.0"` — a data release, independent of the code
39
+ version) for reproducible pipelines. Point
40
+ `structbench-train --data-root` at the snapshot directory.
41
+ Code, benchmark protocol, and leaderboards:
42
+ <https://github.com/qilinli/StructBench>.
43
+
44
+ Autoregressive next-step surrogate of a 2D SPH notched concrete beam under drop-weight impact (ADR-0026). Covers 3 beam widths, 3 impactor shapes, 3 notch positions, and 4 velocities. Three bodies: the K&C concrete beam (part 1) is the predicted deformable; the steel impactor (part 2) and the two support blocks (part 3) are protocol-kinematic (ADR-0026) — driven by ground truth during rollout (both move: the impactor decelerates from its case velocity to ~10-20% on contact, the supports displace a few mm), excluded from the training loss and from position/strain metrics, with both QoIs restricted to concrete particles.
45
+
46
+ - Solver: LS-DYNA (SPH; erosion: no)
47
+ - Loading: drop-weight impact, initial velocity 40-160 m/s, impactor cross-sections plate/disk/rod (case names Rectangular/Sphere/Bullet)
48
+ - Source units: kg-mm-ms (files are strict SI, ADR-0012)
49
+ - Cases: 110 (train 88, val 8, test_interp 12, probe 2)
50
+ - Particles per case: 4264-12966; 502 frames at 0.001 ms
51
+ - Provenance: LS-DYNA parametric sweep (3 widths x 3 shapes x 3 notches x 4 velocities) produced by Curtin collaborators — extends the published 81-specimen drop-weight study (plate/disk/rod impactors at 80/120/160 m/s) with a 40 m/s velocity level; benchmark protocol per ADR-0026.
52
+ - License: CC BY 4.0
53
+
54
+ ## Task
55
+
56
+ autoregressive transition (ADR-0026). Auxiliary target: `max_principal_strain` (-). Models advance the particle state autoregressively from a short ground-truth prefix and are scored on the full predicted rollout.
57
+
58
+ ## Evaluation criteria
59
+
60
+ - Protocol (benchmark-owned, ADR-0032, ADR-0035): 6 input frames, horizon frames [6, 250) of 502 scored (250 µs, ADR-0039); full-length diagnostic, scored at native output times.
61
+ - Protocol rationale: Confirmed (maintainer, 2026-07-20): input_frames = 6 gives C = 5 input velocities (input_frames - 1), the GNS reference history length — the velocity budget is the criterion, not a rigid prefix. The timeline analysis (2026-07-20, on the DUG data copy) shows impact contact from frame 0, so the observed window takes in the first 6 us of contact; accepted. Scored horizon (ADR-0039): rollout metrics and QoIs are scored on frames [input_frames, 250) (250 µs). Internal energy reaches 99% of its final value by frame 77-213 (width-dependent); the remaining frames are ballistic separation and elastic ringing, which dominated full-horizon RMSE (half the final error accrued after frame 301 in baseline rollouts) while adding no fracture physics. The full 502-frame error curve remains a non-leaderboard long-horizon diagnostic. The cracked_fraction QoI threshold 0.01 is a declared protocol definition (ADR-0029, amended 2026-08-06): the SPH source model has no erosion or crack criterion; a 221-case sweep shows the GT fraction shifts ~0.05 mean per case across the factor-2 band [0.005, 0.02], and frame-249 vs frame-501 fractions are nearly identical (0.305 vs 0.317 mean), corroborating the 250 us horizon. Probe split (characterisation, 2026-08-15; height axis added 2026-08-27): the probe cases are out-of-distribution on up to FOUR axes at once — beam width (400/800 mm) and impactor velocity (140/60 m/s) both off the training grids ({320,480,640} mm; {40,80,120,160} m/s), beam height H=100 mm on the 800 mm case (every grid case is H=80; verified from canonical frame-0 extents, 2026-08-27), and, decisively, an OFF-CENTRE impact. All 108 train/val/test_interp cases are struck exactly at midspan (impact offset 0.0 mm, every notch a/b/c variant and width); the probe impacts land ~6% off-centre — a loading mode absent from training entirely. Probe scores therefore measure graceful failure on a genuinely new loading configuration, not ordinary interpolation: global-attention operators mis-localise the response to the learned midspan prior, while relative-position message-passing (MGN/CGN) degrades more gracefully.
62
+ - Metrics: headline is the pooled space+time relative L2 (displacement + max_principal_strain); also reported are position/max_principal_strain RMSE (physical units) and one-step RMSE, plus the quantities of interest below.
63
+ - Quantities of interest: midspan_deflection_peak, cracked_fraction.
64
+
65
+ ## Leaderboard
66
+
67
+ The headline metric is the pooled space+time relative L2 (↓ lower is better); the Scheme column is the prediction scheme. RMSE and quantities of interest are shown for the in-distribution `test_interp` split.
68
+
69
+ _Headline — pooled relative L2 (↓ better)_
70
+
71
+ | Method | Scheme | interp·disp | interp·aux | probe·disp | probe·aux |
72
+ |---|---|---|---|---|---|
73
+ | MGN | autoregressive | 0.2245 | 0.6988 | 0.7672 | 1.255 |
74
+ | CGN | autoregressive | 0.2827 | 0.5876 | 0.5905 | 0.8535 |
75
+ | Transolver | time-conditioned | 0.03517 | 0.2294 | 1.047 | 1.236 |
76
+ | Transolver++ | time-conditioned | 0.03699 | 0.2396 | 1.01 | 1.263 |
77
+
78
+ _Trajectory error — RMSE_
79
+
80
+ | Method | Scheme | interp·pos (mm) | interp·strain |
81
+ |---|---|---|---|
82
+ | MGN | autoregressive | 0.1984 | 0.01997 |
83
+ | CGN | autoregressive | 0.2497 | 0.01697 |
84
+ | Transolver | time-conditioned | 0.03365 | 0.006467 |
85
+ | Transolver++ | time-conditioned | 0.03509 | 0.006714 |
86
+
87
+ _Quantities of interest (MAE)_
88
+
89
+ | Method | Scheme | interp·midspan_deflection_peak (mm) | interp·cracked_fraction |
90
+ |---|---|---|---|
91
+ | MGN | autoregressive | 0.5842 | 0.1081 |
92
+ | CGN | autoregressive | 0.5843 | 0.1892 |
93
+ | Transolver | time-conditioned | 0.04067 | 0.0212 |
94
+ | Transolver++ | time-conditioned | 0.04696 | 0.02438 |
95
+
96
+ ## Baseline details
97
+
98
+ **MGN** (mgn, 2026-08-17, commit `59d5786`)
99
+
100
+ *Native MeshGraphNets (ADR-0047 baseline, ADR-0049 repair): autoregressive next-step on the SPH particle set as a mesh, val-selected model-best-212000.pt, seed 2 of the s1-s2 pair, run notch-mgn-base-s2. PROVISIONAL (ADR-0044/0045). On test_interp it roughly matches CGN (displacement relative L2 0.22 vs 0.28) and both trail Transolver ~6x; on the off-centre triple-OOD PROBE the relative-position message passing degrades more gracefully than the global-attention operator (disp 0.77 vs Transolver 1.05), though CGN is best there (0.59). Pooled relative L2 headline (ADR-0055) from the 2026-08-17 re-eval.*
101
+
102
+ **CGN** (cgn, 2026-07-24, commit `5956d81`, checkpoint: `models/notch_beam_2d_impact/cgn-5956d81/model-best-186000.pt` — private archive; publication parked)
103
+
104
+ *Single-scale CGN (ADR-0034) on the ADR-0039 §4 truncated recipe with the ADR-0038 strain knobs (train_frames 250, aux_tail_weight 3, asinh aux transform at scale 0.01; hidden 192 / 15 MP steps / 2-layer node MLP, noise_std 0.01, batch 4) at 250k steps; seed 1 of the 2026-07-24 h250c pair (seeds 1-2), val-selected checkpoint model-best-186000.pt (186k), one A100-80GB, ~80 h. Extending the same recipe from 200k to 250k steps cut seed-mean test rollout position RMSE 21% and deflection MAE 30% while validation strain RMSE stayed flat (0.0173 -> 0.0163): the extra budget buys kinematics, not damage-field quality. Caveats: the model over-predicts cracked fraction on the reviewed cases (crack MAE 0.19 vs sibling seed s2's 0.13, the one metric s2 wins); the off-grid probe case S_80_400_V140 is this seed's worst rollout (0.59 mm scored vs 0.40 for s2); predictions break the mirror symmetry of centered-notch cases while the ground truth stays symmetric (2026-07-24 finding); full-horizon (502-frame) rollout position RMSE is 0.87 mm on test_interp - diagnostic only, not scored. Relative L2 (rollout_rel_l2_disp/aux) is the pooled space+time headline (ADR-0055), added 2026-08-16 from a re-eval on this checkpoint; RMSE reproduced to <1%, so the blessed RMSE/QoI values are unchanged.*
105
+
106
+ **Transolver** (transolver, 2026-08-16, commit `59d5786`)
107
+
108
+ *Native time-conditioned Transolver (ADR-0054): history-free independent-time-query, no rollout accumulation, so one-step is N/A. Seed 2 of the s1-s2 pair, val-selected model-best-230000.pt, run notch-transolver-tc-s2. PROVISIONAL (ADR-0044/0045). On test_interp it is the strongest baseline (~8x lower displacement relative L2 than CGN); on the PROBE it fails hard (relative L2 > 1) - the off-centre triple-OOD case (ADR-0026 amendment) where the global-attention operator mis-localises the response to the learned midspan prior, while CGN's relative-position message passing degrades more gracefully (probe disp 0.59 vs 1.05). Pooled relative L2 headline (ADR-0055) from the 2026-08-16 re-eval.*
109
+
110
+ **Transolver++** (transolver_plus, 2026-08-18, commit `5b84119`)
111
+
112
+ *Transolver++ (ADR-0057): the native time-conditioned Transolver with both eidetic-state knobs ON - per-point adaptive slice temperature + train-only Gumbel Rep-Slice reparameterisation. Seed 2 of the s1-s2 pair (val-selected model-best-250000.pt, run notch-transolver-tcpp-s2), seed-matched to the plain Transolver entry above (also seed 2). PROVISIONAL method comparison (ADR-0046), not a blessed baseline. It is neutral-to-worse than plain Transolver: test_interp displacement is ~5% worse (0.03699 vs 0.03517) and strain ~4% worse; on the off-centre triple-OOD PROBE displacement is marginally better (1.01 vs 1.05) but strain is worse (1.263 vs 1.236) and both still fail hard (relative L2 > 1). No robust generalisation gain. Consistent with Transolver++ being designed for million-scale geometries: its eidetic-state edits do not help these O(10^4)-particle impact meshes. Pooled relative L2 headline (ADR-0055).*
113
+
114
+ ## Using this archive
115
+
116
+ ```bash
117
+ pip install structbench # or: pip install -e . from the repo
118
+ structbench-train --mode train --config configs/notch_beam_2d_impact/cgn.toml \
119
+ --data-root /path/to/this/folder --out runs/notch_beam_2d_impact-cgn
120
+ ```
121
+
122
+ Machine-readable metadata: `card.json` alongside this file.
123
+ `sph/stress` and `sph/strain` are 6-component Voigt tensors; scalar targets are loader-derived (see the card's aux field).
124
+ Consume with `structbench.datasets.load_case_trajectory` or any HDF5 reader (layout per ADR-0013).
125
+
126
+ ## References
127
+
128
+ - **MGN** — Pfaff, T., Fortunato, M., Sanchez-Gonzalez, A., & Battaglia, P. W. (2021). Learning Mesh-Based Simulation with Graph Networks. *ICLR*. https://arxiv.org/abs/2010.03409
129
+ - **CGN** — Li, Q., Wang, Z., Li, L., Hao, H., Chen, W., & Shao, Y. (2023). Machine learning prediction of structural dynamic responses using graph neural networks. *Computers & Structures*, 289, 107188. https://doi.org/10.1016/j.compstruc.2023.107188
130
+ - **Transolver** — Wu, H., Luo, H., Wang, H., Wang, J., & Long, M. (2024). Transolver: A Fast Transformer Solver for PDEs on General Geometries. *ICML*. https://arxiv.org/abs/2402.02366
131
+ - **Transolver++** — Luo, H., Wu, H., Zhou, H., Wang, J., & Long, M. (2025). Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometries. https://arxiv.org/abs/2502.02414. Adapted per ADR-0057 (thuml reference implementation github.com/thuml/Transolver_plus).
132
+
133
+ ## Citation
134
+
135
+ The data and the code are released together — cite the software
136
+ (`CITATION.cff` in the code repository):
137
+
138
+ ```bibtex
139
+ @software{structbench,
140
+ author = {Li, Qilin},
141
+ title = {StructBench: standardized benchmarks for machine learning on structural simulation},
142
+ year = {2026},
143
+ version = {0.3.0},
144
+ url = {https://github.com/qilinli/StructBench},
145
+ }
146
+ ```
card.json ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "NotchBeam2D-Impact",
3
+ "version": "0.1",
4
+ "description": "Autoregressive next-step surrogate of a 2D SPH notched concrete beam under drop-weight impact (ADR-0026). Covers 3 beam widths, 3 impactor shapes, 3 notch positions, and 4 velocities. Three bodies: the K&C concrete beam (part 1) is the predicted deformable; the steel impactor (part 2) and the two support blocks (part 3) are protocol-kinematic (ADR-0026) \u2014 driven by ground truth during rollout (both move: the impactor decelerates from its case velocity to ~10-20% on contact, the supports displace a few mm), excluded from the training loss and from position/strain metrics, with both QoIs restricted to concrete particles.",
5
+ "provenance": "LS-DYNA parametric sweep (3 widths x 3 shapes x 3 notches x 4 velocities) produced by Curtin collaborators \u2014 extends the published 81-specimen drop-weight study (plate/disk/rod impactors at 80/120/160 m/s) with a 40 m/s velocity level; benchmark protocol per ADR-0026.",
6
+ "data_license": "CC BY 4.0",
7
+ "solver": "LS-DYNA",
8
+ "discretisation": "SPH",
9
+ "materials": [
10
+ "*MAT_CONCRETE_DAMAGE_REL3 (K&C; density 2.4e-6 kg/mm3)",
11
+ "*MAT_PLASTIC_KINEMATIC"
12
+ ],
13
+ "erosion": false,
14
+ "loading": "drop-weight impact, initial velocity 40-160 m/s, impactor cross-sections plate/disk/rod (case names Rectangular/Sphere/Bullet)",
15
+ "source_units": "kg-mm-ms",
16
+ "geometry": "2D SPH notched beam, H 80 x W {320,480,640} mm",
17
+ "n_cases": 110,
18
+ "splits": {
19
+ "train": 88,
20
+ "val": 8,
21
+ "test_interp": 12,
22
+ "probe": 2
23
+ },
24
+ "task": "autoregressive transition (ADR-0026)",
25
+ "aux_field": "max_principal_strain",
26
+ "aux_unit": "-",
27
+ "qois": [
28
+ "midspan_deflection_peak",
29
+ "cracked_fraction"
30
+ ],
31
+ "fields": [
32
+ "node/displacement",
33
+ "node/velocity",
34
+ "node/acceleration",
35
+ "sph/stress",
36
+ "sph/strain",
37
+ "sph/strain_rate",
38
+ "sph/effective_plastic_strain",
39
+ "sph/pressure",
40
+ "sph/density",
41
+ "sph/internal_energy",
42
+ "sph/mass",
43
+ "sph/radius",
44
+ "sph/n_neighbors",
45
+ "sph/deletion",
46
+ "global/kinetic_energy",
47
+ "global/internal_energy",
48
+ "global/total_energy"
49
+ ],
50
+ "particles_per_case": "4264-12966",
51
+ "n_frames": 502,
52
+ "output_dt_ms": 0.001,
53
+ "input_frames": 6,
54
+ "protocol_rationale": "Confirmed (maintainer, 2026-07-20): input_frames = 6 gives C = 5 input velocities (input_frames - 1), the GNS reference history length \u2014 the velocity budget is the criterion, not a rigid prefix. The timeline analysis (2026-07-20, on the DUG data copy) shows impact contact from frame 0, so the observed window takes in the first 6 us of contact; accepted. Scored horizon (ADR-0039): rollout metrics and QoIs are scored on frames [input_frames, 250) (250 \u00b5s). Internal energy reaches 99% of its final value by frame 77-213 (width-dependent); the remaining frames are ballistic separation and elastic ringing, which dominated full-horizon RMSE (half the final error accrued after frame 301 in baseline rollouts) while adding no fracture physics. The full 502-frame error curve remains a non-leaderboard long-horizon diagnostic. The cracked_fraction QoI threshold 0.01 is a declared protocol definition (ADR-0029, amended 2026-08-06): the SPH source model has no erosion or crack criterion; a 221-case sweep shows the GT fraction shifts ~0.05 mean per case across the factor-2 band [0.005, 0.02], and frame-249 vs frame-501 fractions are nearly identical (0.305 vs 0.317 mean), corroborating the 250 us horizon. Probe split (characterisation, 2026-08-15; height axis added 2026-08-27): the probe cases are out-of-distribution on up to FOUR axes at once \u2014 beam width (400/800 mm) and impactor velocity (140/60 m/s) both off the training grids ({320,480,640} mm; {40,80,120,160} m/s), beam height H=100 mm on the 800 mm case (every grid case is H=80; verified from canonical frame-0 extents, 2026-08-27), and, decisively, an OFF-CENTRE impact. All 108 train/val/test_interp cases are struck exactly at midspan (impact offset 0.0 mm, every notch a/b/c variant and width); the probe impacts land ~6% off-centre \u2014 a loading mode absent from training entirely. Probe scores therefore measure graceful failure on a genuinely new loading configuration, not ordinary interpolation: global-attention operators mis-localise the response to the learned midspan prior, while relative-position message-passing (MGN/CGN) degrades more gracefully.",
55
+ "size_gb": 24.9,
56
+ "horizon": "frames [6, 250) of 502 scored (250 \u00b5s, ADR-0039); full-length diagnostic",
57
+ "eval_times": "native",
58
+ "overview": "## The problem\n\nA steel weight drops onto a **notched concrete beam** and the beam cracks: a\nshear wedge punches down under the impactor while flexural cracks open from the\nnotch and race up through the section. Concrete under high-rate impact is a\nbrutal surrogate target \u2014 a quasi-brittle material that softens as it damages,\ndiscrete cracks that localise strain, and a fracture pattern that *is* the\nscientific quantity of interest rather than a by-product of the motion.\n\nStructBench ships the 2D SPH version: an LS-DYNA `*MAT_CONCRETE_DAMAGE_REL3`\n(K&C) C50 notched beam \u2014 height H = 80 mm, width W = 320 / 480 / 640 mm \u2014\ndiscretised by\n2.5 mm SPH particles, struck at midspan by a drop weight of three cross-sections\n\u2014 plate 'P' (10 \u00d7 70 mm), disk 'D' (R = 15 mm), rod 'R' (70 \u00d7 10 mm); the\ndataset's case names call them Rectangular / Sphere / Bullet \u2014 at 40\u2013160 m/s,\nwith no element erosion. The task is an **autoregressive next-step surrogate** \u2014\nfrom a short ground-truth prefix the model advances the SPH particle state one\noutput step at a time, predicting both position and the per-particle **max\nprincipal strain**, the field that carries the crack pattern.\n\n![Schematic of the notch-beam impact setup: a drop weight above a simply-supported notched concrete beam, with the swept parameter ranges.](../../assets/problem_notch_beam_impact.png)\n\n*Problem setup: a drop weight \u2014 plate 'P', disk 'D', or rod 'R' cross-section\n(dataset case names Rectangular / Sphere / Bullet) \u2014 strikes the C50 notched\nbeam at midspan position A; beam height H = 80 mm, width W = 320 / 480 /\n640 mm, notch positions a / b / c. The figure shows the published three-level\nvelocity sweep (80 / 120 / 160 m/s); the benchmark grid adds a fourth level at\n40 m/s.*\n\n## Interpolation vs. the off-centre probe\n\n`test_interp` holds out interior combinations of beam width, impactor shape,\nnotch position and velocity \u2014 every factor level still appears in training, so\nit measures ordinary interpolation. The separate **probe** set is deliberately\nharder: it is out-of-distribution on up to *four* axes at once \u2014 a new width\n(400 / 800 mm), a new height on the 800 mm case (H = 100 mm; every training\nbeam is H = 80), an off-grid velocity (140 / 60 m/s), and, decisively, an\n**off-centre impact** (every in-distribution case is struck exactly at\nmidspan). It measures graceful failure\non a genuinely new loading mode, not interpolation \u2014 and it is where the method\nordering flips (a global-attention operator that wins in-distribution\nmis-localises the response there, while relative-position message passing\ndegrades more gracefully). Everything is scored over the 250 \u00b5s window (ADR-0039)\nin physical units \u2014 position RMSE in mm, strain RMSE \u2014 plus two quantities of\ninterest: peak mid-span deflection and the end-state cracked fraction. The\nnumbers, and the cross-method comparison, are below.",
59
+ "figures": [
60
+ {
61
+ "path": "assets/notch_rollout_methods.gif",
62
+ "caption": "Ground truth vs the three baselines (MGN, CGN, Transolver) on held-out NB-I-640-Sphere-c-120 (test_interp): a W = 640 mm beam under 120 m/s disk ('Sphere') impact, coloured by max principal strain (fringe capped at 0.05, 5x the 1% crack threshold) over the 250 \u00b5s scored window. Transolver (time-conditioned) tracks the central shear wedge and the discrete flexural cracks most closely (rollout strain RMSE 0.004); CGN diffuses them into streaky bands (0.013); MGN is the diffusest (0.020).",
63
+ "alt": "Stacked animation of ground-truth, MGN, CGN, and Transolver strain fringes on a notched concrete beam under drop-weight impact."
64
+ },
65
+ {
66
+ "path": "assets/notch_strain_methods_640_c_120.png",
67
+ "caption": "In-distribution max-principal-strain snapshots (test_interp, W = 640 mm, disk at 120 m/s) at 12 / 72 / 132 / 192 / 249 \u00b5s across the scored window: ground truth vs MGN vs CGN vs Transolver. Transolver (rollout strain RMSE 0.004) reproduces the shear wedge and the discrete flexural cracks closely, while CGN (0.013) and MGN (0.020) smear them into diffuse streaky bands \u2014 the damage field, not the kinematics, is the open gap. MGN and Transolver are provisional native baselines (ADR-0044/0045).",
68
+ "alt": "Grid of strain fringe snapshots comparing ground truth, MGN, CGN, and Transolver at five times."
69
+ }
70
+ ],
71
+ "data_access": "the canonical archive is public on Hugging Face \u2014 [StructBench/notch-beam-2d-impact](https://huggingface.co/datasets/StructBench/notch-beam-2d-impact) (CC BY 4.0): one `.h5` per case, `cases.csv` (split, loading/geometry parameters, SHA-256 manifest) and the LS-DYNA input decks under `decks/`. Fetch one case with `hf_hub_download` or the whole archive with `snapshot_download` (resumable) and point `--data-root` at it; pin the dataset repo's `v0.1.0` tag (`revision=\"v0.1.0\"` \u2014 a data release, independent of the code version) for reproducible pipelines. The maintainer's OneDrive copy remains the master (ADR-0040, amended 2026-08-28).",
72
+ "data_access_label": "public on Hugging Face \u2014 [StructBench/notch-beam-2d-impact](https://huggingface.co/datasets/StructBench/notch-beam-2d-impact)"
73
+ }
cases.csv ADDED
@@ -0,0 +1,111 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ case_id,split,beam_height_mm,beam_width_mm,impactor,impactor_cross_section,notch_position,impact_speed_ms,probe_label,n_nodes,n_frames,file_bytes,sha256
2
+ NB-I-320-Bullet-a-120,train,80,320,Bullet,rod,a,120,,4264,502,164392878,20859b3cc6b25763150ecf1c5755f6d91b734a221779b982d77428943e33e1b3
3
+ NB-I-320-Bullet-a-160,train,80,320,Bullet,rod,a,160,,4264,502,163482189,de958c4a5d047bd5aa8328b25e49936fcdb2eb20cab497f0e668d04b7c44bde8
4
+ NB-I-320-Bullet-a-40,train,80,320,Bullet,rod,a,40,,4264,502,161004199,fbcacae5afe6fa4358fafd879ac6db6367555a44ab64e79a45dae87d3d990d1e
5
+ NB-I-320-Bullet-b-160,train,80,320,Bullet,rod,b,160,,4264,502,164536344,ddde448e481ce0865a696d9dc0151e6d92b1d8aa467dd05873aef0e2009c708e
6
+ NB-I-320-Bullet-b-40,train,80,320,Bullet,rod,b,40,,4264,502,162540731,640b14321489170d75d83b33e9e17507189aae56c82260e4272d5c3afb7cbb53
7
+ NB-I-320-Bullet-c-120,train,80,320,Bullet,rod,c,120,,4264,502,163401186,9b49e215acd45d63c7b879436424c26232563814e56c624c2505d6446909cb52
8
+ NB-I-320-Bullet-c-160,train,80,320,Bullet,rod,c,160,,4264,502,164036539,a18f3a71e0780faa00fdd2b80d62511084600272efe47873c64fe703f024f72c
9
+ NB-I-320-Bullet-c-40,train,80,320,Bullet,rod,c,40,,4264,502,162636183,f2782e292a4bb8d1e534853b73934fb55699f64869116a6ceb5a242bfce3afbd
10
+ NB-I-320-Bullet-c-80,train,80,320,Bullet,rod,c,80,,4264,502,163527975,7d2eb1f6cbe8e0afd0cafba7f98943ecae749a79aa33ac95e71583095f1e3482
11
+ NB-I-320-Rectangular-a-120,train,80,320,Rectangular,plate,a,120,,4264,502,162701086,2d3fe4e9b363e48029bbe9caf113e72aa11cda863954827277c03f9b790381f7
12
+ NB-I-320-Rectangular-a-160,train,80,320,Rectangular,plate,a,160,,4264,502,165178013,c3e42f8ef4cf64a8738605da396c0a8579850a37d4456a50842ed2aa3dc05b59
13
+ NB-I-320-Rectangular-a-40,train,80,320,Rectangular,plate,a,40,,4264,502,165184066,3456e8fcbbfddd9f5766f99e7e84ce607f265dcb88de8f1c111af4c35d135419
14
+ NB-I-320-Rectangular-a-80,train,80,320,Rectangular,plate,a,80,,4264,502,163642494,0a6d8a2c39780b4e9d48f20ef67edd4cf621670b8f073c3517fd586cbdc582a0
15
+ NB-I-320-Rectangular-b-160,train,80,320,Rectangular,plate,b,160,,4264,502,164709697,2461f6a70e973432ad3274de92882d582d4cac3127d888bf9a92584fdefab8f6
16
+ NB-I-320-Rectangular-b-40,train,80,320,Rectangular,plate,b,40,,4264,502,164625381,47ffc06d951002967e33b6f76a64691b222c8276e702a677a614c90310a1178b
17
+ NB-I-320-Rectangular-b-80,train,80,320,Rectangular,plate,b,80,,4264,502,165331864,4453abdb969f753febbce546f30395b8955ea114a22376e9ef7b0cb7f0966b3a
18
+ NB-I-320-Rectangular-c-160,train,80,320,Rectangular,plate,c,160,,4264,502,165015576,99dd4051a96381dece821b6b3dde58cd3b01f0e8d30ef47bdd682d1419217629
19
+ NB-I-320-Rectangular-c-40,train,80,320,Rectangular,plate,c,40,,4264,502,164555822,969f88e5e2aa49fc09f905d81056c9a19ca88f37cd92d06f717fdb9104d59da0
20
+ NB-I-320-Rectangular-c-80,train,80,320,Rectangular,plate,c,80,,4264,502,165035458,f2f04cb3e34ceb6246dd3c2b6b2e759fc23090c57b4be0744811f4dc0d23ca08
21
+ NB-I-320-Sphere-a-160,train,80,320,Sphere,disk,a,160,,4264,502,162275499,1ce62cc9725adc5aad04425e6ddd040b03246854f07e055af701b6dacd142610
22
+ NB-I-320-Sphere-a-40,train,80,320,Sphere,disk,a,40,,4264,502,162779154,52ca855ba9a83075d45e22aa493da989450d8726b7465e9a1692056b95e4f8d0
23
+ NB-I-320-Sphere-b-160,train,80,320,Sphere,disk,b,160,,4264,502,163348968,b38feff7be2dc0cbc6e9f433ba3e36d07cccd6f4e4b2bcbeaa36198f444befba
24
+ NB-I-320-Sphere-b-40,train,80,320,Sphere,disk,b,40,,4264,502,163020173,13f2d02913593cc758f2e8634d6bd165494f26113faaff79114868263e74dced
25
+ NB-I-320-Sphere-c-120,train,80,320,Sphere,disk,c,120,,4264,502,163513999,1c4ab03674fe78afae293480b69c01397245769086e707ff37171f6076047da5
26
+ NB-I-320-Sphere-c-160,train,80,320,Sphere,disk,c,160,,4264,502,163219135,65e438e6b5872f3efe066127826895be8e44e08b1aafa3f958033843e8866ad5
27
+ NB-I-320-Sphere-c-40,train,80,320,Sphere,disk,c,40,,4264,502,163446510,fcecf8a6a7d281856031e5788449edb5c220820547f8115058650b2feeb14852
28
+ NB-I-320-Sphere-c-80,train,80,320,Sphere,disk,c,80,,4264,502,163359053,8b611af9892f024cb1a0d1f54fb0715d2f946f4ffa7fe4102f64487da9a03864
29
+ NB-I-480-Bullet-a-120,train,80,480,Bullet,rod,a,120,,6312,502,237406329,657c59a4e479ea03c27bcf07daf03bae1dacc2de6e080be80c889e7941279a6c
30
+ NB-I-480-Bullet-a-160,train,80,480,Bullet,rod,a,160,,6312,502,237318440,4f0b0aaa5719cfc75fe757278093da052d971f19bf100e6860936b239f407777
31
+ NB-I-480-Bullet-a-40,train,80,480,Bullet,rod,a,40,,6312,502,235425488,58fde9b49053730e5a7e01a20c980e2d88f4cc30c8f97109164723b3c9e614c3
32
+ NB-I-480-Bullet-a-80,train,80,480,Bullet,rod,a,80,,6312,502,236283260,2625b3b91c4fa08f056d22bc90b5b0a66ccb75e33d459b23168fd18a66747fa5
33
+ NB-I-480-Bullet-b-120,train,80,480,Bullet,rod,b,120,,6312,502,239851067,3fea4d997df3e8f4c694ec0d363f503e41b65af9b7e24eca4e14f67e7b38c589
34
+ NB-I-480-Bullet-b-160,train,80,480,Bullet,rod,b,160,,6312,502,239344687,20c6c1bc0b586b8b344105d11a9b0a89a5209f116535e1940837725446882ae1
35
+ NB-I-480-Bullet-b-40,train,80,480,Bullet,rod,b,40,,6312,502,236811941,477ac91dc1f163bc7523b102345d63a83f118e6b788082480cd0fe15c764e7e8
36
+ NB-I-480-Bullet-b-80,train,80,480,Bullet,rod,b,80,,6312,502,237864254,beeaabb4e9802f602972bba0a42b547729d886ebab51c209804fc4c0565c7378
37
+ NB-I-480-Bullet-c-120,train,80,480,Bullet,rod,c,120,,6312,502,239820396,1675c9bf26efffbb59bc51683dddb957c8eb1ae0666ffbd34c9ba76c7296e214
38
+ NB-I-480-Bullet-c-160,train,80,480,Bullet,rod,c,160,,6312,502,238950403,d4923a5f91407da492f80c203e113f15f07860b76948abb9d1feb2232f530aac
39
+ NB-I-480-Bullet-c-40,train,80,480,Bullet,rod,c,40,,6312,502,237416063,b9538d541de597ce34547d8c5f6748898c03088b0f80265f8beb26f2d8bac837
40
+ NB-I-480-Bullet-c-80,train,80,480,Bullet,rod,c,80,,6312,502,237557576,17c1d3443c5836f0b2ca2b286e90d6860f4633427db67b1ffc6af4f8816eeff1
41
+ NB-I-480-Rectangular-a-120,train,80,480,Rectangular,plate,a,120,,6312,502,241537581,ac39b7d1697a3f0c1c5a9990bc06262fd4699d3de694c9d9bdf2688ea12c5465
42
+ NB-I-480-Rectangular-a-160,train,80,480,Rectangular,plate,a,160,,6312,502,240105181,dfd6d657e33227f447b3c85058481e3c3512416d0b4d539627f728e4e55c065e
43
+ NB-I-480-Rectangular-a-40,train,80,480,Rectangular,plate,a,40,,6312,502,238704397,941c89ddf8cc06f0e19f24254bffee2eedd4b94123ef746718c4046babae79c7
44
+ NB-I-480-Rectangular-a-80,train,80,480,Rectangular,plate,a,80,,6312,502,238981489,c3a99592cf84f7d01ac4da75c56837cd04e697722d8eb6479367a1a383529e87
45
+ NB-I-480-Rectangular-b-160,train,80,480,Rectangular,plate,b,160,,6312,502,241598648,19b797f00aed4c86d9bac9f1e8639ac4784a4998b9e97e8d13799c8f98f26944
46
+ NB-I-480-Rectangular-b-40,train,80,480,Rectangular,plate,b,40,,6312,502,240190121,2a8ee5e4590073d6f48a375b16d1bef2b8a38f783d02264618a25211a10550b4
47
+ NB-I-480-Rectangular-c-120,train,80,480,Rectangular,plate,c,120,,6312,502,241381717,c39fbb167130975f2173ff7a3aaddb5828c246614e8bcd53bb59856ab9f5b462
48
+ NB-I-480-Rectangular-c-160,train,80,480,Rectangular,plate,c,160,,6312,502,240607833,4de300fba0cf87e3d3562939df81cf5100c7c4e656d0d6a5a2498de3f6c1a887
49
+ NB-I-480-Rectangular-c-40,train,80,480,Rectangular,plate,c,40,,6312,502,240530586,3e4b33695215bb2944fc8226d4a5b2907bdb84a72f7fe24b71d75b8a828cfc3b
50
+ NB-I-480-Rectangular-c-80,train,80,480,Rectangular,plate,c,80,,6312,502,239898843,4a9de440a8988f07483ca6f85ca271aa11e8b4a1e89ab83c9e368ef7a2d5f88e
51
+ NB-I-480-Sphere-a-160,train,80,480,Sphere,disk,a,160,,6312,502,237097801,ae464ac59e8dca78297055a1be2fce7915ab3155e6cdd9472811e1a70168ee04
52
+ NB-I-480-Sphere-a-40,train,80,480,Sphere,disk,a,40,,6312,502,236848949,2b4ec547f99287eb1854c6ca791042b6275148a36b2de61ec1aabb891ad56007
53
+ NB-I-480-Sphere-a-80,train,80,480,Sphere,disk,a,80,,6312,502,236546733,5b7a3d1e9eee7f2fba320e254f523c73089c9e5da54515f0349be0c83989aac0
54
+ NB-I-480-Sphere-b-160,train,80,480,Sphere,disk,b,160,,6312,502,239313479,0df7cd3b14fd2d4a51c3f4ec16c7154269544af14290d06092c25cb841c1f41a
55
+ NB-I-480-Sphere-b-40,train,80,480,Sphere,disk,b,40,,6312,502,237584912,dba722dcbd94b21a2a849a63a861c9dd1c7b269060a44cc0d514284e5a4c7d84
56
+ NB-I-480-Sphere-b-80,train,80,480,Sphere,disk,b,80,,6312,502,238245076,d23e7ca991e69ab4468e8bc170507315f224cb42581c4132e95389a964b5d3d3
57
+ NB-I-480-Sphere-c-160,train,80,480,Sphere,disk,c,160,,6312,502,238894952,9daff3fdc79d98bdc39730672468bd2d6b323e3ffd772eace6eaa6b96f4773f9
58
+ NB-I-480-Sphere-c-40,train,80,480,Sphere,disk,c,40,,6312,502,236930708,cc9203db27fa51f9384b0f40f25cc4a66bb057221187c586be08123e37ec9675
59
+ NB-I-480-Sphere-c-80,train,80,480,Sphere,disk,c,80,,6312,502,237975409,5d1c4aac11b73c0522744b0385d5ee6d591653c703b0ee1a2fd8a8133790153f
60
+ NB-I-640-Bullet-a-120,train,80,640,Bullet,rod,a,120,,8360,502,321239630,1f01ef02596fd340a7e25765515213c341d337a32fea03d76487c10d759d737a
61
+ NB-I-640-Bullet-a-160,train,80,640,Bullet,rod,a,160,,8360,502,317786506,6d22264a1144f68dccca9490a04c07534223e75e48bb37c9c11a3d51ef095e8d
62
+ NB-I-640-Bullet-a-40,train,80,640,Bullet,rod,a,40,,8360,502,315976413,5638ff8ba895fda841626760f9049e7f42bdc2069fcd60931d31f1a1588a10b3
63
+ NB-I-640-Bullet-a-80,train,80,640,Bullet,rod,a,80,,8360,502,316536879,300032ef099127c8a6cfe5e790fbbdb00ac52db2c0f6b2fa275b9fbd90fa6893
64
+ NB-I-640-Bullet-b-160,train,80,640,Bullet,rod,b,160,,8360,502,320324820,7b7adcfbc39333ffc63488c7e3b15d883c4da38387eed5de686443a7201e6aaa
65
+ NB-I-640-Bullet-b-40,train,80,640,Bullet,rod,b,40,,8360,502,318869857,940f517117222ada854cc84dd0830da1dd117aa008b9565c3bf11c09f56df72c
66
+ NB-I-640-Bullet-c-120,train,80,640,Bullet,rod,c,120,,8360,502,321824573,e37dbfbba4a7fcafee57af2641bdc0be79d770108c437234255545506a49fed2
67
+ NB-I-640-Bullet-c-160,train,80,640,Bullet,rod,c,160,,8360,502,319912073,f3b1d8a43a24815dda3b8bb531da758cad1e4588ac16c751be9835cf09303af9
68
+ NB-I-640-Bullet-c-40,train,80,640,Bullet,rod,c,40,,8360,502,317414789,ee6563b5e628a4bea8177417866f8bec98f3ca0af3ccaacdf2bed3023161f858
69
+ NB-I-640-Bullet-c-80,train,80,640,Bullet,rod,c,80,,8360,502,318524993,5caa83189025461452cbc50b97a5c41d322806b5c78b8a13a084b6e301bbd7a4
70
+ NB-I-640-Rectangular-a-160,train,80,640,Rectangular,plate,a,160,,8360,502,323250596,64a116d4848542e20dea0db98e5d45d99e7d69fd1d97f88073e427102469e88b
71
+ NB-I-640-Rectangular-a-40,train,80,640,Rectangular,plate,a,40,,8360,502,319950929,26c1e55389ff5f46e37de04d3f67dc01b8de883fe6c080d39e62bc7db99703e1
72
+ NB-I-640-Rectangular-a-80,train,80,640,Rectangular,plate,a,80,,8360,502,321615171,f2fa0d9a36a2b9b2f8e5484650ffd09ee1f7b8acfd509a26e22ed2e46df8d668
73
+ NB-I-640-Rectangular-b-120,train,80,640,Rectangular,plate,b,120,,8360,502,322907764,d5503557b059bfeae7eab3feb7f1d914a133ec069c78505b811183e8a58faba3
74
+ NB-I-640-Rectangular-b-160,train,80,640,Rectangular,plate,b,160,,8360,502,322758535,6f2281237ccdd6d6c2074d6102b7a76c5b916c770928d0fb8ed36a9010731919
75
+ NB-I-640-Rectangular-b-40,train,80,640,Rectangular,plate,b,40,,8360,502,321001573,4ca9c67f4f9c0c4f0d5fd7d44296dd7977529401947826b47513d64d73f2b1a2
76
+ NB-I-640-Rectangular-b-80,train,80,640,Rectangular,plate,b,80,,8360,502,323226150,c2fd577821357cc5553b41ebcded5161b4f3eb11491926ce7bbe11f2f3af463a
77
+ NB-I-640-Rectangular-c-120,train,80,640,Rectangular,plate,c,120,,8360,502,324449558,ba1a882068dc560ea763535a25fdbf0ce71de9074efaf7fff3920ac13f47b204
78
+ NB-I-640-Rectangular-c-160,train,80,640,Rectangular,plate,c,160,,8360,502,323304946,20e275ff11aee6bee572c2ffcee5f91edb78347b92007c54ff04bc920f2c0544
79
+ NB-I-640-Rectangular-c-40,train,80,640,Rectangular,plate,c,40,,8360,502,321063483,5252f1f43172a4159e0d02e355ce1ec1071ae120b5a1a73853e72fb3fad47ab1
80
+ NB-I-640-Sphere-a-120,train,80,640,Sphere,disk,a,120,,8360,502,318020663,d0a5210ceb4767cb41ffef736c82f0a20c380c4b1aa7eb7431b7f9094c7ed08a
81
+ NB-I-640-Sphere-a-160,train,80,640,Sphere,disk,a,160,,8360,502,319320476,217f6ad439585dec98f005beded2d2c26aeae7eeda61bba4b0c1052b57762916
82
+ NB-I-640-Sphere-a-40,train,80,640,Sphere,disk,a,40,,8360,502,319042866,050f217c8738141afed085b8b8474e9691757500329763087f591934d96c350c
83
+ NB-I-640-Sphere-b-120,train,80,640,Sphere,disk,b,120,,8360,502,320783017,caf91d9e00b66f4a97d6963ee46caa99b456e3f49852945fc4ce00ae909ad6b2
84
+ NB-I-640-Sphere-b-160,train,80,640,Sphere,disk,b,160,,8360,502,321206306,a3566889e6d4a9e26ddff510580c07982a12a972b22ed3c302799e131093313d
85
+ NB-I-640-Sphere-b-40,train,80,640,Sphere,disk,b,40,,8360,502,318894880,b8a82804b5e01a30931ea114e1fe17d35de64b49e1bd1ad0d656e307739df173
86
+ NB-I-640-Sphere-b-80,train,80,640,Sphere,disk,b,80,,8360,502,320244683,dba16f4f74413828e597c250d9a419747e9d6f8c3e044b6e20e1eb5a9e49e699
87
+ NB-I-640-Sphere-c-160,train,80,640,Sphere,disk,c,160,,8360,502,321164030,0ddbe0ce73ac14fb5c8842687270bd4325e3aeea92e279fe3ca2a6dd357ee1db
88
+ NB-I-640-Sphere-c-40,train,80,640,Sphere,disk,c,40,,8360,502,318810153,ae45ecebbd7d4c75317afcd3c7f56bd109ed0e30f1baa5c8da722401b0194658
89
+ NB-I-640-Sphere-c-80,train,80,640,Sphere,disk,c,80,,8360,502,319728577,74c6ad30e0812a456308ddc12f3258f671fbf29aac2845748d5647ca124c8abf
90
+ NB-I-320-Bullet-b-80,val,80,320,Bullet,rod,b,80,,4264,502,163154871,d6238186aa72d26f86e063be6c4f989b00d5645a2e8f1d1ecb4fdfab46f879b8
91
+ NB-I-320-Sphere-a-120,val,80,320,Sphere,disk,a,120,,4264,502,161461473,ec243a74dbe3416f63c7bd4fc25409281d739d36789131081c298e4f263fe5cb
92
+ NB-I-320-Sphere-a-80,val,80,320,Sphere,disk,a,80,,4264,502,161859634,5241311f83b47f5a78c612b6ef361a3cb288c8ad408a8e14b9a299957089fd5d
93
+ NB-I-480-Rectangular-b-80,val,80,480,Rectangular,plate,b,80,,6312,502,239926838,4d3aa543b3459cdc4d1300ab67659ac033ffa45c845f9d6502fac07380c91bd0
94
+ NB-I-480-Sphere-c-120,val,80,480,Sphere,disk,c,120,,6312,502,238324664,c85631a4308dcc9964d426b6031f046108a9cec4b585955265fd1ca387ce39ac
95
+ NB-I-640-Bullet-b-120,val,80,640,Bullet,rod,b,120,,8360,502,321772794,b953f01941491fb07b1fabda3585a31bf35dbb34dc83cf8eb057b6b2029b98ca
96
+ NB-I-640-Rectangular-a-120,val,80,640,Rectangular,plate,a,120,,8360,502,323208637,6208fd0a6b44732ac7ec3b505c2908b5cad2e6527c3fbb8dcc408a331d3f3962
97
+ NB-I-640-Rectangular-c-80,val,80,640,Rectangular,plate,c,80,,8360,502,322528743,239d4807bce57e3f6efdf64464b58081c3e0f7e20b02fc673cb3d5a76ad3393a
98
+ NB-I-320-Bullet-a-80,test_interp,80,320,Bullet,rod,a,80,,4264,502,161171921,e998a1f4f6a1be2525ae7f0595bdeb4ffe59422ddca418043da03358558faca2
99
+ NB-I-320-Bullet-b-120,test_interp,80,320,Bullet,rod,b,120,,4264,502,163465439,dd6e10ddc940c62e82272dfdee1894f1dfec80fe724058e1705fc4c4550201aa
100
+ NB-I-320-Rectangular-b-120,test_interp,80,320,Rectangular,plate,b,120,,4264,502,164687842,5ecb1766f631f5f7cd6c8527170de843fe3e26b86205c58901048366bbf84d6b
101
+ NB-I-320-Rectangular-c-120,test_interp,80,320,Rectangular,plate,c,120,,4264,502,164457587,1b9d373b82d4dab20b41d7ba35691bc2a51a9bf5c8f0b02975ddf95886ba6c73
102
+ NB-I-320-Sphere-b-120,test_interp,80,320,Sphere,disk,b,120,,4264,502,163535563,a54396c7fb6e4f5d01a65be46d05ad050fbfd098b5e1cb324c4e9d48bb9f9ce1
103
+ NB-I-320-Sphere-b-80,test_interp,80,320,Sphere,disk,b,80,,4264,502,163640279,15fa155290ed9cb703ce831fde6bcb439c9393253b17e3df03a788f51805d537
104
+ NB-I-480-Rectangular-b-120,test_interp,80,480,Rectangular,plate,b,120,,6312,502,241506697,ab65994b7a4b2c1cc718781019b9e11f2df33e3211d2e66c31bb84f830ca075a
105
+ NB-I-480-Sphere-a-120,test_interp,80,480,Sphere,disk,a,120,,6312,502,236720251,37a0477c77395240e21bfd52ac95e78f3ff4ffcdb4f410f2171ccc3801f86a8e
106
+ NB-I-480-Sphere-b-120,test_interp,80,480,Sphere,disk,b,120,,6312,502,239079244,d70e690c4d816e87ab3eb69b6509302ec40c1615e6359723a3b5c2a1a9b9b6cb
107
+ NB-I-640-Bullet-b-80,test_interp,80,640,Bullet,rod,b,80,,8360,502,320277460,75cecec772bfc7739e9f46342df5b853ae21618a6ad6449830b0bb0009180ddc
108
+ NB-I-640-Sphere-a-80,test_interp,80,640,Sphere,disk,a,80,,8360,502,317328129,4720d8027d23b31a570c90d054043c6269ed8d56bec48cf20222f9646bc57b51
109
+ NB-I-640-Sphere-c-120,test_interp,80,640,Sphere,disk,c,120,,8360,502,319999641,e816e9f7a6ebc2739a887da6444fe9e68aca559ac6c47b04c0406940574d22c7
110
+ S_100_800_V60_extrapolation,probe,100,800,Sphere,disk,,60,extrapolation,12966,502,469777436,669b3793baa3f86cf815a1cb4e393bf2bb83ae8206067d7aaad5ef3d20b412f9
111
+ S_80_400_V140_intrapolation,probe,80,400,Sphere,disk,,140,intrapolation,5288,502,202326919,c298c0c3043c3bf38853989d30ceb86e4b8fdb7e3d876cf2c617756d48eed4bb
decks/NB-I-320-Bullet-a-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-a-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-a-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-a-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-b-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-b-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-b-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-b-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-c-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-c-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-c-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Bullet-c-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-a-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-a-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-a-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-a-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-b-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-b-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-b-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-b-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-c-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-c-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-c-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Rectangular-c-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-a-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-a-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-a-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-a-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-b-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-b-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-b-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-b-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-c-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-c-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-c-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-320-Sphere-c-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-a-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-a-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-a-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-a-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-b-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-b-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-b-40.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-b-80.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-c-120.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-c-160.k ADDED
The diff for this file is too large to render. See raw diff
 
decks/NB-I-480-Bullet-c-40.k ADDED
The diff for this file is too large to render. See raw diff