perifit 2.0.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- perifit-2.0.0/PKG-INFO +292 -0
- perifit-2.0.0/README.md +253 -0
- perifit-2.0.0/perifit/__init__.py +184 -0
- perifit-2.0.0/perifit/cli.py +289 -0
- perifit-2.0.0/perifit/config.py +87 -0
- perifit-2.0.0/perifit/core/__init__.py +90 -0
- perifit-2.0.0/perifit/core/local_system.py +218 -0
- perifit-2.0.0/perifit/core/local_system_2d.py +149 -0
- perifit-2.0.0/perifit/core/moments.py +84 -0
- perifit-2.0.0/perifit/core/moments_2d.py +77 -0
- perifit-2.0.0/perifit/core/targets.py +122 -0
- perifit-2.0.0/perifit/core/targets_2d.py +78 -0
- perifit-2.0.0/perifit/core/weights.py +157 -0
- perifit-2.0.0/perifit/core/weights_2d.py +120 -0
- perifit-2.0.0/perifit/data/__init__.py +0 -0
- perifit-2.0.0/perifit/data/test_mesh_cube_1000.csv +1003 -0
- perifit-2.0.0/perifit/io/__init__.py +0 -0
- perifit-2.0.0/perifit/io/writers.py +533 -0
- perifit-2.0.0/perifit/mesh/__init__.py +0 -0
- perifit-2.0.0/perifit/mesh/readers.py +399 -0
- perifit-2.0.0/perifit/quality/__init__.py +2 -0
- perifit-2.0.0/perifit/quality/cleaning.py +58 -0
- perifit-2.0.0/perifit/quality/diagnostics.py +67 -0
- perifit-2.0.0/perifit.egg-info/PKG-INFO +292 -0
- perifit-2.0.0/perifit.egg-info/SOURCES.txt +36 -0
- perifit-2.0.0/perifit.egg-info/dependency_links.txt +1 -0
- perifit-2.0.0/perifit.egg-info/entry_points.txt +2 -0
- perifit-2.0.0/perifit.egg-info/requires.txt +24 -0
- perifit-2.0.0/perifit.egg-info/top_level.txt +1 -0
- perifit-2.0.0/pyproject.toml +57 -0
- perifit-2.0.0/setup.cfg +4 -0
- perifit-2.0.0/tests/test_bbpd.py +205 -0
- perifit-2.0.0/tests/test_cli.py +47 -0
- perifit-2.0.0/tests/test_moments.py +60 -0
- perifit-2.0.0/tests/test_nosb.py +145 -0
- perifit-2.0.0/tests/test_quality.py +128 -0
- perifit-2.0.0/tests/test_weights.py +367 -0
- perifit-2.0.0/tests/test_weights_reference.py +578 -0
perifit-2.0.0/PKG-INFO
ADDED
|
@@ -0,0 +1,292 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: perifit
|
|
3
|
+
Version: 2.0.0
|
|
4
|
+
Summary: Surface correction weights for 3-D peridynamics (bond-based, state-based LPS, and correspondence NOSB-PD) with mesh quality diagnostics
|
|
5
|
+
License: MIT
|
|
6
|
+
Project-URL: Homepage, https://github.com/Hereon-InstituteMS/perifit
|
|
7
|
+
Project-URL: Repository, https://github.com/Hereon-InstituteMS/perifit
|
|
8
|
+
Project-URL: Issues, https://github.com/Hereon-InstituteMS/perifit/issues
|
|
9
|
+
Keywords: peridynamics,surface correction,LPS,NOSB-PD,correspondence,state-based peridynamics,bond-based peridynamics,mesh quality,meshfree,nonlocal mechanics
|
|
10
|
+
Classifier: Development Status :: 5 - Production/Stable
|
|
11
|
+
Classifier: Intended Audience :: Science/Research
|
|
12
|
+
Classifier: License :: OSI Approved :: MIT License
|
|
13
|
+
Classifier: Programming Language :: Python :: 3
|
|
14
|
+
Classifier: Programming Language :: Python :: 3.9
|
|
15
|
+
Classifier: Programming Language :: Python :: 3.10
|
|
16
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
17
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
18
|
+
Classifier: Topic :: Scientific/Engineering :: Physics
|
|
19
|
+
Requires-Python: >=3.9
|
|
20
|
+
Description-Content-Type: text/markdown
|
|
21
|
+
Requires-Dist: numpy>=1.24
|
|
22
|
+
Requires-Dist: scipy>=1.10
|
|
23
|
+
Provides-Extra: vtk
|
|
24
|
+
Requires-Dist: pyvista>=0.42; extra == "vtk"
|
|
25
|
+
Provides-Extra: exodus
|
|
26
|
+
Requires-Dist: netCDF4>=1.6; extra == "exodus"
|
|
27
|
+
Provides-Extra: hdf5
|
|
28
|
+
Requires-Dist: h5py>=3.8; extra == "hdf5"
|
|
29
|
+
Provides-Extra: gmsh
|
|
30
|
+
Requires-Dist: meshio>=5.3; extra == "gmsh"
|
|
31
|
+
Provides-Extra: all
|
|
32
|
+
Requires-Dist: pyvista>=0.42; extra == "all"
|
|
33
|
+
Requires-Dist: netCDF4>=1.6; extra == "all"
|
|
34
|
+
Requires-Dist: h5py>=3.8; extra == "all"
|
|
35
|
+
Requires-Dist: meshio>=5.3; extra == "all"
|
|
36
|
+
Provides-Extra: dev
|
|
37
|
+
Requires-Dist: pytest>=7; extra == "dev"
|
|
38
|
+
Requires-Dist: pytest-cov; extra == "dev"
|
|
39
|
+
|
|
40
|
+
# perifit
|
|
41
|
+
|
|
42
|
+
**Surface correction weights for 3-D peridynamics (state-based LPS, bond-based, and correspondence NOSB-PD) with mesh quality diagnostics**
|
|
43
|
+
|
|
44
|
+
[](https://github.com/Hereon-InstituteMS/perifit/actions)
|
|
45
|
+
[](LICENSE)
|
|
46
|
+
|
|
47
|
+
`perifit` computes optimised nodal influence weights that eliminate
|
|
48
|
+
the **surface/boundary effect** in three-dimensional peridynamics.
|
|
49
|
+
|
|
50
|
+
Supported models:
|
|
51
|
+
| Model | CLI flag | Operator rows | Notes |
|
|
52
|
+
|-------|----------|---------------|-------|
|
|
53
|
+
| **OSB-PD** (ordinary state-based, LPS) | `--model osb` | 109 | Poisson-unrestricted |
|
|
54
|
+
| **BB-PD** (bond-based) | `--model bb` | 54 | ν = 1/4 in 3D |
|
|
55
|
+
| **NOSB-PD** (non-ordinary, correspondence) | `--model nosb` | 36 | Material-independent |
|
|
56
|
+
|
|
57
|
+
Provide any 3-D mesh → get per-node weights in formats ready for
|
|
58
|
+
**Peridigm**, **PeriLab**, or generic CSV / DAT / VTK.
|
|
59
|
+
|
|
60
|
+
---
|
|
61
|
+
|
|
62
|
+
## Background
|
|
63
|
+
|
|
64
|
+
In peridynamics, nodes near the domain boundary have truncated horizon
|
|
65
|
+
neighbourhoods. This reduces their effective stiffness and causes
|
|
66
|
+
displacement errors that do **not** vanish under mesh refinement at fixed
|
|
67
|
+
horizon-to-spacing ratio *m = δ/Δx* — the so-called *surface effect*.
|
|
68
|
+
|
|
69
|
+
`perifit` implements an operator-matching framework for state-based (LPS), bond-based
|
|
70
|
+
(solid), and correspondence (NOSB-PD) models. For each node, a local overdetermined least-squares
|
|
71
|
+
problem is solved to find a scalar influence weight *wᵢ* such that the
|
|
72
|
+
truncated neighbourhood operators reproduce their full-sphere continuum
|
|
73
|
+
counterparts. The per-node weights are coupled through a global sparse
|
|
74
|
+
linear system solved by BiCGSTAB.
|
|
75
|
+
|
|
76
|
+
> **The weights are purely geometric.**
|
|
77
|
+
> Material constants (*G* for state-based, *c* for bond-based) appear as
|
|
78
|
+
> common prefactors in both the local ML matrix and the analytical targets
|
|
79
|
+
> and cancel exactly after per-row normalisation. For the correspondence
|
|
80
|
+
> model, the operator rows are material-independent by construction (no
|
|
81
|
+
> radial singularity, no constitutive constants). Consequently, *E*, *ν*,
|
|
82
|
+
> and any other material constant are **not required** — only the mesh
|
|
83
|
+
> geometry and the horizon matter.
|
|
84
|
+
|
|
85
|
+
---
|
|
86
|
+
|
|
87
|
+
## Results
|
|
88
|
+
|
|
89
|
+
### Convergence and preprocessing time (state-based LPS)
|
|
90
|
+
|
|
91
|
+

|
|
92
|
+
|
|
93
|
+
**(a)** Displacement error convergence for the **state-based LPS** model on a unit bar under uniaxial tension (*m* = 3).
|
|
94
|
+
The corrected weights (solid lines) recover first-order convergence across all
|
|
95
|
+
Poisson ratios, while standard LPS (dashed) stagnates due to the surface effect.
|
|
96
|
+
**(b)** Weight computation time scales near-linearly with the number of nodes.
|
|
97
|
+
|
|
98
|
+
### Mesh quality diagnostics (state-based LPS)
|
|
99
|
+
|
|
100
|
+

|
|
101
|
+
|
|
102
|
+
**(a)** Rocker-arm geometry (1,699 nodes) colored by **LPS** weight values.
|
|
103
|
+
**(b)** Outlier nodes flagged by the built-in diagnostics (2*σ* threshold).
|
|
104
|
+
**(c)** Cleaned mesh after removing the 69 flagged nodes — weight distribution
|
|
105
|
+
is now uniform and well-conditioned.
|
|
106
|
+
|
|
107
|
+
---
|
|
108
|
+
|
|
109
|
+
## Installation
|
|
110
|
+
|
|
111
|
+
```bash
|
|
112
|
+
pip install perifit
|
|
113
|
+
```
|
|
114
|
+
|
|
115
|
+
With optional dependencies for specific mesh formats:
|
|
116
|
+
|
|
117
|
+
```bash
|
|
118
|
+
pip install "perifit[hdf5]" # PeriLab HDF5 output
|
|
119
|
+
pip install "perifit[exodus]" # Exodus mesh input
|
|
120
|
+
pip install "perifit[gmsh]" # Gmsh / Abaqus / VTU input
|
|
121
|
+
pip install "perifit[all]" # everything
|
|
122
|
+
```
|
|
123
|
+
|
|
124
|
+
---
|
|
125
|
+
|
|
126
|
+
## Quick start
|
|
127
|
+
|
|
128
|
+
### Built-in demo (no mesh needed)
|
|
129
|
+
|
|
130
|
+
```bash
|
|
131
|
+
perifit --demo
|
|
132
|
+
```
|
|
133
|
+
|
|
134
|
+
Runs on the bundled 1000-node structured cube mesh and writes
|
|
135
|
+
`csv`, `dat`, and `vtk` output to `./perifit_output/`.
|
|
136
|
+
|
|
137
|
+
### Python API
|
|
138
|
+
|
|
139
|
+
```python
|
|
140
|
+
import numpy as np
|
|
141
|
+
from perifit import compute_weights, write_weights
|
|
142
|
+
|
|
143
|
+
# Build your mesh (or load it — see below)
|
|
144
|
+
dx = 0.05
|
|
145
|
+
xs = np.arange(dx / 2, 1.0, dx)
|
|
146
|
+
gx, gy, gz = np.meshgrid(xs, xs, xs, indexing='ij')
|
|
147
|
+
coords = np.column_stack([gx.ravel(), gy.ravel(), gz.ravel()])
|
|
148
|
+
|
|
149
|
+
# OSB-PD / LPS (default model is "bb"):
|
|
150
|
+
volumes = np.full(len(coords), dx**3)
|
|
151
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='osb')
|
|
152
|
+
|
|
153
|
+
# BB-PD:
|
|
154
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='bb')
|
|
155
|
+
|
|
156
|
+
# NOSB-PD (material-independent):
|
|
157
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='nosb')
|
|
158
|
+
|
|
159
|
+
# Export to any format
|
|
160
|
+
write_weights(
|
|
161
|
+
coords, volumes, weights,
|
|
162
|
+
outdir='./output',
|
|
163
|
+
formats=['csv', 'vtk', 'peridigm', 'perilab'],
|
|
164
|
+
)
|
|
165
|
+
```
|
|
166
|
+
|
|
167
|
+
### Load from a mesh file
|
|
168
|
+
|
|
169
|
+
```python
|
|
170
|
+
from perifit import load_mesh, compute_weights, write_weights
|
|
171
|
+
|
|
172
|
+
# Supported: .csv .dat .txt .npy .npz .vtk .vtu .exo .g .msh .inp
|
|
173
|
+
coords, volumes = load_mesh('my_mesh.exo')
|
|
174
|
+
|
|
175
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='osb')
|
|
176
|
+
|
|
177
|
+
write_weights(coords, volumes, weights,
|
|
178
|
+
outdir='./weights',
|
|
179
|
+
formats=['peridigm', 'perilab', 'csv'])
|
|
180
|
+
```
|
|
181
|
+
|
|
182
|
+
### Mesh quality diagnostics
|
|
183
|
+
|
|
184
|
+
```python
|
|
185
|
+
from perifit import compute_weights
|
|
186
|
+
from perifit.quality import diagnose_weights, clean_mesh
|
|
187
|
+
|
|
188
|
+
weights = compute_weights(coords, volumes, horizon=delta)
|
|
189
|
+
|
|
190
|
+
# Diagnose: flag negative weights and statistical outliers
|
|
191
|
+
diag = diagnose_weights(weights, threshold_sigma=2.0)
|
|
192
|
+
print(diag.summary)
|
|
193
|
+
|
|
194
|
+
# Clean: remove flagged nodes
|
|
195
|
+
result = clean_mesh(coords, volumes, weights, threshold_sigma=2.0)
|
|
196
|
+
print(f"Removed {result.original_n - result.cleaned_n} nodes")
|
|
197
|
+
```
|
|
198
|
+
|
|
199
|
+
### Command-line interface
|
|
200
|
+
|
|
201
|
+
```bash
|
|
202
|
+
# OSB-PD / LPS
|
|
203
|
+
perifit --mesh nodes.csv --model osb
|
|
204
|
+
|
|
205
|
+
# BB-PD (default)
|
|
206
|
+
perifit --mesh nodes.csv --model bb
|
|
207
|
+
|
|
208
|
+
# NOSB-PD (material-independent weights)
|
|
209
|
+
perifit --mesh nodes.csv --model nosb
|
|
210
|
+
|
|
211
|
+
# Specify horizon-to-spacing ratio (default is 3)
|
|
212
|
+
perifit --mesh nodes.csv --m-ratio 4 --output csv,vtk
|
|
213
|
+
|
|
214
|
+
# Provide the horizon explicitly instead
|
|
215
|
+
perifit --mesh solid.exo --horizon 0.005 --output peridigm
|
|
216
|
+
|
|
217
|
+
# Diagnose mesh quality after computing weights
|
|
218
|
+
perifit --mesh cube.csv --diagnose
|
|
219
|
+
|
|
220
|
+
# Clean mesh (remove negative/outlier weight nodes)
|
|
221
|
+
perifit --mesh cube.csv --clean --threshold-sigma 2.0
|
|
222
|
+
|
|
223
|
+
# Show all options
|
|
224
|
+
perifit --help
|
|
225
|
+
```
|
|
226
|
+
|
|
227
|
+
---
|
|
228
|
+
|
|
229
|
+
## Supported input formats
|
|
230
|
+
|
|
231
|
+
| Extension | Format | Notes |
|
|
232
|
+
|-----------|--------|-------|
|
|
233
|
+
| `.csv`, `.txt`, `.dat` | Plain text | Columns: x, y, z[, volume] |
|
|
234
|
+
| `.npy` | NumPy array | Shape (N, 3) or (N, 4) |
|
|
235
|
+
| `.npz` | NumPy archive | Keys: `coords`[, `volumes`] |
|
|
236
|
+
| `.vtk` | Legacy VTK ASCII | Unstructured or structured |
|
|
237
|
+
| `.vtu` | VTK XML | Requires `pyvista` or `meshio` |
|
|
238
|
+
| `.exo`, `.g` | Exodus II | Requires `netCDF4` or `meshio` |
|
|
239
|
+
| `.msh` | Gmsh v2/v4 | Requires `meshio` |
|
|
240
|
+
| `.inp` | Abaqus | Requires `meshio` |
|
|
241
|
+
|
|
242
|
+
## Supported output formats
|
|
243
|
+
|
|
244
|
+
| Format | Description |
|
|
245
|
+
|--------|-------------|
|
|
246
|
+
| `csv` | `node_id, x, y, z, volume, weight` |
|
|
247
|
+
| `dat` | Space-delimited text |
|
|
248
|
+
| `vtk` | Legacy ASCII VTK for ParaView / VisIt |
|
|
249
|
+
| `peridigm` | Text mesh + weight CSV + YAML input snippet |
|
|
250
|
+
| `perilab` | HDF5 mesh with `Nodal_Surface_Correction` field + YAML snippet |
|
|
251
|
+
|
|
252
|
+
---
|
|
253
|
+
|
|
254
|
+
## Applying the weights in your solver
|
|
255
|
+
|
|
256
|
+
The bond-averaged influence weight is
|
|
257
|
+
|
|
258
|
+
```
|
|
259
|
+
w̄ᵢⱼ = (wᵢ + wⱼ) / 2
|
|
260
|
+
```
|
|
261
|
+
|
|
262
|
+
Replace every occurrence of the unit influence function `ω(ξ) = 1`
|
|
263
|
+
in your PD discretisation with `w̄ᵢⱼ`. The Peridigm and PeriLab
|
|
264
|
+
output files include a YAML template showing exactly where to hook in
|
|
265
|
+
the field.
|
|
266
|
+
|
|
267
|
+
---
|
|
268
|
+
|
|
269
|
+
## Auto-inference of dx, volumes, and horizon
|
|
270
|
+
|
|
271
|
+
The nodal spacing `dx` is always estimated automatically as the
|
|
272
|
+
**median nearest-neighbour distance** (KD-tree, O(N log N)).
|
|
273
|
+
This is used to:
|
|
274
|
+
|
|
275
|
+
1. Set nodal volumes to `dx³` when not provided.
|
|
276
|
+
2. Compute the horizon as `delta = m_ratio × dx` when `--horizon` is not given.
|
|
277
|
+
3. Check consistency when **both** `--horizon` and `--m-ratio` are supplied.
|
|
278
|
+
|
|
279
|
+
### Input modes
|
|
280
|
+
|
|
281
|
+
| `--horizon` | `--m-ratio` | Behaviour |
|
|
282
|
+
|-------------|-------------|-----------|
|
|
283
|
+
| not given | not given | horizon = 3 × dx (default m=3) |
|
|
284
|
+
| not given | given | horizon = m_ratio × dx |
|
|
285
|
+
| given | not given | horizon used directly |
|
|
286
|
+
| given | given | horizon used; **warning** if inconsistent with m_ratio × dx |
|
|
287
|
+
|
|
288
|
+
---
|
|
289
|
+
|
|
290
|
+
## License
|
|
291
|
+
|
|
292
|
+
MIT — see [LICENSE](LICENSE).
|
perifit-2.0.0/README.md
ADDED
|
@@ -0,0 +1,253 @@
|
|
|
1
|
+
# perifit
|
|
2
|
+
|
|
3
|
+
**Surface correction weights for 3-D peridynamics (state-based LPS, bond-based, and correspondence NOSB-PD) with mesh quality diagnostics**
|
|
4
|
+
|
|
5
|
+
[](https://github.com/Hereon-InstituteMS/perifit/actions)
|
|
6
|
+
[](LICENSE)
|
|
7
|
+
|
|
8
|
+
`perifit` computes optimised nodal influence weights that eliminate
|
|
9
|
+
the **surface/boundary effect** in three-dimensional peridynamics.
|
|
10
|
+
|
|
11
|
+
Supported models:
|
|
12
|
+
| Model | CLI flag | Operator rows | Notes |
|
|
13
|
+
|-------|----------|---------------|-------|
|
|
14
|
+
| **OSB-PD** (ordinary state-based, LPS) | `--model osb` | 109 | Poisson-unrestricted |
|
|
15
|
+
| **BB-PD** (bond-based) | `--model bb` | 54 | ν = 1/4 in 3D |
|
|
16
|
+
| **NOSB-PD** (non-ordinary, correspondence) | `--model nosb` | 36 | Material-independent |
|
|
17
|
+
|
|
18
|
+
Provide any 3-D mesh → get per-node weights in formats ready for
|
|
19
|
+
**Peridigm**, **PeriLab**, or generic CSV / DAT / VTK.
|
|
20
|
+
|
|
21
|
+
---
|
|
22
|
+
|
|
23
|
+
## Background
|
|
24
|
+
|
|
25
|
+
In peridynamics, nodes near the domain boundary have truncated horizon
|
|
26
|
+
neighbourhoods. This reduces their effective stiffness and causes
|
|
27
|
+
displacement errors that do **not** vanish under mesh refinement at fixed
|
|
28
|
+
horizon-to-spacing ratio *m = δ/Δx* — the so-called *surface effect*.
|
|
29
|
+
|
|
30
|
+
`perifit` implements an operator-matching framework for state-based (LPS), bond-based
|
|
31
|
+
(solid), and correspondence (NOSB-PD) models. For each node, a local overdetermined least-squares
|
|
32
|
+
problem is solved to find a scalar influence weight *wᵢ* such that the
|
|
33
|
+
truncated neighbourhood operators reproduce their full-sphere continuum
|
|
34
|
+
counterparts. The per-node weights are coupled through a global sparse
|
|
35
|
+
linear system solved by BiCGSTAB.
|
|
36
|
+
|
|
37
|
+
> **The weights are purely geometric.**
|
|
38
|
+
> Material constants (*G* for state-based, *c* for bond-based) appear as
|
|
39
|
+
> common prefactors in both the local ML matrix and the analytical targets
|
|
40
|
+
> and cancel exactly after per-row normalisation. For the correspondence
|
|
41
|
+
> model, the operator rows are material-independent by construction (no
|
|
42
|
+
> radial singularity, no constitutive constants). Consequently, *E*, *ν*,
|
|
43
|
+
> and any other material constant are **not required** — only the mesh
|
|
44
|
+
> geometry and the horizon matter.
|
|
45
|
+
|
|
46
|
+
---
|
|
47
|
+
|
|
48
|
+
## Results
|
|
49
|
+
|
|
50
|
+
### Convergence and preprocessing time (state-based LPS)
|
|
51
|
+
|
|
52
|
+

|
|
53
|
+
|
|
54
|
+
**(a)** Displacement error convergence for the **state-based LPS** model on a unit bar under uniaxial tension (*m* = 3).
|
|
55
|
+
The corrected weights (solid lines) recover first-order convergence across all
|
|
56
|
+
Poisson ratios, while standard LPS (dashed) stagnates due to the surface effect.
|
|
57
|
+
**(b)** Weight computation time scales near-linearly with the number of nodes.
|
|
58
|
+
|
|
59
|
+
### Mesh quality diagnostics (state-based LPS)
|
|
60
|
+
|
|
61
|
+

|
|
62
|
+
|
|
63
|
+
**(a)** Rocker-arm geometry (1,699 nodes) colored by **LPS** weight values.
|
|
64
|
+
**(b)** Outlier nodes flagged by the built-in diagnostics (2*σ* threshold).
|
|
65
|
+
**(c)** Cleaned mesh after removing the 69 flagged nodes — weight distribution
|
|
66
|
+
is now uniform and well-conditioned.
|
|
67
|
+
|
|
68
|
+
---
|
|
69
|
+
|
|
70
|
+
## Installation
|
|
71
|
+
|
|
72
|
+
```bash
|
|
73
|
+
pip install perifit
|
|
74
|
+
```
|
|
75
|
+
|
|
76
|
+
With optional dependencies for specific mesh formats:
|
|
77
|
+
|
|
78
|
+
```bash
|
|
79
|
+
pip install "perifit[hdf5]" # PeriLab HDF5 output
|
|
80
|
+
pip install "perifit[exodus]" # Exodus mesh input
|
|
81
|
+
pip install "perifit[gmsh]" # Gmsh / Abaqus / VTU input
|
|
82
|
+
pip install "perifit[all]" # everything
|
|
83
|
+
```
|
|
84
|
+
|
|
85
|
+
---
|
|
86
|
+
|
|
87
|
+
## Quick start
|
|
88
|
+
|
|
89
|
+
### Built-in demo (no mesh needed)
|
|
90
|
+
|
|
91
|
+
```bash
|
|
92
|
+
perifit --demo
|
|
93
|
+
```
|
|
94
|
+
|
|
95
|
+
Runs on the bundled 1000-node structured cube mesh and writes
|
|
96
|
+
`csv`, `dat`, and `vtk` output to `./perifit_output/`.
|
|
97
|
+
|
|
98
|
+
### Python API
|
|
99
|
+
|
|
100
|
+
```python
|
|
101
|
+
import numpy as np
|
|
102
|
+
from perifit import compute_weights, write_weights
|
|
103
|
+
|
|
104
|
+
# Build your mesh (or load it — see below)
|
|
105
|
+
dx = 0.05
|
|
106
|
+
xs = np.arange(dx / 2, 1.0, dx)
|
|
107
|
+
gx, gy, gz = np.meshgrid(xs, xs, xs, indexing='ij')
|
|
108
|
+
coords = np.column_stack([gx.ravel(), gy.ravel(), gz.ravel()])
|
|
109
|
+
|
|
110
|
+
# OSB-PD / LPS (default model is "bb"):
|
|
111
|
+
volumes = np.full(len(coords), dx**3)
|
|
112
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='osb')
|
|
113
|
+
|
|
114
|
+
# BB-PD:
|
|
115
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='bb')
|
|
116
|
+
|
|
117
|
+
# NOSB-PD (material-independent):
|
|
118
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='nosb')
|
|
119
|
+
|
|
120
|
+
# Export to any format
|
|
121
|
+
write_weights(
|
|
122
|
+
coords, volumes, weights,
|
|
123
|
+
outdir='./output',
|
|
124
|
+
formats=['csv', 'vtk', 'peridigm', 'perilab'],
|
|
125
|
+
)
|
|
126
|
+
```
|
|
127
|
+
|
|
128
|
+
### Load from a mesh file
|
|
129
|
+
|
|
130
|
+
```python
|
|
131
|
+
from perifit import load_mesh, compute_weights, write_weights
|
|
132
|
+
|
|
133
|
+
# Supported: .csv .dat .txt .npy .npz .vtk .vtu .exo .g .msh .inp
|
|
134
|
+
coords, volumes = load_mesh('my_mesh.exo')
|
|
135
|
+
|
|
136
|
+
weights = compute_weights(coords, volumes, horizon=3 * dx, model='osb')
|
|
137
|
+
|
|
138
|
+
write_weights(coords, volumes, weights,
|
|
139
|
+
outdir='./weights',
|
|
140
|
+
formats=['peridigm', 'perilab', 'csv'])
|
|
141
|
+
```
|
|
142
|
+
|
|
143
|
+
### Mesh quality diagnostics
|
|
144
|
+
|
|
145
|
+
```python
|
|
146
|
+
from perifit import compute_weights
|
|
147
|
+
from perifit.quality import diagnose_weights, clean_mesh
|
|
148
|
+
|
|
149
|
+
weights = compute_weights(coords, volumes, horizon=delta)
|
|
150
|
+
|
|
151
|
+
# Diagnose: flag negative weights and statistical outliers
|
|
152
|
+
diag = diagnose_weights(weights, threshold_sigma=2.0)
|
|
153
|
+
print(diag.summary)
|
|
154
|
+
|
|
155
|
+
# Clean: remove flagged nodes
|
|
156
|
+
result = clean_mesh(coords, volumes, weights, threshold_sigma=2.0)
|
|
157
|
+
print(f"Removed {result.original_n - result.cleaned_n} nodes")
|
|
158
|
+
```
|
|
159
|
+
|
|
160
|
+
### Command-line interface
|
|
161
|
+
|
|
162
|
+
```bash
|
|
163
|
+
# OSB-PD / LPS
|
|
164
|
+
perifit --mesh nodes.csv --model osb
|
|
165
|
+
|
|
166
|
+
# BB-PD (default)
|
|
167
|
+
perifit --mesh nodes.csv --model bb
|
|
168
|
+
|
|
169
|
+
# NOSB-PD (material-independent weights)
|
|
170
|
+
perifit --mesh nodes.csv --model nosb
|
|
171
|
+
|
|
172
|
+
# Specify horizon-to-spacing ratio (default is 3)
|
|
173
|
+
perifit --mesh nodes.csv --m-ratio 4 --output csv,vtk
|
|
174
|
+
|
|
175
|
+
# Provide the horizon explicitly instead
|
|
176
|
+
perifit --mesh solid.exo --horizon 0.005 --output peridigm
|
|
177
|
+
|
|
178
|
+
# Diagnose mesh quality after computing weights
|
|
179
|
+
perifit --mesh cube.csv --diagnose
|
|
180
|
+
|
|
181
|
+
# Clean mesh (remove negative/outlier weight nodes)
|
|
182
|
+
perifit --mesh cube.csv --clean --threshold-sigma 2.0
|
|
183
|
+
|
|
184
|
+
# Show all options
|
|
185
|
+
perifit --help
|
|
186
|
+
```
|
|
187
|
+
|
|
188
|
+
---
|
|
189
|
+
|
|
190
|
+
## Supported input formats
|
|
191
|
+
|
|
192
|
+
| Extension | Format | Notes |
|
|
193
|
+
|-----------|--------|-------|
|
|
194
|
+
| `.csv`, `.txt`, `.dat` | Plain text | Columns: x, y, z[, volume] |
|
|
195
|
+
| `.npy` | NumPy array | Shape (N, 3) or (N, 4) |
|
|
196
|
+
| `.npz` | NumPy archive | Keys: `coords`[, `volumes`] |
|
|
197
|
+
| `.vtk` | Legacy VTK ASCII | Unstructured or structured |
|
|
198
|
+
| `.vtu` | VTK XML | Requires `pyvista` or `meshio` |
|
|
199
|
+
| `.exo`, `.g` | Exodus II | Requires `netCDF4` or `meshio` |
|
|
200
|
+
| `.msh` | Gmsh v2/v4 | Requires `meshio` |
|
|
201
|
+
| `.inp` | Abaqus | Requires `meshio` |
|
|
202
|
+
|
|
203
|
+
## Supported output formats
|
|
204
|
+
|
|
205
|
+
| Format | Description |
|
|
206
|
+
|--------|-------------|
|
|
207
|
+
| `csv` | `node_id, x, y, z, volume, weight` |
|
|
208
|
+
| `dat` | Space-delimited text |
|
|
209
|
+
| `vtk` | Legacy ASCII VTK for ParaView / VisIt |
|
|
210
|
+
| `peridigm` | Text mesh + weight CSV + YAML input snippet |
|
|
211
|
+
| `perilab` | HDF5 mesh with `Nodal_Surface_Correction` field + YAML snippet |
|
|
212
|
+
|
|
213
|
+
---
|
|
214
|
+
|
|
215
|
+
## Applying the weights in your solver
|
|
216
|
+
|
|
217
|
+
The bond-averaged influence weight is
|
|
218
|
+
|
|
219
|
+
```
|
|
220
|
+
w̄ᵢⱼ = (wᵢ + wⱼ) / 2
|
|
221
|
+
```
|
|
222
|
+
|
|
223
|
+
Replace every occurrence of the unit influence function `ω(ξ) = 1`
|
|
224
|
+
in your PD discretisation with `w̄ᵢⱼ`. The Peridigm and PeriLab
|
|
225
|
+
output files include a YAML template showing exactly where to hook in
|
|
226
|
+
the field.
|
|
227
|
+
|
|
228
|
+
---
|
|
229
|
+
|
|
230
|
+
## Auto-inference of dx, volumes, and horizon
|
|
231
|
+
|
|
232
|
+
The nodal spacing `dx` is always estimated automatically as the
|
|
233
|
+
**median nearest-neighbour distance** (KD-tree, O(N log N)).
|
|
234
|
+
This is used to:
|
|
235
|
+
|
|
236
|
+
1. Set nodal volumes to `dx³` when not provided.
|
|
237
|
+
2. Compute the horizon as `delta = m_ratio × dx` when `--horizon` is not given.
|
|
238
|
+
3. Check consistency when **both** `--horizon` and `--m-ratio` are supplied.
|
|
239
|
+
|
|
240
|
+
### Input modes
|
|
241
|
+
|
|
242
|
+
| `--horizon` | `--m-ratio` | Behaviour |
|
|
243
|
+
|-------------|-------------|-----------|
|
|
244
|
+
| not given | not given | horizon = 3 × dx (default m=3) |
|
|
245
|
+
| not given | given | horizon = m_ratio × dx |
|
|
246
|
+
| given | not given | horizon used directly |
|
|
247
|
+
| given | given | horizon used; **warning** if inconsistent with m_ratio × dx |
|
|
248
|
+
|
|
249
|
+
---
|
|
250
|
+
|
|
251
|
+
## License
|
|
252
|
+
|
|
253
|
+
MIT — see [LICENSE](LICENSE).
|