4.2.0 · Python 3.9+ · MIT

FiberNet

Fiber-network generation, analysis, and mechanical modeling

Connect network generation, structural analysis, and mechanical modeling in Python. Use FiberScope to inspect, configure, and export networks on the desktop.

中文版本

Python library

FiberNet represents structures through nodes and edges, connecting generation, analysis, simulation, and export in reusable Python workflows.

Generate 2D and 3D structures and manage nodes, edges, and fiber radii.
Analyze planar snapshots, pore distributions, and tensile recruitment on supplied trajectories.
Use beam-frame FEM and additional computational backends installed as needed.
Save native graphs and results for analysis with the provided examples.

FiberScope desktop interface

FiberScope 3.1.0 supports geometry configuration, structural inspection, trajectory playback, and export. The Windows download includes its runtime.

Install 4.2.0

The core installation includes NumPy, SciPy, NetworkX, and Matplotlib. Add machine-learning, 3D visualization, or geometry-export dependencies for the workflow you choose.

python -m pip install "fibernet==4.2.0"

Optional dependencies

python -m pip install "fibernet[ml]==4.2.0"  # Machine learning
python -m pip install "fibernet[viz]==4.2.0"  # 3D display
python -m pip install "fibernet[manufacturing]==4.2.0"  # Geometry and solid export

ml supplies machine-learning tools; viz supplies PyVista for 3D display; manufacturing supplies solid-geometry and export tools. The Taichi backend has a separate accel extra.

Generate and inspect a structure

import fibernet as fn

# Coordinates and radius: meters
graph = fn.pattern_2d(
    unit="honeycomb", box=(0.01, 0.01), grid=(2, 2),
    radius=0.00005, seed=23,
)
fn.show(graph)

The example uses SI units: coordinates and radii are in meters, and the elastic modulus is in Pa.

Run a small-deformation model and save

Use graph from the preceding example for a linear beam-frame FEM solve, then save the structure and results.

# Elastic modulus: Pa; strain is the target stretch ratio
result = fn.simulate(
    graph, backend="fem", mode="stretch", strain=1.01,
    nonlinear=False, E=1e9, nu=0.3,
)
print(result.max_displacement)

graph.save_json("network.json")
result.save("fem_result.json")

In the 4.2.0 interface, strain denotes the target stretch ratio: 1.01 means 1% extension. nonlinear=False selects a linear solve, and max_displacement is in meters.

Continue the workflow

Structural analysis and learning

SnapshotFeatureExtractor analyzes planar snapshots; analyze_tensile_recruitment processes supplied edge-strain trajectories. Physical-label generation, regression, and target-curve search each have dedicated examples.

Snapshot example Physical-learning example Target-curve search

Geometric routes and export

The manufacturing geometry interfaces construct continuous routes on planar or curved surfaces and support custom cells. Solid and route exports use the corresponding optional dependencies, with coordinate units and route order managed alongside the inputs.

Route-generation example Solid-export example

Structure and computation examples

2D structure gallery

Twelve 2D structures from the v3 gallery. Source figure

Mechanical-model figures

Eight saved honeycomb frames, colored by edge stretch ratio. Source figure

Beam-frame FEM example, colored by sigma_total: absolute axial stress plus bending stress. Source figure

Voronoi-network example at a target stretch ratio of 1.5. Source figure

Learning and search figures

Classification example: confusion matrix, ROC, feature importance, and learning curve. Source figure

CEM search example: per-iteration objective values and the best value found so far. Source figure

Documentation and examples

PyPI 4.2.0 4.2.0 release notes Quick start Beam-FEM model and units Examples Complete notebook tutorial