The Data Processing Framework (DPF) is designed to provide numerical simulation users/engineers with a toolbox for accessing and transforming simulation data. DPF can access data from solver result files as well as several neutral formats (csv, hdf5, vtk, etc.). Various operators are available allowing the manipulation and the transformation of this data.
DPF is a workflow-based framework which allows simple and/or complex evaluations by chaining operators. The data in DPF is defined based on physics agnostic mathematical quantities described in a self-sufficient entity called field. This allows DPF to be a modular and easy to use tool with a large range of capabilities. It's a product designed to handle large amount of data.
The Python ansys.dpf.core
module provides a Python interface to
the powerful DPF framework enabling rapid post-processing of a variety
of Ansys file formats and physics solutions without ever leaving a
Python environment.
Visit the DPF-Core Documentation for a detailed description of the library, or see the Examples Gallery for more detailed examples.
Install this repository with:
pip install ansys-dpf-core
You can also clone and install this repository with:
git clone https://github.com/pyansys/pydpf-core
cd pydpf-core
pip install . --user
See the example scripts in the examples folder for some basic example. More will be added later.
Provided you have ANSYS 2021R1 or higher installed, a DPF server will start automatically once you start using DPF.
To open a result file and explore what's inside, do:
>>> from ansys.dpf import core as dpf
>>> from ansys.dpf.core import examples
>>> model = dpf.Model(examples.simple_bar)
>>> print(model)
DPF Model
------------------------------
Static analysis
Unit system: Metric (m, kg, N, s, V, A)
Physics Type: Mecanic
Available results:
- displacement: Nodal Displacement
- element_nodal_forces: ElementalNodal Element nodal Forces
- elemental_volume: Elemental Volume
- stiffness_matrix_energy: Elemental Energy-stiffness matrix
- artificial_hourglass_energy: Elemental Hourglass Energy
- thermal_dissipation_energy: Elemental thermal dissipation energy
- kinetic_energy: Elemental Kinetic Energy
- co_energy: Elemental co-energy
- incremental_energy: Elemental incremental energy
- structural_temperature: ElementalNodal Temperature
------------------------------
DPF Meshed Region:
3751 nodes
3000 elements
Unit: m
With solid (3D) elements
------------------------------
DPF Time/Freq Support:
Number of sets: 1
Cumulative Time (s) LoadStep Substep
1 1.000000 1 1
Read a result with:
>>> result = model.results.displacement.eval()
Then start connecting operators with:
>>> from ansys.dpf.core import operators as ops
>>> norm = ops.math.norm(model.results.displacement())
The ansys.dpf.core
automatically starts a local instance of the DPF service in the
background and connects to it. If you need to connect to an existing
remote or local DPF instance, use the connect_to_server
function:
>>> from ansys.dpf import core as dpf
>>> dpf.connect_to_server(ip='10.0.0.22', port=50054)
Once connected, this connection will remain for the duration of the module until you exit python or connect to a different server.