Pressure Coefficients

As stated before, the pressure data is relevant for structural engineers to understand the wind induced stress.

However, it comes in handy to adimensionalize the pressure data into coefficients. Therefore, the analysis is independent of the scale or unit system. It is a way to generalize the pressure data.

Available Coefficients

There are several different coefficients, that are commonly used in the wind industry, such as:

  • Pressure Coefficient: It is a fundamental adimensionalization of the pressure data. It is obtained by dividing a pressure difference by the dynamic pressure.

  • Shape Coefficient: It is equivalent to a resulting pressure coefficient over an area of interest. It is used to combine the pressure effects over the area, in a way to sum the exerted force in each triangle inside this area. Some peaks in each triangle may cancel each other in when calculating the shape coefficient.

  • Force Coefficient: It is a general adimensionalization of the resulting wind induced force over a body. It is calculated by summing the resulting force of each triangle, and dividing it by a representative area.

  • Momentum Coefficient: It is a general adimensionalization of the resulting wind induced momentum over a body. It is calculated by summing the resulting momentum of each triangle, and dividing it by a representative volume. The anchor point to define the momentum lever is an arbitrary point for the whole body.

Geometry Artifact

Each coefficient has its own artifacts dependencies. However, a common artifact shared between all of them, is a description of the structure geometry.

The pipeline accepts geometry in any of the following formats and dispatches by suffix via cfdmod.io.load_mesh():

  • .lnas – the AeroSim native LNAS file with authored surfaces.

  • .stl – a triangle mesh; treated as a single "all" surface.

  • .h5 – an XDMF+H5 with embedded /Triangles + /Geometry; single "all" surface.

  • .xdmf – redirects to its sibling .h5.

The mesh_path argument is optional on every run_* entry point. When omitted, the geometry is read from the source H5 itself (the body H5 for run_cp, the cp timeseries H5 for run_cf/run_cm/run_ce) – so a single-body pipeline does not need a separate mesh file.

When the same building is simulated at several wind directions, each solver run produces a body H5 in its own wind-aligned (“spun”) coordinate frame. Pass a single fixed-frame mesh as mesh_path to run_cp and the Cp output (and everything downstream from it: Cf, Cm, Ce) is written in that shared frame. Triangle ordering must match the body H5 – only vertex coordinates may differ. run_cf / run_cm / run_ce then inherit the reference frame through cp_h5 automatically; no need to repeat mesh_path= on each call.

Note

For LNAS-specific details, see the documentation inside the LNAS repository.

Filtering between coefficients

A coefficient timeseries (Cp, Cf, Cm, Ce) is a normal XDMF+H5 file that can be passed through a chain of signal-processing filters using cfdmod.pressure.filters.apply_filters(). Filters are an opt-in pipeline step; the output is another timeseries H5 with the same on-disk shape, ready to feed the next stage. The applied chain is recorded in /processing_metadata so the lineage stays self- describing.

from cfdmod import MovingAverageFilter, apply_filters

apply_filters(
    input_h5="output/cp.default.time_series.h5",
    output_h5="output/cp.default.smoothed.time_series.h5",
    filters=[MovingAverageFilter(window=3.0)],   # in input time units
    group="cp",
)

See Filters (smoothing as a pipeline step) for details and the rationale behind moving smoothing out of the statistics block.

v2 Quickstart

A worked end-to-end example lives at notebooks/process_container_pack.ipynb in the repo. It

  • loads geometry directly from the body H5 (no LNAS authoring),

  • auto-detects container partition via a triangle-centroid gap sweep,

  • runs Cp / Cf / Cm and writes everything flat under ./output/,

  • uses lever_strategy="region_bbox_corners_xy" to produce four overturning-moment scans per container.

Outputs land flat in output/ (no nested per-coefficient subfolders); per-coefficient stats are merged into a single stats.{h5,xdmf} and every output H5 carries the post-processing config under /processing_metadata/ for downstream debugging.