F13LD.sweep
Sweeps the parameters of a design recipe, homogenizes every sample, and returns a property-tagged batch. It is the forward search across a design space.
What it does.
F13LD.sweep takes one design recipe and varies its parameters across the ranges it detects from the recipe, computing the effective stiffness of each sample with FFT-CG homogenization. The result is a batch of designs, each tagged with its properties.
It is the forward half of design-space search — many designs in, ranked by property. The inverse half, a target in and designs out, is synth.
Where it fits.
Sweep reads a recipe emitted by any design tool and produces a property-tagged batch. Those batches enter the library through ingest — sweep does not write to the vault itself.
Worked example: a sweep.
Load a recipe
Bring in a design recipe (from a tool, a file, or a link). Sweep reads it and auto-detects which parameters can be swept and over what range — there is nothing to set up by hand.
Run the sweep
Each sample is homogenized with FFT-CG, batched across a worker pool. Beam recipes use a faster analytical stiffness.
Filter the batch
Results are tagged with their computed properties — stiffness, anisotropy, relative density. Filter to the region you want.
Store the result
Hand the batch to ingest for the vault, or export it.
The stiffness on each sample is a homogenization estimate, not a full solve. Treat it as a ranking signal and confirm the finalists in lab.
Output.
A property-tagged sweep batch — each entry is a recipe plus its computed properties. The batch is the unit ingest writes into the vault. What the stiffness numbers assume is covered in homogenization.