CEMFORGE Datapacks

A CEMFORGE Datapack is a synthetic data source for researchers and data teams: on-demand access to the CEMFORGE prediction engine. You send a target strength and the materials you can actually source, and you get back a complete single-point printable-concrete formulation with predicted properties and a confidence score. The data comes in the open Open3DCP format, so it loads straight into your tools. There is no location or logistics layer here, which is what LOGiMIX Reports add; a Datapack is the raw prediction. This page shows exactly what a record contains, the range we support today, and five complete example formulations you can download for free.

What you get per record

Each formulation is returned as a flat Open3DCP row: every material and property is a named column, materials are given as mass percent of the total wet mix, units are SI, and unused fields are left empty. Because it is a flat table, it loads directly into pandas, R, Excel, or any database with no parsing. Columns are either canonical Open3DCP fields or clearly namespaced cemforge_ extensions.

Column(s) Meaning
mix_id, name, open3dcp_schema_version, is_3d_printed Identity, context, and schema version
cement_type_1l, slag, fly_ash, silica_fume, concrete_sand, fine_sand, water, superplasticizer, hpmc Materials, as mass percent of total wet mix (columns sum to roughly 100)
w_b_ratio Water-to-binder ratio
compressive_strength_mpa, flexural_strength_mpa, spread_mm, yield_stress_pa Predicted properties (spread is slump-flow spread). Each is a single-point estimate, not a range.
is_synthetic, measurement_confidence Flags marking the value as model-generated and estimated, not lab-measured
cemforge_target_strength_mpa, cemforge_confidence, cemforge_confidence_band The target you requested and the per-result confidence (0 to 1, plus a High / Moderate / Exploratory band)
cemforge_model_source, cemforge_disclaimer Model source and a per-row disclaimer

A JSON example is included that shows one formulation in the exact shape the API returns, including the canonical Open3DCP envelope plus top-level compatibility fields.

Supported range and how to read a prediction

  • The supported range today is the band where we have enough validated data to predict with reasonable confidence: roughly 25 to 75 MPa, 3D-printable concrete. We deliberately do not sell predictions outside this band; it will widen as we add validated data.
  • Every value is a predicted (synthetic) result from a statistical model. It is not a laboratory measurement and not a guarantee. Each row is flagged as synthetic and carries its own confidence score and a disclaimer.
  • Always verify by physical testing (trial batching and the appropriate test methods) before any structural or production use. Use is governed by the CEMFORGE EULA, and the engineer of record owns code applicability.

On the roadmap: proprietary datasets whose results are validated by third-party laboratories, a measured complement to today’s purely synthetic predictions. If lab-validated data is what you need, tell us your target band and we will let you know as it becomes available.

Five free example formulations

These five predictions span the supported strength range. They are free to download, with no account or API key, so you can see the exact shape of the data before you request a larger pack. All five are High-confidence examples. Materials are shown as mass percent of total wet mix.

mix_id name Type 1L slag fly ash silica fume sand water w/b compressive (MPa) flexural (MPa) spread (mm) confidence
CF-SAMPLE-001 Eco 3DCP 30 MPa 18 13 8 47.1 13 0.33 30 4.4 210 0.85 High
CF-SAMPLE-002 Standard 3DCP 40 MPa 25 10 2 48.2 14 0.38 40 5.2 205 0.86 High
CF-SAMPLE-003 Standard 3DCP 50 MPa 29 7 3 46.6 13.5 0.35 50 6.0 195 0.85 High
CF-SAMPLE-004 High-Strength 3DCP 60 MPa 32 4 6 44.4 12.5 0.30 60 6.9 185 0.83 High
CF-SAMPLE-005 High-Strength 3DCP 70 MPa 34 3 8 42.0 11.8 0.26 70 7.6 178 0.81 High

Every value above is a predicted (synthetic) result, not a measurement or guarantee. Verify by physical testing before real-world use. The sand column combines concrete sand and fine sand for display; the downloadable data keeps them as separate Open3DCP columns, and superplasticizer and HPMC are present in the full records.

Download the free sample pack (CSV, API JSON example, and README). No account or API key needed.

Get a larger CEMFORGE Datapack

When you are ready for live predictions against your own targets and materials, buy a prepaid CEMFORGE Datapack directly at datapacks.cemforge.ai: a self-serve checkout and a simple prediction API, with a machine-readable offer feed so an authorized AI agent can purchase and call it on your behalf. To scope a larger or custom pack, email info@sunn3d.com. For in-house teams that want subscription access to the formulation engine directly, see app.cemforge.ai. If you need the location-aware, sourced version for a real project, that is a LOGiMIX Report, not a Datapack.

Request a custom mix

Need a formulation tailored to your own materials and targets rather than a prepaid exploration pack? Request a custom CEMFORGE mix and we will scope it to your project. A custom mix is $199 — the standard $99 Datapack plus a $100 validation step in which we confirm the formulation is sound before we deliver it. Request a custom mix → All custom-mix sales are final. This is an experimental, first-run trial — we are refining the process and will be making these high-quality and beautiful.

Sales, updates, and early access

All sales are final, including standard CEMFORGE Datapacks. We deliver a best-effort product from our current work, and we keep every customer up to date as the models and datapacks improve.

Are you a creator or reviewer? If you would like an early version to evaluate, email info@sunn3d.com with references — links to your other videos — showing that your followers would find CEMFORGE useful.

Important notice

CEMFORGE Datapacks deliver informational performance predictions for engineer-of-record review. Sunnyday Technologies LLC does not certify code compliance and does not provide professional engineering services. Predicted values must be verified by physical testing before any structural or production use.