ForgeCast AI is in public Beta. Outputs are model-based and not yet calibrated to first-party test data. Treat them as illustrative for early design screening, not as a substitute for coupon testing.Full disclaimer →

Skip to main content
Public beta

Rank every viable AM process & material for your legacy part.

Upload a legacy part, set your goals and resources, and ForgeCast AI ranks every viable additive manufacturing (AM) process & material combination with model-based confidence bands and citations down to the datasheet row.

MMPDS-grounded · Walker-corrected fatigue · scatter-derived uncertainty

US-export-compliant data & vendors Your CAD file stays in your browser
Normalized trade-off radar
3 candidates · higher = better · Walker-corrected fatigue
Ti-6Al-4V LPBF
AlSi10Mg LPBF
IN718 DED-LB/p
RECOMMENDED · example
Ti-6Al-4V · LPBF · M2 Series 5
  • ISO/ASTM 52900 (AM terminology)
  • ISO/ASTM 52901 (AM purchase requirements)
  • ASTM F3001 (Ti-6Al-4V ELI PBF)
  • ASTM F2924 (Ti-6Al-4V PBF)
  • ASTM F3055 (IN718 PBF)
  • ASTM F3301 (AM post-processing)
  • ASTM E8 (tensile testing)
  • ASTM E466 (force-controlled fatigue)
  • ASTM E606 (strain-controlled fatigue)
  • ASTM E647 (fatigue crack growth)
120k+
Material × process × parameter combos scored per project
8
Metal AM process variants scored across the 7 ISO/ASTM 52900 categories
180+
MMPDS & vendor datasheet rows
< 3s
Ranker compute, from features to ranked shortlist
Built for

Three audiences, one source of truth.

AM design engineers

Stop guessing on process selection.

  • Compare every viable process × material in seconds, not weeks.
  • See fatigue and strength with scatter-derived 1σ uncertainty bands.
  • Export a build recipe with parameters, supports, and lead-time.
Qualification engineers

Auditable, citation-grounded outputs.

  • Every prediction links back to MMPDS, CMH-17, or vendor datasheet.
  • Walker equivalent-stress correction with documented exponent assumptions, per alloy family.
  • US-export-compliant alloys & vendors only; restricted technical data never leaves your tenant.
Primes & service bureaus

Quote legacy castings without the back-and-forth.

  • Upload a CAD/STL and get unit-cost, lead-time, and risk in one view.
  • Filter by your installed machines, gas, and feedstock inventory.
  • Save projects, share with reviewers, export PDF reviewer packet.
The shift

Process selection used to take weeks. Now it takes a coffee break.

Status quo
2–6 weeks
  • Engineers stitch together MMPDS PDFs, vendor datasheets, and tribal Excel sheets
  • Fatigue numbers used as-is, ignoring R-ratio of the actual service load
  • Process candidates eliminated by gut feel before they're ever costed
  • Spreadsheet trade-offs take 2–6 weeks per legacy part
With ForgeCast AI
seconds, not weeks
  • Every process × material combo ranked from a single CAD upload
  • Walker equivalent-stress correction applied to mean-stress-sensitive fatigue cases
  • Hard constraints + installed-equipment filter narrow the field automatically
  • Ranked shortlist with cited recipes in seconds — review in minutes, exportable to PDF
Use cases

Real parts, real decisions.

Turbomachinery

Hot-section bracket requalification

A legacy IN718 investment casting needs an AM replacement path. ForgeCast evaluates LPBF, DED-LB/M, DED-Arc (WAAM), and PBF-EB against the cast baseline, with HIP + solution + age post-processing assumed and Walker correction applied for the actual R-ratio seen in service.

Result
IN718 LPBF (HIP + aged) ranked first — full parameter recipe (illustrative)
Landing systems

Secondary structure bracket replacement

An obsoleted 7075-T7351 wrought bracket needs a drop-in AM replacement. The tool filters by US-export-compliant vendors, your installed machines, and available Ti-6Al-4V and AlSi10Mg feedstock — surface finish and HIP assumptions are pinned upfront.

Result
5 viable recipes ranked in seconds; HIP + machining of fatigue-critical surfaces included
Airframe

Structural bracket consolidation

Multiple machined brackets consolidated into a single topology-optimized AM part. Fatigue and mass targets are set as hard constraints; cost is secondary.

Result
AlSi10Mg LPBF wins on mass; Ti-64 LPBF (HIP + machined) wins on fatigue — both cited
How it works

From CAD to qualified recipe in five steps.

01

Upload geometry

Drop in your STL/STEP. Critical thickness, surface area, and overhang risk are computed in the browser. Your CAD never leaves the device.

02

Define goals

Slide weights for strength, fatigue, mass, cost, lead-time. Pin must-haves (e.g. minimum UTS) and the optimizer respects them as hard constraints.

03

Set environment

Operating temperature triggers a property derate vs. MMPDS / vendor elevated-temperature curves. Mean-stress ratio R drives a Walker equivalent-stress correction. Spectrum loads are reduced via rainflow + Palmgren–Miner damage summation.

04

Allocate resources & pin process assumptions

Select installed machines, gas, and feedstock. Pin build orientation, support strategy, surface finish (as-built vs. machined), HIP, and powder-reuse / oxygen-pickup limits — assumptions that dominate AM fatigue life. Candidates you can't run are filtered out, not silently scored low.

05

Get ranked recommendations

A ranked list with confidence chips, parameter recipes, and one-click compare. Export the top picks to PDF for review.

What you get

Real outputs, not slides.

These are live components from the app rendered with seed data. The same charts you'll generate from your own geometry.

Fatigue S–N curve · Walker corrected
Ti-6Al-4V LPBF (HIP + machined) · R = 0.1 · 25 °C · illustrative
Unit cost breakdown
Per-part — powder, machine, post-processing · illustrative
Confidence by configuration
Model-based 1σ band across top recipes · illustrative
Methodology

Auditable, not just plausible.

Fatigue scoring & Walker correction

For mean-stress-sensitive fatigue cases, equivalent stress amplitude is computed using a Walker exponent fit per alloy and temper (not just per family). The exponent and a scatter-derived uncertainty band travel with every prediction. Alloys without a true endurance limit (e.g. AM Ti and Al) are scored on finite-life cycles, not a fictional fatigue limit.

Temperature derating & spectrum loading

Operating temperature triggers a property derate against MMPDS / vendor elevated-temperature curves. Variable-amplitude service loads are reduced via rainflow counting and Palmgren–Miner damage summation before fatigue scoring.

MMPDS-grounded, with AM caveats

Wrought and cast baselines reference MMPDS-2024 rows. AM-specific properties pull from vendor datasheets and peer-reviewed literature, since MMPDS AM coverage is limited and most published rows are wrought or cast. CMH-17 is used only for narrow metal-to-composite substitution screening.

Scatter-derived uncertainty bands

Every prediction ships with a 1σ band fit from datasheet and literature scatter — not a hand-wavy ±10%. These bands are model-based and will be re-calibrated as first-party coupon data is ingested.

Security & compliance

Built for regulated work.

Defense primes and government program offices have specific guardrails. ForgeCast AI was designed around them from day one.

US export-compliance aware

The vendor and machine catalog is restricted to US-export-compliant entries; project tags drive a server-side filter, not a UI toggle a user can flip. ITAR / EAR controls hinge on end use and technical data — restricted technical data never leaves your tenant.

Geometry stays in your browser

STEP & STL parsing runs locally via WebAssembly. Your raw CAD file never leaves the device. Only scalar derived features and your project metadata are sent server-side for ranking (see next).

Exactly what the server sees

Server-side ranking receives only scalar feature summaries: mass (g), bounding box (mm), max & thinnest wall (mm), overhang area fraction, and projected build area (mm²). No mesh, no thickness map, no surface data — nothing reconstructible into your CAD.

Cited, reproducible outputs

Every number links to its MMPDS / CMH-17 source row with revision date. Exported PDFs include the full citation chain for stage-gate reviews.

Documented methodology

Built by Quantum Diffusion Corp. with a documented, reviewable methodology. Every assumption and correction is traceable to its source.

Data sources

Every recommendation is grounded in a citable source.

No black-box scoring. The catalog is built on the same handbooks and standards your qualification reviewers already trust.

MMPDS-2024

Metallic materials handbook

Allowables for aerospace alloys. Every mechanical property in ForgeCast cites a specific MMPDS row. The Ti-6Al-4V standard-grade (ASTM F2924) vs. ELI (ASTM F3001) split is automatic: ELI is selected for fracture-critical or cryogenic service tags; standard-grade is the default otherwise.

Coverage
Ti-6Al-4V (std + ELI), IN718, Al 7075, 4340, 15-5PH

CMH-17

Composite materials handbook (screening only)

Used narrowly for metal-to-composite substitution screening on legacy airframe brackets. Not used for AM metallic allowables.

Coverage
Tape & fabric laminates, hybrid layups

NIST AM-Bench

Process benchmark dataset

Independent measurements of LPBF residual stress, melt-pool geometry, and distortion.

Coverage
IN625, IN718, 17-4PH

ISO/ASTM 52900-series

AM terminology & test standards (ASTM Committee F42 / ISO TC 261)

Process classification (52900), purchase requirements (52901), and qualification test methods used to score vendor capability. F42 is the ASTM committee that co-publishes these with ISO TC 261; we cite the standards, not the committee, in exported reports.

Coverage
All seven ISO/ASTM 52900 process categories

Vendor capability matrix

Internally curated

14 US-based service bureaus with audited machine inventory, build volume, and US-export-compliance status.

Coverage
LPBF, DED-LB/M, DED-Arc (WAAM), PBF-EB, binder jet

Beta estimates are model-based and not yet calibrated to measured test data. Confirm critical values with physical testing.

Roadmap

What's shipped, what's next.

Built in public. Beta users vote on what we prioritize next.

Shipped

Ti-6Al-4V (std + ELI) & IN718 catalogs

MMPDS-2024 baselines plus AM-specific vendor datasheets, with the standard / ELI Ti-64 split tracked separately (per ASTM F2924 vs F3001). 14 US-export-compliant LPBF and DED-LB/M bureaus, with lead-time signals.

Shipped

In-browser STEP/STL parser

WASM-based geometry analysis. CAD never leaves your machine.

In beta now

Citation-backed PDF export

One-click reviewer packet with full MMPDS/CMH-17 citation chain.

Q4 2026

AlSi10Mg & IN625 fatigue allowables

Promoting AlSi10Mg LPBF and Inconel 625 LPBF from screening-only to full fatigue-curve coverage with vendor-validated S–N data.

Q4 2026

Vendor-specific cost calibration

Promote the current parametric cost model (machine-hour, powder, post-processing) from generic to per-vendor quotes, pulled directly from bureau rate cards.

2027

Coupon-fed Bayesian calibration

Ingest first-party coupon test results to update recommendation priors per facility and per machine. (Not in-situ closed-loop control — that lives on the machine.)

FAQ

Common questions.

What is the public beta?

ForgeCast AI is in active development. The beta is free, no credit card. Expect rough edges and incomplete catalog coverage — please report bugs to beta@qdlabs.tech.

What's actually "AI" about ForgeCast AI?

Honest answer: today, very little. The current ranker is a deterministic weighted sum over physics-based scores (Walker-corrected fatigue, MMPDS-derated strength, mass, cost, build-time) plus a geometry-fit heuristic and a confidence multiplier — there is no learned model in production. The "AI" in the name refers to the roadmap product: a Bayesian update layer (planned 2027) that recalibrates per-facility priors as first-party coupon data is ingested, and a learning-to-rank step trained on reviewer accept/reject signals collected during the beta. We'd rather be upfront about this than imply ML where there isn't any. Mechanical-property numbers are, and will remain, handbook lookups with documented corrections — never LLM-generated.

Is my CAD data uploaded to a server?

Your raw CAD file never leaves your device — STEP/STL parsing runs in the browser via WebAssembly. Only scalar derived features (mass, bounding box, max/thinnest wall, overhang area fraction, projected build area) and your project metadata (alloy, process, goals) are sent server-side so the ranker can run and so you can come back to your projects later. No mesh, thickness map, or surface data is transmitted.

What does it cost after beta?

Pricing is not yet finalized, but we commit to two things: (1) every beta account is grandfathered into a paid tier at a permanent discount, and (2) the free tier will remain usable for single-part exploration after GA — you won't be paywalled out of the tool you learned on.

Are export-controlled alloys included?

The vendor and machine catalog is restricted to US-export-compliant entries; restricted entries are excluded server-side, not just hidden in the UI. Note that ITAR/EAR controls generally hinge on end use and technical data, not on the alloys themselves — so we also commit that restricted technical data you enter never leaves your tenant.

Who is behind ForgeCast AI?

Quantum Diffusion Corp., focused on legacy-part qualification via additive manufacturing.

How do I share feedback?

Email beta@qdlabs.tech with your project ID and a short description. We triage bug reports within one business day during the beta.

Beta access

Free during beta. Five minutes to your first ranked recommendation.

No credit card. Email confirmation required to verify your account.

Start free beta