Live engine
A traceable scan-to-manufacture pipeline where measurement source, assumptions, solver questions and verification survive all the way to the output.
Phone photos, video and LiDAR are placed against a printed ArUco photogrammetry mat with a 1 mm grid. Structure-from-motion handles general geometry; silhouette/revolve reconstruction handles turned parts; planform-plus-depth handles thin plates; caliper readings anchor dimensions that must be hard numbers.
PicoGK and signed-distance fields handle lattices, channels and biomimetic negative space. CadQuery, build123d and FreeCAD produce parametric source geometry, STEP and drawings instead of a dead mesh.
Before a solver runs, ScanIQ states the engineering question, the decision it will support and the pass criterion. Available analyses include static FEA, modal/resonance, fatigue, OpenFOAM k-ω SST CFD, 1-D acoustics, electromagnetic field maps and material screening.
Deterministic rules, two independent AI models and real-world reference values answer the same question. Agreement raises confidence. Disagreement is escalated to a human rather than averaged into false certainty.
Plain language becomes a measurement plan rather than a hallucinated CAD model.
Scan: capture the cap, section, pitch gauge and calipers. Plan: separate hard dimensions from image-derived shape and unresolved groove depth. Solve: build an editable revolved/parametric model. Verify: re-check Ø41.9 thread crest, 1.5 mm pitch, 55.62 mm overall length and stem OD 16 / ID 11 × 22.8 mm. Make: STEP/STL after assumptions are either measured or explicitly accepted.
The digital twin is reconstructed from photos after the physical part is gone, then the model is interrogated by static and modal physics.
Implicit geometry is useful only if the engineering loop can still explain what changed and why.
The conch/log-spiral hub and helical blades can be generated as signed-distance fields, while blade count, 17.4° tip pitch and the five-vane straightener stay explicit parameters. Electromagnetic field mapping is already part of the toroidal concept; CFD validation for the helical configuration is queued.
Open maritime work