hr-studioGeometry & Structure · 05
A computational design primer · Panelization

Cutting a surface into buildable pieces.

A freeform roof is a beautiful lie: nobody manufactures curved glass by the acre. What gets built is a panelization — hundreds of small, flat pieces approximating the dream. This chapter is about that translation: how many panels, how flat can they be, and the quiet theorem that decides — before any optimizer runs — whether your surface will panel gracefully or fight you.

Scroll to begin
i.

How many pieces?

Lay a grid over a curved surface and each cell becomes a candidate panel. Coarse grids give few, large, affordable panels that visibly miss the surface; fine grids hug it at the cost of seams, joints and money. The error shrinks fast — quarter the deviation for every doubling — so the game is finding the coarsest grid the eye forgives.

Slide the density and watch the deviation fall.

Fewer panels, bigger lie · the view turns slowly
ii.

When is a quad flat?

Triangles are always flat — three points define a plane. A quad's four corners have no such guarantee: lift one and the panel warps, and warped means expensive: cold-bent glass, twisted frames, gaskets that leak. The standard measure is the gap between the two diagonals; zero means planar, and fabricators live by fractions of a percent.

Drag the corners in 3D and watch the warp gauge.

Diagonal gap = warp · drag corners, drag space to orbit
iii.

Surfaces that panel themselves

Here's the theorem that shapes real roofs: slide one curve along another — a translational surface, height = f(u) + g(v) — and every quad in its grid comes out exactly planar, automatically, forever. The cross-terms cancel; the algebra does the fabrication. Morph the same grid toward a dome, and the flatness quietly dies.

Slide between the vault and the dome; the colours are warp.

Vault: exactly flat · dome: warped
iv.

Pushing panels flat

For surfaces that don't panel themselves, you optimize: project each quad onto its own best-fit plane, average the answers, repeat — relaxation again, aimed at flatness. And here's the honest part: it takes the dome a good way down and then plateaus, and it barely dents a twisted grid at all. The optimizer polishes; the choice of surface and grid decides.

Run the planarizer on each surface and read the gauge.

Optimization helps — layout decides
v.

The facade studio

Your roof, your call. Pick the surface — the friendly vault, the plateau-bound dome, the stubborn twist — set the panel density, and choose triangles or quads. Colour the panels by warp and every fabrication problem lights up; run the planarizer and watch how far it can and can't take you. Orbit with a drag. The readout keeps score: panel count against worst warp — the two numbers every facade meeting is really about.

Drag to orbit · colours = warp (blue flat → red warped)
Every freeform surface eventually meets a factory. Density buys fidelity, triangles buy certainty, and the right surface — one curve slid along another — buys flatness for free. Panelization is where geometry signs the construction contract.