3D block facade panels are drawn as solids and manufactured as flat sheets with lines marked on them. Everything between those two states is folding, and the folding decides whether a panel arrives at its intended depth with a face the eye will accept. Depth as a design variable and fold logic at system level belong to the complete 3D block facade guide. This article starts at the press brake.
A flat panel gives a shop one surface to keep true. A folded block gives it several, tied to one another by bends: each fold rotates every facet after it, so an error made at the second hit is still present at the last, enlarged.
The SOGA 3D Block Facade is a 3D block facade system developed by SOGA Design Studio for Indian fabrication. The sequence below is the one the studio settled with Indian fabricators while taking deep-folded panels into batch production.
3D block facade panels start as flat blanks generated by unfolding the modelled block geometry with bend allowances applied. The blank is cut, folded in a planned sequence, stiffened with returns or a back-frame, joined into a cassette, finished and then dimensionally inspected. Depth accuracy depends on controlling springback consistently across the batch.
From a flat blank to a three-dimensional block
The route runs unfold, cut and pierce, break out, fold, stiffen, join and form the cassette, finish, inspect, pack. Most of it is ordinary sheet work. Three operations behave differently once the part has depth: folding, because every bend changes the space the machine needs for the next; stiffening, because heat or a hard fastener behind a wide facet shows on the front; and packing, because blocks of differing depth do not stack.
Unfolding: the geometry step that decides everything
The flat pattern is not the folded block flattened out. Material outside a bend stretches, material inside compresses, and a layer within the thickness does neither. That layer is not at the middle: its position moves with the ratio of bend radius to thickness, and the developed length follows from it. A small error repeated at each bend leaves the last flange short.
Two things get assumed and should not be. A drawing shows a corner; the floor produces a radius, set by the tooling and by what the stock will take, so an unfold built on a sharper radius than the press achieves makes every part wrong the same way. And generic bend allowance tables are approximately right, which is the worst thing they could be, so the bend data SOGA works from is developed on the fabricator’s own press. Grain matters too: bending with the outer fibres along the rolling direction is harder on the material, and at a tight radius it shows as crazing.
Bend allowance, relief cuts and corner closure
Where two bends meet, material at the intersection is asked to be in two places. Without a relief cut it tears at the end of the bend, or bunches and lifts the flange. The relief is sized against the radius the press will achieve and runs past the tangent point of both bends, a dimension that exists only once the tooling is agreed.
Corner closure decides how the panel reads at close range and how long the finish lasts at its most exposed line. A mitred and welded corner is cleanest and introduces heat, which shrinks the seam and pulls the adjacent facets. A riveted lap adds no heat and shows a fastener head; a folded interlocking corner adds neither and costs blank area. Each implies a different blank, so the choice belongs at unfolding.
Flat pattern data required per panel
The package fails by omission rather than error. Each panel needs:
- Panel identifier with elevation, bay and grid position, and its depth step
- Developed blank outline, with piercing, relief cuts and drain positions
- Bend lines dimensioned from the visible face, with the assumed inside radius
- Bend sequence, with the gauging feature nominated for every hit
- Permitted bend orientation relative to grain, and the show face
- Corner closure, stiffener and cleat positions, cassette type and hooks
- Depth at the face, the flatness requirement, and the datum for both
The entries most often missing are the gauging feature and the assumed radius. Both are then decided at the machine, by somebody nobody will ask afterwards.
Folding sequence and tool access
The blank is set against a backgauge, the ram descends, and a flange rotates up into the space above the bed. Every bend after the first is made on a part that is no longer flat. Three collisions end a sequence: the formed part fouls the punch as it rotates, it fouls the machine frame, or the flange the operator must hold now sits behind a facet. A gooseneck punch and a narrow die buy clearance, but achievable depth is settled against the fabricator’s machine, not assumed. An unsupported flange dropped at the end of a bend also leaves the part with a set.
Why fold order changes what is possible
Order does three things at once. It decides reachability. It decides how much the part resists the next bend, since a flange already turned stiffens the region beside it and shifts the springback of what follows. And it decides where error accumulates, because each hit is gauged from a feature, and the available features change as the part forms.
The last costs the most. Gauge early bends from a sheared edge and later ones from a formed flange, and the reference has migrated, so the accumulated deviation lands wherever the sequence ends. The rule SOGA carries onto the drawing is one nominated gauging feature for the whole sequence, with a notch cut into the blank where a fold would destroy it. Face-establishing bends go early; returns and closures go last, where take-up is absorbed at a joint nobody measures.
Springback and depth accuracy
Springback is elastic recovery. Past the point where a bend becomes permanent, some deformation stays and some relaxes when the tool lifts, so the achieved angle is always more open than the angle formed. It is not a defect, cannot be removed, and drifts.
Depth is where the drift becomes expensive. Projection is a product of flange length and fold angle, so an angular deviation is amplified along the flange before it reaches the face: the longer the flange, the more a fraction of a degree is worth at the front. Deep panels are harder to hold than shallow ones.
Air bending, where the punch never presses the material fully into the die, is the most flexible method and the most springback-sensitive, since the achieved angle depends on how far the ram descends. Bottoming reduces the variation; coining nearly removes it, at a tonnage facade-sized flanges make impractical. Facade panels are air bent, so compensation is managed against known drift: stock thickness varying within its own tolerance, strength varying by lot, tools warming, dies wearing.
Holding depth within tolerance across a batch
Depth is verified as depth, not inferred from an angle. A profile template made from the approved first article, referenced against the plane of the cassette perimeter, gives the operator a pass or a fail rather than a reading to interpret. One template per depth step, kept at the brake, is the cheapest control there is.
More useful than a checking interval is the list of events that trigger a check regardless: the first part of a run, a new stock lot, a tool change, a shift change, any interruption long enough for the tooling to cool. A fail stops the run rather than triggering a sort, since sorting removes the bad parts and leaves the cause in place. That rule came out of SOGA’s forming trials.
The tolerance is set by what is visible: depth is read at the joint, where two neighbouring faces sit side by side and a difference reads as a step in a line, so the permitted deviation is agreed against a mock-up at the real viewing distance. Drift within a run is gradual, which is why consecutive panels off the machine go to consecutive positions on the elevation; allocated randomly, the same drift becomes a step between neighbours.
Rigidity: stiffeners, returns and back-frames
Folds stiffen a panel about the axes they run along and do little across them. The floor deals with the consequence: the largest flat facet, usually the visible face, has the stiffness of bare sheet and buckles into shallow waves under in-plane compression. Oil-canning appears there, not on the folds, and cannot be corrected afterwards. The cheapest answer is formed rather than added — a shallow secondary fold across the wide facet, made in the same sequence.
Where the face cannot take a visible crease, stiffening is attached. Welding a rib puts heat into a small area; the weld and the material around it shrink as they cool, pulling the surface into a shallow depression invisible on a bench and unmistakable on a wall in raking sun. Stitching, backstepping and clamping to a jig until cool reduce that print-through but do not remove it. Bonded stiffening avoids the heat and asks for prepared surfaces and a full undisturbed cure, so fixture time is what makes it expensive. Mechanically fixed cleats add neither, though each fastener is a point load that can dimple a face.
| Strategy | Depth achievable | Face flatness | Added weight | Fabrication time | Cost impact | Suitability |
|---|---|---|---|---|---|---|
| Simple returns | Limited by brake reach | Face carries the risk | None | Shortest; one setup | Lowest | Shallow panels, far elevations |
| Returns plus stiffener | Return depth governs | Good one way only | Small, at the back | One added operation | Modest | Medium depth, wide face |
| Back-frame bonded | Free of return depth | Best, if cured face-down | Frame weight | Longest; cure occupies the fixture | High | Deep panels, graded families |
| Welded rib | Free of forming limits | Most at risk; print-through | Small | Moderate; rectification varies | Unpredictable | Concealed areas only |
| Composite backing | Set by the panel | Very good; resists buckling | Highest | Moderate; cure governs | Highest | Where flatness governs |
Joining, sealing and cassette formation
Joining and cassette formation both happen on a fixture, and the fixture is what holds the depth. The setting fixture is a flat table with hard stops locating the panel by its visible face, face-down, so returns, closures, stiffening and frame are added to a part whose most important surface lies on a known plane. Assemble the same panel face-up and the accumulated error goes to the face instead. Fixtures drift too, so each is re-verified against the retained first article.
Sealing here is not a weather line, since the assembly is drained and back-ventilated, but the panel must not hold water inside it. Any upward-tilted facet is a tray, so drains go into the blank before forming; drilled after finishing, they leave an unprotected edge where water sits. Where the cassette is an applied frame, it is built on its own jig with hook geometry set once for the whole depth family, which keeps cassette installation and panel flatness a single detail.
Finishing without damaging fold lines
A faceted part behaves differently in a finishing line from a flat one, in three ways that have to be designed for rather than discovered.
Drainage is the first. The block holds solution in its upward facets and returns; carried forward it contaminates the next tank, and carried into the oven it dries as a tide line under the coating. Racking angle is set so every facet drains, and checked on a real part.
Film distribution is the second. Electrostatic charge concentrates on convex edges and is weakest in concave corners, so film builds heavily along every fold arris and thinly where two facets meet. Thickness is measured in the internal corners, because the open face is never the thin place, and blocks are lifted from designated points rather than by the arrises.
Bend radius is the third, a finishing decision as much as a forming one: a tight radius puts the outer surface into tension and gives the coating least material to sit on, so a radius that risks crazing the stock also leaves a thin line of film along every fold.
Dimensional inspection and first-article approval
A flat panel is checked for size, squareness and flatness on a table with a straightedge. A block has several facets, each with a position and an angle relative to the others, and they cannot all be reached from one setup with hand tools. The first article is therefore measured once, properly, with equipment able to locate points in space, which establishes what the process produces, not what the drawing asked for. The run is then checked against that part with gauges giving pass or fail: a check frame with hard stops at the cassette perimeter, a profile template for depth, and a straightedge across the diagonal in raking light.
Two disciplines make the numbers mean anything. The panel is measured from the datum it was made from, which here is the visible face. And the first article stays a physical part at the bench, the reference for depth, for tone and for adjudication.
Packing panels that must arrive undented
A graded family of blocks does not stack, and that is a manufacturing problem, not a logistics one, because a panel damaged in a crate has consumed every operation above. Load must pass through the cassette perimeter, the only flat, in-plane, stiff element the panel has, so dunnage profiled per depth step, or panels paired face-to-face so projections interlock, keeps the stack off the face.
A dented arris cannot be dressed out: it is a formed feature with a defined radius, and pressing it back leaves a flat spot and cracked film. Crates are sized by depth step and released in erection order, and since a deep panel takes large volume for small mass, crate design is where freight is controlled or paid.
Common defects when 3D block facade panels are manufactured
| Defect | Where it originates | How it is detected | How it is prevented |
|---|---|---|---|
| Blank short at the last flange | Allowance from a generic table | First article in the check frame | Bend data from the fabricator’s own press |
| Crazing along the arris | Bend wrong to the grain, radius too tight | Visible after forming, worse after coating | Bend orientation locked at flat pattern |
| Depth drift through a run | Springback shifting with lot, heat, die wear | Go/no-go template at the brake | Re-gauge at every trigger event; a fail stops the run |
| Facets out of true at the last hit | Gauging reference migrating between bends | Panel will not seat in the check frame | One gauging feature, kept alive by a notch |
| Oil-canning on the visible face | Wide unstiffened facet in compression | Straightedge on the diagonal, raking light | Secondary fold, or bonded stiffening |
| Dented fold arris | Packing; load taken on the face | On unpacking, once it is a site problem | Load through the perimeter; dunnage per step |
How SOGA developed this production sequence
The problem that started the work was narrow and stubborn. Indian shops fold sheet competently, so forming a faceted panel was never in question. Holding the visible face true on a deep panel, across a run rather than on a sample, was. A block accumulates the error of every bend before it, springback shifts with lot and tool temperature, and a face out of plane reads as waviness from across a road. Nobody could state the depth a folded panel would be delivered at, so the geometry could be drawn but not quoted.
The studio worked through it by producing panels. Prototypes established how many folds a panel could carry before the last facet became unpredictable, and fabrication experiments took the work onto a production floor at real batch sizes, where assumptions about tool access and gauging failed first. Structural studies covered face stiffness and hook eccentricity, installation testing on a mock-up settled the joint, and manufacturing optimisation followed once the depth family closed.
Two things changed between the first production run and the settled sequence. The first was assembly orientation: that run built panels face-up, which delivered every accumulated fold error to the visible face. Panels are now built face-down into a fixture holding the face against a flat table while returns, stiffening and frame are added. The second was gauging. That run let the operator take each bend off whichever edge was convenient, so the reference migrated and depth drifted. The blank now carries a gauging feature that survives to the last hit, and the fold order was rearranged so the face-establishing bends happen early.
Development sat inside the six-months-to-a-year range these systems take. It is manufacturable in India because nothing in it needs a machine an Indian sheet fabrication shop does not already run: a cutting bed, a press brake, a bench, a fixture and a coating line, gauged against one retained part.
SOGA Design Studio is a parametric and computational architecture practice based in Gurugram, working across India, Dubai and Singapore, specialising in facade systems engineered for local fabrication.
Frequently asked questions
What is springback and why does it matter?
Springback is the elastic recovery of the material after a bend, so the achieved angle is always more open than the angle formed. It matters on a folded block because depth is a product of flange length and fold angle, which turns a small angular deviation into a much larger one at the face. Tooling is calibrated against it at first-article trials.
How is panel depth tolerance controlled?
Through calibrated tooling, consistent stock lots, and depth gauged against a template made from the approved first article rather than measured with a tape. The permitted deviation is agreed with the fabricator against a mock-up at the real viewing distance, since depth is read at the joint between two faces.
Why do folded panels distort?
Usually because the fold sequence delivers accumulated error to the visible face, or because a wide facet is not stiffened enough and buckles under in-plane stress. Welded stiffening adds a third cause, since local shrinkage pulls the face into a shallow depression. All three are settled at design and fixture stage.
Are 3D panels stiffer than flat panels?
Generally yes: folds act as stiffeners, which is why depth is a structural decision as much as a visual one. The qualification is that a fold stiffens the panel about its own axis and little across it, so a block can be rigid overall and still wave on the surface people look at.
What the shop floor is actually controlling
A 3D block facade panel is one flat blank with lines on it, made several thousand times, and each time a machine acts on one line while the rest of the part waits. The object still has to behave as though every line were made at once, which is why the drawing states a radius, a gauging feature and a datum, not a shape.
The failures follow the same logic: depth drifts because springback was compensated once instead of monitored, and faces arrive out of plane because the reference moved between hits. What that costs sits with 3D parametric facade panels and cost in India; carrying a radius and a gauging feature from model to machine is the subject of algorithm to fabrication drawing.
The engineering here was done in India for Indian production, on Indian presses, against Indian stock tolerances, with Indian fabricators arguing over the inspection plan. It carries abroad because it asks nothing of a floor elsewhere that it has not already asked of a floor here.
Flat-pattern data and inspection tolerances travel with every panel drawing, so the shop floor is never guessing at depth.
