Sageware Textile Upcycling System

Offcuts and remnants are among the least recoverable parts of the textile stream. They are mixed in composition, small, inconsistent, and cheap — which makes them expensive to sort and easy to landfill.

Sageware approached that from the product end: find an output worth making from difficult input. Jewelry beads are small, tolerate variation in source material, and carry enough value per gram to justify a process. The engineering question became whether a compact, reliable machine could produce them from scrap consistently.

Fabric is not a machining material. It is compliant, anisotropic, variable in thickness, and it frays — so it deflects away from an edge instead of shearing cleanly, and no two pieces load into a fixture the same way.

The process therefore stabilizes the material before touching it with tooling. Once hardened, the fabric behaves closer to a sheet material: it supports a cutting edge, holds a stamped form, and can be located in a fixture repeatably. Nearly every tooling decision downstream followed from that change of state.

The brief called for a machine that fits on a bench and runs without an operator managing each piece. Compactness is not just packaging here — it sets the stroke lengths, the tooling size, and how much room there is for the material to move between stations.

Working within that envelope meant combining functions. Stations that located, cut and formed in fewer motions won over sequences that were easier to design but needed more travel and more transfers, since every transfer is a chance for a compliant blank to shift.

My contribution was the tooling and fixturing: the parts that actually touch the material. Cutting hardened fabric needs a genuinely sharp edge with enough support behind it that the blank cannot fold rather than shear, and clearance that lets the cut piece release instead of wedging.

Fixturing had to locate a part that is nominally flat but never quite. The approach was to constrain the blank where the geometry mattered for the cut, and deliberately leave it free elsewhere — over-constraining a compliant part just moves the error somewhere less visible.

The fixture combined two manufacturing methods for two different reasons. Sharp steel inserts do the cutting and stamping, because nothing printable holds an edge against fabric for a useful number of cycles. The surrounding fixture body was printed in high-temperature SLA resin, which tolerated the process temperatures while letting the geometry iterate in a day.

That split was the enabling decision: wear-critical features in machined steel, everything else in a printed body that could change between trials. It kept iteration fast without accepting tooling that would dull after a handful of parts.

Tooling for a variable material gets designed by trial. Each fixture revision ran a batch across a deliberate spread of fabrics — different weaves, thicknesses and fiber blends — and the failures were more informative than the successes.

Frayed edges, incomplete cuts, blanks lifting and pieces sticking in the cavity each pointed at a specific geometric cause. Iterating the printed body against those modes, while keeping the steel edges fixed, was the fastest available loop.

The three requirements pulled against each other. Reliability wanted generous clearances and simple motions; compactness wanted short strokes and combined stations; safety wanted cutting edges enclosed and hands away from anything that moves.

Where they conflicted, reliability and safety won. A machine that jams once every twenty parts is not automated in any meaningful sense, and sharp tooling in a benchtop product has to be guarded by geometry rather than by instructions.

By the end of the term the team had a compact automated concept that produced beads from fabric scrap, with tooling and fixturing that handled a realistic spread of input material.

The lesson I carried forward is about sequencing: with a difficult material, the highest-leverage move is usually to change the material's state so ordinary mechanical design can work on it — rather than designing ever more clever tooling around the raw behavior.