Fiber Organizer Robot — 250-micron fiber, aligned to 5 microns
viaPhoton’s patented fiber organizer, on a factory floor in Aurora, Illinois. Machine learning and computer vision took fiber sorting from three minutes to eight seconds per cable, inside an automation effort that lifted capacity forty times.
- Capacity increase
- 40x
- Sorting accuracy
- 96%
- Alignment precision
- 5µm
- Cables per month
- 9,600
Problem
An Amazon-partnered fiber optics supplier in Aurora, Illinois was building about 240 cables a month. Before a cable can be terminated, its 250-micron fibers have to be in the right order and precisely aligned, and sorting the fibers for one cable took about three minutes.
Constraint
Each fiber is a quarter of a millimetre across and alignment had to hold to 5 micrometres — on a production line run by operators rather than data scientists, where every second of latency is paid on every cable.
Decision
I treated the model as the small part of the problem — maybe 30% of the work. The rest was everything around it: image capture that behaves the same on every shift, inference that fits the machine’s timing, and the edge cases a clean dataset never contains. A 96% model an operator cannot run is worth less than a 90% model they can.
Outcome
Fiber sorting fell from about three minutes to eight seconds per cable at 96% accuracy, with alignment held to 5 micrometres. On the organizer line, AWS cable output went from one an hour to six; across the wider factory automation, monthly capacity went from 240 cables to 9,600 — 40x. The machine is covered by viaPhoton’s US Patent 12,001,073 B2.
viaPhoton’s fiber organizer takes optical fibers loaded in a random order and moves them into a predetermined sequence, with a camera and a neural network checking the result. The machine is covered by viaPhoton’s US Patent 12,001,073 B2 and runs in production in Aurora, Illinois. As a machine learning engineer at Cowlar Design Studio, I worked on its machine learning and computer vision.
The problem
An Amazon-partnered fiber optics supplier was building about 240 cables a month. Before a cable can be terminated, its 250-micron fibers have to be in the right order and precisely aligned, and sorting the fibers for one cable took about three minutes. Throughput was the bottleneck.
What the work covered
- Computer vision and machine learning to sort 250-micron fibers, at 96% accuracy
- Fiber alignment to 5-micrometre precision
- Several fiber termination steps integrated into one automated line, at a 96% success rate
The result
Sorting time fell from about three minutes to eight seconds per cable. On the organizer line, AWS cable output went from one cable an hour to six — about fifty a day. Across the wider factory automation effort, monthly capacity went from 240 cables to 9,600, a 40x increase.
The machine is covered by viaPhoton’s US Patent 12,001,073 B2 and runs in production in Aurora, Illinois.
What I learned
Production ML is not research ML. The model was maybe 30% of the work. The rest was hardware integration, edge cases, and building a system that a factory operator could run without a data scientist in the room.