BRENNEN

Industries / Food, farming and forestry / Food and beverage processing

Sugar refining

Sugar refining is the work of crushing sugar cane and slicing sugar beet, cleaning and boiling down their juice, growing sugar crystals from it, and refining the raw sugar into white sugar, while cane mills burn the leftover fiber to power themselves.

About 100 CE

The Arthaśāstra, an Indian treatise on running a state, lists the processed products of sugar cane as syrup, jaggery, massecuite (sugar crystals still in their syrup), soft brown sugar and crystal sugar, in James McHugh's translation of the passage. Patrick Olivelle dates the core of the text to between 50 and 125 CE and its final form to 175 to 300 CE. Popular histories give dates from about 500 BCE to 350 CE for the first crystal sugar without citing direct evidence, so this file starts from the dated text.

journals.library.ualberta.cascroll.in

35

Milestones achieved across 6 domains.

8

To watch for: announced, not yet confirmed outside.

13

Open deliverables. Nobody has shipped them.

In one minute

Where it started

By the early centuries CE, people in India boiled cane juice until crystals formed and drained off the dark syrup, molasses. European-owned plantations worked by enslaved Africans later spread cane sugar from Madeira and other Atlantic islands to the Caribbean, and their raw sugar was shipped to refineries in Europe. In the 1800s the work became an industry: the first beet sugar factory opened in 1801, the vacuum pan was patented in 1813, multiple-effect evaporation came in 1843, and a centrifuge for spinning syrup off crystals was shown in 1851.

Where it stands

The International Sugar Organization counted 178.9 million tonnes of sugar made in 2022, 80.1% of it from cane and 19.9% from beet; Brazil and India made about 36 million tonnes each, and the European Union led in beet sugar. In November 2026 the US Department of Agriculture forecast 178.1 million tonnes for 2026/27, about 5% below the year before. Cane mills burn bagasse, the crushed cane fiber, for their own steam and sell spare power to the grid, while beet factories, which burn gas and coal, are trying heat pumps, electric vapor compressors and biogas.

What is open

Much of the work in a sugar factory is still judged by eye or by slow laboratory tests: how finely cane is shredded, how well beet is cut, how crystals grow in the pan, how much wash water a centrifuge needs, and how much leaf and soil arrive with the cane. Factories keep their images and readings to themselves, and no public set of labeled crystal photos, centrifuge frames or cane samples was found. Cheap cameras, small open AI models that can be retrained on one graphics chip, and public emissions registers now let a small team build these test sets, along with a free map of refineries and a carbon figure for each factory.

Domains

AchievedWatch forOpen

Advances to build with

1 research note

Open in sugar refining

3 of 13 open deliverables

Open · Crushing and extraction

Prepared cane graded from photos

A camera and model that grade shredded cane from a photograph, predicting the laboratory measures of how well it has been prepared: how fast juice drains through it (percolation rate), its packed density, and the Displacement Rate Index, a timed test of how quickly water displaces juice in a sample.

Built with what exists

A phone or industrial camera with fixed lighting; the percolation, density and Displacement Rate Index tests SMRI used; an open image model such as a ResNet or a Vision Transformer, fine-tuned with PyTorch on one graphics chip (see the Note "One GPU to fine-tune").

Credit · unclaimed

Open · Clarifying and evaporating

Settling tests read from video

Software that reads a time-lapse video of a juice settling test, in which limed juice and a flocculant (a chemical that clumps fine particles) settle in a glass cylinder while a technician notes the height of the mud line every minute, and returns the settling curve, the settling rate and the final mud volume, so that flocculant doses can be compared on more tests than a technician can watch.

Built with what exists

A phone on a tripod with a lit white background; OpenCV and Python to find the mud line in each frame; optionally a segmentation model such as Meta's Segment Anything for murky interfaces.

Credit · unclaimed

Open · Crystallizing

A public crystal photo benchmark

A public set of microscope photos of sugar crystals, each labeled with crystal size measured by sieving (the mean aperture and the spread around it), with a scoring program and a baseline model that reads both from a photo.

Built with what exists

a laboratory microscope with a camera, a stack of test sieves, open segmentation models such as Meta's Segment Anything or a U-Net in PyTorch, and a small open vision model fine-tuned on one graphics chip (see the Note "One GPU to fine-tune").

Credit · unclaimed

Where it sits