Sugar and starch plants: changing beet, evaporation by feel

Sugar and starch content changes over the season

What happens. Beet, potatoes and other crops arrive with a different sugar, starch or fat content as the season goes on. The plant is set up for an average, and the operators correct after the fact, once the first lab value or a drifting process value shows that the raw material has changed.

What it costs. Yield and quality move with the feed. Part of the sugar or starch ends up in the by-products, and energy is spent on a product that was not the one the settings were made for.

Why the usual fix fails. More margin on the setpoints keeps the product in specification but costs yield every day. A lab value once per shift arrives after the material has already passed through, so the correction comes too late for the load that caused it.

How a fixed-price remote diagnosis checks it. From your process logs and the lab results of the last seasons I look at how the feed quality shows up in flows, temperatures and densities along the line. The fixed-price remote diagnosis tells you whether the raw-material quality can be computed early enough to steer the plant ahead of the change, and what that would take.

Background: Computing unmeasured states for control

Massecuite consistency is judged by the operators

What happens. In the crystallisation pans the operators judge the massecuite (the thick mix of crystals and syrup) by eye, by the feel of the stirrer and by experience. How far the sugar has really supersaturated is not measured directly, so each shift boils a little differently.

What it costs. Crystals of uneven size, sugar lost to the molasses, steam spent on over-long boiling and pans that are blocked longer than necessary. All of it shows up in yield and energy.

Why the usual fix fails. A refractometer or a conductivity probe in the pan gives a number, but it reacts to impurities and to the batch itself, so operators still trust their eyes. Fixed boiling recipes do not follow the changing syrup quality.

How a fixed-price remote diagnosis checks it. From pan logs (vacuum, temperature, steam, stirrer load, feed) and the crystal size results of the lab I look at whether the supersaturation can be computed during the boil. The fixed-price remote diagnosis answers whether a predictive controller could run the pan from that value, and what you would have to provide.

Background: Computing unmeasured states for control

The moisture end point of starch drying comes from the lab

What happens. The dryer is steered on the lab result for residual moisture, which arrives some hours after the sample was taken. In between, operators hold the dryer settings that were right for the last result.

What it costs. Starch dried too much costs energy and sometimes quality; starch dried too little is out of specification and may have to be run again. Either way the dryer works a shift behind the product.

Why the usual fix fails. Taking more samples ties up lab staff and still lags behind. An online moisture probe at the dryer outlet measures one spot, and it needs a calibration against exactly the lab values that arrive late.

How a fixed-price remote diagnosis checks it. From dryer logs (air temperatures, flows, feed rate, outlet readings) and the lab moisture values I check whether the moisture can be computed continuously. The fixed-price remote diagnosis shows whether that value is good enough to lead the dryer, and which of today’s measurements it would rely on.

Background: Computing unmeasured states for control

Fewer sugar plants mean longer campaigns for the rest

What happens. After recent closures of sugar plants, the remaining factories take in more beet and run longer campaigns. The plant was designed for a shorter season, so evaporators, pans and dryers run for weeks in a state where small losses add up.

What it costs. Energy per tonne of sugar and the cost of keeping a plant in operation for longer both matter more. A longer campaign also means more days of beet that has been stored and has changed.

Why the usual fix fails. Running the same recipes for longer does not help: the beet is different in week ten than in week two. Extra staff for a longer season are hard to find, and building new capacity is out of the question.

How a fixed-price remote diagnosis checks it. From campaign logs (evaporator steam, juice densities, pan data, energy meters) I look at where the plant uses more energy than the process needs and how it changes over the weeks. The fixed-price remote diagnosis tells you whether predictive control of evaporation and boiling could cut that, and what it needs from your data.

Background: Economic predictive control