Aluminium plants: cells and furnaces run on habit and margin

Smelter cells draw the same power whatever it costs

What happens. The potline runs at constant amperage around the clock. Power prices swing from hour to hour, but the cells take the same load either way. Changing the amperage feels risky: the heat balance of every cell depends on it.

What it costs. Electricity is one of the largest costs of a smelter. A load that ignores the price pays the expensive hours in full, and the plant cannot offer the grid any flexibility it might be paid for.

Why the usual fix fails. Manual load cuts in a price spike are cautious and late; nobody knows how far the cells could really go. Fixed rules (“reduce by this much for at most so long”) are set for the worst cell in the line.

How a fixed-price remote diagnosis checks it. From your amperage, voltage, bath temperature and price history I estimate how far each cell’s heat balance can move before the ledge or the bath gets into trouble. The fixed-price remote diagnosis shows whether a load plan that follows the price inside those thermal limits is feasible, and what it would take.

Background: Economic predictive control

Alumina concentration in the bath is not measured

What happens. The point feeders add alumina on a cycle of overfeeding and underfeeding, steered by the cell resistance. The actual concentration in the bath is not measured; it is inferred from how the voltage reacts. When the inference is wrong, the cell runs too lean and goes into an anode effect, or too rich and builds sludge on the cathode.

What it costs. Every anode effect means a voltage spike, lost current efficiency and a burst of perfluorocarbon gases, which weigh heavily in the plant’s greenhouse-gas balance. Sludge raises the cathode voltage drop and disturbs the cell for days.

Why the usual fix fails. Retuning the feed thresholds by hand helps one cell and upsets another, because cells age differently. Bath samples for the laboratory arrive hours later and show one moment of one cell.

How a fixed-price remote diagnosis checks it. From your feeder events, pseudo-resistance curves, anode-effect log and bath analyses I check how well the resistance signal tracks the real concentration. The fixed-price remote diagnosis tells you whether a computed concentration per cell can lead the feeding, and what it would take.

Background: Computing unmeasured states for control

Nobody knows how thick the side ledge is

What happens. The frozen bath on the side walls protects the lining. Too thin, and the side wall is attacked; too thick, and the bath volume shrinks. It is checked now and then with a rod and otherwise guessed from bath temperature and superheat.

What it costs. A lining that wears through ends the cell’s life early, in the worst case with a tap-out. Operators keep the cells on the cautious side, which costs energy and blocks any change in amperage.

Why the usual fix fails. More manual measurements cost staff time and show a few points at one moment. Side-wall thermocouples or shell temperature scans help, but they need a model to turn a temperature into a thickness.

How a fixed-price remote diagnosis checks it. From your shell and bath temperatures, superheat, amperage and voltage I compute the heat balance of the cells and the ledge it implies, as I once computed unmeasured states of the superfluid-helium cooling of the Large Hadron Collider at CERN, the European particle-physics laboratory. The fixed-price remote diagnosis shows whether such a computed ledge is reliable enough to guide operation.

Remelting furnaces are run by experience

What happens. The melter decides when to fire hard, when to stir and when the bath is hot enough, from experience and a dip thermocouple now and then. Charges differ in size, scrap type and moisture, so every melt behaves a little differently, and nobody sees the temperature of the solid charge while it melts.

What it costs. Overheating the bath to be safe burns gas and thickens the oxide skin and the dross. Melting too slowly holds up the casting line. Both depend on who is on shift.

Why the usual fix fails. A fixed burner programme is written for the heaviest charge and wastes gas on light ones. Extra thermocouples measure the bath, not the charge, and suffer in the furnace atmosphere.

How a fixed-price remote diagnosis checks it. From your burner logs, flue gas temperatures, charge weights and bath temperature readings I estimate how much solid metal is left and how hot the bath really is during the melt. The fixed-price remote diagnosis tells you whether a computed melt state can guide the burners, and how much gas and dross that could save.

Alloy composition drifts with the scrap mix

What happens. Scrap arrives with a composition that is only roughly known: profiles, cuttings, used beverage cans, mixed post-consumer scrap. The charge is put together from a list and experience; the spectrometer sample after melting shows whether it hit the alloy window. Often it did not, and alloying elements or primary metal go in to correct it.

What it costs. Correction loops cost furnace time and gas, primary metal is expensive, and a melt that cannot be brought back into specification is downgraded. Many plants buy cleaner scrap than they would need, just to stay safe.

Why the usual fix fails. A spreadsheet charge calculation assumes fixed compositions for each scrap type. When the real scrap differs, the calculation does too, and the margins grow.

How a fixed-price remote diagnosis checks it. From your charge records, scrap categories and spectrometer results I estimate what each scrap type really contains and how much it varies. The fixed-price remote diagnosis shows whether an optimised charge plan, updated with every analysis, can hit the window with cheaper scrap, and what that would take.

Ingots crack at the start of the cast

What happens. In vertical casting the first part of the ingot is the critical one. Casting speed, water flow and metal level are ramped up by a recipe. When the starting block, the metal temperature or the alloy are slightly different, the butt curls or a start crack opens.

What it costs. A cracked ingot is scrap or needs the butt sawn off, a crack that spreads can abort the whole drop, and cracks cause safety incidents with water and molten metal.

Why the usual fix fails. The start recipe is adjusted after each bad cast and slowly becomes very cautious, which costs time on every drop. One recipe for several alloys and formats is always a compromise.

How a fixed-price remote diagnosis checks it. From your casting logs (speeds, water flows, metal levels, temperatures) and the record of cracked and good starts I check which conditions go with cracking. The fixed-price remote diagnosis tells you whether an optimised start-up path per alloy and format can reduce cracks, and what it would take.

Background: Optimal paths and setpoint plans

Strip cools along the coil and properties vary

What happens. During hot rolling the tail end of the strip waits longer than the head and cools down. The finishing temperature falls along the coil, and with it the grain structure and the mechanical properties. A pyrometer shows the surface temperature at one point, not the temperature profile through the coil.

What it costs. Properties that vary along the coil mean cut-offs at the tail, downgrades and complaints from customers who need uniform forming behaviour.

Why the usual fix fails. Rolling the whole coil hotter or faster for the cold end overheats the head. Fixed speed-up tables are set once for a product and do not follow changes in entry temperature, alloy or gauge.

How a fixed-price remote diagnosis checks it. From your mill logs (entry and finishing temperatures, speeds, coolant flows, gauges) I compute how the temperature develops along the coil and what part of it a speed or coolant plan could compensate. The fixed-price remote diagnosis shows whether a computed temperature can lead the compensation, and what that would take.

Press speed is held back by the exit temperature

What happens. The profile heats up during the stroke from the work of deformation. If the exit temperature gets too high, the surface tears or shows speed cracks. So the press runs at a speed that is safe for the hottest moment of the stroke, which is too slow for the rest.

What it costs. Throughput on every billet, and a press that is slower than it could be. When the exit temperature still overshoots, the profile is scrap.

Why the usual fix fails. The operator adjusts the ram speed by hand from the pyrometer, which reacts late. A fixed speed profile per die does not follow changes in billet temperature, alloy or die wear.

How a fixed-price remote diagnosis checks it. From your press logs (ram speed, pressure, billet and exit temperatures) and your scrap records I estimate how the exit temperature builds up during the stroke. The fixed-price remote diagnosis tells you whether predictive control of the ram speed, holding the exit temperature close to its limit, could raise the speed, and what it would take.

Background: Predictive control of process plants

Billets are heated by a fixed profile

What happens. The billet furnace heats every billet by the same programme. The press, however, does not always need the same temperature: dies, alloys and stops vary. When the press waits, billets soak too long; when it runs faster, they leave too cold. The core temperature of the billet is not measured.

What it costs. Gas for billets that were hotter than necessary, a lower press speed when billets are too hot, and pressure peaks or stops when they are too cold.

Why the usual fix fails. Raising the furnace setpoint gives a reserve but costs gas and press speed. Contact thermocouples at the exit show the surface, not the core or the taper along the billet.

How a fixed-price remote diagnosis checks it. From your furnace temperatures, gas consumption, billet throughput and press data I compute the temperature inside the billets as they leave the furnace. The fixed-price remote diagnosis shows whether a computed billet temperature can steer the furnace to the temperature the press actually needs, and what that would take.

Homogenising cycles are long and fixed

What happens. Homogenising furnaces run a fixed time-and-temperature programme per alloy, set for the heaviest load. Whether the coldest billet in the middle of the load reached its temperature hours ago or only just now is not known; a few thermocouples in test billets are the only measurement.

What it costs. Gas and furnace time for loads that were ready early, and a furnace that becomes the bottleneck between casting and extrusion.

Why the usual fix fails. Shortening the cycle by hand is a risk to the metallurgy, so it is rarely done. More test billets cost handling time and still show only a few places in the load.

How a fixed-price remote diagnosis checks it. From your furnace logs (air temperatures, fan speeds, burner power), load weights and test-billet temperatures I compute the metal temperature across the load. The fixed-price remote diagnosis tells you whether cycles can end when the coldest billet has had its time at temperature, and what it would take to run them that way.

Melt losses stay invisible until the month-end balance

What happens. Metal goes into the furnace and less comes out. The difference sits in the dross and is known only when the metal balance is drawn up, often weeks later. Which charges, scrap types or furnace practices caused the losses can no longer be traced.

What it costs. Lost metal is among the most expensive losses in a remelting plant. Without knowing where it occurs, nobody can act on it.

Why the usual fix fails. Weighing the dross per melt helps, but dross holds varying amounts of metal. Reviewing the month-end balance comes too late to change the practice that caused it.

How a fixed-price remote diagnosis checks it. From your charge weights, scrap categories, energy use, tapping weights and dross weights I check whether a mass and energy balance per melt can show where metal is lost. The fixed-price remote diagnosis tells you whether a computed loss per melt is reliable enough to steer charging and furnace practice, and what it would take.

Every furnace is run on its own

What happens. Melting furnaces, holding furnaces and homogenising furnaces each have their own burner control and their own operator. Holding furnaces stay hot while waiting for metal; several furnaces fire hard at the same time and drive the gas peak.

What it costs. Gas for furnaces that wait hot, peak charges and contract capacity sized for overlapping peaks, and time lost when one furnace waits for another.

Why the usual fix fails. A shift plan on paper sets the order but cannot follow delays. Each furnace controller optimises itself and knows nothing about the others.

How a fixed-price remote diagnosis checks it. From your furnace logs, gas metering and production records I check how much time and gas go into waiting and overlapping peaks. The fixed-price remote diagnosis shows whether scheduling the furnaces as one fleet would pay off, and what it would take.

Background: Economic predictive control

Fluoride and dust analysers fail and the emission record has gaps

What happens. The permit requires continuous measurement of fluoride and dust after the fume treatment. The analysers drift, need calibration and fail, often at the wrong moment. Then the record has gaps, and the environmental officer has to explain them.

What it costs. Compliance risk, staff time for maintenance and reports, and in the worst case an argument with the authority over data that is missing just when something went wrong.

Why the usual fix fails. A second analyser doubles the cost and shares many of the same weaknesses. Filling gaps by hand with estimates is hard to defend.

How a fixed-price remote diagnosis checks it. From your analyser history, scrubber data (alumina feed, fan flow, pressure drops) and potline data I check whether the emissions can be computed from process values well enough to flag drifting analysers and bridge short failures. The fixed-price remote diagnosis tells you how good such a computed value would be and what it would take; whether the authority accepts it is a separate question.

Background: Computed measurements

A new line inherits the old line's control

What happens. A new potline, casting line or press line gets the control concept of the old one, often with the old parameters copied across. But the new line is larger, faster or built differently. During commissioning the loops oscillate, the cells or furnaces run unstable, and the start takes much longer than planned.

What it costs. A long ramp-up with lower output, scrap and wear in the first months, and engineers tied up on site.

Why the usual fix fails. Tuning on site by trial and error is slow and stops when the line runs well enough, not well. The equipment supplier’s commissioning team leaves once the acceptance tests pass.

How a fixed-price remote diagnosis checks it. From the drawings, design data and, where it exists, the data of the old line I check which control loops and settings will not carry over and why. The fixed-price remote diagnosis tells you what the control concept of the new line needs before it starts up, and what it would take.

Casting and rolling are scheduled by hand

What happens. The casting shop and the rolling mill each have their own plan, put together by hand. Slabs wait, cool down and are reheated; at other times the mill waits for metal. When an order changes or a cast is delayed, the plan is redone on a whiteboard.

What it costs. Gas for reheating slabs that could have been rolled warmer, time in the reheating furnace, and capacity lost while one stage waits for the other.

Why the usual fix fails. Experienced planners do it well but cannot weigh every option in minutes. Planning software for orders often knows nothing about temperatures and furnace times.

How a fixed-price remote diagnosis checks it. From your production records, furnace logs and order data I check how much heat and time are lost between casting and rolling. The fixed-price remote diagnosis shows whether a joint schedule for both stages would pay off, and what it would take.

Background: Optimal paths and setpoint plans