Reheat Furnace and Rolling Mill Optimization in Steel

Steel carries a heavy energy bill, and most of it is settled in the hot end. According to the IEA's Iron and Steel Technology Roadmap, the sector consumed 845 Mtoe in 2019, about a fifth of all industrial energy use, and it emits 2.6 Gt of CO2 a year, roughly 7% of energy-system emissions. The average crude-steel intensity sits near 19 GJ per tonne on the IEA's accounting. A good slice of that goes into heating solid steel back up so it can be shaped: reheating slabs, billets, and blooms to rolling temperature, then driving them through a mill.

This piece is about that hot end, narrowly. Not ironmaking, not the melt shop. The reheat furnace and the rolling line, where energy, yield, and quality trade against each other every minute, and where instrumentation earns its keep. The U.S. DOE's bandwidth study on iron and steel manufacturing puts current-typical hot rolling at about 3.0 GJ per tonne of product, against a state-of-the-art figure near 1.66 GJ per tonne for a hot strip mill. That gap is the prize. Almost all of it lives in the furnace, not the mill stands.

Where the heat actually goes

Start with a single number every furnace engineer already feels in their gut: available heat. It's the fraction of the fuel's energy that's left to do useful work after the flue gases carry their share out of the stack. A burner firing into a furnace running at 2,400°F doesn't get to keep most of what it burns. The combustion products leave near furnace temperature, and the hotter the furnace, the more energy walks out the exhaust. That single fact governs almost everything else in furnace optimization.

Process heating is where this energy concentrates. According to the EIA's steel industry analysis brief, process heating accounts for roughly 30% of the industry's fuel use even before counting the coke burned in the blast furnace, and the reheat furnace is a large part of that for any mill that rolls. So the levers are narrow and well understood. Burn with the right amount of air. Recover heat from the flue before it leaves. Don't reheat metal you didn't have to cool in the first place. Each lever has a number attached, and each number is measurable with sensors a plant can actually mount on a reheat furnace.

The hard part isn't knowing the levers. It's holding them at their setpoints across shift changes, fuel-quality drift, and a rolling schedule that never quite matches the plan. The DOE bandwidth study frames the headroom at sector scale: bringing existing plant up to the best technology and practice available represents, according to the study, roughly 197 TBtu a year of savings across the processes it covers, about 39% against the thermodynamic baseline, with further R&D-stage gains on top. Reheating alone takes a large share of rolling energy — the DOE work cites a range of 22% to 32% of rolling energy going into reheat. That's where to point the instruments first.

Combustion control: the cheapest tonne of fuel you'll ever save

Before you spend a cent on heat-recovery hardware, check the air. Most furnaces run with more excess air than they need, and every extra pound of air is mass you heat to furnace temperature and then throw away up the stack. Excess air shows up directly as oxygen in the flue gas, which is why a zirconia O2 probe in the exhaust duct is the single most useful sensor on the furnace. It reads the one variable that tells you, in real time, whether the burners are tuned or bleeding fuel.

The DOE process-heating guidance shows the size of this. Per its air-to-fuel ratio tip sheet, a furnace at 2,400°F flue-gas temperature running rich at 25% excess air sits at only 22% available heat. The rest leaves up the stack as heated air and combustion products, bought and wasted in the same breath.

Trim the burners toward the optimum, near 10% excess air, and available heat climbs to about 29%. According to the DOE example, that single ratio change saves roughly 24% of the furnace's fuel, not merely a quarter of the excess air, and the only cost is measurement and control. No new hardware, no outage, just an honest air-fuel ratio held where it belongs.

Why doesn't every furnace already run there? Because the optimum isn't a fixed setpoint. The right excess-air level drifts with firing rate, with burner wear, with damper position, and with the calorific value of the fuel (mixed coke-oven and blast-furnace gas wanders more than pipeline natural gas). A ratio that was clean at high fire goes rich or lean at turndown. So the practical answer is a control loop, not a one-time tune: meter air and fuel, trim against measured flue-gas oxygen, and log the result so drift is visible before it costs a week of fuel. That's an O2-trim loop, and it's table stakes for any furnace that matters.

Furnace pressure belongs in the same conversation. Run the furnace too far below atmospheric and cold air leaks in through doors, skids, and charge openings, chilling the load and forcing the burners to make up the difference. Run it too high and hot gas and flame puff out of every opening, wasting fuel and punishing the steelwork. A furnace held slightly positive at hearth level, with pressure measured and controlled, keeps infiltration down and combustion stable, and it protects the very flue-gas readings the O2 loop depends on. Air that leaks in past the burners corrupts the stack oxygen measurement and sends the trim loop chasing a number that isn't real.

Recovering the heat in the flue

After the air is right, the next tonne of fuel comes from the exhaust itself. The flue gas leaving a reheat furnace is the largest single loss on the heat balance, and the most direct way to claw some of it back is to use it to preheat the incoming combustion air. Warm air carries energy into the furnace that you'd otherwise buy as fuel, and it does so without touching the process at all.

The numbers are large enough to change a capital plan. The DOE's preheated combustion air tip sheet tabulates the savings: a furnace with 2,000°F flue gas that preheats its combustion air to 800°F cuts fuel use by about 26%, and pushing the preheat higher on a hotter furnace reaches into the 40s of percent. Two pieces of hardware do this. A recuperator is a gas-to-gas heat exchanger sitting on the stack, transferring heat from exhaust to incoming air through tube or plate walls. A regenerator stores heat in a packed medium that combustion air and flue gas pass through in turns, which is how regenerative burner pairs reach very high preheat temperatures by firing one side while the other recharges.

There's a caveat worth stating before anyone orders steel. Air preheat justifies itself on energy saved, not on temperature alone, and the DOE guidance puts processes above 1,600°F as good candidates while calling preheat hard to justify below 1,000°F. Dirty or scale-laden exhaust fouls and attacks heat exchangers, so a reheat furnace's particulate-heavy flue is a real engineering constraint, not a footnote. Recuperator metallurgy and cleaning access have to be designed for the gas you actually have, and a unit specified for clean air on a dirty stack will plug or corrode long before it pays back.

Hot charging: don't cool what you'll only reheat

The largest single energy move in the hot end isn't in the furnace at all. It's in scheduling. If a slab goes from the caster to the reheat furnace still hot, the furnace has far less work to do. The DOE bandwidth study's theoretical-minimum tables make the point cleanly: rolling a flat carbon slab charged cold, from around room temperature, needs on the order of 0.83 GJ per tonne of heat, while charging the same slab hot, near 1,170 K, drops the heat requirement to roughly 0.27 GJ per tonne. That's about a threefold cut in the reheat duty, won by not letting the steel cool in the first place.

Hot and direct charging are the reason caster and rolling schedules have to be treated as one optimization, not two. The benefit is real but conditional: it depends on slab availability matching mill demand in time, on temperature uniformity surviving the transfer, and on logistics that keep a hot slab moving. Miss the window and the slab cools, the furnace pays the full cold-charge bill, and the scheduling gain evaporates. This coupling is exactly what telemetry and a shared data model are good at: knowing slab temperature, position, and the mill's next demand at the same moment, so dispatch is decided on data rather than habit.

Yield rides along with this. Every minute a slab spends hot in an oxidizing atmosphere grows scale, and scale is steel you bought, oxidized, and will knock off as waste before rolling. The bandwidth study flags scale-free and low-oxidation reheating among the technologies that improve both energy and product yield, because the furnace atmosphere and residence time that minimize fuel also tend to minimize the metal lost to oxide. Shorter, better-controlled reheat is good for the fuel bill and the yield account at once. And a slab that comes out of the furnace at the right temperature, evenly soaked, rolls into fewer off-gauge metres at the mill, which is yield of a second kind.

Instrumenting the furnace: from setpoints to a model

None of these levers hold themselves. A reheat furnace is a distributed thermal system with multiple zones, each with its own burners, its own setpoint, and its own response time, heating a moving load whose mass and target temperature change with every product. The classic answer is zone temperature control with thermocouples and a basic combustion loop per zone. The better answer adds a model.

Model-based, or Level 2, furnace control tracks each piece through the furnace and predicts its internal temperature from a thermal model rather than from the furnace gas temperature alone. The target is discharge temperature and through-thickness uniformity, not zone air temperature. A slab can read hot on the surface and still be cold in the core, and the mill feels the difference as roll-force variation and off-gauge product. A model that knows residence time, charge temperature, and load geometry can set zone targets to hit the right discharge condition with the least fuel, and it can ramp down cleanly when the mill stalls instead of soaking steel at full temperature while it waits. That delay handling matters: an unplanned mill stop with the furnace held hot burns fuel and overcooks the steel nearest the discharge, costing both energy and metallurgy.

A model is only as good as the data feeding it, and that puts the instrumentation layer on the critical path. It takes ruggedized thermocouples and pyrometers that survive a furnace environment, flue-gas oxygen and pressure measurement, fuel and air flow metering, and drive and roll-force signals off the mill — plus a transport layer, typically OPC-UA or Modbus off the PLCs and Level 1 controllers, that carries all of it to a common, timestamped store. Pyrometry deserves particular care: a coated or sooted lens reads low, the model trusts it, and the furnace quietly overshoots to compensate. We build exactly this kind of edge telemetry and analytics layer in our work; the approach is described on our edge telemetry and analytics platform. Once furnace and mill data share a clock and a model, optimization stops being a quarterly audit and becomes a continuous loop.

A model also has to earn the operator's trust, or it gets switched to manual on the first odd shift and never switched back. That means validating it against discharge pyrometer readings and against what the mill actually sees in roll force, and showing the operator why it set a zone the way it did. A black box that lowers a zone setpoint during a slow run, with no visible reason, looks like a fault to the person on the pulpit. The same telemetry that feeds the model should feed the screen in front of the operator, so the model's decisions are legible and a drifting sensor is caught by a human before it quietly skews the optimization. Optimization that the crew doesn't believe is optimization that doesn't run.

The standards world has a name for that loop. ISO 50001:2018, the energy management systems standard, is built on a plan-do-check-act cycle around energy performance indicators and a measured baseline. For a reheat furnace, the natural indicator is gigajoules of fuel per tonne discharged, normalized for product mix and charge temperature, and the standard's discipline is what keeps a hard-won combustion tune from quietly drifting back over a year. The IEA roadmap is blunt about how much this ordinary, available work is worth: improving operational efficiency and adopting best available technology across a plant's units can cut energy by around 20% per tonne of crude steel, before any new process chemistry.

Where this breaks, and where to start

Honesty about limits keeps these projects from stalling. Sensors in a reheat furnace live hard; thermocouple drift and pyrometer fouling are constant, and a model fed by a drifting sensor will confidently optimize toward the wrong number. Heat-recovery hardware fails when the flue chemistry wasn't respected in the design. Hot charging collapses the moment caster and mill fall out of sync, which on a real shop floor is often. And combustion tuning isn't permanent: burners wear, dampers stick, fuel quality wanders, so an O2-trim loop without logging and review reverts to where it started. None of these levers is a one-time fix. They're things you hold.

So where do you start? Measure before you spend. A furnace with a working flue-gas oxygen measurement, honest fuel and air metering, and a logged energy performance indicator already has the data to find its own cheapest savings, and combustion tuning usually pays back before any heat-recovery capital is even quoted. From there the order tends to follow the heat balance: get the air-fuel ratio under closed-loop control, hold furnace pressure, recover flue-gas heat where the gas chemistry allows, couple the caster and furnace schedules to capture hot charging, and only then layer a thermal model over a furnace whose inputs you trust. Each step is grounded in a number you can measure on your own furnace, which is the only kind of optimization worth committing to. Done in that order, every stage funds the next, and the furnace ends up cheaper to run and steadier on quality than the day you started.