Anyone who runs a steam cracker knows the math changes when the feed changes. Switching from naphtha to ethane can raise ethylene yield substantially, but it also sharply reduces coproduct revenue from propylene and butadiene. The price spread between naphtha and ethane can be wide, particularly when oil prices are high.
That spread has made feedstock switching a competitive necessity. Capturing it without losing yield during transitions remains one of the hardest process control problems in petrochemicals.
Why feedstock transitions cost more than the spreadsheet shows
A feedstock transition is an operating regime with its own duration and its own losses. The spreadsheet may show a lighter feed producing more ethylene than a heavier one. What it rarely captures is the margin surrendered while the unit moves from one optimum to the next.
In practice, many sites do not switch feedstocks with a clean step change. They blend in storage, ramp at the battery limits, or chase availability across pipeline batches and ship parcels. That creates a moving target for furnace severity, dilution steam and coil balancing. When the feed changes faster than the plant can validate it, crews choose stability first, and that yield giveaway can last longer than the planning model assumes.
Residence times in the pyrolysis section are measured in tenths of a second, so small deviations in feed properties show up immediately in selectivity. Coil outlet temperature, the steam-to-hydrocarbon ratio and firing distribution all need coordinated adjustment, and the sequence matters:
- If firing is pushed before dilution steam and coil balancing catch up, tube metal temperatures can climb quickly.
- If dilution steam moves too early, the convection section and downstream compression see swings that create their own constraints.
Coproduct ripple effects
The coproduct effect compounds the problem. Shifting to lighter feeds generally reduces propylene and C4 availability, which can pinch downstream units that count on those streams. A propylene shortfall can reduce load on a polymer unit, while a C4 shortfall can change extraction rates and inventory strategy for butadiene or raffinate.
Those effects show up as sitewide margin shifts, beyond furnace economics. Planning teams working from static LP models may not see the cascade until the transition is underway. Teams make better flexibility decisions when the transition plan is tied to routing, storage and downstream constraints. The value of flexibility depends on how well a plant manages feed variability, as much as on which feedstock is cheapest on paper.
Where conventional process control reaches its limits
Advanced process control (APC) gets stressed during feed transitions for a straightforward reason. The control model was built around a narrower operating envelope than the transition creates.
When feed composition changes mid-run, operators often see it in the divergence between what the APC predicts and what the analyzers report. APC models rely on linear dynamic relationships identified for a specific feedstock, and as composition shifts, the gap between model output and actual process behavior widens.
Steam cracking amplifies this because several relationships shift at once. Feed composition changes the furnace heat absorption profile, the cracking kinetics and the coking tendency. A coil outlet temperature move that produced a predictable ethylene gain on one feed can produce a different response on another, especially when dilution steam, coil pressure drop and firing distribution are also moving.
Measurement delays
Many key quality signals are delayed. Online gas chromatographs update on cycles, lab confirmation arrives later, and inferential estimators trained on a narrower feed range can drift in the wrong direction during a transition.
Plants compensate by detuning applications and running with larger safety margins. That protects equipment and product specifications, but it gives up the yield the APC was deployed to deliver.
A mismatch between linear control and nonlinear chemistry
The root cause runs deeper than calibration. Heavier feeds bring aromatics and higher molecular weight species that change cracking kinetics and coke formation rates. A model tuned for light feed cracking will mispredict behavior on heavier material by construction. Retuning parameters does not fix a linear model of nonlinear chemistry, which is why the industry is exploring learning approaches to the problem.
The mismatch erodes operator trust at the worst moment. When model predictions diverge from what the board shows, experienced operators rationally revert to manual control. Their pattern recognition is real and hard-won, but in manual the plant runs on judgment alone, without a model weighing every interacting variable.
Where coordination across teams breaks down during feed changes
Feedstock transitions are more often limited by coordination across teams than by any single unit. Operations sees real-time process data in the DCS. Engineering works with simulation tools and historical data. Planning uses scheduling software with limited real-time feeds. Each team decides on incomplete information, and transition decisions can still depend on manual handoffs because the critical context sits in separate tools.
The lag shows up first in constraint management. Operations may see early warnings such as rising coil pressure drop, a compressor approaching a limit, or a quality swing that calls for a severity pullback. Planning may still be working from assumed yields and a transition duration that does not match the control room. Engineering may be weighing longer-term constraints, such as fouling, run length risk or equipment duty limits, without a direct line to the hour-to-hour transition choices.
Severity changes during a transition also ripple beyond the furnace. A severity push that improves ethylene yield can add load on hydrogenation, alter refrigeration duty or change fractionation cut points. Those sequential dependencies, combined with disconnected information systems, create delays when speed matters most.
How a shared plant model closes the gap
Transitions go better when operations, planning and engineering work from one model of plant behavior. When planning can see how a proposed feed change affects severity requirements and transition time in near real time, and engineering can check the implications before the transition starts, the coordination lag shrinks.
A shared model also changes the tone of the conversation. Instead of debating whose data is right, teams align around the same predicted trade-offs, and that alignment can matter as much as the optimization itself.
Managing severity dynamically through the transition
The central operating decision in any feedstock transition is severity management under uncertainty. Higher coil outlet temperatures usually increase olefin yields but accelerate tube coking and shorten run length. Pulling severity back protects run length at the cost of immediate yield.
Transitions sharpen this trade-off because the coking tendency of the new feed, and the actual cleanliness of the heater and convection section, are not known with confidence when the switch begins. Conventional practice resolves it with fixed conservative assumptions. A site may carry a uniform safety margin on tube metal temperature or pressure drop, because pushing into an unknown coking regime carries real metallurgical risk. That is how the best achievable severity in steady operation becomes safe but suboptimal during a transition.
Learning coking behavior under changing feeds
A model trained on plant operating data does not have to assume a static coking rate. It can learn how coking proxies move under specific feed blends and furnace conditions, then hold severity closer to the economic optimum for longer stretches of the run.
During a feed change, live tube metal temperature modeling can keep the furnace within safe metallurgical limits while narrowing the gap to the optimum. How fast the optimization adapts to the new feed properties determines how much yield the cracker recovers during the transition. A model that learns continuously from operating data can detect when the plant's response no longer matches the prior feed assumption and recommend coordinated moves across severity, dilution steam and constraints, instead of waiting for periodic retuning.
Guardrails and authority
The model sees only what the instrumentation sees. If key sensors are unreliable, or if disturbances go unmeasured, the optimization needs tighter guardrails and more operator judgment. Implementations that hold up in operations treat the model as a decision tool inside an operating envelope. Metallurgical limits and process safety limits stay where they are.
Plants usually start in advisory mode, where the model recommends setpoint changes and operators decide what to implement and when. Teams get what-if analysis when constraints conflict, such as yield against run length or ethylene against coproducts, and can rehearse a feed transition before making it live. Each crew starts from the same recommended operating point with the same explanation of the trade-offs, which can reduce variation from shift to shift.
Over time, crews see which recommendations hold up across shifts and feed blends, and sites that move to closed-loop control carry the same boundaries and learnings forward.
How Imubit applies this
Imubit applies Map, Model and Control to steam crackers in chemicals and polymers. Map traces transition losses to the recurring decisions behind them, such as severity ramps, the sequence of dilution steam and firing moves, and the yield assumptions planning works from. Model uses Deep Learning Process Control® (DLPC®) to learn a nonlinear dynamic model of the cracker from its own history, guided by first principles, so the effect of a feed change can be seen before it is made. Control takes those decisions to closed loop, writing setpoints to the existing APC or DCS minute by minute, inside limits set by plant engineers.
