Every furnace severity decision in an ethylene plant sets the operating envelope for everything downstream. Coil outlet temperature (COT) determines column loading in the fractionation train, compression duty, refrigeration demand, and how many days remain until the next decoke. With margin pressure persisting across the chemical sector, the difference between a plant that recovers margin and one that does not often comes down to how well it manages these dependencies between sections.
Severity, dilution steam and energy intensity carry direct financial weight for teams facing structural oversupply and tighter emissions constraints. Feed choices matter as much as furnace tuning, because they shift the best settings across every downstream unit.
The hot section sets the pace for the plant
The cracking furnace is the economic center of the plant. Hydrocarbon feed is preheated and cracked in the presence of dilution steam at roughly 750 to 900°C, with residence times measured in fractions of a second. The temperature and time chosen together determine the volume and composition of everything entering the downstream train.
That is not a single-variable decision. Higher COT lifts ethylene yield at the cost of more fuel. It also speeds coke deposition on the tube walls and pushes tube metal temperatures toward the metallurgical limits that end a furnace run. A sound furnace efficiency strategy treats yield, fuel cost, run length and tube life as one coupled trade-off.
The transfer line exchanger and quench section stop secondary reactions after cracking. In naphtha-fed crackers, the oil quench and primary fractionator that follow bring their own constraints, and fouling in the primary fractionator can cut into production and reliability. Severity decisions that push tubes toward their thermal limits also shrink the margin against thermal fatigue across the fired equipment.
Severity sets composition, and composition sets how hard the cracked gas compressor works, how much refrigeration the cold section needs, and how load is distributed through the separation train. No downstream unit can fully compensate for a poor severity decision made hours earlier.
Compression, treating and separation as one coupled system
After the hot section, the plant moves into compression, treating, drying and cryogenic separation. These units do not behave as isolated assets. Disturbances pass forward through the train and feed back to the furnaces through recycle loops, so chemical process optimization has to work across the train.
The cold section and fractionation train absorb the effects of upstream severity decisions. When cracked gas composition shifts, refrigeration demand and column feed compositions shift with it, and the disturbance carries from one column through the train. Operators often see the downstream effect before they can trace it back to the severity change that caused it.
Compression adds another feedback path. Cracked gas compressors run close to their surge margins, and their efficiency depends on the suction composition set by the quench and primary fractionator. Tightening those inlet conditions lowers compression energy and can relieve constraints in the cold train, but only if upstream severity and dilution steam targets move in the same direction.
From the demethanizer forward, the fractionation train is a coupled system. Columns feed each other, and every upstream disturbance passes slowly through the sequence, so stable control in one column shapes what happens in every column after it. The train is never at true steady state. Through the ethane recycle loop, fractionation performance also affects furnace feed composition, which means separation quality influences cracking yield. For optimization, the hot and cold sections behave as one system with a long time constant.
Performance drivers that cross section boundaries
Three drivers matter most, and all three span more than one section. Treating them as independent problems is usually where the gap between the best-run plants and the rest opens up.
Cracking severity as a dynamic variable
Severity is often managed as a fixed setpoint, but the best target shifts continuously as coke builds, tube skin temperatures rise, feed composition changes, downstream constraints tighten and relative product prices move. A few dollars per ton between ethylene and propylene can change whether severity should lean toward one or the other. Decoke triggers based on a single threshold miss the interaction between these variables and leave run-length value behind.
Energy intensity across the plant
Furnace decisions and cold section operation are tightly linked. Energy management in an ethylene plant is a trade-off problem, because electricity, fuel gas, steam and chilled water interact throughout the process. The IEA counts steam cracking among the most energy-intensive industrial processes, so minimizing total cost means managing those interactions across the plant.
Conservative constraint setting
When constraints are set too conservatively, the operating window shrinks and margin leaks. A standard real-time optimizer finds the best operating point inside the constraint boundaries but does not optimize the constraint values themselves. Engineers usually set those values with static margins, and relaxing them safely is where most debottlenecking work happens.
Why unit-level APC leaves plantwide value behind
Severity control is one of the most sophisticated advanced process control (APC) applications in an ethylene plant. Model predictive control manages charge rate, severity and steam ratio together, balancing pass flows to even out coking rates across the furnace. That unit-scope control remains valuable and should keep doing what it does well.
The plantwide gap comes from how APC models are built. Model predictive control relies on linear dynamic models identified over a narrow operating envelope. The gap between predicted and actual process response widens during feed transitions and late in the run, which is when optimization matters most. Steam-to-hydrocarbon ratio interacts nonlinearly with temperature and residence time, and linear models cannot represent those interaction terms.
The gap also crosses section boundaries. When the ethylene column reaches its capacity limit, the best response may be to shift yield toward propylene and adjust charge rates to maximize total plant value. That decision needs plantwide visibility that individual unit controllers do not have. Cold section controllers managing refrigeration load have no view of the severity decisions driving their demand.
Where a learned model closes the gap
Years of board-level pattern recognition stay essential, and a learned model is built to complement that experience. A model trained on the plant's own data captures the nonlinear response across severity, steam ratio, feed composition and coil age at the same time, which is the basis of a self-optimizing olefins plant.
Many plants begin in advisory mode. The model supports decisions, what-if analysis and consistency across shifts, while operators decide whether to act. As recommendations prove out and the operating boundaries are clear, the plant moves to closed loop, with operators setting the limits the controller works within and able to take over from the board.
A shared model of plant behavior also changes how operations, planning and engineering work together. LP targets built on linear assumptions can be compared against what the plant does. Decoke timing decisions can account for downstream column loading. Engineering proposals for severity changes can carry the full energy cost alongside the yield effect.
Common questions
How does a change in furnace severity affect downstream fractionation stability?
A severity change alters the volume and composition of cracked gas entering compression and separation. Higher severity raises refrigeration load and shifts column feed compositions through the fractionation sequence. Because the columns feed each other, the disturbance propagates through the train over several hours, so managing it needs control that spans the whole train.
Can a learned controller work alongside existing APC and DCS in an ethylene plant?
Yes. It sits on top of the existing APC and DCS. The model reads plant data from the historian and existing instrumentation, then writes setpoints through the control system operators already use. Advisory mode supports human and AI collaboration, so teams can evaluate recommendations before the controller is allowed to act on them.
Why is decoke scheduling a plantwide decision and not a furnace-level one?
Taking one furnace out of service shifts load to the remaining furnaces and changes throughput, yield and downstream column loading. The timing also affects energy use across the cold section and the on-stream factor across the furnace fleet. A single-furnace trigger based on COT or run hours alone misses those trade-offs.
How the Imubit Platform applies
Imubit applies Map, Model and Control to the ethylene plant as part of its work in chemicals and polymers. Map traces lost ethylene margin to the recurring decisions behind it, such as severity, dilution steam, decoke timing and constraint settings across the hot and cold sections. Model uses Deep Learning Process Control® (DLPC®) to learn the plant's nonlinear behavior from furnace to C2 splitter from its own history, guided by first principles. Control takes those decisions to closed loop, writing setpoints to the existing APC and DCS, inside limits set by plant engineers.
