Delayed coking is one of the hardest trade-offs to run day to day. Every hour taken out of the drum cycle adds throughput, and it also adds thermal stress to equipment that is expensive to replace. The operating question is how to push capacity and yield without trading away drum life.
The coker converts the lowest-value stream in the barrel into liquid products that feed the FCC unit and the hydrocracker. Its yield structure drives refinery-wide profitability, and the gap between average and top-quartile coker performance is well documented.
Closing that gap starts with knowing where margin sits in the drum cycle, which yield levers matter at a given site, and how to apply them consistently across shifts.
How drum cycle phases shape coker economics
Cycle time sets coker capacity directly. On a two-drum unit running 16-hour coking cycles, the pair fills a drum every 16 hours, about 548 drums a year (8,760 hours divided by 16). Stretch the cycle to 18 hours and capacity falls to about 89% of the baseline (16 divided by 18). The arithmetic is simple, but holding a shorter cycle depends on what happens in each phase of it.
A delayed coker is a semi-batch process with six phases: online coking, drum switching and steam stripping, water quenching, draining, coke cutting and warm-up. Each phase has its own constraints and failure modes. The bottleneck is usually one or two offline steps that expand slowly over time.
During online coking, the main economic drivers are heater outlet temperature, drum pressure, coil velocity steam and feed rate. Feed rate can be limited by heater duty, fractionator stability or wet gas compressor loading. Heater fouling moves the practical limit over the run, so the operating envelope on day one of a heater cycle is different from the one on day ninety.
Water quenching sharpens the trade-off between throughput and drum integrity. The largest thermal stresses occur in this phase, and the quench rate determines how much fatigue builds up per cycle. The pattern of temperature gradients through the shell, skirt and nozzle regions matters more than any single temperature reading. Poor draining or uneven internal wetting can leave hot spots that complicate cutting.
Cycle studies at well-run sites often find that a few repeatable deviations account for most of the lost capacity:
- slow drains after foaming episodes
- extra quench soak time added without a clear trigger
- switch sequences modified over the years and never re-optimized
Better analytics and tighter control can recover that time without treating every cycle as a special case.
Yield levers that move margin
Inaccurate coker yield estimates in the refinery LP model cost gross refining margin, because they misalign secondary unit loading and crude processing decisions. The levers to recover that margin are well established but unevenly applied, mostly because each one touches several constraints.
Oil partial pressure
Lowering oil partial pressure is widely reported to have the largest single-parameter effect on yield. At constant heater outlet temperature, a lower oil partial pressure lets more of the heavy material leave the drum as vapor before it cokes, which shifts yield from coke toward heavy gas oil.
In practice, oil partial pressure comes down by reducing drum pressure, increasing steam dilution, or both. Each route has knock-on effects that show up at the next drum switch. Lower pressure raises volumetric vapor traffic, which can load the fractionator overhead and the wet gas compressor. More dilution steam changes fractionation behavior and can move the constraint to sour water handling.
Recycle ratio and coking temperature
Recycle ratio is the most accessible lever. A lower recycle ratio reduces coke make and can raise liquid yield by several weight percent, depending on feed and unit constraints. It requires no capital, though operations has to balance it against heater capacity and coking rate. The best recycle ratio is usually a band rather than a single number. It depends on the heater fouling state, the position in the drum cycle and downstream capacity.
Coking temperature offers incremental gains. A higher temperature lowers coke yield and raises liquid recovery, but it can accelerate fouling and shorten heater run length. Well-run sites treat temperature as a trajectory over the heater run rather than a fixed setpoint.
Feed quality
Feed quality underlies every other lever. Higher Conradson carbon residue means more coke and fewer liquids. Higher metals downgrade the coke. Higher sulfur adds to downstream hydrotreating demand.
Feed quality is also rarely steady through a cycle. Tank turnover, blending variability and swings in upstream units can change coker feed mid-cycle, and yield optimization that ignores these shifts usually ends up as conservative operation. Operational tuning cannot fully make up for a poor feed mix, so aligning refinery operations and crude selection with what the coker can handle matters alongside running the levers well.
Why cycle and yield optimization resist conventional control
Cycle management and yield optimization look like separate problems on paper, but in operation they interact constantly. Lowering drum pressure to improve liquid yield raises vapor flow at the next drum switch, which makes the fractionator harder to stabilize. Shortening the cycle to gain throughput narrows the window for adjusting quench when feed quality shifts. Heater fouling steadily erodes the temperature headroom the yield levers depend on, and it changes the thermal profile that drum integrity requires.
Every yield lever adjustment propagates into cycle dynamics, and every cycle decision constrains the yield levers. Optimizing one axis in isolation usually leaves margin on the other.
Where linear APC reaches its limits
Standard advanced process control handles continuous, steady-state optimization well, but a delayed coker is semi-batch. Every drum switch sends a cascade of disturbances through the fractionator, and many of the critical decisions during switching are discrete actions tied to permissives, interlocks and operator timing rather than classic control loops. Physical limits such as minimum pumparound flow force manual intervention at the moment coordinated optimization would add the most value.
Product quality data adds to the difficulty. Lab results arrive hours later, so control relies on tray temperatures as proxies, and operators run away from specification limits to avoid off-spec product. That giveaway is lost margin, and it widens whenever an operator is unsure where feed quality, heater condition and cycle position have moved the true constraint.
Linear models cannot track the time-varying interaction between heater fouling, cycle phase and lever positions. The result is either frequent model retuning, which needs specialist engineers who are already scarce, or an accepted loss of throughput.
What a model learned from cycle data adds
A model trained on a site's own historical cycle data can anticipate drum switch transitions and prepare a coordinated response before the fractionator destabilizes, adjusting reflux, pumparound targets and heater firing as the switch sequence begins. Because it learned from the plant's operating history rather than idealized equations, it reflects site-specific constraints that generic models miss. It also follows heater fouling as it progresses, instead of relying on linear relationships that degrade between retuning intervals.
Real-time quality estimation
Models built from plant data can estimate product quality in real time, which shortens the slow lab feedback loop that keeps operators away from spec limits. When operators can see where the true constraint sits at any point in the heater run, giveaway shrinks without raising off-spec risk.
Cycle-level asset integrity
Industry experience shows that a small fraction of operating cycles can cause a disproportionate share of fatigue damage at critical locations. That changes the throughput and integrity trade-off. Instead of conservative practice on every cycle, teams can make targeted adjustments based on cycle-specific damage signals, separating fast cycles that are safe from cycles that create high stress with no economic benefit.
That distinction depends on cycle operations, yield performance and equipment integrity data feeding the same model, instead of sitting in separate systems nobody has time to cross-reference during a shift.
Closing the gap to top quartile
Top-quartile coker operations outperform the industry average on availability, heater run length and maintenance cost. The difference usually comes from better transition control, tighter constraint management and more consistent execution from shift to shift.
How Imubit applies this
Imubit applies Map, Model and Control to delayed cokers in refining. Map traces lost capacity and yield to the recurring decisions behind them. Model uses Deep Learning Process Control® (DLPC®) to learn a nonlinear dynamic model of the heater, drums and fractionator from the unit's own history, guided by first principles. Control takes those decisions to closed loop across the drum cycle, writing setpoints to the existing APC or DCS, inside limits set by plant engineers. Plants can start in advisory mode and move to closed loop as confidence builds.
