The sulfur recovery unit (SRU) rarely draws attention until it becomes the constraint on everything upstream. When the SRU reaches capacity or underperforms, the refinery cannot run heavier, cheaper sour crudes, and the margin those crudes offer is lost. Refiners that manage SRU constraints together with crude economics and overall operating capability can recover margin that would otherwise go uncaptured.
The Claus process sits at the center of that equation. It is where upstream decisions on crude selection, hydrotreating severity and product specifications meet a single thermodynamic and regulatory constraint.
How Claus chemistry limits recovery
Two acid gas streams meet at the SRU inlet. Amine acid gas, from the amine systems that treat hydrotreater and hydrocracker recycle gas, carries mainly H₂S and CO₂. Sour water stripper gas adds H₂S, ammonia and water vapor.
In the thermal section, about one-third of the incoming H₂S burns with air to form SO₂. EPA's emission factor guidance for sulfur recovery puts combustion chamber temperatures at 980 to 1,540 °C. Part of the SO₂ reacts with the remaining H₂S in the furnace itself. Hot gas from the combustion chamber is quenched in a waste heat boiler that raises medium- to high-pressure steam, then enters the first sulfur condenser, where EPA puts recovery at about 65 to 70% of the sulfur. The rest of the reaction takes place over the catalyst beds.
Combustion air is trimmed to hold a 2:1 molar ratio of H₂S to SO₂, usually measured by a tail gas analyzer. That stoichiometric target is the unit's main control variable, and it sets how close the unit runs to its best recovery.
The recovery ceiling and tail gas treatment
Side reactions form carbonyl sulfide and carbon disulfide (CS₂), and both must be hydrolyzed back to H₂S in the first catalytic converter. That is why the first bed runs at the highest temperature in the catalytic section. If first-bed temperature falls during turndown or a shift in feed composition, unconverted carbonyl sulfide and CS₂ pass to the tail gas and overall recovery falls.
Each catalytic stage recovers roughly two-thirds of the sulfur entering it. EPA gives typical recovery of 92 to 95% for a two-bed Claus plant, 95 to 96% for a three-bed plant and 96 to 97% for a four-bed plant, and notes that at normal operating temperatures and pressures the Claus reaction is thermodynamically limited to 97 to 98% recovery.
The gap between Claus-only recovery and current emission limits is why tail gas treatment exists. The tail gas treating unit (TGTU) closes that gap at a cost. When the Claus section underperforms, the TGTU takes on more sulfur compounds, its chemical consumption rises and its own operating margin narrows.
The equilibrium limit applies to every refinery configuration. Better control can narrow the gap between actual and theoretical recovery at each stage, but the chemistry decides how much sulfur ultimately reaches the tail gas.
How feed variability destabilizes recovery
Upstream operations largely set SRU feed composition, and that composition is rarely stable. As hydrotreating severity rises to meet tighter product sulfur specifications, the ratio of sour water stripper gas to amine acid gas shifts upward. Heavier opportunity crudes and more refractory feed streams add to the shift.
The volumetric effect matters more than most planning models allow for. A refinery can overload its SRU with no increase in sulfur throughput, by moving more of the acid gas load from the amine system to the sour water stripper. The shift changes the operating load and the combustion conditions in the thermal section, even when sulfur mass flow looks manageable in the planning model.
Ammonia and feed contamination
Ammonia adds a complication because furnace conditions and first-bed conversion are tightly linked. Higher furnace severity improves ammonia destruction and lowers the risk of ammonium salt deposition in downstream catalyst beds. Running the furnace hotter than necessary wastes energy and can stress the refractory lining. When temperature control slips in either direction, product quality and recovery both suffer.
Feed contamination affects reliability too. Changes in flame temperature and combustion quality can contribute to catalyst fouling and deactivation, even when the root cause is not obvious from bed temperature profiles alone. Operators who have run the same SRU through several turnaround cycles often develop a feel for these conditions, but that expertise does not transfer easily across shifts or to newer crew members.
How the SRU sets the crude slate ceiling
An SRU at maximum throughput cannot take a higher sour crude fraction, whatever the market is doing. It limits how much sulfur the whole refinery can handle while meeting stack emission limits. When sour-sweet crude differentials widen, every barrel of discount the SRU cannot support is refinery margin left uncaptured.
Regulation has narrowed the room for imprecise SRU control. Federal NSPS Subpart J sets a stack SO₂ limit of 250 ppmv, dry basis at zero percent excess air, for Claus plants controlled by an oxidation system or by a reduction system followed by incineration, and state permits often set limits tighter than the federal floor. Permit conditions apply during startups and restarts as well, so compliance risk peaks when control is hardest.
A restart after an unplanned SRU trip often means running at reduced throughput until the unit stabilizes, and the emission limits apply throughout that recovery.
Where planning models and operating reality diverge
The economics cut both ways. SRU overcapacity means carrying capital charges on idle assets. SRU undercapacity means lost heavy crude opportunities, forced throughput cuts and compliance-constrained restarts that can keep units down longer than the underlying mechanical issue warrants.
Planning teams face this tension every time the LP runs. The crude slate that looks best on margin may push the SRU closer to its limit than operations is comfortable managing, especially during feed transitions or late in the catalyst's life. Without a shared, current view of where that limit sits under present conditions, the plan and the plant diverge. Operations usually responds by running more conservatively than the LP suggests. The buffer protects compliance and leaves SRU capacity unused.
The early signs that the SRU is becoming the bottleneck are indirect. Room for feed changes shrinks, tolerance for control drift tightens, and there is less flexibility to absorb shifts in sour water stripper gas. The unit can limit crude processing before sulfur mass flow alone looks excessive. The practical signal is how close the unit runs to its operating envelope.
Why conventional control struggles in SRU service
PID controllers are single-variable and reactive. Model predictive control handles multivariable interactions but degrades as changes in feed composition shift its process gains. The Claus process presents the conditions where both struggle. Several variables interact at once, constraints move with feed composition, dead time adds to the difficulty, and the SRU depends on upstream units it does not control.
Field experience shows PID control producing oscillations in critical SRU variables when feed composition changes faster than the control response can follow. Because SRU feed changes more often than feed to steadier downstream units, advanced process control models can degrade faster than the maintenance cycle refreshes them. Model upkeep becomes nearly continuous, and the value of the original implementation erodes. SRU APC also often runs in a silo, with limited feedforward from the upstream units whose decisions set its feed.
What a model learned from plant history changes
A model learned from the unit's operating history captures the nonlinear dynamics and multi-unit interactions that linear models miss. Because it is trained on the unit's own process data, guided by first principles, it reflects how the unit responds across the feed quality, catalyst condition and equipment states the plant has run through.
No model can stand in for the pattern recognition that comes from years of running an SRU through upsets. When the model handles the multivariable coordination, experienced operators can focus their judgment on abnormal situations and the decisions a model cannot make.
From advisory mode to closed loop
Deployments that build trust start in advisory mode. The model recommends setpoint moves and operators compare them with their own judgment, a form of human-AI collaboration that builds confidence step by step. That matters in SRU service because feed shifts arrive through several variables at once.
As crude blends change, the model can show when feed composition trends are moving the unit toward a constraint before an alarm would fire. Newer operators see how experienced crews weigh competing constraints, and teams can test what-if scenarios before recommendations reach the control system.
Advisory mode is a step toward closed loop. There, the controller writes setpoints for air demand and bed temperatures through the existing APC or DCS as feed conditions shift, with operators able to take over at any time.
Planning against current SRU headroom
Coordination between functions changes too. Without shared visibility, SRU constraints usually surface after crude purchases are committed and severity targets are set. When planning, operations and engineering share one model of plant behavior:
- Planning can screen crude candidates against current SRU headroom before procurement.
- Operations can see the expected acid gas load from a proposed severity change.
- Engineering can tell whether a bottleneck is a control issue or a capacity limit.
Decisions that used to happen in sequence can happen in parallel, and the SRU becomes a constraint the refinery manages ahead of time instead of one it discovers after the fact.
How the Imubit Platform applies
In refining, Map traces lost sour crude capacity and compliance margin at the SRU to the recurring decisions behind them, such as air demand, bed temperature and severity targets. Model uses Deep Learning Process Control® (DLPC®) to learn the sulfur recovery system's nonlinear dynamics from its own history, across the thermal section, catalytic stages and tail gas treatment. Control takes those decisions to closed loop, writing setpoints to the existing APC and DCS, inside limits set by plant engineers.
