What Is Cascade Control?
One Loop Setting Another's Target

Back to Blog

Primary and Secondary Loops

A cascade scheme has two controllers. The primary (master) loop controls the variable you actually care about - say, the temperature of oil leaving a heater. Instead of driving a valve directly, its output is a setpoint. The secondary (slave) loop controls a faster intermediate variable - the fuel-gas flow or pressure to the burner - and it accepts that setpoint from the primary and drives the final valve.

The key requirement is speed: the inner loop must respond several times faster than the outer loop. Because the inner loop is watching fuel flow directly, it catches a disturbance - a dip in fuel-gas supply pressure - and corrects it in seconds, long before that disturbance would have shown up as a temperature error the slower outer loop could see.

Why and Where Cascade Is Used

Cascade control shines when there is a measurable intermediate variable between the final element and the primary variable, and when disturbances enter through that intermediate variable. Common oil-and-gas examples include heater and heater-treater temperature (temperature-to-fuel cascade), compressor and pump discharge control, and level-to-flow cascades on separators and surge drums where the level loop sets an outflow-rate target.

The payoff is faster, tighter disturbance rejection than a single loop can achieve. The cost is added complexity: two controllers to tune, in the right order (inner loop first, then outer), and the need to handle the inner loop going to manual or hitting a limit so the outer loop does not wind up. When the intermediate variable is well chosen, that added effort is well worth it.

A Worked Heater Cascade, Step by Step

Take the classic heater example with symbolic numbers. The outer loop compares outlet temperature T against its setpoint T_sp and produces an output between 0 and 100 percent. That output is scaled into a fuel-flow setpoint F_sp between the burner's minimum and maximum firing rates, F_min and F_max. The inner flow controller compares measured fuel flow F against F_sp and moves the fuel valve. In steady state, T sits at T_sp, the outer output is steady, F equals F_sp, and the valve rests wherever the current supply pressure requires.

Now let the fuel supply pressure dip. With the valve position unchanged, flow through it falls, so F drops below F_sp. The inner loop sees that error within its own fast cycle and opens the valve until F is restored - the burner never sees a meaningful loss of fuel, and T barely moves. Without the cascade, the same dip would starve the burner at constant valve position, T would sag after the process deadtime, and only then would a single temperature loop start opening the valve - a correction that arrives one full process lag too late and often overshoots on the way back.

The same walk-through explains why the outer output is a setpoint rather than a valve signal: the outer loop is deliberately kept ignorant of supply pressure, valve character, and everything else the inner loop absorbs. Its whole job reduces to one question - how much fuel does the process need right now - while the inner loop answers a different one - how to deliver exactly that flow with this valve at this pressure.

Tuning Order and Loop Speed Separation

Cascade tuning has a fixed order: inner loop first, always. Put the outer controller in manual, give the inner loop setpoint steps, and tune it for fast, well-damped tracking - the inner loop's job is to be a crisp actuator, not to be clever. Only when the inner loop tracks its setpoint cleanly do you close the outer loop and tune it, treating the tuned inner loop as part of the process it sees. General loop tuning practice applies to each loop individually; the cascade adds only the ordering rule.

The speed separation between the loops is the design constraint. If the inner loop cannot respond clearly faster than the outer - because its measurement is slow, its valve is sticky, or the intermediate process is itself sluggish - the two controllers start fighting over the same dynamics, and the cascade performs worse than a well-tuned single loop. When you cannot get clean separation, the honest fix is to abandon the cascade rather than detune both loops into sluggishness. How much separation is enough is a judgment call that depends on the process; the symptom of too little is sustained cycling that appears only when both loops are in automatic.

Initialization, Windup, and Mode Handling

Many cascade problems in the field are mode-handling problems, not tuning problems. When the operator takes the inner loop to local automatic or manual, the outer controller is commanding a setpoint nobody is listening to - it must stop integrating and instead track the inner loop's actual setpoint, so that restoring cascade mode is a bumpless transfer rather than a lurch. The same applies when the inner loop saturates: with the valve wide open, more outer-loop output cannot produce more flow, and an outer controller that keeps integrating will wind up and overshoot badly once the constraint clears. Controllers handle both cases with setpoint tracking and limit feedback from inner to outer - but only if those signals are actually configured.

Watch the inner loop's health as an early warning. An inner setpoint pinned at its high or low limit means the cascade has run out of authority - the process is asking for more than the fuel train, pump, or valve can deliver - and that is worth an alarm long before the primary variable drifts. A cycling inner loop usually means valve stiction, and the cascade will faithfully propagate that cycle upward if it is not fixed at the valve.

When Cascade Is Not the Answer

Cascade only helps against disturbances that enter through the intermediate variable. If the dominant upset comes from somewhere else - feed temperature or throughput changes in the heater example - the inner fuel loop never sees it, and the cascade behaves exactly like a single loop. For measurable upstream disturbances of that kind, feedforward control is the complementary tool, and real heaters frequently run cascade and feedforward together: cascade to reject supply-side disturbances, feedforward to preempt demand-side ones.

Skip the cascade, too, when there is no trustworthy intermediate measurement - a flow signal that is noisy or unreliable at low rates makes a poor inner variable - and when the maintenance reality of the site cannot support two tuned controllers per loop. A cascade nobody understands gets put in manual at the first upset and stays there; a simpler scheme that operators trust and use beats a sophisticated one they bypass.

Frequently Asked Questions