7.1 Feedforward, Cascade, Ratio, Override, and Split-Range

7.1 Feedforward, Cascade, Ratio, Override, and Split-Range

Specification 2.B on the PE Control Systems exam goes well past a single PID on a transmitter and a valve. After you can close basic feedback, the remaining analog strategies are about where the disturbance is measured, which loop is allowed to be slow, and which controller is allowed to move the valve when two objectives collide. Feedforward, cascade, ratio, override (selective) control, split-range, and gap control are that toolkit. None of them retires feedback; they layer on it.

Feedforward: act on the disturbance you can measure

Feedforward computes a corrective output from a measured disturbance so the manipulated variable moves before the controlled variable has fully felt the load. The ideal compensator is the negative of the disturbance-to-PV path divided by the MV-to-PV path. In hardware you almost never get a clean inverse, so the practical form is a static gain plus lead/lag: the gain sets the steady-state MV change per unit disturbance (approximately minus the disturbance gain over the process gain), lag keeps the kick from slamming the valve, and lead recovers lag that lives in the process path but not in the disturbance path.

Feedforward is not a substitute for feedback. Unmeasured loads, model error, bias on the disturbance measurement, and every setpoint change still need a feedback controller. The usual implementation adds the feedforward signal to the feedback output, or to the setpoint of a secondary flow loop. If you put the feedback PID in manual because feedforward is in, the loop walks off on the first unmeasured disturbance — an exam-sized mistake.

Cascade: the inner loop must be the fast one

Cascade uses the primary (outer) controller to write the setpoint of a secondary (inner) controller. The inner loop is almost always flow or valve position (sometimes a fast pressure). It exists to reject valve stiction, actuator supply-air changes, and upstream header disturbances before they integrate into the slow quality variable.

The secondary must be substantially faster than the primary. A common design check is an inner closed-loop response several times faster than the outer — often about three to five times, or a secondary time constant much smaller than the primary. If that speed ratio is missing, the two PIDs fight and cascade is worse than single-loop.

Exam trap: putting a slow temperature, or a gas chromatograph, in the inner loop. Temperature is usually the outer quality loop. An inner temperature loop around a still-slower composition loop can be valid in a triple cascade (analyzer, temperature, flow), but the innermost loop is still flow or valve position. Never invert the nest.

Valve-position control as an inner loop shows up when the flow meter is missing or poor: the primary writes a position setpoint to a positioner. The positioner is already a cascade inside the instrument; you are just exposing it to the DCS.

Ratio: wild flow, controlled flow, limiter

Ratio control keeps controlled flow over wild flow equal to a setpoint ratio, or writes controlled flow as ratio times wild flow. The wild stream is the one you do not set — an upstream process flow, a committed blend stock, sometimes a fuel header. The controlled stream is the one whose flow controller gets that product as setpoint. A ratio limiter clamps the ratio so the mix cannot enter an unsafe region: fuel-rich firing, off-spec blend, or starving a reagent.

Protect the ratio station at low wild flow. Division by a near-zero wild measurement produces nonsense setpoints and can wind up the controlled-flow PID. Use a low-flow cutoff or freeze the last good ratio output.

Override and selective control

Override (constraint) control lets a limit controller take the valve away from the normal controller when a constraint is approached. Analog high-select and low-select blocks choose the more open or more closed demand.

• Low-select on a fuel or feed valve: the tighter of the quality controller and a high-pressure or high-temperature constraint wins, so the valve can only close further.
• High-select on a recycle or coolant valve: the more open of capacity control and a protection controller wins.

The controller that is not selected will saturate and wind up unless it tracks the selected output (external reset / anti-windup). When the constraint clears, a wound-up PID dumps a slug of output and can slam the process back into the constraint. On this exam, override without anti-windup is an incomplete answer.

Split-range, overlap, gap, and gap control

Split-range maps one controller output onto two final elements: typically 0–50% to valve A and 50–100% to valve B, or 4–12 mA and 12–20 mA. Heating versus cooling on a jacket, steam versus vent on pressure, and two-valve capacity are the usual pictures.

Overlap keeps both valves slightly open across the split so there is no dead zone — at the cost of fighting (steam and coolant both on). Gap at the split keeps both closed in a band around 50% so they do not fight — at the cost of a dead zone in regulation.

Fail positions are a pair, not two isolated fail arrows. Exam trap: steam fail-closed and coolant fail-closed. After instrument-air or power failure you have no energy path — neither heat nor cooling. For an exothermic system the coolant usually fails open, or a separate safety dump exists, and steam fails closed. Design the pair so the safe utility remains available.

Gap control is a different tool: a deadband around setpoint where the controller does not move the manipulated variable. Averaging level in a surge tank and some noisy pH loops use it so inventory absorbs swings instead of passing every ripple downstream. Do not confuse gap control with split-range gap.

Worked: fired-heater fuel/air ratio plus cascade on fuel flow

A process heater outlet temperature is slow (coil holdup, brickwork). Fuel-header pressure moves in seconds. Single-loop temperature control to the fuel valve lets every header bump hit the flame, then the coil, then the outlet transmitter.

Cascade: the temperature controller (primary) writes fuel-flow setpoint. The fuel flow controller (secondary) sees the fuel flow transmitter and drives the fuel valve. Header pressure is a disturbance inside the flow loop and is rejected there. The inner loop is fast; the outer loop remains temperature.

Ratio: air-flow setpoint equals limited ratio times measured fuel flow, with the ratio clamped so the mixture cannot go richer than the design maximum. On many process heaters air is the controlled stream following fuel. Cross-limiting (still ratio plus override) raises air before fuel on a load increase and cuts fuel before air on a decrease. That is selector logic, not a second temperature PID.

Optional feedforward from feed rate can be added to the fuel-flow setpoint so a feed increase immediately asks for more fire. The temperature controller stays in automatic. Feedforward does not replace the cascade.

Worked: split-range steam versus coolant

A jacketed-vessel temperature controller uses 0–50% for steam and 50–100% for coolant. As output falls from 50% toward 0%, the steam valve opens (more heat). As output rises from 50% toward 100%, the coolant valve opens. A small gap around 50% keeps both seated so the jacket is not simultaneously heated and cooled.

Steam fails closed. Coolant fails open if the chemistry can run away. If both fail closed, an air failure leaves the batch with no cooling path — the PE trap. Overlap would waste utilities; a wide gap would let temperature wander through the split.