11.2 Feedback Control and PID

11.2 Feedback Control and PID

Quick Answer: Feedback loop: measure PV → compare to SP → error → controller → MV/OP → valve/actuator → process. P reacts now; I kills offset; D damps fast moves (watch noise). Choose fail-open or fail-closed so loss of air/signal leaves the plant safer.

Section 11.1 delivered measurements. Domain E next asks what the plant does with those measurements: automatic regulatory control. On Qatar process units, thousands of loops hold levels, temperatures, pressures, and flows. The UPDA/MMUP Chemical exam expects loop structure, PID roles, and fail-safe valve thinking—not full tuning lab skill.

Feedback Loop Elements

A classical single-loop feedback structure:

1. Process — the equipment and fluid inventory that produce the variable of interest
2. Sensor / transmitter — measures the process variable (PV)
3. Controller — compares PV to setpoint (SP) and computes an output
4. Final control element — usually a control valve (sometimes a variable-speed drive, heater power, etc.) that changes the manipulated variable

| Signal name | Meaning | |---|---|---| | SP | Setpoint — desired value (operator or master cascade) | | PV | Process variable — measured value | | Error e | Difference between SP and PV (sign convention must be consistent with action) | | MV / OP | Manipulated variable / controller output — command to the final element | | CO | Sometimes used for controller output (synonym territory with OP) |

Information flow: disturbance hits the process → PV moves → controller sees error → output changes → valve moves → process responds → PV returns toward SP.

Feedback vs feedforward (awareness): Feedback corrects based on measured PV error after the disturbance affects the process. Feedforward measures a disturbance and acts before PV moves. Licensing stems focus on feedback; feedforward is a bonus concept.

Worked Narrative: Level Control on a Separator

• PV: liquid level from a DP transmitter
• SP: 50% level
• Final element: outlet control valve
• If feed surge raises level (PV > SP), the controller opens the outlet valve (direct or reverse action configured so the valve move is correct) to drain inventory until level returns

If the level transmitter freezes, the loop may drive the valve based on a false PV—measurement integrity is part of control quality (link back to 11.1 and forward to alarms/SIS in 11.3).

Setpoint, Error, PV, and MV/OP

Setpoint is the target. Operators change SP for production modes; advanced controls or cascades may write SP automatically.

Error is typically:

e = SP − PV

If SP = 100 °C and PV = 95 °C, e = +5 °C (need more heating if the loop is temperature with a heater/steam valve configured accordingly).

Some vendors use PV − SP; controller action (direct vs reverse) absorbs the sign so that increasing error drives the valve the correct way. Exam focus: error is the difference that drives corrective action, and PV is measured, not assumed equal to SP during transients.

Manipulated variable is what the controller is allowed to change—fuel gas flow, coolant rate, reflux, etc. The controller output is often 0–100% corresponding to valve position demand or a signal to an I/P converter (current-to-pneumatic) that sets air pressure on a diaphragm actuator.

Proportional, Integral, and Derivative Actions

A PID controller blends three contributions. Exact industrial equations vary (ideal, series, parallel forms; dependent vs independent gains). For UPDA, master what each term does.

Proportional (P)

P action: output contribution proportional to present error.

• Larger error → larger immediate corrective move
• Gain K_c (or proportional band PB = 100%/K_c in some conventions) sets aggressiveness
• Alone, P control often leaves offset: a sustained load may need a sustained valve position that pure P only produces with a permanent nonzero error

Intuition: P is the “present” term—reacts now, but may not finish the job to zero error under load.

Integral (I)

I action: output contribution grows with the accumulation of error over time (integral of e).

• As long as error persists, I keeps winding the output
• Eliminates steady-state offset for many common process types (the classic reason to add I)
• Too much I → overshoot, oscillation, integral windup if the valve saturates (output stuck at 0% or 100% while error still integrates)

Intuition: I is the “past” term—remembers that we have been wrong for a while and keeps pushing until error is gone.

Derivative (D)

D action: output contribution related to the rate of change of error (or of PV in many implementations).

• Anticipates where the error is heading; can reduce overshoot on sluggish loops
• Amplifies noise on noisy PVs (flow is often noisy)—D is used carefully or not at all on some loops
• Derivative kick: a sudden SP step can spike D action if D acts on error; many controllers apply D to PV only to reduce kick on setpoint changes

Intuition: D is the “future/trend” term—slows a rapid dive or climb, but is unkind to noisy signals.

PID Combinations (Exam Language)

Tuning intuition (not full Ziegler–Nichols procedure): increasing gain speeds response but can destabilize; more integral removes offset faster but can oscillate; more derivative can calm overshoot if noise allows. Exam stems ask which term removes offset (I) or which is sensitive to noise (D).

Worked Concept: Offset with P-Only

A cooler must reject more heat after a feed temperature rise. At the new steady state the coolant valve must sit more open. With pure P control, that extra opening only appears if a permanent temperature error remains (valve position tied to error). Adding integral action slowly opens the valve further until temperature error returns to zero—at which point I holds the new bias in the output.

Final Element: The Control Valve

Most chemical-plant loops end at a globe or other throttling valve with an actuator (often pneumatic).

Fail-Open vs Fail-Closed (Safety Choice Intro)

On loss of instrument air, electrical signal, or actuator power, the valve should go to a predetermined fail position.

Rule: Choose the fail position that moves the process toward a safer state for the credible failure (not necessarily “closed is always safe”). Cooling and quenching services often fail open; fuel, feed to hazards, and isolation services often fail closed. Document the choice in design and HAZOP (Chapter 12).

Air-to-open valve typically fails closed on air loss if spring-return is so arranged; air-to-close often fails open—confirm spring and bench set rather than memorizing brand slogans alone. Exam cares about process-safe direction, not only the pneumatic nickname.

Cascade and Other Structures (Light Awareness)

• Cascade: master controller sets SP of a slave loop (e.g., temperature master → flow slave) for faster disturbance rejection
• Ratio: maintains F_A / F_B for blending or burner air–fuel
• Override / selectors: constrain a loop when a constraint PV hits a limit

These appear as vocabulary; single-loop PID remains the core.

UPDA Chemical Exam Workflow

1. Identify PV, SP, final element, and the disturbance in the stem.
2. State whether the question is about structure, P/I/D role, or fail-safe.
3. Map “offset remains” → need I; “noisy flow” → beware D; “fuel trip on air fail” → fail-closed fuel valve.
4. Do not invent numerical tuning constants unless given; reason qualitatively.
5. Connect bad PV (failed transmitter) to why independent safety layers exist (Section 11.3).

Common Traps

• Believing P-only always drives error exactly to zero under a new load
• Thinking integral speeds initial response more than it eliminates offset
• Applying heavy derivative on noisy flow without filtering
• Assuming fail-closed is universally safest
• Confusing setpoint with process variable
• Mixing controller output with the measured PV

Section 11.3 explains why even a well-tuned PID loop is only the basic process control system (BPCS) layer—and how dynamics, alarms, and interlocks/SIS complete Domain E’s control-and-safety picture.