PID Tuning Guide in Distributed Control System

PID Tuning Guide in DCS — Zohaib Jahan

Control Equation

// PI Controller Output
CO(t) = Kp × e(t) + Ki × ∫e(t) dt

// where:
CO(t) = Controller Output (0–100%)
e(t) = SP − PV (Error signal)
Kp = Proportional Gain
Ki = Integral Gain (repeats/min ≡ 1/Ti)

Kp — Proportional Gain

Reacts immediately to error
Higher Kp → faster, more aggressive response
Alone: always leaves a residual offset
Think of it as: how hard you push back on error

Ki — Integral Gain

Accumulates error over time
Eliminates steady-state offset
Too high → integral windup, instability
Think of it as: patience that corrects drift

Gain Behaviour — What Goes Wrong

Kp TOO HIGH

→

Oscillation / Instability — loop hunts continuously around setpoint

Kp TOO LOW

→

Sluggish Response — slow to reach setpoint, large persistent offset

Ki TOO HIGH

→

Integral Windup — output saturates, valve slams open/closed repeatedly

Ki TOO LOW

→

Persistent Offset — PV never reaches SP exactly under steady load

Balanced PI ✓

→

Stable, fast, offset-free — PV tracks SP through load and SP changes

🏭

Field Reality — Why We Skip Derivative

Derivative amplifies high-frequency noise — common in orifice-based flow transmitters and long cable runs
DeltaV, 800xA, CENTUM VP all default to Td = 0 for process loops out of the box
D term only considered for: batch temperature ramps, override control on very fast processes

Start with Ki = 0

Set integral to zero. Switch controller to Auto with a small Kp (e.g. 0.5). Observe baseline behaviour. Confirm loop direction — direct or reverse acting.

Increase Kp — Watch for Oscillation

Increase Kp in steps. Apply a small SP change (+5%) after each step. Stop when sustained oscillation or overshoot > 20% appears.

Back Off Kp by 30–50%

Reduce to 50–70% of the value that caused oscillation. Response should now be stable with some offset — expected at this stage.

Add Ki Slowly — Remove Offset

Start with a small Ki and increase gradually until steady-state offset disappears. If oscillation returns, reduce Ki slightly and hold.

Validate Under Real Conditions

Test with load changes and SP changes (±10%, ±20%). Document all values before and after in DCS audit log.

Validation Checklist — Before You Sign Off

PV reaches SP without sustained oscillation
Overshoot < 10–15% (or per process requirement)
Valve movement is smooth, not hunting

Tested at min, normal, and max load
No integral windup on large SP steps
All values logged in DCS change record / MOC

⚡ Field Notes

Always test in Manual first — confirm valve stroke direction and stiction behaviour before Auto
In DeltaV, use Control Studio Trending. In 800xA, use Aspect Object tuning panel
Ti (min/repeat) = 1/Ki — verify your DCS platform expression before entering any values
For fast loops (flow): take small steps quickly. For slow loops (temperature): wait full response before next step

Typical Starting Parameter Ranges

Loop Type	Dynamics	Kp Range	Ki / Ti	Key Engineering Notes
⚡ Flow	Fast	0.3 – 0.8	0.3–1.0 r/m Ti: 1–3 min	Noisy signal — use input filter. Valve positioner dominates dynamics. Avoid high Kp — noise causes valve wear.
🔵 Pressure	Medium	1.0 – 3.0	0.3–1.5 r/m Ti: 0.7–3 min	Speed depends on vessel volume. Gas header = fast; liquid back-pressure = slow. Confirm self-regulating vs integrating first.
🟢 Level	Slow / Int.	1.0 – 5.0	Very Low Ti: 5–20 min	Integrating by nature — accumulates, never self-corrects. P-only often sufficient. High Ki causes windup very fast.
🌡️ Temperature	Very Slow	1.5 – 5.0	Very Low Ti: 5–30 min	Significant dead time in fired heaters and reactors. Never tune aggressively. Wait full time constant after each change.
⚗️ Density	Very Slow	1.0 – 4.0	Very Low Ti: 10–30 min	Measurement lag from analyser or Coriolis meter. Verify sample conditioning and calibration before tuning.

⚠ Starting ranges only — verify against actual plant response. Do not apply directly without process confirmation.

⚡ Flow Loop

Apply input filter (5–10 sec) before tuning
Usually inner loop of a cascade arrangement
Retune if rangeability or orifice plate changes

🟢 Level Loop

P-only acceptable for most surge drums
Average Level Control (ALC): intentional P-only, wide band
Never add fast integral — causes valve cycling

🌡️ Temperature Loop

Measure dead time before starting any tuning
Heater loops: auto-tune only in stable conditions
Wait 3–5 time constants after each parameter change

🔵 Pressure Loop

Gas pressure: very fast — reduce Kp aggressively
Liquid back-pressure: slower, more forgiving
Check for upstream/downstream loop interaction first

1. Cascade Control

Primary (outer) sets SP of Secondary (inner)
Secondary must be 3–5× faster than primary
Inner loop rejects disturbances before outer PV is affected
Always tune secondary first, then primary

Example: Temperature (primary) → Flow (secondary) on a heat exchanger

SP ──→ PRI CTRL→SP
SEC CTRL→VALVE
PV ←── Temp TX
PV ←── Flow TX

2. Ratio Control

Maintains a fixed ratio between two process streams
One stream is wild (uncontrolled), other is controlled
Ratio = Controlled Flow ÷ Wild Flow
Common: fuel/air combustion, blending, reactant feed

Example: Steam-to-feed ratio in a reformer furnace

Wild Flow → RATIO
STATION→SP
FLOW CTRL→VALVE
Ratio SP ──┘

3. Split-Range Control

One controller drives two final elements in sequence
Output split: 0–50% → Valve A, 50–100% → Valve B
Valves typically act in opposing directions
Common: heating/cooling, pressure vent + supply

Example: Reactor temp — hot oil valve (0–50%), cooling water valve (50–100%)

0–50% → VALVE A (Heat)
SP → CTRL ─┤
50–100% → VALVE B (Cool)
PV ←── Process TX

When to Use What — Quick Reference

Strategy	Use When	Key Requirement	Common Tag Pattern
Simple PID	Single PV, no major upstream disturbance	Stable process gain	FIC, TIC, PIC, LIC
Cascade	Inner disturbances frequently upset outer PV	Inner loop must be 3–5× faster	TIC (primary) / FIC (secondary)
Ratio	Two streams must hold fixed proportion	Reliable flow measurement on both	FFC / FFY (feedforward element)
Split-Range	One loop must handle two opposing actions	Correct valve characterization at split	Single controller, two valve tags

How It Works

Injects a relay or step perturbation into the loop while in Auto or closed-loop mode
Measures process gain, dead time (θ), and time constant (τ) from PV response
Calculates Kp and Ti using internal rules — typically IMC or Lambda based
Presents recommended values — engineer must review and accept, never auto-apply blindly

AUTO TUNER
Injects test signal
      ↓
Measures PV response
      ↓
Calculates: Kp, Ti
      ↓
ENGINEER REVIEW
      ↓
APPLY / REJECT

Platform Auto Tuner Reference

DCS Platform	Auto Tuner Tool	Method	Access Path
Emerson DeltaV	Control Studio Auto-Tune Wizard	Step / Relay feedback	Control Studio → Module → Tune
ABB 800xA	PID Tuner (Aspect Object)	Relay feedback (Åström-Hägglund)	800xA → Aspect Object → PID Tuner tab
Yokogawa CENTUM VP	TDLB Self-Tune / PID Self-Tune block	Step response identification	CENTUM VP → FCS → TDLB tag
Siemens PCS 7	PID_Compact Autotuning	Step response (process identification)	CFC editor → PID_Compact → Commissioning tab

Field Perspective — Pros & Cons

✅ Pros

Fast starting point — saves hours of manual bump testing
Removes guesswork on unfamiliar processes during commissioning
Reliable on clean, simple loops — well-behaved flow and pressure loops
Quantifies process gain and dead time — model data is valuable even if you reject the suggested tuning
DeltaV closed-loop relay method causes less process upset than open-loop step test

❌ Cons

Assumes a linear FOPDT model — not always valid in real process plants
Unreliable on noisy signals — orifice-based flow loops often produce bad results
Fails on integrating processes (level) — relay method struggles with runaway integrators
Result validity tied to conditions at test time — different load, feed, or temperature = different result
Can suggest overly aggressive tuning — IMC lambda too tight for conservative plant operations

⚠️

Before Running Auto Tuner — Non-Negotiable Checks

Verify valve health — stiction or hysteresis corrupts relay test results
Confirm process at normal, stable operating conditions — not during startup or shutdown
Inform Operations — auto tuner intentionally upsets the loop during the test
Always review suggested values before applying — never blind-accept any auto tuner output
For safety-critical or custody transfer loops: obtain MOC (Management of Change) approval first

⛔ Before You Touch Any Parameter

Verify transmitter calibration — zero and span
Check for valve stiction or hysteresis — step test in manual
Confirm control valve direction (ATO / ATC)
Review noise level on PV trend — filter if needed
Ensure loop is at safe, steady-state conditions
Inform Operations before starting any tuning activity

✅ During Tuning

Keep a hand on Manual — be ready to intervene instantly
One change at a time — never change Kp and Ki together
Use small SP steps (5–10%) for evaluation and comparison
Wait for full process response before each next step
Record tuning values before and after in DCS change log
Trend at minimum: SP, PV, CO, and valve feedback (if available)

⚠️ Avoid Aggressive Tuning In These Loops

Level control — over-tuned level hammers the valve and upsets downstream units
Temperature loops — dead time makes overshoot large and recovery very slow

Analyser-based PV — long sample lag makes tight tuning counterproductive
Cascade primary — inner loop instability propagates fast to outer controller

Golden Rules of PID Tuning

Understand the process first. Tuning without process knowledge is guessing.
Test in Manual before Auto. Confirm the valve moves correctly first.
Document everything. Every change must be traceable — MOC.
Never chase noise. Filter the PV — don't tune around a noisy signal.

Stability over speed. A slightly slow loop is safer than an oscillating one.
Retune after major changes. New transmitter, valve, or feed change invalidates old tuning.
Valve health = loop health. No tuning fixes a sticky or oversized valve.
Less Ki is almost always right. Integral is the most over-tuned parameter in the field.

DCS Platform Terminology — Quick Reference

Platform	Proportional	Integral	Note
Emerson DeltaV	GAIN (Kc)	RESET (min/repeat)	Ti = 1/RESET
ABB 800xA	Kp (gain)	Ti (integral time, min)	Standard ISA form
Yokogawa CENTUM VP	P (prop. band %)	I (repeat/min)	PB% = 100/Kp — convert before entering
Siemens PCS 7	KP (gain)	TI (integral time, s)	TI in seconds — confirm units carefully