Temperature, pressure, flow, level: loop specifics

How typical process loops differ: process dynamics, noise, whether D is needed, where to start tuning. Worked examples for flow, temperature and level with one method.

In short. The method is the same: step test → model → tuning (IMC/SIMC). But loop types differ a lot: flow is fast and noisy (PI, no D), temperature is slow with dead time (PI or PID), level is integrating (usually slow averaging control), pressure ranges from "like flow" to "like level" depending on the medium and volume.

Comparison

LoopDynamicsPV noiseControllerWhat to watch
Liquid flowseconds; small θhighPI, no DTs ≤ 0.1–0.2 s; valve stiction; valve non-linearity
Liquid pressurelike flowmediumPIwater hammer; fast valves
Gas pressure in a vesselseconds to minutes; often integratinglowPIvessel volume sets the speed; output limits
Levelintegratingwaves, foamP or PI with slow tuningaveraging or tight control is a process decision
Temperatureminutes; noticeable θlowPI or PID with a D filterheating/cooling non-linearity, split range, sensor lag
Composition, pHlong θ (analyser)variesPI, slow; often cascadeanalyser θ; strong pH non-linearity

Example: flow

Normalised model K = 1.2 %/%, τ = 3 s, θ = 1 s. SIMC, τc = θ:

Kc = 3 / (1.2 · (1 + 1)) = 1.25 %/%        Ti = min(3, 4·2) = 3 s

Sample time: θ/10 = 0.1 s. If the block runs every 0.5 s, that is half the dead time — increase τc or speed up the task (discretisation).

Open in the tuning tool

Example: temperature

Section model: K = 1.8 °C/%, τ = 45 s, θ = 12 s. SIMC PI: Kc = 1.042 %/°C, Ti = 45 s (overshoot 5.5 %, settling 74 s). IMC PID (aggressive): Kc = 1.816, Ti = 51 s, Td = 5.29 s — a faster rise, but D needs a filter and D on measurement (D filter).

Heating and cooling often have different K and τ: tune for the "stronger" direction or use split range with separate gains.

Example: level

Integrating process, k′ = 0.001 1/s, θ = 10 s (integrator model):

GoalτcKc, %/%Ti
Tight (level matters, e.g. in a reactor)10 s = θ1/(0.001·20) = 504·20 = 80 s
Averaging (the vessel smooths the flow)100 s1/(0.001·110) = 9.094·110 = 440 s

With averaging control the level may move within allowed limits, while the flow to the next unit changes smoothly.

Typical mistakes

  • D in a flow loop — noise is amplified, the valve wears.
  • Tight level tuning when a smooth flow is needed: disturbances are passed down the process chain.
  • Ts larger than θ in fast loops (flow, liquid pressure).
  • One setting for heating and cooling with very different K.

Limits

The table describes typical behaviour; a particular plant may differ (a short pipe with a large sensor θ, gas with a small volume, etc.). The numbers in the examples illustrate the calculation and are not recommended values for any plant.

Loading the map…

Sources

  1. Skogestad — Simple analytic rules for model reduction and PID controller tuning, J. Process Control 13 (2003) 291–309: SIMC for self-regulating and integrating processes — NTNU
  2. Åström & Murray — Feedback Systems, 2nd ed. (2020), ch. 11 «PID Control», §11.1–11.3 — Caltech / Lund University

Updated 2026-10-12