6.2 Process Dynamics and Loop Response

6.2 Process Dynamics and Loop Response

Quick Answer: If a step in the valve makes the PV settle at a new value, the process is self-regulating and a first-order-plus-dead-time (FOPDT) model is a fair starting point. If the PV ramps without bound (typical unmatched tank level), the process is integrating — do not steal a flow-loop PI tune.

Process dynamics sit in the Control Systems theory slice of the 2027 exam: loop response, pressure-volume-temperature (PVT) relationships, and the difference between analog, discrete, and sequential behavior. The scoring key is not a vendor's lambda-tuning screenshot. It is whether you named the plant type before you named a controller.

Self-regulating vs integrating

A self-regulating process finds a new steady state after a sustained move in the manipulated variable (MV). Open a cooling-water valve 10% and exchanger outlet temperature falls, then levels off. Open a liquid flow valve and the orifice ΔP (hence inferred flow) rises to a new value set by the hydraulic resistance. The open-loop step response is an S-curve toward a plateau. Those plants have a finite steady-state gain Kp = ΔPV / ΔMV.

An integrating (non-self-regulating) process does not find a new steady state while a material or energy imbalance persists. The textbook picture is liquid level in a tank whose outflow is a constant-speed pump (or a downstream flow controller) that does not care about level. If inlet exceeds outlet by a constant 50 gpm, inventory rises at 50 gpm until the tank overflows. The Laplace picture is Kp/s (or Kp e^(−θs)/s with dead time). There is no finite plateau, so you cannot quote a FOPDT Kp from “final minus initial” the way you can on a flow loop.

Pressure of a closed gas volume with a net mass flow in is integrating in the same way. A liquid flow through a short line is almost never integrating. That distinction is the exam trap in this section: treating an integrating level like a self-regulating flow loop.

First-order-plus-dead-time

Most self-regulating PE items will accept the FOPDT plant

$G_P(s) = \frac{K_p e^{-\theta s}}{\tau s + 1}$

• Kp — process gain, in PV units per MV unit (for example °C per % valve).
• τ — apparent time constant: after the PV starts moving, the time to reach 63.2% of the total change.
• θ — apparent dead time: delay before the PV moves, from transport, sensor lag, valve deadband, or several small lags lumped together.

When θ/τ is small, the loop can tolerate more controller gain. When θ/τ is large (dead-time dominant), high gain oscillates — that fight is finished in the PID section. Here, just extract the three numbers honestly from a step test.

Worked step test (qualitative construction + numbers)

A steam-heated exchanger is at steady 80 °C outlet with the steam valve at 40%. At t = 0 you step the valve to 50% (ΔMV = 10%) and hold it. The recorded outlet temperature:

1. Does not move until t = 1.0 min. That delay is dead time θ = 1.0 min (sensor, piping, and steam-chest lags lumped).
2. Then rises on an S-curve and settles at 100 °C. ΔPV = 20 °C, so Kp = 20 °C / 10% = 2.0 °C/%. Settlement proves self-regulating, not integrating.
3. 63.2% of 20 °C is 12.6 °C, so the 63.2% point is 92.6 °C. That temperature is reached 4.2 min after the PV starts moving (t = 5.2 min on the chart). τ = 4.2 min.

Qualitative construction if the curve is noisy: draw a tangent at the steepest (inflection) point. The tangent's intercept with the original steady value estimates θ; the intercept with the new steady value estimates θ + τ. The 28%/63% two-point shortcut is the same idea: if t28 = 2.4 min and t63 = 5.2 min from the step, τ ≈ 1.5 (t63 − t28) = 4.2 min and θ ≈ t63 − τ = 1.0 min.

Do not call the 63.2% clock from t = 0 if the PV sat still for a minute — that mistake folds dead time into τ and you will overstate how “fast” the plant is.

Interacting lags

Two first-order lags in series with no interaction (a thermowell, then a well-mixed vessel whose outlet does not talk back) multiply: 1/((τ1 s+1)(τ2 s+1)). Two interacting lags — classic liquid levels in series where the downstream head holds up the upstream tank — are slower and more S-shaped than that product. Apparent dead time grows because the early part of the curve is flatter. On a step test you still fit FOPDT; just know the physical plant is not “one tank.”

Inverse response

Inverse response is not a slow lag. The PV starts the wrong way, then turns and goes to the expected steady value (or, on an integrator, eventually ramps the expected way). In transfer-function language there is a right-half-plane zero fighting the eventual gain.

Two PE-favorite plants:

• Boiler drum level (shrink/swell). A sudden steam-demand increase drops drum pressure. Water in the drum flashes, bubbles swell, and indicated level rises even though mass inventory is falling. Feedwater then has to catch up; true inventory was always going down. A step decrease in steam demand does the opposite (shrink). Tight level control on swell fights the bubble, not the mass, and you overfill or trip.
• Some steam-heated exchangers and reboilers. Opening the steam valve can briefly drop outlet temperature (condensate and pressure dynamics) before the duty increase wins. A PID with derivative on that dip will kick the valve the wrong way.

If the curve only sits still and then goes the right way, that is dead time plus lag, not inverse response.

Integrating level, again, as a control implication

Surge tanks, condensate receivers, and many intermediate storage drums exist to absorb flow mismatch. Average level should be mid-scale; instantaneous level is allowed to wander. A self-regulating flow loop should be tight: you want the kg/h that the recipe named. Copying that tightness onto LIC of a surge drum passes every upstream flow bump straight downstream and defeats the vessel. Averaging level, gap/deadband, or a slow P-heavy controller belongs on the surge drum; tight PI belongs on the flow that the drum is protecting.

PVT relationships

Pressure, volume, and temperature are not a separate “chemistry” topic here. They are why a loop’s gain changes with operating point:

• Ideal-gas inventory: PV = nRT. For a nearly constant volume, pressure is a proxy for moles. A temperature swing at fixed mass changes indicated pressure — a P-only pressure loop will offset if you ignore T.
• Gas flow through an orifice is not the liquid-orifice square-root story without compensation. Mass flow depends on density, hence on P and T. Uncompensated FT gain rises when the gas is denser (higher P, lower T).
• Liquid filled systems are nearly incompressible: a small valve step can make a large, fast pressure spike (very small τ, large Kp) — that is why some pressure loops look “too hot” compared with the temperature loop on the same unit.

When an exam item gives you a gas at two P,T states and asks why the pressure controller now overshoots, check density and gain, not a new PID philosophy.

Analog vs discrete vs sequential response

Aliasing belongs with discrete sampling: if you sample slower than the dynamics you care about, a fast analog oscillation looks like a different slow wave. Sequential control that waits on a temperature soak is still a timer plus a comparator, not an LIC.

Process type vs control implication