Chapter 2: Simple Process

Models & PID Control

 2.1 Introduction
• We study simple feedback control loops
(process section + controller).
• Goal: process variable y follows setpoint r as
closely as possible.
• Two core control problems: servo (setpoint
changes) and regulation (disturbance/load
changes).
• Most process-control applications emphasize
regulation.
Simple Feedback Loop
 Control Error Definition

• e (error) drives the controller action.

• Sign conventions must be consistent across
the loop.

e=r−y
 2.2 Step-Response Models —
Experimental Motivation
• Place controller in manual mode and apply a
step change Δu to the control signal.
• Record the output response y(t) over time
(chart recorder or timed readings).
• From the response, estimate key process
characteristics used in tuning.
Typical Step Response
 Static (Process) Gain

• Measures steady-state amplification of the

process.
• Estimated once the process output has
stabilized after the step

Kp = Δy / Δu
 Dominant Time Constant (T)
• T is defined at the 63% point of the total
output change (dominant behavior).
• Many processes are multi-capacitive (multiple
time constants), but a dominant one is often
used for tuning.
 Overshoot and Damping (Step
Response)
• Overshoot is measured relative to the steady-
state change.
• Industrial open-loop processes often show
negligible overshoot (well damped).
• A damping measure can be defined from
peak/valley ratios (as introduced here; refined
later).
Integrating Process Step Response
Velocity Gain (Integrating Process)
Kv = Δy / (Δu · L)

• Used when output does not settle (integrating

behavior).
• Dead time L and Kv often characterize
integrating processes.
 2.3 Common Process Types — Why
Categorize?
• Different process sections have different
dynamics (speed, stability, amplification).
• Cataloging step responses helps select models
and controllers.
• Process section includes everything except the
controller (valves, actuators, transducers,
modules).
Six Canonical Step Responses
 Single-Capacitive Processes (First-
Order Behavior)

• Step response resembles a simple RC circuit.

• Often a good first approximation for many
flow/pressure/temperature subsystems.
RC Circuit Step Response
 RC Time Constant (Example)
T=RC

• Illustrates the idea of a time constant using an

electrical analogy.
• Used as a conceptual reference for first-order
process behavior.
 Multi-Capacitive Processes
• Most common process type; often appears
with dead time.
• Analogous to several RC filters in series.
• Common in flow and pressure processes, and
some temperature processes.
 Integrating Processes
• Output does not naturally settle after a step; it
accumulates (ramps).
• Typical when storage dominates: level,
pressure in closed vessel, batch concentration,
batch temperature.
• Integrating vs stable processes behave very
differently under control.
 Oscillating Processes
• Step response oscillates around the final
value.
• Less common in uncontrolled process
sections; more common when loops are
poorly tuned.
• Occurs in recirculating concentration systems
and many mechanical elastic systems.
 Dead-Time Processes
• Output shows no reaction until a delay (dead
time) has passed.
• Rarely pure dead time alone—usually dead
time + lag dynamics.
• Dead time strongly limits achievable closed-
loop performance.
 Inverse Response
• Output initially moves in the opposite
direction of its final change.
• Often caused by competing dynamic paths
(fast path vs slow path).
• Requires care in identification and controller
tuning.
 2.4 The PID Controller — Motivation
• PID is the most widely used industrial
controller.
• It is a natural extension of the simplest
controller (On/Off).
• We derive P, PI, and PID structures and
interpret their roles.
 On/Off Control Law
• Simplest controller; switches between two
levels.
• Maintains small error by frequent switching →
oscillations.
 Proportional (P) Control — With
Proportional Band
• For small errors, control is proportional to
error (linear region).
• For large errors, saturates similar to On/Off
behavior.
Residual Error Under P Control (within
band)

• Stationary error is generally nonzero under P

control.
• Zero stationary error would require: (1) K → ∞
or (2) u0 = u (idealized).

e = (u − u0) / K
 PI Control Law

• Integral action replaces constant bias u0 with

an accumulated error term.
• Eliminates residual stationary error by
adjusting the bias over time.
 Integral Term as a Time-Varying Bias
u0(t) = (K/Ti) ∫₀ᵗ e(τ) dτ

• If e ≠ 0 persists, u0(t) keeps changing → drives

e toward 0 in steady state.
• Provides “memory” of past errors.
PI Output Equality Illustration
 Why Derivative Action? (Intuition)
• If error decreases rapidly, controller should
avoid overshoot (be cautious).
• If error suddenly starts increasing, controller
should react strongly (be proactive).
• Derivative action uses the error’s rate of
change to anticipate behavior.
 PID Control Law

• P: reacts to current error.

• I: reacts to accumulated error (removes
steady-state error).
• D: reacts to error trend (anticipation).
PID Interpretation
 Practical Notes on D Action (as
introduced here)
• Most beneficial when prediction helps (e.g.,
many temperature control applications).
• In practice, D action is sensitive to
measurement noise.
• Later lectures refine tuning and
implementation details.
 2.5 Chapter 2 Summary (Engineering
Takeaways)
• Step-response experiments provide practical
identification parameters (Kp, T, L, etc.).
• Process types (single-/multi-capacitive,
integrating, dead time, inverse, oscillating)
guide controller choice.
• PID structure emerges from extending On/Off
→ P → PI → PID.
• Key equations for controllers (P/PI/PID) must
be understood before tuning methods.
Step-Response Parameter Summary
(Engineering Reference)
Static gain: Kp = Δy / Δu
Dead time: L (time until output begins to respond)
Time constant: T (time to reach 63% of total Δy)
Velocity gain: Kv = Δy / (Δu · L) (integrating processes)

These parameters are extracted directly from experimental step tests

They form the basis for all classical controller tuning methods
Controller Selection Logic (Process →
Control Mapping)
Single-capacitive, no dead time → P or PI control
Multi-capacitive with dead time → PI control
Integrating processes → PI control mandatory
Processes with strong lag or inertia → PID may be beneficial

Dead time reduces achievable performance regardless of controller choice

Controller complexity must match process dynamics
 Engineering Tuning Intuition (Pre-
Tuning Guidance)
Increase K (gain) → faster response but higher risk of oscillation
Decrease Ti (stronger integral action) → faster offset removal but overshoot risk
Increase Td (derivative action) → improved anticipation but noise sensitivity

Good tuning balances speed, robustness, and actuator limitations

Formal tuning rules are introduced in later chapters
Instructor Notes — How to Teach This
Chapter Online
Pause after step-response identification and ask students to classify the process type
Emphasize why PI control dominates industrial practice
Use the residual-error equation to explain why P control is insufficient
Stress that PID is not 'better by default' — it is context dependent