Process Control Basics: Feedback Loops and Reactor Example

Process control keeps a process variable such as temperature, pressure, flow, or level near a target by measuring the current value, comparing it with the setpoint, and adjusting something it can manipulate to shrink the error. The single distinction that organizes the whole topic is this: you hold the controlled variable steady by moving the manipulated variable, and those are never the same thing.

The Five Terms Side By Side

The error that drives the loop is

where $r(t)$ is the setpoint and $y(t)$ is the measured value. The error $e(t)$ tells the controller how far the process is from the target. When a disturbance pushes the process off target, the controller moves the manipulated variable in the direction meant to reduce $e$. The exact rule depends on the controller design, but the feedback idea, measure, compare, correct, stays the same. The most common confusion is collapsing two of these rows: in a temperature loop you want to hold temperature steady, but you do that by changing steam or coolant flow, not by "moving temperature" directly.

When Control Earns Its Keep

A chemical process rarely stays where you left it, because feed conditions change, utilities fluctuate, and reaction rates respond to temperature. Manual control means a person watches and adjusts by hand; feedback control means the loop runs the compare-and-correct step automatically. Reach for automatic control when disturbances happen faster or more often than a person can correct consistently, or when drifting off target is expensive. A small temperature change might only trim yield in one unit, but in another it could shift selectivity, create off-spec product, or raise safety risk, and that is when control matters most.

Worked Example: Reactor Temperature Control

A jacketed reactor should run at a setpoint of $80^\circ \mathrm{C}$. The measured temperature suddenly drops to $76^\circ \mathrm{C}$ because the incoming feed is colder than usual. The error is

The controlled variable is reactor temperature; a reasonable manipulated variable is steam-valve opening to the jacket, since steam flow sets heat input. For a proportional-only rule over this range, model the valve-signal change as

With controller gain $K_c = 5\% \text{ valve opening per } ^\circ \mathrm{C}$,

so the controller asks for about $20\%$ more valve opening. This is a simplified teaching case; in a real plant the final response also depends on the existing valve position, controller tuning, actuator limits, and any integral or derivative action. Still, the logic is the same: the reactor is too cold, so the loop increases heat input. As temperature climbs back toward $80^\circ \mathrm{C}$ the error shrinks, so once the measured temperature reaches $79^\circ \mathrm{C}$ the same proportional rule calls for only about $5\%$ extra opening. That shrinking correction as the process nears its target is the essence of negative feedback.

Common Confusion Points

These are the loop misconceptions worth flagging:

• Confusing controlled and manipulated variables. In a temperature loop, temperature is controlled while steam or coolant flow is manipulated.
• Expecting feedback to erase error instantly. With process delay or a slow sensor, the loop can still respond sluggishly or oscillate.
• Treating all loops alike. A fast flow loop and a slow composition loop differ greatly in control difficulty.
• Equating process control with PID. PID is common, but on-off, cascade, ratio, feedforward, and model-based methods all belong to process control.

Where Process Control Is Used

Process control appears wherever a variable must stay in a useful range: temperature control in reactors and heat exchangers, pressure control in vessels and gas systems, level control in tanks and separators, flow control in feed and utility lines, and composition or pH control when product quality depends on mixture balance. The payoff is practical, since quality, efficiency, stability, and safety all depend on holding those variables near target. To check the core idea, pick one familiar loop and name four things cleanly: setpoint, controlled variable, manipulated variable, and one likely disturbance.