Process control coursework generally teaches PID tuning solely as a Setpoint tracking problem. Most textbook plots show a sharp step change, modest overshoot, and a rapidly settled response. That framing follows future engineers onto the plant floor, and a great many loops end up tuned to handle dynamic process behavior in a way that is misaligned with the loop’s true control objective – the ability to effectively reject disturbances.

PID controllers spend much of their time rejecting disturbances associated with ambient temperature swings, upstream pressure upsets, feedstock variability, and other routine perturbations. Ironically, control loops tuned for tracking Setpoint typically respond sluggishly to these everyday disturbances which ultimately undermine plant performance and profitability.

Disturbances Come in Many Forms

A disturbance is any unmeasured or uncontrolled input that pulls the Process Variable away from Setpoint. While disturbances generally fall into three groups, PIDs have to handle all of them:

• Slow and gradual – ambient temperature swings, heat exchanger fouling, valve and pump wear over service life.
• Fast and step-like – upstream pressure upsets, sudden throughput changes pushed by operators or by an APC layer.
• Variable and hard to anticipate – feed composition shifts, impurities, catalyst drift.

A loop that’s tuned specifically for handling one type is rarely ready for another. 

When Disturbance Rejection Is the Real Objective

Not every loop needs aggressive disturbance rejection. As an example, a flow loop downstream of an APC layer may be moved every few minutes. The APC itself is effectively the disturbance, and the loop needs to be quick about tracking each adjustment to Setpoint. In contrast, a surge tank level loop may need to absorb upsets as a means of protecting a downstream process. Tracking Setpoint is far less critical than shielding the downstream process and the associated instrumentation. Following are four (4) considerations for determining if tuning for disturbance rejection should be the control objective – no Setpoint tracking:

• Timing – How tight a band must the loop hold, and how quickly must it return to Setpoint after an upset?
• Severity – How significant are the upsets? What does a control failure cost? A quality miss, a flare event, a runaway, an unplanned shutdown?
• Frequency – How often do process disturbances arrive versus Setpoint changes?
• Probability – Which disturbances do operators actually see on this loop, in practice? Operators are often an underused source of process knowledge here.

The answers tell you whether the loop’s real job is tracking a target or rejecting upsets. Tuning for the wrong objective will simply and predictably produce poor results. The same principle is the starting point for What Is the Control Objective?, and it is the prerequisite for every choice that follows.

Which Term Does the Work

The PID controller includes three (3) levers – Proportional, Integral, Derivative – and each behaves differently when under load. Consider the following:

• While Proportional action gives an immediate response to Error, it never fully closes the gap. Integral action is where most disturbance rejection work gets done. 
• Integral eliminates the offset that Proportional alone cannot, and the speed of recovery depends largely on how aggressive the Integral is set. 
• Derivative can accelerate responses to disturbances, but it amplifies measurement noise. Loops that do need Derivative require filtering on the Process Variable to keep noise from being multiplied.

For most loops, a faster Integral is the lever that improves disturbance rejection the most. For high-severity loops where a deviation cascades (e.g. reactor temperature with strong exothermic feedback), Proportional Gain becomes the dominant lever. Recall that Integral acts on accumulated Error while Proportional Gain delivers the initial push needed to arrest the deviation before it gets out of control.

Methods Built for the Objective

Tuning correlations differ in what they prioritize. Internal Model Control and Lambda both let you set the Closed-Loop Time Constant directly, which is useful for matching response speed to the disturbance profile. Still, load-optimized tuning rules can go further by weighting Integral action more aggressively for faster recovery from load disturbances, with a tradeoff in Setpoint response.

The tuning procedure is the same regardless of method: bump the process to capture the dynamics, fit a First-Order-Plus-Dead-Time model, calculate tuning parameters, implement the new tunings, test/validate, and document. Where open loop step tests are not, then Setpoint changes made in automatic can supply enough data for modeling – provided the loop is observed across its Design Level of Operation.

One Tuning Set Will Not Last

A loop tuned at 20 tons per hour should be expected to behave differently when regulating at 40 tons per hour. The calculated Process Gain will shift at different rates of throughput, equipment wear, catalyst changes, seasonal conditions, etc. Even identical reactors at the same site develop their own personalities. Tuning is not a one-time exercise, and a controller that worked effectively at commissioning can drift into oscillation a year later without anyone noticing until it shows up in the product.

Watching loop performance continuously across the many “states” in which your plant actually runs is what catches the drift early. PlantESP flags the small subset of loops in a facility that need attention at any given time, while LOOP-PRO Tuner handles the model identification and tuning on the loops that warrant adjustment. For teams building the underlying judgment in-house, Control Station’s interactive training workshops cover this same approach.

Disturbance rejection is often the objective that many loops are actually targeting. Tuning to that objective is essential to keeping a plant steady when the day does not go according to plan.