When two people argue over which of two tunings is better, the integral of absolute error settles it with a number. IAE and its relatives, ISE and ITAE, are ways of summing up all the error a loop accumulated while responding to a disturbance or a setpoint change, collapsing a whole messy response into a single score. These are the objective functions that auto-tuners minimize and that performance-monitoring tools use to compare loops and tunings. This guide gives plain definitions of the error-integral family, explains when each is preferred, and shows how they turn trend data into an objective verdict.
IAE / ITAE Error Integrals in one line: The integral of absolute error, or IAE, is a single-number score of loop performance found by adding up the absolute size of the control error over the whole response to a disturbance or setpoint change. Smaller IAE means less total error and better performance. It belongs to a family that includes ISE, which squares the error to punish large deviations, and ITAE, which weights error by how long it lingers, and these integrals are the objective functions auto-tuners and monitoring tools minimize.
The problem the error integrals solve is that a loop's response to an upset is a complicated curve, and comparing two such curves by eye is subjective. Did the loop with the bigger overshoot but faster settling do better or worse than the one that crept back slowly with no overshoot? The error integrals answer by accumulating the error over the entire response and reporting the total. Every moment the measurement is off setpoint contributes to the score, so a loop that gets back on target quickly and stays there racks up a small total, while one that lingers off setpoint or swings widely racks up a large one. The winner is the tuning with the smaller integral.
IAE, the integral of absolute error, is the most straightforward member of the family. It simply sums the magnitude of the error, ignoring whether it is above or below setpoint, so overshoot and undershoot both count as error and cannot cancel each other out. This makes it an honest, intuitive measure of total deviation. Because it treats a small persistent error and a brief large one on comparable terms, IAE is often described as balanced, penalizing neither extreme dramatically, which is part of why it is a popular default criterion for tuning.
The power of reducing a response to one number is that it makes comparison objective and automatic. Give an auto-tuner the freedom to adjust controller settings and tell it to minimize IAE, and it will search for the tuning that produces the smallest total error against a test disturbance. Give a monitoring tool two periods of data, before and after a change, and it can compute the error integral for each and report which tuning performed better, with no argument. The single number is what lets machines and monitoring systems judge control quality without a human in the loop.
The reason there is a family rather than one metric is that different ways of accumulating error emphasize different aspects of a response, and the right emphasis depends on what you care about. ISE, the integral of squared error, squares the error before summing it. Squaring makes large errors count enormously more than small ones, so ISE heavily penalizes big deviations and overshoots while barely noticing small lingering ones. A tuning optimized for ISE will fight hard to keep peak error small, at the cost of tolerating a longer tail of small error as the loop finally settles. Choose ISE when large excursions are the thing you most want to avoid.
ITAE, the integral of time-weighted absolute error, multiplies the absolute error by how much time has elapsed before summing. Early error, right after the upset, is weighted lightly because it is somewhat unavoidable, while error that persists as time goes on is weighted more and more heavily. The effect is that ITAE punishes a loop that lingers off setpoint, favoring tunings that settle promptly and completely even if they overshoot a bit at first. ITAE is often preferred when a clean, fully settled response matters more than shaving the very first peak, and it tends to reward well-damped tunings that do not leave a long slow tail.
IAE sits between these, weighting all error equally by magnitude regardless of sign, size relative to other errors, or timing. In practice the choice among them is a choice of what to optimize for: ISE to crush peaks, ITAE to eliminate lingering error, IAE for a balanced compromise. None is universally correct, and a thoughtful engineer picks the criterion that matches the loop's job. A safety-critical loop where large excursions are dangerous might be tuned to ISE; a quality loop where a slow return to spec is costly might be tuned to ITAE. Knowing the family lets you match the yardstick to the goal instead of accepting whatever an auto-tuner defaults to.
Computing an error integral needs only the recorded error over a response, the setpoint and measurement traces that a SCADA historian already stores. That makes IAE, ISE, and ITAE natural metrics for a monitoring platform to calculate from archived trends, scoring how each loop handled the disturbances and setpoint changes it actually experienced. A loop's error-integral score becomes a compact, objective indicator of control quality that can be tracked over time and compared across loops.
In a cloud SCADA platform such as Merobix, this supports the most persuasive use of the metrics: the objective before-and-after comparison. Capture the error integral for a loop over a representative period, retune it, capture the integral again under comparable conditions, and the two numbers settle whether the change helped, with no reliance on impressions. Because the computation runs on historized data across sites, a central engineering team can score loops on remote assets they never watch live and rank which retuning efforts actually paid off.
The approach applies wherever disturbance rejection and setpoint tracking have value, which is essentially every controlled process. Oil and gas facilities, water treatment, power generation, and manufacturing all run loops whose ability to shrug off upsets can be scored with an error integral. Having those scores computed automatically from SCADA data turns loop performance into a measurable, reportable quantity, and lets the choice of criterion, IAE, ISE, or ITAE, reflect what each loop is actually there to do.
IAE is a single-number measure of loop performance found by summing the absolute size of the control error over the entire response to a disturbance or setpoint change. Because it uses the magnitude of the error, overshoot and undershoot both count and cannot cancel out. A smaller IAE means the loop spent less total time and distance off setpoint, so lower is better.
All three sum error over a response, but they weight it differently. IAE sums the plain absolute error, treating all error equally by magnitude. ISE squares the error, so it heavily penalizes large deviations and overshoots while nearly ignoring small ones. ITAE multiplies error by elapsed time, so it punishes error that lingers and favors tunings that settle promptly. The choice reflects whether you most want to crush peaks, eliminate lingering error, or strike a balance.
They serve as objective functions that auto-tuners minimize: an auto-tuner adjusts controller settings to find the tuning that produces the smallest chosen integral against a test disturbance. They are also used to compare tunings objectively, by computing the integral for a before period and an after period and seeing which is smaller. This replaces subjective judgment of a response curve with a single comparable number.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.