Using Average Absolute Error to Uncover Tuning Issues

Average Absolute Error (AAE) Provides an Intuitive Alternative to Variance and Simplifies the Assessment of Changes in PID Controller Performance

Quite simply, things at a manufacturing facility can change for the better or they can change for the worse. When production staff are the agents of change, then more often than not those changes result in improvements to some aspect of production. On the flip side when they’re not involved in overseeing those changes, bad things are usually afoot.

Due to the size and complexity of most manufacturing environments change is a constant. Some of those changes can have a subtle impact like the use of a different feedstock or a shift in environmental factors (e.g. increased humidity due to rain). Other changes can be more profound like the sudden failure of mechanical equipment. Whether subtle or significant, change will impact production efficiency and throughput to one degree or another.

Due to the size and complexity of most manufacturing environments change is a constant. Some of those changes can have a subtle impact like the use of a different feedstock or a shift in environmental factors (e.g. increased humidity due to rain). Other changes can be more profound like the sudden failure of mechanical equipment. Whether subtle or significant, change will impact production efficiency and throughput to one degree or another.

Maintaining awareness of the different changes and assessing their potential impact on performance can present a challenge. Technologies like control loop performance monitoring (CLPM) keep a constant watch on a facility’s PID controllers, using various key performance indices (KPIs) to assess the health of regulatory control systems. Most CLPM systems utilize simple Variance to assess change. Average Absolute Error is an alternative that offers a simpler means of interpreting data and assessing the implication.

Consider the following:

  • Defining the Metric

Average Absolute Error is a metric for assessing change within a given PID control loop. It quantifies the difference between the loop’s Set Point and the corresponding measured Process Variable. As a measure of control loop performance, Average Absolute Error tracks how effective the loop is at maintaining Set Point.

  • Rating the Change

Some degree of variability should be expected within a PID loop as production processes are highly interactive and inherently dynamic. That said, any notable increase in the value for Average Absolute Error generally corresponds with a change that production staff should at least note if not investigate and address. While modest increases of 5%-10% generally correspond with everyday variability, increases greater than 10% point to an issue in need of attention.

  • Seeking a Solution

While tuning is not the answer to all controller issues it can resolve some increases in PID loop variability. An adjustment in feedstock, a load change, and even a shift in seasons are sources of change that can be accommodated with modified tuning parameters that are better suited for the process’ altered dynamics.

Two keys to managing change are awareness and analysis. Metrics like Average Absolute Error can provide timely awareness of changes in production conditions. As part of a plant-wide monitoring solution Average Absolute Error and other KPIs enable production staff to stay on top of their expansive environment. That said, analysis is equally important. More often than not it is determining the source of the change that provides the key to an effective corrective action.

 

These resources offer related content:

Iconic-Mark_Inverted-Color

The Relationship Between Loop Performance Monitoring and Loop Tuning

Imagine that you just purchased software to help with the tuning of your plant’s control loops.  You’re ready to enhance control and to improve production throughput and efficiency, right?  With software successfully installed you consider the 100s if not 1000s of PID controllers regulating your...
stop light yellow caution

The “D” in PID Stands for: Do Not Use (Sometimes)!

The Derivative Term is not only the last letter in PID (i.e. Proportional-Integral-Derivative) it’s also the most maligned of the three. With its big kick, the Proportional Term provides an immediate correction for changes in control and it is clearly the star of the PID...
Iconic-Mark_Inverted-Color

What Effect Does Noise Have on Controller Tuning?

Noise – that random variation in your data signals – is seemingly everywhere. Low frequency process and high frequency signal noise are two sources that directly affect control. There’s no avoiding it. Indeed, when tuning PID controllers practitioners simply need to deal with it. As...

Still looking for more?

Now that you’ve gotten the basics, connect with our team to learn how our people, processes and technologies can help you optimize.