What is Variance? What Metrics are Useful in Identifying Process Change?

Average Absolute Error (AAE): A Value-Added Metric that Puts Change Into Context

Ever walk onto the production floor and just sense that something was different? Whether it was something you saw, heard or even felt, a certain sixth sense kicked in and triggered an alert something’s changed. With what takes place at a production facility change is an everyday occurrence. Considering the typical facility’s size, the interconnected unit operations, the network of piping and instrumentation, and the nonstop throughput, there’s no avoiding change. In fact, a key part of the job is simply staying one step ahead of it.

Generally speaking change is not a good thing in a process manufacturing environment. That’s particularly true when the change causes issues with throughput and quality or when it results in equipment failure and unplanned downtime. But not every change is catastrophic. Some are the result of normal wear and tear. Still, even small changes affect performance and sooner or later they add up to big changes.

Process engineers have historically looked to Variance along with a variety of other metrics as a means of measuring change and staying on top of it. An alternative metric worth considering is called Absolute Average Error (AAE) a unique performance metric introduced by your friends at Control Station. AAE offers different and more application-specific details related to process dynamics and change.

• Details, Details, Details

In the field of Statistics Variance has nearly universal application value. It is simply a calculation of the spread of a given data set, an average of the data’s squared differences from the Mean. From a practical application standpoint that information can be useful whether you’re managing a stock portfolio, polling an electorate, or even forecasting the weather.  Still, Variance is generic as a metric and lacks meaningful context, it’s just a number. Its values are always positive, and they are not expressed in engineering units. Engineers are reliant on details, and Variance doesn’t provide all that’s truly needed.

• Change Put Into Context

AAE calculates change by considering error (i.e. SP-PV) in absolute terms and averaging the values.  And because we know specific aspects of a given control loop such as the loop’s Set Point, AAE values can be expressed both in engineering units and in terms that are highly relevant to the loop’s status. AAE values can be easily trended to show change that highlights change in a loop’s performance over time. AAE may not apply to portfolio theory, but engineers are focused on controlling processes and not on assessing the Dogs of the Dow.

• A Practical Example

Consider a classic shell and tube heat exchanger where the PID controls the rate of cooling liquid flowing through the shell.  That cooling liquid enables the heat exchanger to achieve its control objective of maintaining consistent output exit temperature. However, change in the process’ dynamics whether due to age or differences in feedstock result in output variability.  An engineer is shown a report that lists the Variance at ’12.2’ and the AAE at 5.1^? Fahrenheit/Gallon. Although both values have meaning, the AAE value is more than a number as it puts the change into context.

This blog regularly notes that the average production facility has many, many PID control loops, so many that it can be difficult to put information into the appropriate context. AAE is among numerous metrics and diagnostic tools that enable production staff to analyze large amounts of process data and to uncover valuable insights. It empowers engineers to remain one step ahead.