You’ve probably heard the saying “you can’t control what you can’t measure.” How interesting that many of the things we would like to control are difficult to measure!

With foods, you might want a certain flavor, color, or crunchiness. Some plants work hard to control odors. None of these can be easily measured directly.

To make matters worse, for direct process control we can’t wait around for a lab test or a human to provide the measurement. So what can you do?

There are several techniques that can be employed to provide some measure of process control.

You may be lucky, and have a simple process that can use a related sensor to detect some secondary property. For example, in some solutions, salt content can be inferred from measurements of conductivity. Since conductivity is easily measured online, you can get control over salt content.

In a more complex process, you may have multiple process parameters that correlate with the desired result. In this case, you may be able to develop a “soft sensor” based on these correlations. Most soft sensors are simply a least-squares/best fit model of the correlations.

Another approach is to forget about the end result completely, and just try to control the underlying variables more tightly. For example, in turbines, NOx is often “controlled” by carefully maintaining tight control over the combustion temperatures, and the air/fuel ratios or excess O2 measurements.

Each of these techniques has a place in modern process control.

Keep an eye on the instrument market, too. Recent years have seen an increasing number of on-line real-time alternatives to previously cumbersome measurement techniques. ISA’s magazine, InTech, generally follows the sensor market. You can look through their archives at:

So, maybe we need to modify that old adage a bit… “You can’t control what you can’t measure…unless you can infer…or estimate…or find an alternate measure”. Good luck!