Tutorial: What characteristics of a product are most important, and how do you measure them? A recent Ask Control Engineering blog question dealt with finding a sensor that would be able to tell if the liquid flowing through a pipe was water or orange juice. This question brought up the larger issue of defining product attributes and finding an appropriate approach for measuring them. What kind of sensors should you use?

Peter Welander, Control Engineering

Share

A recent Ask Control Engineering blog question dealt with finding a sensor that would be able to tell if the liquid flowing through a pipe was water or orange juice. This question brought up the larger issue of defining product attributes and finding an appropriate approach for measuring them.

This case is relatively simple. It isn’t hard to tell the difference between orange juice and water. If you think about it, there are several obviously different characteristics that can be measured:

• Color and opacity (using simple machine vision);
• Acidity (using pH or conductivity);
• Sugar content (using specific gravity); and
• Thickness (using viscosity).

If you think long enough, there are probably others you could come up with. The question is to identify with the one of those that is the most positive and easiest to measure. In this particular case the suggested answer was conductivity.

Some applications like this one are fairly straightforward. Others are more subtle and harder to analyze. The issue has gained relevance in process industries over the last few years as manufacturers have had to deal with a growing level of variability in feedstocks. This was a discussion point in the article, Solving Process Instability (Control Engineering, June/July 2010). More companies are finding that their traditional supplies of raw materials and fuels may no longer be available or have become too expensive. The material that is substituted may not be the same, and that can cause instability in a process or change the characteristics of the final product.

The first challenge to solving the problem is to figure out exactly how the product is different. What attribute has changed? What is different about the chemical feedstock from Source B when compared to Source A? Or, how has your longstanding supplier changed its process?

Returning to our orange juice producer, let’s say the company normally buys concentrate from Florida, but, for whatever reason, decides to start buying from California. The resulting juice is a little different, and you have to determine what that difference is. Is it a slight change in acidity? More or less fructose? If you have to compensate for the change, even a very subtle change, you have to find a way to identify and quantify it, and that means finding the right kind of measuring device.

Your initial effort to identify the difference or differences may require the help of lab, but once you know what you have to look for, the choice of a sensor on a production level may be yours. Depending on the specificity of the attribute, there may be more than one option. To complicate things further, more than one attribute may be in play. Getting orange juice concentrate from Brazil may add another to the list. Imagine the number of factors that might be in play when evaluating a product as complex as crude oil.

This specificity of matching sensor and attribute is one of the reasons that there are occasional product contamination scandals. If you have no reason to expect that cyanide has been mixed into your orange juice outside of deliberate and malicious tampering, you probably aren’t going to look for it. Performing broad spectrum analysis to look for every conceivable contaminant is not practical on a day-to-day basis with every product. Unless the cyanide is capable of changing one of the other attributes, you may not realize it’s there. This has been the reason behind some recent incidents in the pharmaceutical industry. We’ll assume that the affected companies expanded their ongoing product testing procedures.

However complicated the situation is, it eventually comes back to basic analysis.

- Peter Welander, is content manager for Control Engineering. Reach him at pwelander(at)cfemedia.