Last month, I talked about vision vs. video and brought up the possibility of confusion between inspection and measurement. Some think that inspection is qualitative and measurement is quantitative. I can’t say I’ll put the matter to rest, but I’ll present a few different considerations to think about.

Where you do it
Incoming inspection is a function or a department where purchased goods are inspected to make sure they meet specifications. This is one place where the difference between inspection and measurement comes into play. It depends on what you’re inspecting.

Many times, the criteria for inspecting a purchased item are cosmetic in nature. Is it the right color? Does it have any blemishes or imperfections? Such assessments may not be quantitative or measurable, so you can argue that such parts are inspected but not measured.

Many times, however, some aspect of that inspection process involves measurement. An obvious example is a part that must have particular dimensions to fit or function properly. Part of the incoming inspection process of such a part is measurement of those dimensions to determine whether they’re within specification. In this case, measurement is part of the inspection process.

Consider an extreme case where every part is measured when received. It’s far more likely that the department where this takes place will be called incoming inspection, not incoming measurement. (And it’s easier to call a person an inspector than a measurer.)

What’s inspected anyway?
Let’s go back to that inspection for blemishes. In the case of optical parts, such as mirrors and lenses, blemishes are called artifacts. Holding an optical part up to a light to look for artifacts is a typical inspection technique. This is where the measurement angle comes into play even if the criterion isn’t a measurement. What’s an acceptable artifact?

One could argue that a “go/no-go” test for artifacts (Yes, there are some./No, there aren’t.) isn’t a measurement. That’s true if you consider that the artifacts weren’t measured (for size, for example). But you could also argue that the simple count of optical parts with and without artifacts is a measure. In this case, it’s a measure of how many parts have artifacts, which is easily turned into a percentage. To show how this is a measurement, consider artifact inspection as a quality control process. As an example, consider inspecting 50 lenses. Let’s say 10 of those lenses have artifacts on them. We know that 20 percent of those lenses weren’t artifact-free. This might constitute a tolerable rejection rate. That needs to be determined.

That simple percentage of parts with problems compared to the entire production is a useful figure, and it becomes more useful as a process measure when tracked over time. If 50 lenses are inspected every day, does the percentage of lenses with artifacts stay at 20 percent? If so, the process would appear to be quite stable. But this is where the issue of what’s being inspected comes into play yet again.

Inspecting parts, people or processes?
Let’s stick with the simple optics inspection for artifacts. Suppose the parts being inspected come from a manufacturing process that is stable—unchanging. Every part is made the exact same way. That 20 percent of the parts have artifacts might point to a problem with that manufacturing process. (If only it were that simple!) On the other hand, how many people inspect the parts? If there’s one inspector who follows a defined process and inspects every part in the exact same manner, one variable is removed. This means, of course, that if there are two inspectors, you need to find out whether one of them is finding more artifacts than the other. Are they finding them in equal proportions? Does one hold the parts differently (vary the inspection process)? Is one inspection station near an air vent that might be dropping dust particles on parts waiting for inspection? There are numerous variables even in this simple example. Human judgment is a major variable.

You might think that the single inspector using the same process every day would remove variables in the inspection process that could affect the percentage of lenses with artifacts. No such luck. Consider the day of the week, for example. Studies have shown that inspections performed on Mondays and Fridays pass parts that would have failed if inspected on Tuesday, Wednesday or Thursday. Theories about the reason include problems with getting started after the weekend and distractions in anticipation of the upcoming weekend. Whatever the cause, such variability leads to another measure of that simple inspection process—tracking the pass/fail rate by day of the week and time of day (The same variation in inspection outcomes occurs over the course of a business day).

This simple inspection may not be measuring anything, but the measurement of the results of this inspection process, over time, can tell a lot about several processes. Suffice it to say there are numerous variables to take into account.

Back to measuring
Continuing the artifact inspection example, the question of measurement comes up in the definition of that blemish or artifact. What constitutes an artifact on the part in question? The presence or absence of something can be the determinant of an artifact-free surface. But how far do you go? For example, you might ignore minor imperfections. The problem with that is that minor is too subjective. What’s minor to one inspector might be major to another. Inspection processes try to avoid this ambiguity by providing examples of acceptable and unacceptable artifacts, usually as photographs or illustrations. That method works well until you find an artifact that is on the margin—is it acceptable or not? A solution to this interpretation dilemma is to measure the artifacts.

This is where it gets murky again. If you decide to measure the artifacts to determine what size or type is acceptable and what is unacceptable, you need to decide how to measure them. If the measurement process relies on the skill of the person doing the measurement, it’s unlikely that every inspector will pass and fail parts identically. The performance of each inspector can vary over a week and during a day. Yes, this is a convoluted way of getting to the point, but it’s meant to show that inspection or measurement in and of itself, no matter what you call it, isn’t 100-percent reliable.

Uncertainties such as these lead manufacturers to automated measurement as the solution. An automatic measuring machine should perform consistently, no matter who’s operating it. Assuming it’s capable of measuring the part and distinguishing bad artifacts, an automatic measuring machine can be the arbiter in the inspection process.

Is this ironic?
As these examples show, it isn’t unusual to use a measuring machine in the inspection process. The key point is that inspection and measurement can be highly operator-dependent if the outcome depends on the skill of the person performing the task. Because one person can perform differently from another person and an individual can perform inconsistently depending on time of day or day of the week, even highly trained personnel can introduce variability into the inspection/measurement process. Automatic measurement devices and systems that remove operator subjectivity can be the key to consistent, reliable inspection outcomes.

Next time, if it’s automatic is it always accurate?