MEASUREMENT SYSTEM ANALYSIS

 MEASUREMENT SYSTEM ANALYSIS

There is one aspect of continuous improvement I simply don’t see practiced enough – measurement systems analysis. So, what I want to do in this article is offer an overview of what measurement systems analysis (MSA) is all about.

What Is Measurement System Analysis? 

Measurement System Analysis (MSA) is used to determine the suitability of a measurement system for use. It is crucial to have a well-functioning measurement system so that the data collected is accurate and precise. There are many factors to consider when conducting a measurement system analysis. This paper will discuss the importance of Measurement System Analysis and how to go about completing one.

 

MSA is a process used to evaluate the suitability of a measuring system for use. A measuring system can be any combination of a transducer, signal conditioner, display, recorder, or data acquisition system used to obtain a measurement. A measuring system is suitable if it meets the required technical performance specifications. MSA is used to identify and quantify the sources of variation in a measuring system.

 What Are the Types of Measurement System Analysis?

There are three main measurement system analysis types: attribute agreement, variable agreement, and stability.

• Attribute agreement is a statistical method that assesses the consistency of ratings between two or more raters.
• Variable agreement evaluates the agreement between two or more measurement systems that generate quantitative data.
• Stability assesses the consistency of measurements over time.

Variation

Being able to attack variation is an extremely important aspect of continuous improvement. But variation is a tricky opponent. The variation we see isn’t always what we think it is.  Allow me to explain with a simple diagram.

 

Gage R&R Variation at the top, we see the observed process variation. In other words, this is the data that we’d use to conduct a Process Capability Study where obviously, we’re interested in understanding how our process variation is behaving. Unfortunately, two things make up our total observed process variation, the actual process variation, and the measurement variation. Put another way, the variation we’re observing may be mostly due to the measurement system, and continuing to attack the actual process variation won’t help at all. Our actual process variation consists of short-term, long-term, and part-to-part variation. Measurement variation consists of several characteristics including accuracy, repeatability, reproducibility, stability, and resolution.

Measurement System Characteristics

Let’s spend some time discussing each of these important measurement system characteristics. First, accuracy is the ability of the gauge to measure the true value of a part on average.  In other words, it’s possible for a measurement system to have high variability but still be accurate so long as the average value of the measurements is close to the true value. Next, repeatability, which is a component of precision, is attained when the same person takes multiple measurements and gets the same, or similar results each time. A close cousin to repeatability is reproducibility, the second component of Precision. Reproducibility is attained when other people get the same, or similar results, as you do when measuring the same item. While repeatability focuses on how well you measure something, reproducibility compares your measurement performance to other people’s measurement performance. Next, stability is attained when measurements taken by the same person, or gauge, vary little over time. In other words, it shouldn’t matter what day of the week or time of day it is. We should always be able to measure in an accurate and repeatable manner. Last, but certainly not least, sufficient resolution means that your measurement system provides at least five, more preferably, distinct values in the range you’re measuring Sufficient Resolution.

For example, let’s say we wanted to measure the heights of three children with a scale that only measures the nearest foot. When we did this, our results were 3 feet for child one, 4 feet for child two, and 5 feet for child three. In other words, we only had three distinct values. As it turns out, the key to ensuring we have an adequate resolution is by determining the amount of discrimination our scale needs.

Discrimination refers to the number of decimal places that can be measured by the system.  Increments of measure should be approximately one‐tenth of the width of the product specification or process variation.

For example, let’s say that we’re working with a process that has an upper customer specification limit of 80 mm and a lower customer specification limit of 60 mm. Discrimination GRR. When we subtract 60 from 80, we learn that our tolerance is 20 mm. In other words, this measurement system needs to be able to discriminate to at least 2 mm since 20 mm divided by 10 is 2 mm.

The procedure of MSA: Gage R&R Study

A software program at a thermal control company is programmed to cut a piece of metal to 12 inches. This piece of metal will eventually become housing for thermal control, so it’s imperative that the first piece of metal measures accurately each time. As part of this company’s quality control, they’ve created a measurement system in which line operators randomly pull pieces of metal off the line to measure them with a digital length gauge. This helps to ensure the machine’s ability to accurately cut the metal.

But how do these operators know that they can rely on their digital length gauge? In this case, the company decides to perform a Gage Repeatability and Reproducibility Study (Gage R&R).

Step 1: Determine the Type of Data Collection 

In this case, the manufacturing company wants to know if there is any variation in each piece of metal’s measurements. This is called variable data, which means the potential exists to have measurements that vary between samples.

Step 2: Sample Collection and Operator Selection

The next step is to collect a random sampling of the sheet metal during any given production run. It’s important to obtain at least 10 samples. Once the samples have been randomly chosen, recruit three operators who routinely complete the measurement system process to participate in the study. Before the study begins, the sampled sheet metal pieces are labeled with their appropriate lengths without the operators being aware of these labels.

Step 3: Measurement Process

For this example, the random sampling includes 10 samples of sheet metal casings. Each operator will measure the sample casings and record their data. Each operator will measure the same random sampling of ten sheet metal casings three times, for a total of thirty measurements. Lastly, the study organizer will rearrange the sample set between each operator to remove any potential bias.

Step 4: Calculations

Once the operators have completed all three rounds of measurement, the study organizer will compare each set of measurements to three evaluation areas. First, the organizer will compare each measurement to a master value. Second, the organizer will compare each operator’s measurements across all three rounds, essentially comparing each operator to themselves. This is called ‘within’ variation. Last, the organizer will compare each operator’s measurements to the other appraiser’s measurements. This is called ‘among’ variation.

When the operator compares each variation measure, they’re looking for any potential measurement error. If the ‘within’ variation varies greatly, there is likely inconsistency in the process the operator uses to measure the sheet metal casings. If the ‘among’ variation varies greatly, there is likely inconsistency in how each operator was trained to measure the sheet metal casings. 

Once the organizer has compared the variation measures, they’ll begin the calculation process to identify the following information:

• Mean readings for each operator
• Standard deviation for each operator
• Differences between each operator’s average and standard deviation 

Here, the organizer is looking at the distribution of the data. If all the numbers stack close to the desired mean, in this case, twelve inches, that means the operator, the measurement process, and the measurement tools are working properly. This is called accuracy and usually means everything is right on track.

When to Use Measurement System Analysis?

Measurement System Analysis is a statistical tool that can be used to assess data quality. This tool can be used to determine measurements' accuracy and identify sources of error. Measurement System Analysis can assess data quality from various sources, including surveys, experiments, and observational studies.

Measurement Systems Analysis

We’ve covered a lot of terms and concepts so far which may make you feel a little overwhelmed. The good news is we have an extremely powerful tool at our disposal that wraps everything that we’ve discussed up into a single statistical tool called Measurement Systems Analysis, or MSA for short.