Why You WANT Variable Gage Analysis (GR&R) Results

TOWARD ZERO DEFECTS COLUMN

BY CHUCK PFEFFER

This past winter, as the season’s first powerful nor’easter began gaining steam off the East Coast, millions of Americans faced the prospect of the first major snow in over two years. The caveat: forecast uncertainty in cities like Washington, DC; Philadelphia; and the coastal boroughs of New York City.

Even the slightest storm track shift east or west meant a radically different outcome of rain versus snow, demonstrating unequivocally that quality forecasting comes down to quality data. The more information meteorologists have about any storm, the lower the data variability and the greater the forecast precision and accuracy.

Data variability is something manufacturing metrologists understand well. Like a weather forecast, a gage analysis is only as accurate as the data fed into it. Without accurate data, the variation in values becomes too great to be of any service. Imagine if the National Weather Service called for two to 24 inches of snow and a high temperature of 28 to 60 degrees F.

Similarly, the variance in a gaged device must be narrow enough and the granularity of information collected deep enough to ensure that what’s being measured is correct. The expression may be “you can’t manage what you don’t measure.” But for metrologists, it’s more precise to say, “You can’t manage what you don’t measure, accurately.”

Rarely do non-metrologists consider this basic measurement need. Whether it’s a gas gage on a car or boat, a blood pressure monitor, or a micrometer or caliper, readout reliability depends on how rigorously it’s been tested. This is their true gage of success for metrologists.

This is where gage repeatability and reproducibility (GR&R) studies earn their worth.

Why Variability Is OK

The American Society for Quality defines GR&R as the process used to evaluate an instrument’s accuracy by ensuring its measurements are repeatable and reproducible. It’s a statistical analysis used by engineers and product specialists to improve measurement consistency and accuracy from three touch points: variability in the product itself, repeat accuracy of the measuring device used by the same operator, and repeat accuracy of the measuring device used by a different operator.

In a typical GR&R study, three operators measure 10 parts two times. Taken in whole, this is what narrows measurement variability in both amplitude and magnitude into acceptable ranges. Thus, the process isn’t some perfunctory pre-step to product production or a box to be checked off a preassembly to-do list; it’s the essential first step in any manufacturing process.

Too often, manufacturers avoid GR&R because customers misinterpret the data. Without understanding how the process works, they focus on the accuracy of the device performing the measurement instead of the process being measured.

However, the difference shouldn’t be difficult to grasp. Think of it as a souped-up version of the carpenter’s adage: “Measure twice and cut once.” By applying proper GR&R principles, manufacturers can ensure their quality control is up to par.

There are several methods for conducting a GR&R study, but there are two widely accepted methods. They won’t generate the same results, but they will (in most cases) be similar:

• Average and Range. M appraisers measure N items R times to estimate quantities such as total variation, precision-to-tolerance ratio, standard deviation of measurement error, and percent of study contribution from various error components. (This is according to a 2017 report by Statgraphics Technologies, a subsidiary of Statpoint Technologies Inc. that develops and markets Windows software for statistical analysis).
• Analysis of Variance (ANOVA). A collection of statistical models and their associated estimation procedures used to analyze the differences among means in a sample. Sources of variation include the process itself, sampling, or the measurement system.

A common misconception by inexperienced analysts is that you should have no variance whatsoever. However, the results of a GR&R study will inevitably vary – but that doesn’t mean they’re unreliable. In fact, a GR&R study that reveals no variance is flawed. Why? Because no matter what you measure and no matter how many environmental controls you account for, variability is an unavoidable state of nature.

Like an experiment, you must control for variables you can account for. Very often it’s the little things that matter most. Something as simple as not tightening the arm on a portable coordinate measuring machine (CMM) table, for instance, can throw off results.

There are established ranges of acceptability confirming that what you’ve measured is accurate, both from an operator perspective and a gage readout perspective. Provided you’ve set it up correctly, you’re looking for “percent tolerance” or “percent process variation.”

Per the Automotive Industry Action Group (AIAG), a not-for-profit organization that sets standards for ensuring quality, ranges under 10% are acceptable. Ranges of 10% to 30% are marginal or acceptable for some applications. Ranges over 30% are unacceptable.

But Not ALL Variability is OK

What you don’t want is a flood of products that flirt with the upper range of acceptable variance. That’s especially true when ramping up output after a period of relative dormancy. After all, whenever you take a process dry from nearly completely shuttered to open again, or from low-speed to high-speed ramp-up, quality control and/or production challenges are inevitable. Before successful ramp-up can occur, a team of engineers should review the prelaunch process.

That’s why, as the world continues to emerge from its socially distanced cocoon and the global economy kicks into a higher gear this spring, the GR&R is more important than ever in ensuring customers receive products that meet quality specifications.

This is true even as the process continues pushing the boundaries of what’s possible in precision measurement with advances in noncontact point cloud data acquisition, such as our Quantum ScanArm, a technology that will require new GR&R studies. Benefits include faster, more accurate, and actionable 3D documentation; faster execution and measurement times; minimized scrap and rework costs; and less product development risk.

With a little luck, as spring turns to summer and summer to fall, the world will be in a vastly better place. Meteorologists will be focused on new data – hurricane models and storm path projections – where the most subtle changes can mean enormous differences in expected outcomes.

From my perspective, it seems that many professionals, in many walks of life, could learn a lot from meteorologists and apply the logical underpinnings of GR&R studies to their work – the unsung heroes of our measurement-obsessed world.