MSA with Jackie Graham:
Setting up a measurement study
This article is the third in a series in which Jackie Graham, Ph.D., will explore measurement issues and measurement systems analysis. Graham is director of PQ Systems Australia, Pty.
Setting up a measurement study involves looking at the measurement system with respect to equipment, technique, environment, and people involved. This column will address the way in which a simple measurement study can be set up.
Let’s set up a simple study, using an example to illustrate the approach. The example examines how to set up a measurement study for assessing the system used to collect the pH of a product. This is a classic style of study. For your very first study, try to choose a measurement system with which you are familiar, and one where the tests are relatively simple and quick to perform. To set up a basic study the following are required:
One piece of equipment has been calibrated before the study. Make sure the equipment is in good working order.
Example: Use one pH meter. Perform a normal calibration prior to the study.
Two or three testers are required.
All testers should be fully trained in the test method. Ideally, they should complete the test on a regular basis.
Testers should be unaware of the study so the tests are conducted in the normal way.
Ensure that the testers are given adequate time to perform the tests. Be sure that they do not become fatigued.
Example: Three testers are chosen. All use a pH meter as a normal part of their working day.
Ideally, 10 different samples should be used. This number can be reduced to as few as 3; however, the greater number of samples used, the more reliable the results.
The samples should be different from one another and reflect the normal operating range of the equipment.
Example: For the pH assessment, 6 samples are assembled. One sample with a low pH, as acidic as would normally be seen for the test, has been selected. Another sample with a high pH, as alkaline as would normally be seen in the test, has also been selected. The other samples are selected between these two extremes.
Sample selection is the same if parts are being measured rather than solutions.
Each sample is presented to each tester two or three times for testing. So, in the case of pH, each of the samples should be made as homogenous as possible and split into the required number of samples. If two testers are measuring each sample twice, this will be four samples. Each sample will need to be uniquely identified in such a way that it can be easily identified for analysis but is not identifiable by the tester.
When using parts, each tester will measure the same parts two or three times, so they need to be identified in such a way that the testers will not realize that they are testing the same parts more than once.
The sample’s true value should be known, if possible. If an assessment is being made of a pH meter, then standard pH solutions can be used. Comparisons can then be made between the known answer and the results from the study. In some applications, known values are not always available.
Example: Three standard pH solutions are purchased for the study. Their known values are 3, 5, and 7.Three production samples with different pH values have also been selected. Since it is relatively easy to test standard samples, but can be more difficult to test product, both standard samples and product are used to assess the measurement system. The standard pH solutions will reveal any bias issues, and the production samples will ensure that product can be reliably measured.
Each tester tests each sample twice, so six sub-samples are taken from each standard solution and six sub-samples from each of the production samples. Great care is taken not to contaminate the sub-samples and alter the pH value. Each sub-sample has been identified with random numbers, and a note kept of each random number and the related standard solution or product.
Normal operating procedures should be used.
Present the samples to the testers in a random order.
Example: The samples are presented to each tester in a random order. Each tester tests all the samples and records the results on a table indicating the sample’s random number and the result.
It is important when conducting a measurement study to avoid complexity. Keep the study as simple as possible. Otherwise, it can be difficult to ascertain meaningful information from the study data.
The next step is to analyze the data. An upcoming column will address some basic analysis techniques. We recommend GAGEpack for all your measurement system needs.
Copyright 2007 PQ Systems.
Please direct questions or problems regarding this web site to the Webmaster.