I am in auto manufacturing and for a small wheel hub we make, we are checking at a frequency of 1 out of every 10 parts. We have 4 spindles doing the same thing, so essentially one operator is constantly gaging. My boss wants to reduce the amount of inspection to perhaps 1 out of 20 or more, and is asking me to put together the data to present to the customer, in order to get their approval. I am trying to go through the logic in my head, and I can't quite get it. Is the frequency of inspection dictated by whether or not you can detect a shift in the process with that specific frequency? My boss, who like to use the word capability a lot, wants me to show data from several months and if capability is good, reduce inspection.
blairhogg 12/14/2004
Try something like this - Select at random several samples of different weights, up to the maximum weight to be measured. Identify these samples (1,2,3,4). Label these a,b,c,d, keeping trakc of which sample (a-d) is which sample (1-4), then have someone weigh them and record the weights. Change the sample id's, again keeping track of the relationship, and repeat. Do this a few times. You should now have several sets of data for each sample. Perform a Gage R&R on each sample set independantly. They may be different, due to the different weights involved. You might find the scale is a percent of maximum in error rather than a percent of actual weight.
firebrew 12/13/2004
Before you fall into the trap of "every problem looking like a loose nail", check out some of the very simple yet effective techniques in the text "Evaluating The Measurement Process" by Wheeler-Lyday. Don Wheeler does an elegantjob of presenting analysis of gaging systems free of specifications. In fact, he proposes that for the true purpose of gaging selection, specifications are not always the best measure. An analysis of gage stability, and bias may be more important than the traditional R&R. In short, decide what you are trying to learn, and select your analyses based on that (those) questions rather than simply doing a "Gage R&R" because some text book or ISO Quality Standard suggests it.
John Bruman
bobdoering 12/11/2004
Gage R&R is a very powerful tool - it collects ALL variation and anayzes the between ("appraiser variation") and within ("equipment variation") variation. By using the various containers of vast variation in weight, the tool will show you have this variation - and will likely mask any other variation from your gaging system. You Between-sample variation (AV) will be pretty big.
Generally, this variation of the measurement system is then compared to the specification (or control limits, if used) to determine if it provides adequate resolution - or answers the question "how do you know it is the right gage for the job". If what you are measuring has no specification, then what is it you are trying to show with the gage R&R? In order to answer this question, we probably need to get a better idea what the scenario is. What is the specification - the amount of material that is required in the process and its tolerance? How do the containers relate to the specification – are several used for the add to the process?
The one thing I can assure you that you should NOT do is use one operator! You should have the variation of measurement skills represented in a Gage R&R - from certified trainee to experienced operator. The only time this tool should be used with one operator is if you are trying to determine the "between" variation of several gages, such as 1 operator, 3 gages, 10 parts instead of 1 gage, 3 operators, 10 parts. This is analytical process that is rarely used correctly, however.