Authored by:
Craig Tallar
Product Cost Engineer
Milwaukee, Wis.
Edited by Leland Teschler

Most engineers are familiar with the practice of Statistical-Process Control (SPC). SPC is good for getting a feel for when a process is under control. But there can be a problem with making parts to limits specified by SPC. The tendency is to manufacture parts to the mean value specified in SPC charts when it would both be less expensive and a better design practice to target Maximum Material Conditions (MMC) instead. In GD&T, the MMC refers to a surface associated with a size dimension that stays within its tolerance zone but contains the greatest amount of material allowable.

SPC started out as a concept that helped operators determine whether processes are under control. There is nothing wrong with using SPC this way. However, use of SPC has led some manufacturers to now make products with all their dimensions targeted at the mean limit derived from SPC studies rather than toward the MMC. There are two key reasons such a practice is problematic. First, it reduces the amount of material in the component. Consequently, there can be a commensurate reduction in the greatest strength the component can exhibit. Second, getting the dimension down to the mean value can require additional machining, which adds to the cost and potential scrap or rework of the component. This extra work is costly and unnecessary when the feature can accept a large variations in its size while its relationship to other features is controlled.

I have witnessed various corporations introduce ISO 9000 along with SPC. I have found many have converted from a feature specification using the MMC and the minimum material condition (LMC) to one using the “mean material” condition with a plus and minus deviation. Use of a mean value rather than maximum and minimum dimensions neatens up the statistical table of dimension data. It also creates presentable charts with even distribution curves. However, it does not help the person who is manufacturing the features, or aid in cost containment.

It has been my experience that organizations begin using this “mean value” with equal plus and minus deviation when they introduce SPC. It is fair to ask who benefited from this change. It seems the only people benefiting are the technicians or engineers who collect the data to produce the quality documents and graphs for presentation.

I have found universally that the people who make the product and the engineers who design and specify the limits of the product prefer the MMC and LMC rather than using a mean value. Granted, presenting a mean value on a drawing does provide a “clear understanding of the limits” and target without any required calculations by the manufacturing operator. The MMC then becomes the target. Especially in machining metal, this practice provides a cushion if, for some reason, the tooling, program, material, and machine variables all happen to be in the direction that would result in not removing the calculated material. It is better to leave extra material rather than remove too much. Yes, this does now require another cut path, but a necessary removal of material is more economical than removing it unnecessarily or reworking or scraping the product.

For parts with tolerances of 0.020 in. or less, there may be a negligible difference between targeting the mean dimension rather than the MMC. However, an MMC target is still beneficial when one considers all the variables that can develop with regard to materials, the manufacturing processes involved, operators, the environment, and so forth.

Consider a product that is extremely large and which can accept a tolerance of 0.200 in. or more, relatively large as tolerances go. The quality of the features can be addressed with GD&T. However, the removal of material in a large area shouldn’t be undertaken lightly. It can be difficult to reach some locations, and it may take repeated cuts by an expensive automated machine to hit the mean of the limit. This not only takes additional time at a high cost rate, but also adds delays and a need to dispose of more material waste.

Of course, the chances of having to scrap the part is slim, with all that material stock to work with. However, the extra cutting also boosts the chance of tooling damage from tool wear and cut time. To reduce the chance of tool breakage, many operators would reduce feeds and speed which only adds to cycle time and lengthens delay.

When I have discussed product specifications with designers and engineers, I have advocated the principles of ISO 9000, manufacturability, quality, and using the greatest range of size that the product will accept. This practice makes it possible to produce a part to spec through multiple methods using operational discretion, rather than through potentially costly processing simply to meet a noncritical constraint. And this all begins with simple feature limitations of MMC and LMC on the drawing.

Use of MMC and LMC entails no misinterpretations or miscalculations. However, SPC charts will not have the preferred bell curve. So what? The bottom line is whether or not the part gets manufactured within specification, whether it is made with greater integrity, and whether it is produced efficiently. It is hard to argue that a part meeting these criteria is incorrect even if it doesn’t come close to target values on an SPC curve.

Most SPC practitioners who chart parameters today use a program that requires only the MMC, LMC and the actual measured values. So why present the mean limit on the SPC document? Is it because the default tolerances are given in plus-and-minus values? If so, that’s a poor reason and only causes greater error. Most, if not all, default block tolerances are in numbers that are simple to add or subtract.

Another factor that can cause incorrect numbers and values on a design is the application of ANSI or ISO limits and fits per ISO 286:1988. (This standard is identical to European standard EN 20286:1993 and defines a system of linear dimension tolerances, deviations and fits. It can be applied to tolerances and deviations of smooth parts and to fits created by their coupling. It is used particularly for cylindrical parts with round sections and in smooth parts of other sections.) These design guidance tables are not in a plusand- minus format. They can have different values in one direction or in both directions. Manufacturing personnel can easily review and understand which fit was selected. Therefore, it is easy to verify the designer’s intent.

So this is what I advocate for all manufacturers: Use the greatest limits the application, situation, or assembly can accept that provides the function the designer intends. Provide the MMC and LMC clearly to avoid miscalculations by other parties who may not have the best lighting, clarity of document, or environment in which to perform the calculations. Process the features with the least amount of material removal which will give the best product. All in all, letting SPC precede MMC is placing the cart before the horse.

Finally, it might be said that SPC practices have colored aspects of manufacturing other than those associated with making parts. Consider, for example, the hiring of personnel. I submit that some companies have stopped using as their hiring target the center of the expertise bell curve. They seek the second or third quartile of the brightest and the best. The average candidate is not good enough.

However, if companies based their hiring strictly on salary requirements, they would seek those willing to work for the lower second or third quartile. The two views are hard to reconcile. The moral: SPC is only good if your total part tolerance is 0.020“in., or 0.5“mm. It doesn’t work well for hiring decisions.

© 2012 Penton Media, Inc.