If you polled a large number of quality gurus around the world and asked them to list the top ten concepts that a quality professional must understand, at or near the top of the list would be the difference between customer limits and process limits.
Tell me if this makes sense to you: A process can be in perfect statistical control but be completely incapable of meeting the requirements of its customers. If you understand this statement, you can stop and browse the web. Just make sure you are helping others to gain this knowledge.
A process will run the way it is set up to run. It has inferent, "common cause" variation that governs its behavior every working day of the year. Every now and then, "special causes" occur that interrupt the normal behavior of the process. These "out of control" conditions must be discovered and eliminated. Once special causes (or as many as can be eliminated) are eliminated, the process is said to be in statistical control and governed by random behavior.
Process limits set the boundary between common cause and special cause variation. These limits can't be arbitrarily set but must come from the process' historical behavior. If you set the limits any other way, you are holding the process to a standard that it was not set up to meet. It is analogous to demanding zero defects from a supplier when you refused to pay them up front for preventive techniques designed to garuantee zero defects. You can ask for zero but the supplier was not set up that way on the purchase order.
Notice I have not mentioned a tolerance or specification yet. Before you can start talking about a process' ability to meet a customer's requirement, you need to know that the process is in control and stable. If it is out of control or constantly moving, how can you truly say that a customer's requirement can be met?
Once the process limits are calculated and taken to be sacred, you now overlay what the customer wants onto the process limits. If the customer's limits are wider than the process limits, this suggests that the process is capable of meeting customer requirements. If the customer's limits are tighter than the process limits, you are in for a mess. This means the process will not be able to consistently meet the customer's requirements.
Golden rule #1 of Quality Engineering: Don't confuse customer limits with process limits. Don't use customer limits to calculate process limits. Let the process speak for itself.
Subscribe to:
Post Comments (Atom)
Posts
2 comments:
Just as you mentioned, the process should speak for itself. However, what would be a standard practice of measuring for process stability? I know that you must demonstrate a trial run prior to collecting data and once this has been completed and the process is now being monitored for the data collection- where is the standard to provide the customer this vital information to look at the costs of quality? If the customer did not ask for zero defects without calculating it into the price of a product, most suppliers would not show that they are not capable of producing a product.
We can venture into a discussion on advanced quality planning (APQP). If a customer wants zero defects, then they must ask for it upfront and pay for it. Asking for it mid stream and expecting it for free is not realistic. If they ask for it at the outset, processes can be equipped to achieve it. (mistake proofing, etc.)
A standard for measuring process stability: I usually want a supplier to run his process for a minimum of four hours. During the four hours, he collects the data that will serve as the voice of the process. Ideally, you want to conduct your study to capture as many sources of variation as possible (part to part variation, different shifts, different operators, tool changes, etc.)
My experience is that you never really eliminate all special causes. So, one could argue that you never reach statistical control. But, the frequency of occurence for special causes should be so low that statistical control (and process stability) is all but in place.
Post a Comment