Statistical process control - Use of control charts

A process transforms inputs, such as raw materials, into outputs. Outputs can be physical products or services. The resulting product, whether physical or not, has characteristics, some of which are important to the customer or user. This transformation may vary depending on the settings of various adjustable parameters in the process or on the level of other factors that may be involved (e.g., outside temperature).

Mastering the process means being able to predict product characteristics given the level of all parameters influencing the process. It also means being able to adjust the adjustable parameters of the process according to the level of influencing factors so that product characteristics meet expectations.

Among the factors affecting product characteristics, whether these factors are inherent or not, some have a very significant influence, others a moderate influence, and still others a very slight influence. When defining a product, we accept a degree of variability in product characteristics (tolerance range), and we accept that certain factors with little influence on product characteristics may not be identified and, due to their unknown fluctuation, may cause variability in product characteristics, provided that the product characteristics remain broadly acceptable.

To ensure that a product complies with specifications, the process operator has two options:

• check all products if possible (in the case of destructive testing, this is not possible);

• having identified all the influencing factors, having made adjustments so that the product characteristics meet expectations, and having verified through sample checks that no other previously unknown factors have emerged, i.e., that the result is consistent with the cause-and-effect relationships already identified.

Based on these definitions, there are two distinct cases for applying statistical process control (SPC):

• The influencing factors have not been fully identified: it is necessary to observe fluctuations in the process, determine when unidentified influencing parameters appear, identify these factors, and find ways to counteract their influence. Technical investigations are required.

• The influencing factors have already been fully identified: it is simply a matter of checking that no new ones appear. A monitoring technique with simple rules for reacting based on observations, without necessarily using control...