Method and system for the statistical control of industrial processes

A method for the control of an industrial process through an electronic processing system associated with a production line, including the performance of a statistical test Φ checking the properties of at least one group of products obtained through the process, whose power provides reproducibility/sensitivity information about the decision making outcome of the test, said method comprising:

• producing at least one pilot sample including a plurality n of exemplars of the product obtained through said process,
• acquiring a plurality of measurement data (X1, . . . , Xn) for at least one observed variable relating to said sample which is indicative of the properties of the product, the measured data relating to the pilot sample having a distribution Fn approximating to the unknown distribution F for the observed variable,
• determining a sample size on which the statistical test should be performed, said sample size being the minimum sample size such that the lower bound
• (
• π
• ^
• F
• n
• ,
• ,
• l
• α1
• ⁡
• (
• m
• )
• )
• of a confidence interval having a predetermined level (1−α1) associated with the bootstrap distribution for the pointwise estimate of the power of the test performed on a sample of that size taken from the pilot sample having a distribution Fn, is at least equal to or greater than a predetermined desired power value, and
• producing at least one sample of products to be subjected to said test Φ, including a plurality m of exemplars equal to said sample size determined, the outcome of the test determining control of the process in accordance with a predetermined criterion.