The advantages of 'confidence/reliability' calculations are explained. Such calculations are demonstrated for Attribute data (pass/fail, yes/no data) as well as for variables data (i.e., measurements). If variables data is 'Normally distributed' the calculations are extremely simple. The webinar explains how to handle 'non-Normal' data, and provides the methods, formulas, and tools to handle non-normality.
The webinar includes a discussion of how one OEM manufacturer has implemented 'confidence/reliability' calculations instead of AQL sampling plans for all of its clients. And suggestions are given for how to use 'confidence/reliability' QC specifications instead of 'AQL' QC specifications. The use of 'reliability plotting' for assessing product reliability during R&D is also discussed. The webinar also includes an examination of ISO and FDA regulations and guidelines regarding the use of statistics, especially in regards to Sampling Plans.
Why You Should Attend: Almost all manufacturing companies spend time and money to inspect purchased parts upon receipt, in order to evaluate part quality before the parts Supplier is paid. 'AQL' sampling plans are used almost universally for such inspections. However, AQL plans actually provide very little information about part quality.
A better way to assess the quality of purchased parts is to use 'confidence/reliability' calculations. Such calculations are very easy to perform using tables and/or an electronic spreadsheet. ISO 9001 and ISO 13485 requirements include establishing 'processes needed to demonstrate [product] conformity'; FDA's GMP (21CFR820) requires that 'sampling methods are ad-equate for their use'. An AQL sampling plan does not provide what is needed to meet either of those requirements. FDA guidelines state that 'A manufacturer shall be prepared to demonstrate the statistical rationale for any sampling plan used' --- it is not possible to 'demonstrate' that an AQL sampling plan ensures product quality.
On the other hand, confidence/reliability calculations can be easily shown to provide evidence of product quality, and the statistical rationale for such calculations is easy to explain and demonstrate.
Areas Covered in the Session:
AQL and LQL sampling plans
ANSI Z1. 4
Confidence/Reliability calculations for
Normally-distributed variables data
Transformations to Normality
Normal Probability Plot
John N. Zorich,