ENBIS: European Network for Business and Industrial Statistics

Chance, assignable causes, and statistical control in reliability

21 September 2009, 16:05 – 16:25

Abstract

Submitted by

Thomas Svensson

Authors

Åke Lönnqvist, Thomas Svensson

Affiliation

SP and Volvo Cars, Sweden

Abstract

The great success of the Shewhart concept of statistical process control may be symbolized by the three words: chance, assignable causes and statistical control. The idea to distinguish between chance and assignable causes made it possible to invent simple, but powerful, tools for controlling production processes. The idea of statistical control introduced a new way of thinking that allows for rational treatment of processes that are known to be non-deterministic.

In the problem of design for reliability it is important to take two types of uncertainties into account, namely those caused by scatter and those caused by lack of prior knowledge. The scatter leads to variation in life to failure, but the lack of knowledge results in biased reliability measures. Can scatter be interpreted as chance causes, and lack of knowledge as assignable causes? Can the idea of statistical control be interpreted as the design situation when uncertainties from lack of knowledge are essentially eliminated?

We will discuss if reliability problems such as scatter, model uncertainties, statistical uncertainties, and human errors in design and usage can gain something from the successful statistical thinking represented by Shewhart and his successors.

View paper

Return to programme