Process capability studies in theory and practice

Abstract:

The existence of variation has been a major problem within industry since the early days of the industrial revolution and perhaps even earlier. The fact that two parts not ever will be identical, forces every organisation to find a strategy for how to master variation. Process capability studies, a method designed to judge whether a process is capable or not, often plays an important part in such a strategy.

The concept of process capability studies has received both positive and negative criticism during the last decade. For instance, the supporters of process capability studies emphasise the importance of using the method to identify improvement priorities to be focused in the overall improvement process within an organisation.

However, as all methods, process capability studies has its limitations. Actually, it is not principally the method as such that has been criticised, but rather the measures of capability used when conducting process capability studies, the so called process capability indices. All existing process capability indices have some weaknesses, even the most sophisticated indices have relatively poor statistical properties which might lead the user to make incorrect decisions, even if most theoretical aspects of how to conduct process capability studies are known by the user. The use of process capability indices is for instance partly based on the assumption that the process output is normally distributed, a condition that is often not fulfilled in practice, where it is common that the process output is more or less skewed.

This thesis focuses on process capability studies in both theory and practice. In part 1 of the thesis some theoretical aspects of how to conduct process capability studies are identified and then the adherence to these aspects within Swedish industry is investigated. This study reveals that there are certain gaps between how process capability studies are supposed to be conducted according to theory and the way they actually are carried out in practice. The study also tries to explain why these gaps exist, by analysing common obstacles when implementing and conducting process capability studies.

In part 2, a simulation study focusing on the effects of skewness on estimates of some process capability indices belonging to the family of indices named Cis presented. The effects of skewness are studied in three different cases, one incapable case, one case just capable and one very capable case. In all cases, four lognormal distributions with different skewness are used. The results from the simulation study indicate that the effect of skewness is relatively systematic, and therefore there are some hope that future investigations might use these results when formulating some practical solution to the problem of how to use process capability indices when the process monitored has a skewly distributed output.

Finally, the results are summarised and discussed and some suggestions for future research are given.