The greatest challenge for any analysis that includes geometric imperfections is to decide which imperfection might be the appropriate imperfection for a certain problem. The ‘‘worst’’ imperfection (i.e. imperfection which leads to the lowest failure load) is usually considered to be the appropriate one. While this assumption is always on the safe side, there is always some uncertainty about whether another imperfection may exist that could lead to a worse outcome, and the possibility that this worse imperfection could possibly arise in the fabrication process. On the other hand, the ‘‘worst’’ imperfection of all might well be quite unrealistic, leading to very uneconomic designs if implemented (Rotter, 2004). At the same time the ‘‘worst’’ imperfection must be identified as having a specific form with a corresponding amplitude. This means that a certain imperfection shape might only produce low failure loads for certain amplitudes and another imperfection shape could take over as the ‘‘worst’’ imperfection shape for other amplitudes (Song et al., 2004).