Statistical characterisation and modelling of random geometric imperfections in cylindrical shells

Thin cylindrical shells are the most prevalent and important structural component of vessels across the process industries. Such structures are prone to accidental buckling due to inadvertently induced vacuum. Minor deviations in the nominal geometry of the shell can affect the apparent initial buckling load. One common deviation is that the radius of the vessel is not constant but rather varies randomly with location on the shell. This paper presents extensive experimental data permitting a full statistical characterisation of defects of this nature. The data was obtained from detailed measurements of 39 replicate test vessels at the laboratory scale. Both amplitude and frequency content of this type of imperfection is quantified. Furthermore a methodology whereby the variation in radius is characterised as a two dimensional random field is outlined. An algorithm to generate realisations of this field is developed and the output is shown to be consistent with the measured results.