STATISTICAL-ANALYSIS OF IMPERFECTIONS IN STIFFENED CYLINDERS

Despite several recent experimental and analytical programs on shell buckling, there is a lack of studies examining the characteristics of imperfection patterns produced by manufacturing. As a result, differing conclusions can be reached in strength assessment and in evaluating code recommendations, depending on imperfection-modeling assumptions. A number of small-scale stringer-stiffened cylinders, all manufactured by the same method, were tested at Imperial College, London, England. Prior to loading, these models were carefully measured to record their initial imperfections. Data-reduction techniques were subsequently used to obtain Fourier-series representations of these surfaces. The purpose of this paper is to present the results of a detailed statistical analysis carried out on the Fourier coefficients. Both univariate and correlation statistics are presented for the variables characterizing the imperfection modes. In addition, the fitting of probability distributions is discussed and relevant parameters are presented for the dominant modes. The results show that because of the common manufacturing process several trends exist in the imperfection patterns. Thus, guidelines are developed that improve imperfection-modeling assumptions in bucking-strength analyses.