A procedure of approximate evaluation of the critical loads of shells proposed earlier on the basis of the nonlinear theory of shells is used to analyze the influence of initial imperfections of any shape on the parameters of the critical loads. The numerical results are compared with the available experimental and theoretical data corresponding to the complete loss of stability of shells and exhaustion of their load-carrying capacity.Keywords: nonlinear theory, models of shells, imperfections of the shape, critical loads.
Introduction.A significant difference between the theoretical and experimentally established critical loads explains the necessity of development of new procedures of investigation of the process of loss of stability.The first procedures used for the numerical analysis of smooth cylindrical shells in the linear statement were based on an idealized computational scheme [1][2][3][4]. Thus, the shell was regarded as geometrically perfect and perfectly elastic and its initial state was assumed to be momentless.As indicated in [5], the application of the concepts of nonlinear theory can be traced back to S. P. Timoshenko and C. B. Biezeno who used these concepts in the problems of clicking the rods and spherical domes. Later, L. H. Donnell [6] and K. Marguerre [7] finally formulated the foundations of geometrically nonlinear theory for smooth shells. The following important feature of the behavior of compressed cylinders was established in [8,9]: The load decreases in the supercritical stage of deformation.In [10,11], an attempt was made to establish the improved experimental and theoretical values of the upper critical loads by using the linear momentless theory of shells. Despite the application of fairly accurately prepared models, the difference between the theoretical and experimental critical loads was quite large [10, Table 28]. Thus, in particular, the experimental values of critical loads for smooth shells differed from the theoretical values by a factor of 2-3 [10]. Ribbed shells were studied in [10] to get the required agreement between the theoretical and experimental results (see Tables 28 and 29 in [10]). As a result, the difference between the theoretical and experimental data for shells with four frames and increasing number of stringers (24, 32, 40, and 48) was greater than 70% (greater than 60% as compared with the mean experimental values). This means that the minimum theoretical critical loads were, as before, much higher than even the mean experimental values.The attempts to remove this difference were made in analyzing a batch of shells with stringers and frames with much larger cross-sectional areas than earlier [10, Table 29]. As a result, the minimum theoretical values of stresses for smooth and framed shells were 2-3 times higher than the experimental values. For shells with stringers, the experimental values were 1.9 (k =16 stringers) and 1.3 (k = 64) times lower than the theoretical values. For shells with stringers and frames, the theoretical values were twice higher tha...