This paper has developed an efficient nonlinear finite element method that covers both initial deformations and initial stresses of general distribution in calculating the ultimate strength of ring-stiffened cylinders. The developed method and two world-widely used commercial codes(NASTRAN and ABAQUS) were simultaneously applied to the same analysis model within the extent those commercial codes can cover, to check the validity of the present method. After the validity check, it was used for parametric studies for more general case of initial stress distribution, which drew some useful information about the imperfection sensitivity of the ultimate strength of ring-stiffened cylinders.
While ring-stiffened cylinders are very useful structural elements used as a submarine pressure hull or members of offshore structures, their strength formulas are largely based on limited test results rather than theoretical background because their complex structural characteristics make the theoretical analysis difficult. The most prominent characteristics are that the ultimate strength calculation requires highly nonlinear analysis and that the ultimate strength is highly affected by initial imperfections like initial deformations or initial stresses inevitably involved during the process of fabrication. This paper has developed a nonlinear finite element method that can enhance its calculation efficiency by using axisymmetricity of circular cylinders while maintaining the function of handling both initial deformations and initial stresses of general distribution in calculating the ultimate strength of ring-stiffened cylinders. In the conventional axisymmetrical analyses(Subbiah, 1988), circumferential integration in making the element stiffness matrix used to be carried out analytically, by which material nonlinearity could not be covered. This way, the distribution of material properties, varying with the yielding process over the surface and through the thickness, can be properly taken into account and also the initial imperfections of any distribution can be easily considered by putting these as an initial condition.
