Generalized probabilistic response surfaces for the buckling strength assessment of stiffened panels

The ultimate compressive strength of stiffened panels has been well studied and a variety of simplified buckling formulas are available. However, there is a significant degree of uncertainty that is associated with the parameters that affect the buckling problem and are currently not accommodated within the deterministic response surfaces. This work proposes a computationally efficient framework that is based on probabilistic methods, experimental design and inferential statistics for the construction of generalized buckling response surfaces that are founded on an acceptable risk level defined by the user. The process is demonstrated on a marine stiffened panel that includes Tstiffeners. Finite Element models have been employed as the mapping function between the uncertain input variables and the response quantity. It is shown that surrogate modeling could not be used to accelerate the problem and therefore a limited number of computationally demanding simulations combined with formal statistical theory may be efficiently used. The constructed response fronts are associated with a probability of non-exceedance and may be further used within a reliability-based analysis and design scheme.

Automated summary

This work proposes a computationally efficient framework based on probabilistic methods, experimental design, and inferential statistics for constructing generalized buckling response surfaces. Finite Element models are used to map the uncertain input variables to the response quantity, and it is shown that surrogate modeling is not applicable for accelerating the problem. The constructed response surfaces are associated with a probability of non-exceedance and can be used in reliability-based analyses and design schemes.