Enhanced Deterministic Performance of Panels Using Stochastic Variations of Geometric and Material Parameters

Abstract

The effect of stochastic variation in material and geometric properties on structural performance is important for robust design. Knowledge of such effects can be acquired by applying variation patterns to a structure using random fields through a Monte Carlo analysis. The output is postprocessed to show the correlation pattern between the stochastic variation of a structural property and a chosen mechanical response measure. The resulting patterns are used to identify areas most susceptible to variations, as well as areas which have the most potential to increase structural performance by varying the material parameter or geometry.By using these maps of local sensitivity to variations with respect to the structural response it is possible to redistribute material properties or geometry to promote certain behavior. This is demonstrated on a flat plate and curved panel, by either varying the Young's modulus, or thickness of the structure to increase the linear buckling load. In both of these variations the average property is set to remain the same as the original structure. Applying the redistribution increased the linear buckling load by up to 29%.

Buckling Sensitivity of Tow-Steered Plates Subjected to Multiscale Defects by High-Order Finite Elements and Polynomial Chaos Expansion

“…In this context, during the last decades, increasing interest in spatially distributing the uncertainty has appeared. These techniques are known as random or stochastic fields [ 12 ] and have already been applied for the study of aerospace structures, as devised in the works by van den Broek et al [ 13 , 14 ], and Scarth et al [ 15 ]. In the cited articles, material property and geometric variations were induced to investigate how they may affect the free vibration and buckling performance of the structures.…”

“…It is worth commenting that in those cited works, different approaches to stochastic field implementation were considered. For instance, in references [ 13 , 14 , 15 ] the Covariance Matrix Decomposition (CMD) was used, whilst [ 16 ] utilised the Karhunen-Loève Expansion (KLE). More information regarding different implementations can be found in the article by Spanos and Zeldin [ 17 ].…”