Overcoming the curse of dimensionality in composite laminate blending

Abstract

The specific properties of fibre reinforced laminated composite plates can be optimized by tailoring the local stacking sequences to approach the locally required stiffness and strength. Blending enforces continuation of some or all plies from one stacking sequence to another to ensure structural integrity and manufacturability. Effectivity in blending becomes important when the number of variables increases, also referred to as the curse of dimensionality. A CA-based blending strategy is proposed which evolves the local optima to a blended configuration. The optimal design provided by the CA-based algorithm, for a horse shoe pattern benchmark case, is comparable to the state-of-the art blending methods in terms of weight. A new benchmark problem is proposed, to prove the effectivity for blending a large number of sections. The computational effort of the CA-based algorithm increases linearly with the amount of sections, which solves the curse of dimensionality.