Multi-step blended stacking sequence design of panel assemblies with buckling constraints

In this paper, we propose a multi-step framework for design of composite panel assemblies and subsequent blending of the designs to ensure laminate continuity across multi-panel configurations. Multilevel optimisation is frequently used for solving complex optimisation problems. In composite design this approach leads to stacking sequence mismatch among adjacent structural components which is generally referred to as blending problem. To overcome stacking sequence mismatch, a guide-based genetic algorithm (GA) is used which in essence forces the design to be completely blended at any step in the design process. A serious drawback of guide based approach is that it necessitates repeated analysis of the entire structure within the GA iterations. A multi-step framework is proposed where the structure is first optimised using panel thickness and lamination parameters as continuous design variables. The continuous optimisation is performed using a successive convex approximation scheme. In the second step, discrete blended stacking sequences are obtained using a guide-based genetic algorithm. The fitness function in the guide-based GA is evaluated using convex approximations of the response. In this fashion, the cost of evaluating structural response within the GA optimisation is eliminated. The proposed framework is demonstrated via design of an eighteen panel horseshoe configuration, where each panel is optimised individually subject to a local buckling constraint. Numerical results indicate that the present algorithm is capable of producing near-optimal fully blended designs at a small fraction of the computational cost of traditional blending algorithms.