Multi-Objective Optimization of Composite Structure Using Rule-Based Parametrization

Ultra-light unmanned aerial vehicles (UAV) benefit from use of composite materials due to weight savings and ease of manufacturing for piece production. However, complex shape and internal structure of wings limits the use of hand calculations that can be employed to optimize the structure. Therefore, design process can greatly benefit from the FEM (Finite Element Method) calculations coupled with CFD (Computational Fluid Dynamics), composite mechanics and composite failure theories. This multi-physics approach allows to accurately describe behavior of a wing during flight. Due to nonlinear response of a system after changes in wing design and flight characteristics, we present a method of optimization using artificial neural networks. This allows to accurately describe influence of given parameters on the stress and strain distribution as well as reduce number of design points. Material data has been gathered from experimental tests of simple specimens. Based on this data more complex elements were designed using FEM and tested experimentally in order to validate numerical calculations in a transdisciplinary rapid prototyping exercise. Advanced failure criteria not only predicted failure but also failure mechanism, thus catastrophic failure can be prevented. In the future this multi-physics approach can incorporate numerical analysis of manufacturing and curing process therefore reducing need for experimental validations. Further development of neural networks will lead to them being directly implemented into FEM codes.