An Adaptive Optimization System in Multi-Scale and Multi-Objective Wing Structure Using Deep Learning Algorithms

In the aerospace industry, wing design involves multidisciplinary considerations such as material science, aerodynamics, structural architecture, and manufacturing feasibility. The balance between these primary issues and corresponding design parameters is usually hard to achieve. Optimization is hence an essential procedure to assure maximum design efficiency. The conventional optimization method relies heavily on iterative modeling, testing, analysis, and improvement cycle. This process can be time-consuming, costly, and imprecise due to the complexity of the operative conditions. Here we propose an adaptive optimization system built upon a digitized wing structure database and machine learning algorithms, which enable the rapid formulation of multi-scale<b> </b>wing designs for specific performance demands. This system unfolds as two stages of work. The first stage is oriented toward aerodynamic optimization, where Computational Fluid Dynamics (CFD) simulation is exploited to attain performance outputs corresponding to an extensive collection of geometric inputs. A neural network is subsequently utilized for parametric study and finding the optimal wing configuration. The optimization has successfully improved the aerodynamic efficiency of a fixed-wing aircraft by over 4% compared to the average performance of initial geometric inputs when cruising at 0.48 Mach. The second stage work concentrates on the definition of best stacking sequences with composite materials adopted and on the wing box layout. The multi-objective optimization adapts a deep learning schema that lies upon a series of critical structural requirements, such as specific strength, shear flow, cost, margins of safety, etc. To this end, the proposed methodology aims to significantly reduce the computational cost and contribute to better flight performance. With all the merits addressed, the proposed approach will possess great potential for optimizing several aircraft’s wing-like sub-components when concerned with different mission environments and load conditions.