Optimization Design Method for High-Aspect-Ratio Composite Wing Based on Neural Network and Genetic Algorithm

The optimization of structural parameters is crucial to achieve enhanced performance and efficiency in the composite wing. Nowadays, the neural network algorithm offers a promising approach for capturing intricate relationships within the complex design space. This paper introduces an optimization design method for the guidance of the high-aspect-ratio composite wing based on neural network and genetic algorithm. Based on a typical configuration of composite wing, eight internal structure layouts are considered during optimization for maximizing the strength-to-weight ratio. A two-stage optimization method is proposed with different machine learning methods to assess the predictive efficacy concerning the failure load and weight of a composite wing. The response surface model established by the back-propagation (BP) neural network shows the highest prediction accuracy of 94%. The stress analysis shows that the double back C-beam structure obtains the highest loading efficiency, which is selected as the layout scheme in subsequent optimization. The developed response surfaces, integrated with conventional constraints, constitute an objective function for a design optimization model through genetic algorithms. The optimization process focuses on the positioning of the web, along with the thickness and stacking sequence of the upper and lower edge plates, leading edge, and skin. Design optimization instances are examined to validate the constructed framework and demonstrate the reduction in the structural weight of the composite wing. It reveals a significant improvement of 146% in the strength-to-weight ratio, which facilitates an efficient decision-making process in pursuit of superior aerodynamic and structural characteristics.