Multi-objective optimization of composite stiffened panels for mass and buckling load using PNN-NSGA-III algorithm and TOPSIS method

A novel multi-objective optimization framework for composite stiffened panels is proposed in this study, leveraging a combination of the Parallel Neural Network (PNN), Non-dominated Sorting Genetic Algorithm-III (NSGA-III), and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method. This framework demonstrates high efficiency and accuracy in obtaining the optimal design for intricate optimization challenges. The PNN in this framework, leveraging data-driven methods, addresses the limitations of Classical Laminate Theory (CLT) in constructing optimization surrogate models, such as challenges in parameter range determination, lack of independence, and the necessity for secondary inverse problem solving. In contrast to NSGA-II, NSGA-III which uses reference points and correlation operators achieves more uniform and rich Pareto fronts under stacking sequence constraints. Additionally, to minimize the required effort and expert knowledge in selecting optimal design parameters, this framework incorporates the Entropy Weight Method (EWM) and TOPSIS method. EWM calculates the entropy of optimization objectives from all alternatives in the Pareto front, assigns weights accordingly, and employs TOPSIS to rank the closeness of each alternative to the ideal solution, thereby identifying the optimal design.