Purpose The study aims to develop an optimization methodology for reinforcing panel-type structures with the directed energy deposition (DED) additive manufacturing process. The primary objective is to minimize the mass of the panel-type structure while satisfying bending stiffness and natural frequency constraints, providing a decision-support method for determining the optimum stiffener design for DED applications. Design/methodology/approach A representative panel-type structure was selected, and stiffener paths were planned and represented using Seniw’s Matrix. The matrix is flattened immediately for processing through a fully connected neural network (FCNN). FCNNs were trained to predict key mechanical responses, such as mass, bending stiffness and natural frequency, based on the stiffener configurations. A genetic algorithm (GA) was employed to perform the optimization, using the trained neural networks (NNs) to compute the objective and constraint functions efficiently. Additionally, a DED process simulation was conducted to evaluate distortion resulting from the thermal input during stiffener manufacturing. Findings The proposed optimization framework successfully identified the optimum stiffener path that met both bending stiffness and natural frequency constraints while minimizing mass. Moreover, the framework also serves as a surrogate model in applications where finite element models cannot be easily parameterized and analytical solutions are not available. This results in a substantial computational time advantage. Whereas the NN-based optimization completes in roughly 30 min, direct FEA-based optimization would require nearly 300 min per generation. Implementation of the proposed optimization approach led to a 43.48% improvement in stiffness with a marginal 5.28% increase in mass. Originality/value This study introduces a hybrid optimization approach combining artificial NNs and GAs for optimum stiffener path in DED-based additive manufacturing. By representing each stiffener configuration with a method such as Seniw’s Matrix, it becomes possible to process these combinations through NNs with high accuracy, thus becoming a useable surrogate model for structural optimization. The proposed framework provides a valuable tool for lightweight and performance-oriented design of panel-type structures in aerospace, automotive and defense applications.
Aydoğdu et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: