Akademik Veri Yönetim Sistemi
Engineering Computations (Swansea, Wales), ss.1-18, 2025 (SCI-Expanded, Scopus)
Purpose – This research introduces the novel use of “parametric design” and “Voronoi diagrams” for generating synthetic 3D data to enhance the training of deep learning models in engineering and manufacturing sectors. The primary aim is to develop a methodology that facilitates the creation of complex geometries and improves the model’s capacity to understand and replicate intricate structures. Design/methodology/approach – A method is devised that simplifies the production of complex geometries through parametric design and Voronoi diagrams. This approach enables the generation of synthetic datasets that are used to train deep learning models, enhancing their ability to process and reproduce complex structures efficiently. Findings – The efficacy of the proposed method was validated by training a deep learning model with the generated synthetic data and comparing its performance against models trained with real data. The trained model demonstrated generalization ability akin to those trained with authentic datasets. Furthermore, models trained with this synthetic data exhibited up to a 50% increase in relative safety factor when evaluated using physics-based simulations. Research limitations/implications – While promising, the research is limited by its initial application scope and the synthetic nature of the data. Future work should explore the application of this method across diverse engineering fields and with varying data scales to fully understand its broader implications and potential. Originality/value – This study is the first to integrate parametric design and Voronoi diagrams for synthetic data creation in deep learning training within engineering contexts. It provides a significant advancement in the use of artificial intelligence in rapid prototyping and complex engineering tasks, offering a substantial increase in the capacity to generate and process intricate data structures.