Researchers Noah J. Bagazinski and Faez Ahmed have developed a groundbreaking approach to ship design, leveraging generative artificial intelligence to streamline the traditionally lengthy process. Their work, titled “ShipGen: A Diffusion Model for Parametric Ship Hull Generation with Multiple Objectives and Constraints,” introduces a diffusion model that can generate parametric ship hull designs while considering multiple objectives and constraints, potentially revolutionizing the maritime industry.
Ship design is a complex, time-consuming endeavor that involves balancing numerous trade-offs to create efficient and effective vessels. Traditional methods can take years, but Bagazinski and Ahmed’s research offers a promising alternative. By employing a denoising diffusion probabilistic model (DDPM), they have developed a system capable of generating parametric design vectors for ship hulls. This innovative approach not only reduces design cycle time but also creates novel, high-performing designs that could lead to significant cost savings in both shipbuilding and operation.
The researchers’ study focuses on the generation of parametric ship hull designs using a parametric diffusion model. This model considers multiple objectives and constraints, addressing the complexity of ship design where the hull must meet numerous performance criteria. The DDPM generates tabular parametric design vectors, which are then evaluated for their feasibility and performance. By incorporating classifier guidance, the DDPM produced feasible parametric ship hulls that maintained a 99.5% coverage rate of the initial training dataset, a substantial improvement over random sampling of the design vector parameters across the design space.
In addition to feasibility, the researchers also focused on performance improvements. By leveraging performance guidance, the DDPM-generated ship hulls demonstrated an average reduction of 91.4% in wave drag coefficients and an average increase of 47.9 times in the total displaced volume compared to the mean performance of the hulls in the training dataset. These improvements are significant as they directly impact the operational efficiency of ships. Lower drag coefficients mean reduced fuel consumption and operational costs, while increased displaced volume can enhance a ship’s revenue-generating potential.
The practical applications of this research are vast. By reducing design time and generating high-performing hull designs, the maritime industry can accelerate the development of new vessels that are more efficient and cost-effective. This could lead to substantial savings in fuel costs and operational expenses, making shipping more sustainable and economically viable. Furthermore, the ability to quickly generate and evaluate multiple design options allows for greater innovation and optimization in ship design, pushing the boundaries of what is possible in maritime engineering.
Bagazinski and Ahmed’s work represents a significant advancement in the field of ship design. Their use of generative artificial intelligence and diffusion models offers a powerful tool for maritime engineers and designers, enabling them to create better ships in less time. As the maritime industry continues to evolve, the integration of such advanced technologies will be crucial in meeting the demands for efficiency, sustainability, and innovation. Read the original research paper here.