DARTS-NETGAB: Design Automation and Real-Time Simulation Using Neural Networks Ensembles for Turbocompressors on Gas-Bearings
Year: 2024
Type: conference-paper
Abstract: In the domain of engineering design, where efficiency in simulation and precision in modeling are paramount, this study introduces DARTS-NETGAB, a pioneering platform uniquely designed for real-time simulation and automated design. Specifically tailored for gas-bearing supported turbocompressors, DARTS-NETGAB integrates neural network ensembles with a parametric CAD construction library to deliver unprecedented prediction speeds and modeling precision across various engineering systems. This integration allows for seamless, real-time performance evaluations of complex, multidisciplinary systems and automated CAD model generation. This framework streamlines the design process, reduces cycles times and enhances adaptability to manufacturing imperfection. DARTS-NETGAB features a user-centric interface developed using the advanced Panel-Bokeh Python libraries, facilitating dynamic and interactive design modifications directly within a web browser. This capability enables immediate visualization and adjustment of a comprehensive turbocompressor model, thereby streamlining the transition from theoretical design to practical application. The paper details how the combination of DARTS-NETGAB’s rapid, accurate predictive capabilities with its robust design tools not only advances micro-turbocompressor design but also revolutionizes engineering processes across diverse systems. By merging cutting-edge computational techniques with practical, user-friendly tools, DARTS-NETGAB offers a significant improvement over traditional methods, fostering more efficient and innovative engineering solutions.
Authors: Soheyl Massoudi, Joseph Bejjani, Timothy Horvath, Dogukan Üstun, Jürg Schiffmann
Institution: École Polytechnique Fédérale de Lausanne
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4404333800