Luminary Cloud, a physics AI platform, announced Shift-Wing, its newest physics AI model in the Shift family.
This model will enable aircraft designers and aerospace engineers to explore more designs, earlier in the conceptual design phase and avoid costly late-stage rework by performing real-time, high-speed aerodynamic analysis of transonic wings with Physics AI inference.
The SHIFT-Wing model and dataset were developed with the Nvidia PhysicsNeMo framework, in collaboration with Otto Aviation, an advanced aircraft manufacturer pioneering laminar flow technology.
“At Otto, we believe the future of aircraft design lies at the intersection of first principles and artificial intelligence,” said Obi Ndu, chief information and digital officer at Otto Aviation, in a statement. “This partnership with Luminary Cloud enables us to unlock scale in generation of physics-based simulation data, making it possible for our engineers to explore, optimize, and validate aerodynamic concepts faster than ever before. By combining our flight sciences expertise with their simulation platform, we’re not only accelerating innovation but also co-developing next-generation surrogate AI models that bring physics-informed intelligence into every stage of design.”
In today’s environment, where guarantees of higher performance are required earlier in the design stage, aircraft designers are limited by their ability to search vast design spaces with sufficient fidelity for advanced aircraft wings.
Due to limitations of existing engineering simulation tools and reliance on simplified physics models, potentially superior designs remain unexplored and aero-structural-control interactions are often ignored early in design.
Shift-Wing opens up the design space for aircraft manufacturers by enabling aerospace engineers to experiment with new designs with high fidelity and low risk. Development and redesign costs for commercial aviation can be substantial, with large cost overruns being common. Combined with regulatory compliance and certification costing millions and taking years, these expenses have stifled breakthrough innovation.
“It is a dream come true to release a Physics AI model that can help aerospace companies transform how they push the boundaries of conceptual wing design,” said Juan J. Alonso, CTO of Luminary Cloud, in a statement. “Shift-Wing unlocks AI-driven innovation for the next generation of aircraft by allowing aerospace companies to feasibly explore more designs than previously possible and to use the almost-instantaneous aerodynamic predictions to introduce interactions with other elements of the design, including structural analysis and actuator and control system design.”
Alonso is also the Chair of the Department of Aeronautics & Astronautics at Stanford University and former Director of NASA’s Fundamental Aeronautics Program.
One of the biggest bottlenecks in building accurate Physics AI models has been access to large-scale, high-fidelity simulation data — until now. The Shift-Wing aerodynamic dataset was generated by parametrically modifying and simulating several thousand geometry variants of the NASA Common Research Model wing-body geometry (similar to the Boeing 777) using Luminary’s high-fidelity Computational Fluid Dynamics (CFD) platform.
Solutions were created using Luminary Mesh Adaptation to guarantee the highest accuracy and superior resolution of all flow features including shock waves. The model is trained on this dataset using Nvidia PhysicsNeMo, leveraging its DoMINO architecture for external-aerodynamics, enabling Physics AI predictions. Like Shift-SUV, Luminary Cloud’s first model for automotive SUV predictions, the Shift-Wing pretrained model and dataset is open source for non-commercial use and can be licensed for commercial use.
The AI-based workflow will augment engineers by giving them more time and space to think creatively. By leveraging AI inference and high accuracy within 2% for drag coefficient predictions for in-distribution samples Shift-Wing enables interactive design exploration, inverse design, load calculations, and design optimization using realistic problem setups including all necessary constraints. Unlike traditionally slow and expensive simulations, Physics AI models provide instant and accurate aerodynamic performance insights in a fraction of the time and cost — especially valuable in the early conceptual design phase.
The Physics AI model offers several key benefits for the aerospace industry, such as:
- Rapidly apply Physics AI models to optimize wing designs. Achieve mission-aligned wing designs in a fraction of the traditional development time by reducing the number of simulations needed and cost of developing surrogate models from scratch.
- Accurately span the entire speed regime, from subsonic to transonic, with confidence. The SHIFT-Wing database combines highly-accurate simulations using Luminary’s mesh adaptation to properly resolve and capture shocks, boundary layers, and shear layers.
- Gain access to a large-scale, extensively validated simulation dataset. The model is based on the trusted and widely known NASA CRM geometry, trained on thousands of high-fidelity CFD simulations across varied angles of attack, configurations, and flow conditions, and parameterized to describe a broad range of realistic wing geometries with guidance from Otto Aviation and Luminary experts.
- Adapt SHIFT-Wing to your application. Customize the SHIFT-Wing model by leveraging the provided large dataset and Physics AI model and adding high-fidelity simulations that capture your design intent. Customization is critical for accurate and effective use of Physics AI in high speed flows, enabling better model fine-tuning to meet engineering discipline requirements.
“The aerospace industry is understandably cautious with new configurations,” said Scott Drennan, president at Otto Aviation, in a statement. “But, at Otto, we’ve discovered how to manufacture, design, and support an aircraft that performs in laminar flow. We proved it on the Celera 500 and are now applying those lessons to our transonic Phantom 3500 aircraft. Our work on SHIFT-Wing continues that focus, bringing engineering discipline to new aerodynamic designs where smart choices can lead to measurable gains.”
“AI-Physics and Accelerated Computing promise to revolutionize the aerospace industry by accelerating design cycles and reducing cost while pushing the boundaries of engineering design,” said Tim Costa, senior director of CAE & CUDA-X at Nvidia, in a statement. “With Nvidia PhysicsNeMo, Luminary Cloud is building the tools that will transform the industry with near real-time aerodynamic simulation and design optimization.”
Available now, aerospace companies, and research institutions that are interested in the Shift-Wing model and dataset can access it directly on Hugging Face.