The fourth and final part of our Age of AI series hears from NVIDIA on how generative AI is influencing product design
In the fourth and final part of our Age of AI series, NVIDIA’s John Linford gives his take on how generative AI is influencing product design and engineering
Having pioneered accelerated computing since the early 1990s, NVIDIA has become one of the largest and most influential companies in the artificial intelligence (AI) space. With arguably the world’s most advanced platform for generative AI, the firm is enabling industries across the globe to deploy generative AI applications into production at scale.
As such, there are few better placed to provide insight on AI’s current and future impact on product design than NVIDIA’s principal technical product manager, John Linford. International Design Engineer picked his brains on the progress made so far in AI design tools and their application.
“Designers are now using AI-enhanced tools to create products for industries like automotive, manufacturing, aerospace and energy,” he says. “AI is helping optimise material layouts for maximum strength and minimal weight while taking into account physical constraints like stress, strain, fatigue and thermal limits. AI surrogate models for thermal diffusion, fluid flow and linear deformation provide real-time product performance feedback as the design evolves and enable designers to anticipate physical figures-of-merit like drag or surface pressure before high fidelity simulations are performed.”
Simulation-based design processes are also undergoing transformation, with AI reducing computation time and improving accuracy in simulations of complex physical systems, he adds: “Everything manufactured is first simulated, and physics-informed neural networks (PINNs) enable simulations that are orders-of-magnitude faster than traditional approaches. PINNs enhance AI models by incorporating the governing equations of physics directly into the machine learning framework. This produces surrogate models that combine physics-based causality with simulation and observed data, enabling real-time prediction. NVIDIA PhysicsNeMo includes PINN architectures appropriate for external aerodynamics, fluid flow and other applications.”
AI DESIGN TOOL ADVANCEMENTS
The capabilities of generative AI – the ability of algorithms to create new text, images, sounds, animations, 3D models and computer code – are moving at incredible speed. By employing large language models (LLMs), the technology can substantially reduce the time people devote to manual tasks like searching for and compiling information, for instance.
Within more industry-focused scenarios, “Generative AI is changing the game in semiconductor manufacturing, for instance LLMs enhancing code generation in the design space,” Linford offers. “For example, NVIDIA uses ChipNeMo, a domain specific foundation model trained on NVIDIA data, to help our internal design work. Engineering assistant copilots are helping design chips, design systems and optimise fab operations. Agentic workflows combine multiple AI agents to execute complex tasks, like bug summarisation and analysis on multi-modal data such as text, images and video.”
He continues, “At NVIDIA, we build AI to build chips for AI! Over the last 6 years, we have done many AI for chip design research projects and several of these are in production use today. Our projects cover a wide spectrum of design areas, such as architecture design, RTL, verification, synthesis, cell design, physical design – including placement, route and optimisation – as well as analogue design and lithography.”
INDUSTRY OUTLOOK
Generative AI is rapidly finding its feet in a wide range of industrial sectors, Linford says. For instance, streamlining vehicle design and manufacturing in the automotive industry, powering a new wave of healthcare innovation through drug discovery and enhancing the operational efficiency of telecommunications.
“Physics-inspired AI is transforming the automotive, manufacturing, aerospace and energy industries,” he continues. “For example, Siemens Gamesa used the NVIDIA PhysicsNeMo framework to train a PINN that led to 4,000 times faster wind turbine wake optimisation compared with traditional approaches. This speed increase enables large-scale, detailed wind farm layouts that optimise turbine placement and maximise energy output.”
Real-time digital twins (RTDTs) are the cutting edge of computer-aided engineering (CAE) simulation, because they enable immediate feedback in the engineering design loop. RTDTs have soared in demand within the aerospace, automotive and electronic design industries in particular.
“Siemens Energy uses digital twins to maximise uptime for heat recovery steam generators,” Linford says. “These massive machines use hot exhaust gases to boil water. The exhaust gases can cause corrosion, leading to downtime for system maintenance. High-fidelity simulations of multiphase turbulent flow help predict where and when corrosion occurs. Physics-informed AI can infer this flow in seconds, making these simulations feasible. With simulation, unplanned downtime is reduced by up to 70%, saving the industry $1.7 billion per year.
A LOOK INTO THE FUTURE
According to Linford, generative AI will have a pivotal role to play in how design engineering processes of the future will operate.
“Advanced warehouses and factories use fleets of hundreds of autonomous mobile robots, robotic arm manipulators and humanoid robots working alongside people,” he says. “Implementations of increasingly complex systems of sensor and robot autonomy require coordinated AI training in simulation to optimise operations, help ensure safety and avoid disruptions.”
And what about from NVIDIA’s perspective? “NVIDIA offers enterprises a reference architecture of NVIDIA accelerated computing, AI NVIDIA Isaac and NVIDIA Omniverse technologies to develop and test digital twins for A-powered robot ‘brains’ that drive robots, video analytics AI agents, equipment and more for handling enormous complexity and scale,” he offers. “This framework brings software-defined capabilities to physical facilities, enabling continuous development, testing, optimisation and deployment.”
“Accelerated computing enables AI, and AI enables quantum computing. AI models trained on quantum data generated from simulators and physical hardware are expected to unlock useful quantum computing. Broad adoption of AI ultimately leads to data generation and model training at scales that were previously impossible, transforming industries and enabling new opportunities.”
