142 INTERVIEW Accelerating engineering innovation with AI Neural Concept’s Thomas von Tschammer explains to Richard Humphreys how physics-aware AI and Microsoft Azure are helping manufacturers explore, evaluate and optimise more product designs in less time For manufacturers, designing a new product can involve countless iterations across computer-aided design (CAD), computer-aided engineering (CAE), product lifecycle management (PLM) and simulation tools. Neural Concept is aiming to make that process more exploratory by placing an AI layer over the systems engineers already use. “We are moving toward a world where AI becomes part of the engineering process itself, connecting design, simulation and optimisation so engineers can explore more possibilities and make better product decisions earlier,” says Thomas von Tschammer, co-founder and US managing director of Neural Concept. The company’s platform uses engineeringspecific AI to generate designs, predict performance and optimise products with near-real-time feedback. It has been adopted by more than 70 original equipment manufacturers and tier one suppliers across industries including automotive, aerospace, electronics and semiconductors. Microsoft Azure has been central to scaling the platform since Neural Concept was founded in 2019. The company uses Azure high-performance computing and AI infrastructure for managing intensive workloads, while Azure Kubernetes Service dynamically scales the resources needed to run its AI workloads and Azure Storage provides access to the datasets used to train and operate engineering models. “The next challenge for manufacturers is taking engineering AI beyond individual use cases and scaling it across teams, programmes and geographies,” says von Tschammer. “Microsoft provides an enterprise environment that supports that transition.” That infrastructure is important because engineering organisations can work with terabytes of simulation data spread across physics domains, software tools, teams and locations. Neural Concept can use this data to train models that predict the physical performance of products and components across areas such as aerodynamics, thermal management, structural mechanics and electromagnetics. Rather than replacing existing engineering systems, the platform sits above them. CAD, CAE and PLM tools remain the sources of record for product data, detailed engineering and final validation, while Neural Concept applies AI to the iterative work between simulation and validation. That distinction is important because general-purpose AI models are not designed for the constraints of physical engineering. “Engineering AI has to reason within the realities of the physical world,” says von Tschammer. “Geometry, physics, materials, performance requirements and engineering “ Engineering AI has to reason within the realities of the physical world”
RkJQdWJsaXNoZXIy NzQ1NTk=