Search

AI model accelerates sheet metal incremental forming simulations - Sciencemag.cz

Translated title. Original: AI model urychluje simulace inkrementálního tváření plechů - Sciencemag.cz

Researchers created a neural surrogate for incremental sheet metal forming that learns from extensive FEM data to predict the sheet state quickly, enabling faster evaluation of process variants and geometry compensation while reducing reliance on time-intensive simulations.

Vehicle Repairability Design with Virtual Twin Simulation - Dassault Systèmes blog

Aftersales and Dassault Systèmes promote using virtual twins early in vehicle design to evaluate repairability. This approach reduces warranty costs and repair times by simulating service scenarios, tool access, and dismantling steps before physical vehicles exist, while aligning engineering data and requirements across the design and service domains.

[2610.31113] Can AI Co-Design Structures? An AI Agent for Early-Stage Structural Design

This paper presents a proof-of-concept workflow where an AI agent, using natural language prompts, generates and iterates structural models via a COMPAS-based API and a graph-based data representation, aiming at AI-assisted structural co-design. The results are preliminary but show feasibility for integrating language, agents, and domain data in early-stage structural design.

How AI will create the Formula 1 car of the future even before putting it on the tracks - Canaltech

Translated title. Original: Como a IA vai criar o carro de F1 do futuro antes mesmo de colocá-lo nas pistas - Canaltech

AWS and FIA are collaborating to use data and AI to digitally simulate Formula 1 regulations and car design, enabling faster rule adjustments, cost reductions, and enhanced race strategy through extensive telemetry analysis. The approach extends from digital Aerodynamics to pre-season rule testing, with Fernanda Spinardi of AWS explaining the impact on teams and fans.

Mazda designs the "face" of its cars with AI, conducts joint research with OpenAI to explore balancing air resistance and design - CNET Japan

Translated title. Original: マツダ、クルマの「顔」をAIで設計　OpenAIと共同研究、空気抵抗とデザインの両立探る - CNET Japan

OpenAI and Mazda are jointly researching AI-assisted car design to optimize front-end aerodynamics, safety, and cost. The effort uses AI models to model and simulate shapes, aiming to balance drag reduction with design goals; findings will be shared in a few months, with no details on mass-market adoption yet.

Systems Thinking Is Vital in the Age of AI - IEEE Spectrum

Engineering combines human judgment with AI to search, design options, and simulate outcomes while experts validate safety, feasibility, and user impact; the engineer coordinates diverse knowledge and adapts goals as the system evolves.

Dream • Built World Models

Dream develops built world models that automate design and engineering for commercial buildings, taking a project from layout to a build-ready drawing set while checking geometry, constraints, and code across multiple sites. The approach targets fast, site-adapted designs for structures like data centers, stores, and offices, functioning as architect of record for projects in California and Texas.

Paper page - PLE: Physical-Engineering's Last Exam

PLE tests language-model agents on 85 realistic, long-horizon engineering tasks across four disciplines. The best tested setup scores 50.6% on average, pointing to a substantial gap between current agents and autonomous engineering work.

ROAM-AI Launches Virtual Engineer for ESP Optimization | The Nati

ROAM-AI announces Virtual Engineer, an AI-powered capability within its ESP Optimization that delivers on-demand, physics-informed engineering analysis via a chat interface, validated on hundreds of wells. It combines machine learning with engineering physics to diagnose performance, generate analytics, and provide actionable guidance with safeguards, aiding bandwidth-constrained teams across North American fields.

AI Can Draw Blueprints Now - Stephen Gallagher

An architect tests 13 AI drafting models from 2024–2026 and finds dramatic improvements by 2026, with Claude Opus 5.5 producing permit-quality construction drawings for a simple house. The study suggests a shift in professional practice toward supervising AI-generated technical drawings rather than handcrafting them.