On 21 May 2026, the Society of Automotive Engineers organised a seminar on AI and simulation..
To mark the occasion, Julien Jourdan and Gireg Lanoë travelled to the Forum Armand Peugeot in Poissy to attend the SIA study day. The event confirmed a growing trend: AI is becoming a key driver not only for enhancing our competitiveness but also for transforming our engineering practices.
Here are the key points from the various morning talks:
- Stellantis – AI Vision for Engineering by David Routier
AI is a strategic tool for accelerating decision-making, simulation and access to knowledge, using an approach based on specialised agents connected across the group’s IT infrastructure. - Mistral AI – Custom AI vs off-the-shelf solutions by Clément Auguy and Benoît Iweins
The real impact comes from solutions tailored to each client’s specific needs, particularly to speed up simulation and create real-time digital twins. - Dassault Systèmes – Agent-based AI by Delphine Genouvrier
An overview of Dassault’s applications as they transition from being mere ‘tools’ to AI agents capable of intervening in engineering processes and orchestrating complex tasks. - Stellantis – Quality of requirements & SysML models by Guanrui Hou
A practical example of how AI is used to improve the quality of drafting and interpreting requirements, whilst also highlighting the limitations involved in implementing key stages: human validation remains essential. - Renault – From bug to automated testing by Rémy Protat, Maëlle Brassier and Amine Menad
End-to-end automation pipeline: bug analysis, knowledge capture and automated test generation, delivering a significant return on investment. - Michelin – LLM & engineering by François Deheeger
Above all, AI is transforming the management and utilisation of knowledge, with a shift towards domain-specific agent-based systems that support R&D teams in the design and development of new products by drawing on the company’s historical data.
In the afternoon, the talks showed a clear shift towards integrated systems focused on specific disciplines (mechanical engineering, electrical engineering, etc.) and use cases (documentation, requirements, development, testing, etc.), with a strong emphasis on data structuring to make the most of AI agents. Here is a detailed look at the talks in question:
- Adagos – Model parsimony by Mohamed Masmoudi
An approach aimed at simpler but more effective models, with a significant impact on computation times and the data required. - Michelin – Digital twin of the tyre by Maxime Boulanger and Jeanne Macias
Presentation of diagnostic tools that combine physical measurements and data to create on-board models designed to improve vehicle safety: sensorless tyre pressure monitoring, wear detection, ADAS data, predictive maintenance, and more… - Stellantis – Virtual sensors by Miguel Dinis
Data analysis and simulation studies to estimate unmeasurable variables (soft sensors), with practical application to the EDM (Electric Drive Module = inverter + motor + gearbox). - Alpine – Virtual sensors in competition by Thomas Brichard
A study of the virtualisation of a brake temperature sensor to enhance data reliability and improve decision-making and performance in a racing car (24 Hours of Le Mans). A real-world case study illustrating both the benefits and limitations of machine learning under extreme conditions and in real time. - Mathworks and Schaeffler – AI informed by physics through Moubarak Gado and Nicolas Lamarque
A hybrid approach combining physical models and AI to create explainable, robust and deployable systems.
Discussions on the impact of AI in engineering are set to continue, and in the meantime, come and meet Acsystème at SIA Powertrain from 17 to 18 June in Lille.