Using insight from the FISITA Digitalisation Expert Group white paper and real-life case studies, experts from Applus+IDIADA, GM, and Stellantis illustrate how AI is transforming CAE, writes James Scoltock
Artificial intelligence has the power to transform computer‑aided engineering (CAE), and the FISITA Digitalisation Expert Group’s recent webinar AI for CAE – Opportunities and Challenges in Automotive Engineering explored both the promise and the pitfalls of this emerging technology, with case studies and key findings from the working group’s upcoming white paper of the same name.
The session featured insights from Enric Aramburu, Fluid Engineering Product Manager at Applus+IDIADA, Laurent Di Valentin, CAE Senior Fellow at Stellantis, and Simon Xu, Tech Fellow for Vehicle Optimisation and Machine Learning at General Motors. Together, they painted a nuanced picture of how AI could reshape simulation and design processes in automotive engineering, and shared insights from their own work and the collective work of the FISITA Digitalisation Expert Group.
Accelerating simulation with AI
Aramburu opened the discussion by stressing the need for breakthroughs beyond traditional CAE. “CAE has become very mature and it’s a very good tool, but we need something else and AI has shown promise,” he said. He described how reduced‑order models and graph neural networks are already being used to replicate complex simulations, noting: “Instead of simulating injection moulding in detail, we can now predict results in a matter of seconds.”
“By building a machine learning model trained on both simulation and physical test results, we can enhance the quality of crash signal predictions and improve calibration of restraint systems” – Laurent Di Valentin, Stellantis, and Chair, FISITA Digitalisation Expert Group
Xu highlighted the potential of AI to speed up crash simulations. “These studies are CPU‑intensive and turnaround time is not satisfactory. We try to use a graph neural network to predict dynamic crash behaviours… the goal is to replace resource‑heavy simulations in the early design stage, so we can use AI‑driven predictions and give decisions such as yes or no, good or bad,” he explained.
Reliability and trust
Xu also noted caution, as reliability remains a critical challenge. “In safety‑critical domains like crash or ADAS, we cannot simply be happy with black‑box AI. We need to know for sure what is the basis of these proposals. Explainable AI through sensitivity analysis and physics‑linked surrogate models is essential,” he said.
Di Valentin, who chairs the FISITA Digitalisation Expert Group, emphasised the importance of combining simulation with experimental data to improve fidelity. “The hybrid twin strategy benefits from the best state‑of‑the‑art simulation and mixed with experimental data coming from previous projects,” he explained. “By building a machine learning model trained on both simulation and physical test results, we can enhance the quality of crash signal predictions – especially in high‑frequency ranges – and improve calibration of restraint systems.”
In every scenario data is king. Xu reinforced the importance of robust data pipelines. “The dataset is the key. Today we probably throw away much of the data we run on a daily basis. We need to streamline data collection, storage, and access across simulation and test workflows,” he noted.
Success stories and practical applications
Despite these challenges, the webinar showcased examples where AI is already delivering value. Aramburu described how machine learning has been applied to material card generation: “Traditionally this was a long, costly process. By training on thousands of simulations, we achieved 90 percent accuracy in predicting material card parameters – enough for concept phases.”
“Success requires more than model accuracy. It requires reliable and consistent data pipelines, transparent and interpretable models, and strong integration, validation and governance” – Simon Xu, General Motors
Xu pointed to pedestrian impact modelling as another success. “Overall, it is able to achieve a good prediction, even with highly complex component interactions… this is a pretty successful case and is going to be deployed going forward in production,” he said.
Towards responsible adoption
The overarching message was one of cautious optimism. AI offers accelerated innovation, reduced costs, and potential environmental benefits, but its adoption must be carefully managed. “Success requires more than model accuracy,” said Xu, adding, “It requires reliable and consistent data pipelines, transparent and interpretable models, and strong integration, validation and governance.”
FISITA’s Digitalisation Expert Group is preparing a White Paper that consolidates these insights, reviewing the current state of AI in CAE and projecting its potential for broader adoption. For automotive engineers, the challenge now is to engage critically with these technologies — balancing enthusiasm with caution, and ensuring that AI becomes a trusted partner in the design and development of tomorrow’s vehicles.
Watch the AI for CAE – opportunities and challenges in automotive engineering FISITA webinar in full.
Use the form below to register your interest to be notified when the Digitalisation Expert Group AI for CAE white paper is available, and to be kept up to date with other FISITA activities.
Stay up to date with FISITA content
We invite you to register your interest to be notified when the Digitalisation Expert Group white paper is available, and to be kept up to date with other FISITA activities.