PhysicsML & AI for Manufacturing in CAE Airbag Design
Accelerating Airbag Design with Past CAE Data — Cut Time and Costs
In automotive safety, each design choice involves balancing speed, precision, and cost. Nowhere is this more evident than in airbag development, where the inflated shape of the airbag must precisely conform to both vehicle geometry and occupant physiology while meeting regulatory standards and crash performance targets.
A leading automotive airbag manufacturer approached us with a familiar challenge:
- Design iterations were taking months
- Engineering teams were tied up for weeks
- Simulation costs were climbing due to repeated use of expensive FEA solvers
And yet, they were sitting on a goldmine: a rich repository of validated FEA simulations spanning hundreds of design variants. The question was, can we put that data to work to accelerate the next round of designs?
Before: Design Cycles Measured in Months
The process started with the Application Engineer visiting the OEM, sketching potential airbag designs as 2D drawings, accounting for car interior geometry and passenger profiles. These sketches were returned to the engineering team for further review, refinement, and validation.
- CAD modeling
- FEA simulation of the inflated shape
- Back-and-forth adjustments with the OEM and AE
This loop, repeated over multiple designs, consumed weeks of engineering effort and solver hours often for routine variations.
After: Instant Feedback, Smarter Use of Resources
We developed a lightweight App that enables the AE to predict the 3D inflated airbag shape directly from their 2D sketch without waiting on CAD teams or solver runs.
What’s under the hood?
- 2D shape processing from sketch
- Shape adjustment based on car profile and fabric parameters
- 3D inflated shape prediction using ML models trained on past FEA simulations
- Geometry and physics constraints to ensure predictions reflect real-world behavior
The App acts like a digital assistant, providing fast, reliable, and explainable results.
The Results
- AEs now close initial design loops in days instead of months
- Engineering time is preserved for advanced simulations and final validation
- CAE solver usage is focused on new, non-routine scenarios
- Overall design cost and lead time are significantly reduced
A Broader Lesson
This is not just about airbags. It is about how manufacturers across industries can use existing simulation data to train AI models that augment their design process.
AI for manufacturing enables teams to extract more value from historical CAE data, turning past simulations into predictive tools that drive smarter decisions. By embedding physics-informed AI into early-stage workflows, companies can:
- Compress the design cycle
- Reduce simulation costs
- Enable engineers to focus on high-value challenges
At Intelimek, we specialize in building these hybrid AI solutions by combining domain expertise, validated physics, and practical ML. Our goal is to make AI for manufacturing not just a concept—but a day-to-day productivity tool that delivers measurable impact on the ground.