All in the Placement

New research pinpoints just how much posture and seatbelt alignment matter in a car crash.

Written by Louise Poirier

No two car crashes are alike, even during independent vehicle testing. Engineers have known this for decades, but the usual explanations, from manufacturing tolerances to airbag timing, don’t tell the whole story.

To get a better picture, a team of researchers from Porsche and the Technical University of Munich looked at two variables that haven’t been properly modeled: the seated posture of a test dummy and the seatbelt’s path across its body.

Crash Test Variations

Even identical crash tests produce different results each time. Despite that, automotive engineers account for this before the first physical prototype is ever built. But design changes late in development are costly and constrained, so the solution is virtual robustness testing. Engineers will run stochastic simulations to understand how small variations in setup affect safety outcomes.

Two Sources of Uncertainty

Before every crash test, the dummy is seated and the seatbelt is routed across its body. Its position is never the same, as the hip, torso, and leg angles always vary slightly, and the belt drapes differently across the chest and pelvis. Individually, these variations seem minor, but in a highly nonlinear crash event, small geometric differences at the start can produce meaningfully different injury readings.

Learning from Data

The researchers used measurement data from 32 physical sled tests conducted by the Partnership for Dummy Technology and Biomechanics. Each test included 3D surface scans of the dummy posture and seatbelt routing. The team built statistical models directly from this data, capturing the real-world correlations in body position, limb angles, and belt placement.

A Probabilistic Skeleton

A test dummy’s posture is represented as a skeletal linkage that runs along angles and joints from head to hip. The researchers then used a Linear Gaussian Bayesian Network, which is a structured statistical model that captures how each body segment’s position depends on the one below it. From this model, the team generated 100 physically consistent dummy postures to feed into crash simulations.

Where the Belt Lies

A seatbelt’s path across the chest and pelvis depends on the test subject’s posture and varies based on how the belt is fastened. The team modeled lap and diagonal belt sections separately using Multitask Gaussian Process models, which predict the most likely belt path for any given posture, while also capturing inherent variability. The found that realistic belt routing is consistent with the dummy’s seated position.

Virtual Crash Tests

With the two-stage model trained, the team developed 100 stochastic simulations of dummy posture and belt routing. Each combination was automatically transferred into an adapted finite-element model and run as a virtual frontal crash simulation. Other variables, such as an airbag or retractor, were held constant to isolate the effect of geometric uncertainty alone.

Where Geometry Matters Most

Chest compressions were most sensitive to diagonal belt routing. Pelvis and femur forces were dominated by the dummy’s horizontal position, but seat placements further back also increased knee impact velocity and lower-body loading. Some injury criteria varied enough across simulations that regulatory limits were breached in 13 of 100 runs, with no change to the vehicle hardware.

Moving Toward Safer Cars

Traditional crash development treats setup parameters as fixed values, but this work shows that ignoring geometric variability can hide real safety risks. By modeling uncertainty early and propagating it through simulation, it’s possible to identify dangerous input combinations before physical prototypes exist. This reduces costly late-stage redesigns and helps ensure every car’s safety is optimized.


Louise Poirier is managing editor.

All photos: Getty Images

INTERESTED IN LEARNING MORE?

“Two-Stage Probabilistic Modeling to Study the Influence of Uncertainties in Occupant Seating Position and Seatbelt Routing on Occupant Loads” was published in the June 2026 issue of the ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering.

READ FOR FREE*

*Access this article for free through Sept. 30, 2026.

© 2026 The American Society of Mechanical Engineers. All rights reserved.

About ASME

Privacy and Security Policy

Preference Center

ASME Membership

Access your Benefits

Renew your Membership

Advertising & Partnerships

Terms of Use

Contact Us