Related Experiment Video
Updated: May 14, 2026
09:32
Subject-specific Musculoskeletal Model for Studying Bone Strain During Dynamic Motion
Published on: April 11, 2018
A data-driven surrogate model to connect scales between multi-domain biomechanics simulations
Gavin Paiva1, Sampath Bhashyam, Ganesh Thiagarajan
1University of Missouri-Kansas City, Kansas City, MO 64110, USA.
Summary
A new data-driven surrogate model accurately predicts cartilage stress using neural networks. This bridges finite element and multibody modeling gaps for enhanced biomechanical simulations.
Area of Science:
- Biomechanics
- Computational Modeling
- Biomedical Engineering
Background:
- Finite element (FE) and multibody (MB) modeling are crucial for simulating cartilage behavior.
- Bridging the gap between these modeling paradigms can enhance simulation accuracy and data availability.
- Current rigid multibody cartilage simulations offer limited detailed stress information.
Purpose of the Study:
- To develop a data-driven surrogate model.
- To bridge the gap between FE and MB modeling for cartilage simulation.
- To expand the information obtainable from rigid MB cartilage simulations.
Main Methods:
- Modeled a canine stifle cartilage indentation experiment using both FE and MB analysis.
- Developed a surrogate model using neural networks.
- Trained the neural network on force data from the MB model and von Mises stress from the FE model.
Main Results:
- The surrogate model adequately approximated von Mises stress calculated by the FE model.
- The approximation was based on force values derived from the MB model.
- A correlation coefficient exceeding 0.96 was achieved, indicating high accuracy.
Conclusions:
- A data-driven surrogate model can effectively bridge FE and MB modeling.
- Neural networks provide a viable method for approximating complex stress distributions.
- This approach enhances the utility of MB simulations for cartilage biomechanics.
