Model order reduction techniques to identify submarining risk in a simplified human body model.

Area of Science:

• Computational mechanics
• Biomechanical engineering
• Machine learning applications

Background:

• High-fidelity human body model (HBM) simulations are computationally intensive, limiting advanced analyses.
• Conventional methods create barriers for uncertainty quantification, sensitivity analysis, and optimization studies.

Purpose of the Study:

• To investigate linear and non-linear parametric reduced order models (ROMs) as non-intrusive replacements for HBM simulations.
• To enable faster and more profound analyses in biomechanical engineering.

Main Methods:

• Developed a non-intrusive framework coupling dimensionality reduction with machine learning surrogate models.
• Compared linear and non-linear dimensionality reduction techniques for HBM analysis.

Main Results:

• Achieved speed-ups of several orders of magnitude compared to traditional simulations.
• Demonstrated accurate generalization across the design space using data-driven black-box models.
• Validated the effectiveness of both linear and non-linear ROM approaches.

Conclusions:

• Reduced order models (ROMs) offer a computationally efficient alternative to high-fidelity HBM simulations.
• The non-intrusive framework accelerates complex engineering analyses, including optimization and uncertainty quantification.
• ROMs are valuable tools for engineers, enhancing the feasibility of in-depth studies.