Accelerated simulation methodologies for computational vascular flow modelling.

Area of Science:

• Computational fluid dynamics
• Biomedical engineering
• Medical device development

Background:

• Vascular flow modeling is essential for understanding vascular pathologies and designing medical devices.
• Current models, based on Navier-Stokes equations, are computationally intensive due to multi-physics and multi-scale complexities.
• This complexity leads to high costs and excessive simulation times, hindering practical application.

Purpose of the Study:

• To review and analyze accelerated simulation methodologies for computational vascular flow modeling.
• To assess the applicability and effectiveness of various reduced order modeling (ROM) and machine learning (ML) techniques in accelerating vascular flow simulations.
• To identify challenges and propose future research directions for optimizing vascular flow modeling.

Main Methods:

• Review of reduced order modeling (ROM) techniques, including zero-/one-dimensional and modal decomposition-based approaches.
• Exploration of machine learning (ML) methods such as ML-augmented ROMs, ML-based ROMs, and physics-informed ML models.
• Discussion of the strengths, limitations, and domain-specific challenges of each acceleration method.

Main Results:

• Various ROM and ML techniques show promise for accelerating vascular flow simulations.
• Each method presents unique advantages and disadvantages depending on the specific application and anatomical complexity.
• Accuracy and speed-up factors were analyzed for different vascular flow simulation acceleration applications.

Conclusions:

• Accelerated simulation methods are crucial for advancing vascular flow modeling.
• The choice of acceleration technique depends on the specific requirements of the vascular flow problem.
• Future research should focus on multi-scale acceleration methods to address geometric variability in complex vascular geometries.