A Data-Driven Space-Time-Parameter Reduced-Order Model with Manifold Learning for Coupled Problems: Application to...

Area of Science:

• Computational fluid dynamics
• Biomedical engineering
• Materials science

Background:

• Microcapsules are crucial for drug delivery, requiring accurate modeling in microfluidic systems.
• Simulating microcapsule dynamics in microchannels is computationally intensive.
• Understanding fluid-structure interactions is key for optimizing drug vehicle performance.

Purpose of the Study:

• To develop an innovative data-driven model-order reduction (MOR) technique for microcapsule suspensions.
• To accurately simulate microcapsule behavior in microfluidic channels across varying parameters.
• To enhance the efficiency of modeling microcapsule-based drug delivery systems.

Main Methods:

• A data-driven MOR technique using global Proper Orthogonal Decomposition (POD) reduced bases for space and parameter variables.
• Offline computation of reduced bases followed by online computation for new parameter instances.
• Identification of the nonlinear low-order manifold of reduced variables using diffuse approximation for time evolution.

Main Results:

• The proposed MOR technique accurately and stably models microcapsule dynamics.
• Validation through numerical comparisons with a full-order fluid-structure interaction model.
• Successful reduction of a complex space-time-parameter problem.

Conclusions:

• The developed MOR technique offers an efficient and accurate approach for modeling microcapsule suspensions.
• This method has potential applications in healthcare, particularly for drug delivery vehicles.
• The approach is broadly applicable to coupled problems, especially those involving quasistatic structural mechanics models.