Reduced Order Modeling of Stokes Flow in Lattice Structures

Development of a reduced order model for the stokes flow for geometry optimization of static mixers.

Short Description

A reduced order model for the stokes flow in lattice structures was built to enable an accurate and efficient evaluation of the velocity and pressure fields in the domain, which can be used in the process of geometry optimization.

Workflow

A bunch of simulation results, calculated with IGA, are collected in the solution vector. On this solution vector, a Proper Orthogonal Decomposition is performed to reduce the dimensionality and different regression models are trained on this reduced dimensional data. The used regression models are:

• Linear Regression
• Gaussian Process Regression
• Radial Basis Interpolation

The relative error of the Proper Orthogonal Decomposition and regression models are analyzed for different numbers of reduced dimensions and the training and evaluation time of the regression models are compared.

Exemplary prediction of the x-component of the velocity field