Numerical Mesh Truncation Boundary Conditions Optimized via Machine Learning

Mesh truncation in Finite Element Method (FEM) is critical and presents conflicting options. Boundary Integral (BI) option yields very accurate results but results in non-sparse solution matrix while Artificial Absorbers (AAs) or Absorbing Boundary Conditions (ABCs) preserves sparsity of the matrix, they need to be placed sufficiently far away from the structure to yield accurate data, resulting in large number of unknowns. Paper presents a method to optimize the placement of the truncation boundary close to the object being modeled to minimize the size of the problem and the run time Optimization is object-dependent as it should be and is achieved using Machine Learning (ML). The proposed scheme is promising but needs further study to fully explore its potential to be of practical use.