Efficient Low-Frequency Human Exposure Assessment With the Maximum Entropy Snapshot Sampling

Abstract

Numerical dosimetry simulations of human exposure to low-frequency magnetic fields, according to International Commission on Non-Ionizing Radiation Protection (ICNIRP) recommendations, are typically computationally and memory-intensive. By employing reduced-order models (ROMs) for the high-fidelity linear systems to be solved, simulation efficiency can be significantly enhanced, thereby enabling a comprehensive numerical assessment of human exposure. For model generation, snapshot-based reduced basis methods (RBMs) as the proper orthogonal decomposition (POD), which rely on the singular value decomposition (SVD) of a matrix whose columns are the solution vectors of a high-fidelity system, are commonly used in the context of POD. Due to the recurrence of redundant information in most solution vectors, SVD becomes a computationally and memory-intensive step. With the maximum entropy snapshot sampling (MESS) strategy, the number of solution vectors can be efficiently reduced to the essential ones. This work presents a reduced basis for efficient human exposure assessment in a computationally and memory-efficient manner using this information-theoretic framework.