Efficient Physical Truncation of Low-Frequency ATEM Problems in Specific Geometries by Using Random Forest Regression Based PMM Model

In addressing the challenges posed by low-frequency airborne transient electromagnetics (ATEM), it is necessary to take into account the considerations of accuracy, computational efficiency, and the scale and intricacy of the physical domain. This becomes particularly crucial when dealing with large-scale, complex issues, with the aim of mitigating the computational resource burden associated with managing such complexities. In order to further meet the aforementioned criteria, a Perfectly Matched Monolayer (PMM) model has been introduced into the Random Forest Regression (RFR) framework. The RFR-based PMM model has demonstrated exceptional accuracy through the utilization of Bagging's integrated learning methodology, while also reducing the computational resource requirements for processing time. In comparison to traditional machine learning models, our model has exhibited significant advantages in terms of training stability, model efficiency, and parallelization capabilities. To verify and establish the reliability of this approach, three-dimensional numerical simulations of the ATEM problem were conducted. The proposed model in this study has exhibited superior accuracy, efficiency, and versatility in addressing the low-frequency ATEM problem, integrating with the FDTD method.