Surrogate modeling for fast experimental assessment of specific absorption rate

Experimental dosimetry of electromagnetic fields (EMFs) in biological tissue is important for validating numerical techniques, designing electromagnetic exposure systems, and compliance testing of wireless devices. Compliance standards specify a two-step procedure to determine the peak spatial-averaged SAR in 1 g or 10 g of tissue in which the measurement locations lie on a rectilinear grid (selected up-front). In this chapter, we show the potential of surrogate modeling techniques to significantly reduce the duration of experimental dosimetry of EMF by using a sequential design. A sequential design or adaptive sampling differs from a traditional design of experiments as data and models from previous iterations are used to optimally select new samples resulting in a more efficient distribution of samples as compared with the traditional design of experiments. Based on a data set of about 100 dosimetric measurements, we show that the adaptive sampling of surrogate modeling is suitable to speed up the determination of the peak SAR location in an area scan by up to 43 and 64% compared with the standardized area scan on a rectilinear grid (IEC 622090, IEEE Std 1528:2013) for the LOLA-Voronoi-error and the LOLA-max surrogate model, respectively.

ISBN:9783030048709

Publication year:2019

Accessibility:Open