The assessment of the exposure to electromagnetic waves is nowadays a key question. Dealing with the relationship between exposure and incident field, most of previous investigations have been performed with a single plane wave. Realistic exposure in the far field can be modeled as multiple plane waves with random direction of arrival, random amplitude, and random phase. This paper, based on numerical investigations, studies the whole body specific absorption rate (SAR) linked to the exposure induced by five random plane waves having uniformly distributed angles of arrival in the horizontal plane, lognormal distributed amplitudes, and uniformly distributed phases. A first result shows that this random heterogeneous exposure generates maximal variations of ±25% for the whole body specific absorption. An important observation is that the exposure to a single plane wave arriving face to the body, used for the guidelines, does not constitute the worst case. We propose a surrogate model to assess the distribution of the whole body SAR in the case of an exposure to multiple plane waves. For a sample of 30 values of whole body SAR induced by five plane waves at 2.4 GHz, this simple approach, considering the resulting SAR as the sum of the SAR induced by each isolated plane wave, leads to an estimated distribution of whole body SAR following the real distribution with a p value of 76% according to the Kolmogorov statistical test.

In this paper, we analyze the variation of Whole Body Specific Absorption Rate (WBSAR) induced in Visible Human by five plane waves with random direction of azimuth angles, amplitudes and phases. A new and fast numerical method presented in a first paper [1] is used to perform this study. Considering a finite set of 36 azimuth angles, the Latin Hypercube Sampling is used to design an experiment plan. Two distributions of amplitudes are used to design the plan and the different results of WBSAR for each case are compared. The response Y (WBSAR) is modeled as a combination of input parameters influencing the exposure. Using input parameters, the azimuth angles described by the surface projected and the WBSAR, a regression estimates the model coefficients. The influence of inputs variables of the function is analyzed.