Electromagnetic Uncertainty Analysis Using Stochastic FDFD Method

This paper presents a novel stochastic finite-difference frequency-domain (S-FDFD) method that incorporates the statistical variations of electrical properties directly into the FDFD algorithm by using the delta method approximation. The method is validated through two examples-a 1-D bioelectromagnetic layered model and specific absorption rate in a 2-D slice of a human head model. Comparisons between the single run S-FDFD results and those of the Monte Carlo method (the benchmark in statistical simulations) show that the S-FDFD method can predict the mean and variance of electromagnetic fields with a fair degree of accuracy in a computationally efficient way. Our numerical experiment also shows that the accuracy is highly dependent on the approximation for the correlation coefficients of the variance equations.