A new methodology to analyze electrical systems in the time and frequency domain containing a large number of uncertain parameters is presented. The methodology is based on screening and design of experiments to build surrogate reduced models at a well specified set of frequency values. As an example, the conducted interferences of a Power Converter with uncertainties in its component values are analyzed. The model has a large number of parameters, which are described by Probability Density Functions (PDF). The output considered is a standard measurement of conducted interferences and its PDF is rapidly determined, if compared to the Monte Carlo (MC) approach.
Electromagnetic Compatibility; Monte Carlo; Parametric Uncertainty; Power Electronics; Probability Density Functions