In this paper, a modeling approach is developed for analysis, composes of voltage or current impulse functions internal circuits of electronic devices. Two types of recurrent with different magnitude and different rise/fall time in neural networks (RNN), both with and without time sequence, frequency domain [4], [5], as shown in Fig. 1 (a). However, are trained to learn the dynamic responses of interferences in among the interferences in time domain used for analysing frequency and time domain respectively. After training, the RNN on theitrfere ces time don use foranalyin model provides fast evaluation of interference responses of the EMSeofelenctroni devices,therei vonte of 'traezid original internal circuits, which is useful for electromagnetic interference with sharp slew rates whose voltage magnitude iS susceptibility (EMS) analysis and optimization of electronic unchangedwithin almost a halfcycle period, as shown in Fig. devices. Two examples are provided to demonstrate the validity 1 (b). For these two types of interferences with different of the proposed modeling approach. waveform characteristics, two kinds of neural-network-based models have been constructed, shown in Fig. 2 and Fig. 3.