Artificial neural network modeling of electromagnetic interference caused by nonlinear devices inside a metal enclosure

摘要

The artificial neural network (ANN) modeling of electromagnetic interference from a nonlinear device connected with transmission lines exposed to electromagnetic field to the linear device inside a metal enclosure is investigated in this paper. A rectangular loop antenna operating at 2.5GHz is used to excite electromagnetic field inside an enclosure with dimension 99x49x44cm(3). Another wideband loop antenna welded with a HSMS-282C Schottky diode pair is used as the nonlinear device connected with the transmission line in order to enhance the coupling and radiating effects inside the enclosure. The linear devices are three monopole probes connected with spectrum analyzer to measure the output power. Experimental results show that the output power from probes near the wideband loop welding with a pair of diodes contains not only fundamental but harmonic components, which are caused by the nonlinear response of diodes excited by induced current. A three-layer ANN model is proposed to predict this nonlinear electromagnetic interference inside the metal enclosure. Results show that the trained BL-MLP model using 576 measured data can predict the response of probes (linear device) interfered by the loop antenna welding with diodes (nonlinear device) inside the cavity well.