Experimental Prediction of the Radiated Emission and Final Measurement Process Optimization based on Deep Neural Networks According to EN 55032

Electromagnetic interference (EMI) is the presence of unwanted electromagnetic emission which has the potential to cause disturbances in electronic and electronic devices. Therefore, any equipment must be certified that it meets electromagnetic compatibility (EMC) requirements. To meet these requirements the equipment must be tested for conducted and radiated emissions in a certified EMC testing company. Most of these tests are time-consuming, thus cost and test time are big challenges when performing an EMI test. In this paper, an approach is proposed to find effectively the worth-case positions during the final measurement phase on critical frequencies in EMI measurements according to the norm EN 55032 in the range 30MHz to 1GHz by using a developed measurement software and deep neural networks (DNN). Firstly, because of its advantage of relatively simple model structure and strong data features extraction, a one-dimensional convolution neural network (1D CNN) was used to predict the positions that meet the maximum radiation emission level. The effectiveness and prediction accuracy of CNNs for the high input variance emission levels were low, therefore a hybrid deep learning neural network framework, that combines CNN with long short term memory(LSTM) was adopted to forecast the worst-case of the high variance emission levels. The DNNs were trained using real EMI measurements for different types of equipment under test (EUTs) in a Semi Anechoic Chamber (SAC) by Cetecom GmbH in Essen, Germany. Secondly, the results from our developed software were compared with the target labeled results from Rode& Schwarz EMC32 Software to evaluate our proposed measurement. By determining the worth-case position by predicting the azimuth of the turntable and the height of the antenna, the required time to perform the final measurement phase is effectively decreased.