This paper presents a comparative study of various artificial intelligence methods, including artificial neural networks (ANNs), recurrent neural networks (RNNs), k-nearest neighbors (KNN), random forests (RFs), and particle swarm optimization (PSO) techniques, to see which one can predict conducted electromagnetic interference (CEMI) better. The DC/DC converter simulations and experimental results demonstrated a high level of matching. According to the simulation results, the datasets were highlighted by varying key parameters related to the supply voltage, load current, switching frequency, duty cycle, component choice, PCB layout, filter capacitance, and gate resistance in a systematic way. During the assessment, each AI technique is checked regarding prediction accuracy, computational efficiency, and error rates using different metrics such as mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²). It is observed that KNN performs better than the other methods, giving only the lowest error in predictions and showing very fast computing speed. Furthermore, KNN gave the best results with R² above 0.97, MAE below 5.9 dBµV, and RMSE under 7.3 dBµV. This method worked better than others in all test cases. According to the measurements, the predicted and actual EMI levels match very well and show that the proposed method is strong and reliable. Further, basically, these results show that KNN has the same potential to work as an effective and efficient tool for predicting CEMI in power electronics. Its strong performance can further help in developing better and more reliable power systems for practical use, while the system itself provides valuable insights to engineers for electromagnetic compatibility design and compliance.

This paper presents an innovative reliability bench of aging life tests designed to high power RF applications for device lifetime under pulse conditions. The temperature effect on the parameters of power RF LDMOS (Radio Frequency Laterally Diffused-Metal–Oxide–Semiconductor) devices is highlighted. Indeed, the acceleration of the degradation mechanisms is related, directly or indirectly, to the temperature variation. The tests carried out on the power amplifier will be "Life-test RF" type (accelerated aging under constant constraints) over a period of 1500 h to quantify the drifts of the parameters measured (mainly P_OUT and I_DSS) under reliability bench life-test at different temperatures. The parameters of devices have been characterized i.e. static, dynamic and RF before and after testing. This allows us to quantify the degradation, of the shift, of a certain number of electrical quantities (V_TH, G_M, R_DSON, C_RSS, etc.). The analysis of the physical results has been presented (simulator 2D ATLAS-SILVACO) to explain and observe the physical review of temperature impacts on power RF LDMOS performance. Finally, initial impacts analysis have been discussed.

This paper presents a time-domain (TD) approach based on hybrid artificial intelligence (AI) to speed up convergence of radiating sources characterization in power electronics. To obtain a representative equivalent model of device under test, a dedicated optimization framework has been developed in TD using a particle swarm optimization (PSO) toolbox. In addition, for elementary feature extraction, a pseudo-Zernike moment invariant (PZMI) descriptor has been defined. Finally, with the aim of identifying remaining dipole parameters and classification problems, artificial neural networks (ANN) have been implemented. A coupling of TD electromagnetic (EM) inverse method based on a PSO algorithm along with PZMI and ANN application has been investigated and applied to a real test case. Experimental measurements have been conducted using the near-field scanning technique above an alternating current (AC)/direct current (DC) converter. Obtained results are discussed based on a comparison between measured and estimated EM field distributions using both the hybrid AI method and a conventional TD inverse method based on genetic algorithms (GA) only. This study confirms that, compared with those given by non-hybrid method, the proposed algorithm further improves the convergence speed while maintaining high accuracy. Hence, the present work offers an impressive perspective for radiated emissions characterization using hybrid AI algorithms.

Gallium Nitride (GaN) High Electron Mobility Transistors (HEMTs) are emerging as the technology of choice for next-generation power conversion systems, offering switching speeds, on-state resistance, and power density unattainable with silicon or even silicon carbide (SiC) devices. However, the fast switching transients that make GaN attractive also make it a demanding source of electromagnetic interference (EMI), so credible electromagnetic compatibility (EMC) analysis requires an accurate functional device model. This paper addresses the functional electrothermal modeling of a commercial 650 V GaN HEMT (GS66504B) as a prerequisite to EMC validation. The manufacturer-supplied Level 3 SPICE model is evaluated against experimental static (I-V) and dynamic (C-V) measurements. Significant discrepancies motivate an optimization methodology in which an initial manual procedure is superseded by a fully automated pipeline coupling LTspice with a Genetic Algorithm in MATLAB R2025b. A forward/reverse and dual-temperature-segment strategy reduces the mean absolute relative error to below 7% (forward I-V) and 13% (reverse I-V) over 25–100 °C, while a dedicated two-stage C–V optimization reduces the reverse-transfer capacitance error from 95.4% to 2.89%. The resulting compact, unified, and fully validated model underpins the ongoing EMC validation phase, where it will be combined with extracted parasitic and cable models in a DC-DC converter topology.