A Neural Network Based Calibratable Arc Model for Air Electrostatic Discharge

This article introduces an innovative calibratable modeling approach to effectively capture intricate arc behavior in air electrostatic discharge (ESD). The proposed method incorporates a compact electric arc resistance model rooted in the Rompe–Weizel law, calibrating by a physics informed neural network (PINN). The systematic uncertainties in the air discharge behavior due to environmental factors and measurement procedure are succinctly and effectively quantified through introducing an equivalent arc length in the compact model. A dedicated electrostatic air discharge behavior library is developed by a reduced-order partial element equivalent circuit model for training the neural network. The PINN calibrates the arc model according to the measured discharge currents on a standard ESD calibration set. The fidelity of the calibrated compact electric arc model is verified by the simulation and measurement. The efficacy of the proposed approach is observed through a case study. This new environment-aware modeling method provides deeper insights into air discharge phenomena and proves its promising potential in characterizing noncontact electromagnetic discharge.