Estimation of Material Properties Through Charge Transport Simulation Using Machine Learning

The increasing use of power electronics equipment in various operating environments has raised concerns about the space charge accumulation in insulation materials, especially in high-radiation environments. In order to ensure more robust reliability of power electronics equipment, it is important to have a better understanding of the space charge phenomena in insulation materials. In this study, we investigated the estimation of material properties related to space charge accumulation through the space charge transport simulation using machine learning. As results, the inverse analysis technique for estimating material properties related to charging through the space charge transport simulation was developed. However, some estimated values of material properties with low sensitivity to the space charge simulation had errors as large as approximately 10% compared to the true values. Therefore, it is necessary to consider improving the mathematical model of the space charge transport simulation as well as applying the inverse analysis technique to actual insulation materials in the future.