Classifying Radiation Degradation of Epoxy Molding Compound by Using Machine Learning and its Effect on Thermal and Mechanical Properties

Abstract

Power semiconductors play a crucial role in power conversion applications within nuclear power plants. These semiconductors are enclosed using polymeric materials for cost-effectiveness. Researchers have substantiated that polymeric materials are subject to radiation-induced degradation in nuclear power plants, prompting reliability studies. Consequently, investigating the radiation degradation behavior of polymeric materials becomes imperative to ensure their reliability and stability. This study focuses on the degradation of epoxy molding compound (EMC), a type of polymeric material, under the influence of total ionizing dose (TID). To analyze the effects of TID conditions on EMC, data was collected and subjected to various tests, including FTIR (Fourier Transform Infrared Spectroscopy) spectroscopy, thermal conductivity testing, and nanoindentation testing. These tests were conducted to assess chemical changes, thermal properties, and mechanical properties of EMC as a consequence of TID exposure. TID induces random ionization damage of EMC. Five EMC samples with different total cumulative doses were produced by varying the TID exposure time. Spectral data were obtained from the fabricated EMC samples by FTIR spectroscopy. FTIR spectral data was used to build a machine learning model, and the degree of EMC performance degradation due to TID exposure was determined. In our study, we selected an optimal algorithm among six machine learning algorithms. Dimensionality reduction methods such as ReliefF and PCA were also applied to build a more simplified discriminant model. As a result, it was confirmed that radiation changed the thermal properties of EMC materials. However, no change in the mechanical properties of EMC was observed under our test conditions.