Presentation information

General Session

General Session » GS-10 AI application

[4F4-GS-10o] AI応用:マテリアル情報

Fri. Jun 11, 2021 3:40 PM - 5:20 PM Room F (GS room 1)

座長:清原 慎(東京工業大学)

3:40 PM - 4:00 PM

[4F4-GS-10o-01] Investigation of the Life Estimation for the Deterioration by Void Discharge in Epoxy Resins Using Machine Learning

〇Iori Ooki1, Takahiro Nishigaki1, Takashi Onoda1, Yasutomo Otake2, Takahiro Umemoto2, Ken-ichi Suga2 (1. Aoyama Gakuin University, 2. Mitsubishi Electric Corporation)

Keywords:Void Discharge, Degradation estimation, Support Vector Machine

In the insulation diagnosis of electric power equipment, partial discharge, which is a precursor phenomenon of dielectric breakdown, is generally detected.
In addition, the generation distribution in the AC voltage phase of partial discharge correlates with the discharge generation factor, so it is used as an index for maintenance.
This time, we focused on the void discharge deterioration in the epoxy resin, which is a factor of aging deterioration, and investigated whether the degree of deterioration can be estimated by learning the phase characteristics of partial discharge generation using various machine learning methods.
In order to estimate the degree of deterioration from the partial discharge, what is used as the feature quantity from the partial discharge pattern is important.
In this study, the maximum absolute value of the partial discharge voltage and the average absolute value of the partial discharge voltage that can be obtained from the partial discharge pattern were used as the characteristic quantities of the partial discharge used for estimating the degree of deterioration.
Using these features, we estimated the degree of deterioration by applying multiple machine learning methods including support vector machines.
As a result, it was shown that the degree of deterioration can be estimated at about 85\% in the support vector machine.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.