Spectrogram-based Aging Assessment for Insulations on Stator Coil Bar

Based on four kinds of spectrograms, this paper studies how to effectively predict the age of insultations on stator coil bar. The prediction of using time can guide the assessment of stator coil bar. In this work, some samples with different ages and temperatures of epoxy insulations are generated in an accelerated aging test. Then the microscopic morphology spectrograms are obtained for the samples. We use the histogram of oriented gradient (HOG) and gray-level co-occurrence matrix (GLCM) to extract the features of the spectrograms. Finally, support vector machine (SVM) is used to classify and predict the using time of samples based on HOG and GLCM features. This method with accuracy 90.3333% on the classification and prediction for the age of insulations on stator coil bar.