진동 데이터 특징 추출 과정 내 Segment 설정에 따른 고장 진단율 영향도 분석

Abstract

With the advent of the Fourth Industrial Revolution, PHM technology is being developed to diagnosis failures of mechanical equipment and predict the remaining useful life of equipment. Acquiring quality data at a major step of PHM is an important process that has a significant impact on future diagnosis and prediction processes. In the process of acquiring and preprocess ing data, the feature extraction process is a process of extracting meaningful features of the data. This study analyzed the effect on the failure diagnosis rate according to the segment setting in the feature extraction process of vibration data. Therefore, the noise of the vibration data was removed, features were extracted for each segment, and the failure diagnosis rate was calcu lated using a classification algorithm. Based on the finally calculated results, the performance of the classification algorithm was evaluated using F1 - Score. As a result of the final analysis, the process of setting the segment size of the vibration data is also important, but the process of applying the noise reduction methods and feature values suitable for the characteristics of the data is more important.