기어 이 파손 정도에 따른 진동신호의 특징기반 경향 감시

Abstract

Gear systems are widely used in various industries. However, they fail due to different reasons suchas poor manufacturing and assembly processes. Currently, preventive maintenance (PM) is periodicallyperformed to ensure that a gearbox system is safely operating. However, unnecessary PM results indefects and maintenance cost. Therefore, a method to diagnose defects is developed using the featuresof machine learning. In this paper, lab-scaled gearbox is used as the experimental model, which canbe simulated in four stages: normal and 10 %, 20 %, and 30 % of tooth breakage. The calculated featureswere selected using the genetic algorithm. Three features were used to diagnose the limitationsof the gear system. Consequently, the severity of tooth breakage of the gearbox was classified for fourstages by the three selected features. In addition, the increasing or decreasing trend of the value offeatures was identified according to the severity of the defect.