차량설비의 알람 유형 분류를 위한 오픈셋 인식 심층 신경망 (Open Set Recogniton Dep Neural Network for Clasifcation of Alarm Types in Automobile Facilty)

서지정보

· 발행기관 : 대한산업공학회
· 수록지 정보 : 대한산업공학회지 / 47권 / 2호 / 224 ~ 231페이지
· 저자명 : 김상훈, 홍현욱, 정의현, 김성범

초록

It is importanto predict critcal alarms in the manufacturing proces that can reduce the utilzation rate of theproduction facilty. In recent years, dep neural networks have ben widely used for prediction of alarm types in themanufacturing proceses. However, thexisting dep neural network clasifers folow aclosed-set asumption thatal predictable categorieshould be learned in the training stage. For this reason, when an unknown alarm type comesinto the clasifer, it should be clasifed as one of the predefined alarm types. In the actual manufacturing proceses,it is extremely dificulto colect al posible types of alarm data. Therefore, a model with open set recogniton isrequired to identify unknown alarm types. In aditon, because the alarm type data colected from productionfacilties ocursimultaneously, a multi-label clasifcation model is necesary. In thistudy, we propose a multi label open set recogniton model combined with background data that can improve the abilty to identify unknownalarm types. We demonstrated the usefulnes and aplicabilty of the proposed method by comparing it with existingopen set clasifcation methods using real proces data obtained from an automobile industry.

영어초록

It is importanto predict critcal alarms in the manufacturing proces that can reduce the utilzation rate of theproduction facilty. In recent years, dep neural networks have ben widely used for prediction of alarm types in themanufacturing proceses. However, thexisting dep neural network clasifers folow aclosed-set asumption thatal predictable categorieshould be learned in the training stage. For this reason, when an unknown alarm type comesinto the clasifer, it should be clasifed as one of the predefined alarm types. In the actual manufacturing proceses,it is extremely dificulto colect al posible types of alarm data. Therefore, a model with open set recogniton isrequired to identify unknown alarm types. In aditon, because the alarm type data colected from productionfacilties ocursimultaneously, a multi-label clasifcation model is necesary. In thistudy, we propose a multi label open set recogniton model combined with background data that can improve the abilty to identify unknownalarm types. We demonstrated the usefulnes and aplicabilty of the proposed method by comparing it with existingopen set clasifcation methods using real proces data obtained from an automobile industry.

참고자료

· 없음