검색 상세

WPCA 알고리즘을 이용한 LDPE 생산 공정 중 2차 압축기의 이상 진단 모델 개발

Fault-detection Model using Weighted Principal Component Analysis for the Hyper-compressor in a Plant Manufacturing Low-Density Polyethylene

김지선 (Ji Seon Kim, 포항공과대학교)

원문보기

초록 moremore
This paper presents a data-based monitoring method to detect faults in the hyper-compressor of a plant that manufactures low-density polyethylene (LDPE). Previous fault detection studies for the LDPE manufacturing plant focused on detecting reactor faults. However, since the compressor is operated in considerably high pressure, it could incur critical damage compared to the fault that happened in the reactor. The model uses weighted principal component analysis based on operation data of an LDPE-manufacturing plant, measured hourly for 4 years. The principal component control limit (PCCL) detected shutdown better than did Hotelling’s statistic. A PCCL index based on WPCA model with balancing factor of 0.6 achieved a detection rate of 80%.
This paper presents a data-based monitoring method to detect faults in the hyper-compressor of a plant that manufactures low-density polyethylene (LDPE). Previous fault detection studies for the LDPE manufacturing plant focused on detecting reactor faults. However, since the compressor is operated in considerably high pressure, it could incur critical damage compared to the fault that happened in the reactor. The model uses weighted principal component analysis based on operation data of an LDPE-manufacturing plant, measured hourly for 4 years. The principal component control limit (PCCL) detected shutdown better than did Hotelling’s statistic. A PCCL index based on WPCA model with balancing factor of 0.6 achieved a detection rate of 80%.
목차 moremore
1. INTRODUCTION
2. METHODS
2.1. WPCA algorithm and monitoring statistics based on WPCA
...
1. INTRODUCTION
2. METHODS
2.1. WPCA algorithm and monitoring statistics based on WPCA
3. FAULT DETECTION USING WPCA APPROACH
3.1. The target data and occurrence history of shut down
3.2. Fault detection model based on WPCA
4. RESULTS
5. CONCLUSION