Fault Diagnosis for Natural Gas Dehydration Device Based on the Combination of PCA and SDG
Received:April 16, 2021  Revised:June 04, 2021
View Full Text  View/Add Comment  Download reader
DOI:10.7643/issn.1672-9242.2021.09.016
KeyWord:dehydration device  fault diagnosis  PCA  SDG
              
AuthorInstitution
XIONG Wei Chongqing Gas Mine of Southwest Oil and Gas Branch, PCL Chongqing , China
MA Hao School of Mechanical Engineering, Chongqing University, Chongqing , China
ZHANG Bo Chongqing Gas Mine of Southwest Oil and Gas Branch, PCL Chongqing , China
LONG Yan-li Chongqing Gas Mine of Southwest Oil and Gas Branch, PCL Chongqing , China
YIN Ai-jun School of Mechanical Engineering, Chongqing University, Chongqing , China
Hits:
Download times:
Abstract:
      The paper aims to quickly locate the fault source of the natural gas dehydration device. Principal component analysis was performed on the monitoring parameters that reflected the running state of the dehydration system so as to identify the abnormal parameters. Combined with the dehydration process flow chart, the causal analysis of all parameters and potential fault sources was carried out to establish the overall SDG model of the dehydration device. The symbols of each abnormal parameter were substituted into the model, and the equipment faults that caused these parameters were determined according to the two-way reasoning rule. The SPE and T2 statistics constructed by principal component analysis remained within the control limits of 20 and 141 during normal operation of the device, and both increased sharply at the same time when a fault occurred, and then fluctuated in a range well above the control limit. The main fault parameters identified by the contribution diagram method were TEG circulation, reboiler temperature and buffer tank level. The combination of PCA and SDG can integrate the advantages of the two methods, improve the efficiency of petrochemical equipment inspection and maintenance, and ensure the reliable operation of the equipment.
Close