Using Smart Diagnostic Technology to Monitor Operations of Critical Equipments in Refinery

• Main Authors: Jenn-Yeu Chen, 陳震宇
• Other Authors: San-Yih, Hwang
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/7c7487

碩士 === 國立中山大學 === 管理學院高階經營碩士學程在職專班 === 107 === The key process variables of critical equipments in the petroleum refining industry shall be fully monitored or controlled. Once an abnormality occurs, the rates of equipment damage mechanisms may be affected, resulting in a loss of control. A major hazard accident occurs often due to the operator/ operators failure to comply with procedures, technical capabilities fault, the communication failure, fatigue, unclear handover, lack of manpower for operation, operation alarms flooding, failure to track the process operation condition, taking the wrong coutermeasures, etc.. This study takes the rotating and static critical equipments of the pilot unit in refinery as an example. To develop a trend prediction model for the key process variable, the historical operational data of the critical equipment were fed into the exponential smoothing method. The condition-dependent analysis and sensitivity analysis of Bayesian Network methodology is presented to analyze the dependency between the relevant variables and key process variable, and the disposal steps during an abnormal event in order to prevent any safety holes in the workplace caused by human error. Based on the estimation, modeling complexity, applicability in future and through the Parsimony Principle, One-parameter double exponential smoothing with forecasting model for key variable is identified to achieve the best accuracy. No matter the dynamic or static equipments in this study, Bayesian network method allows variables feature screening to verify the most related variables to the key variables when every offset occured. The study also find the causes of each offset may be different and the variables that have the greatest influence on the key variables my not be the same. The influence of each variable on the key variables will also change over time. During the abnormal operation period of the equipment, it is not easy to judge the causal relationship of all variables in the system only by the expert domain knowledge. The proposed method in this study provides a more convincing process of causal reasoning. It will help the management in refinery to understand the causes of equipment anomalies and undertake the response in advance.