Classificação on-line de situações anormais em operação de processos industriais baseada em processamento de alarmes e variáveis de processos

Abstract

Industrial processes are subject to failures in their thousands of components at any time and can lead to shutdowns, loss of product quality, equipment damage or even accidents. In this sense, the alarm system is necessary to aid in the identification of process abnormalities. However, during a process failure it is common for the operator to be subjected to hundreds of alarms causing overload beyond the human processing capacity. This phenomenon is known as alarm flood and to treat them properly is a challenge for the modern alarms systems. Thus, the present work aims at the development of an online alarm processing methodology capable of assisting the operator in the identification and classification of abnormal situations of the process, especially in moments of alarm overload. To validate the proposal, a case study was carried out on a process simulator widely used and accepted by the scientific community called Tennessee Eastman Process. The results indicate that it is important to identify and monitor the abnormality scenarios underway in industrial processes. The results show that the methodology is efficient to identify and follow the abnormality scenarios in progress in industrial processes.