An activated inference method for fault diagnosis of three-phase separator based on rare fault data

The three-phase separator is the core equipment in offshore production platform, which may cause significant economic losses in case of abnormalities. Therefore, it is critical to identify the cause of three-phase separator abnormalities. However, in three-phase separator systems, the available fault samples are usually limited, resulting in the inability of conventional diagnostic methods to accurately identify the causes. In this paper, we propose a diagnostic algorithm called the abnormality inference network. Firstly, the algorithm determines the network structure based on process mechanisms and alarm records. Second, expert knowledge is processed using fuzzy theory to determine the likelihood rating between nodes. In addition, the network structure is intricate and complex, and ternary groups are used for network deconstruction in order to facilitate computation and storage. Finally, this paper uses test cases combined with PSO algorithm to determine the minimum activated value of nodes. Meanwhile, this paper designs an activated inference algorithm to diagnose the abnormality cause online. The three-phase separator is used for validation, and the results show that the algorithm can accurately diagnose the working condition and equipment abnormality of the three-phase separator.