Fault Diagnosis of a Crude Oil Pre-Treatment Systems Based on DOOBNs

Abstract The crude oil pre-treatment system plays a crucial role in the domain of oil and gas gathering and transportation, serving as a linchpin in ensuring the efficient and secure flow of petroleum resources. Its main function is to process the wellhead production, and complete the dehydration and degassing of the production, bringing it up to the standards required for subsequent processes. The crude oil pre-treatment system is intricate, and its fault could potentially result in significant economic losses and catastrophic consequences. Therefore, the precise and expeditious identification of both the location and the reasons for the occurrence of faults within the crude oil pre-treatment system is of paramount importance. To address this imperative, a fault diagnosis model based on dynamic object-oriented Bayesian networks is designed to conduct fault attribution analysis on the system. The model is mainly composed of the static Bayesian networks and the Markov model. The static Bayesian network consists of Fault Layer, Fault Symptom Layer, and External Variable Layer, used to identify the causes of system faults. The Markov model is employed to determine the temporal evolution of the system’s states. In this study, static Bayesian networks is utilized not only to diagnose individual faults but also to successfully identify the causes of concurrent faults. Then, static Bayesian networks is extended in the time domain by 20 iterations. By inputting fault symptoms into the first static model, predictions of the system’s state are achieved, allowing the time of system maintenance completion to be determined. By inputting fault symptoms into the twentieth static model, a retrospective analysis of the system state is conducted, thereby determining the initiation time of the fault. The research indicates that the model’s results maintain a high level of consistency with on-site fault logs, providing crucial support for on-site production. Furthermore, this also validates the effectiveness, practicality, and accuracy of the model.