Centrifugal Compressor Maintenance Using Fault Tree and a Bayesian Network Methods for System Reliability Analysis and Dependability

Oil and gas refining units are considered among the important establishments in the economy of countries, as they depend on them for the production and export of energy. These units base their work on rotating machines such as gas turbines (GT), gearboxes (GB) and centrifugal compressors (CC), the CC is considered the basic pillar of the unit, it is also the first responsible of the unit shutdown at times, as its damage leads to great economic losses such as a decrease in oil and gas production, and this is what makes it a focus and concern for the producing countries. Therefore, maintaining the safety of centrifugal compressors increases the reliability of the system and the continuity of the unit's work permanently and continuously, and this can only be achieved with good monitoring and continuous maintenance, and with all this care from the manufacturers and the preventive measures used, they may be exposed to some defects and risks that cause damage to the internal parts such as bearings, shaft, … etc. Therefore, to raise the reliability and extend the life cycle of compressors, we must mainly rely on continuous inspection and accurate diagnosis to reduce various defects. In this paper, we studied two different cases of two compressors, one of which was subjected to vibrations and the other to high temperatures in the bearings at AL-MERK and BRN units in in southern Algeria. where the unit was completely stopped to fix this defect, this made us think about how to know the main causes of this malfunction. Where we adopted in this paper a hybrid approach, which is a methodology that evaluates the reliability of the system by combining two methods, one traditional is the Fault Tree (FT) to identify events and find out the type of defects that cause the main failure, and the second is a Bayesian Network (BN) relies on artificial intelligence to know the impact of each event on the system, This study helps in analyzing the system reliability, and predicting all the causes that would impede the continuity of the compressors work. Through pre-stored data collected by maintenance experts and monitoring workers, relying on a Bayesian network to identify the fragile components in the system and the most influential ones and calculate the extent of their impact on the system. Finally, by performing this analysis, we found that vibration is the main cause of compressor failure.