Abstract

The fault diagnosis of centrifugal compressors is widely discussed in the literature and in industrial companies. Vibration analysis, NDT techniques and other techniques represent the majority of the techniques used. Sometimes and in specific situations these techniques are unfortunately not effective for decision making. However, these techniques need a decision making support tool. In this context and in order to alleviate this constraint we presented in this article a contribution for fault diagnosis in centrifugal compressors. This contribution consists mainly in the development of a probabilistic graphical model called the Bayesian network. Through its structure, our network has made it possible to define the causal links between causes and effect. Furthermore, its parameters give the possibility to classify the corrective action priorities. These actions are directly related to the a posteriori probabilities calculated by inference in the developed Bayesian network. The results found show that the proposed method is very efficient and the decision-making becomes more precise than in the case where the conventional and traditional methods of monitoring are used alone.

Information

Item Type:	Conference
Divisions:	» Faculty of Science and Technology
ePrint ID:	1734
Date Deposited:	2019-05-10
Further Information:	Google Scholar
URI:	https://www.univ-soukahras.dz/en/publication/article/1734

BibTex