Multi Sources Information Fusion Based on Bayesian Network Method to Improve the Fault Prediction of Centrifugal Compressor

Abstract The centrifugal compressor is an important machine in the oil and gas industry, so the fault prediction of these machines is widely discussed in the literature. Several techniques can and should be used in fault prediction of centrifugal compressors: vibration analysis, non-destructive testing techniques, operating parameters, and other techniques. But in particular cases, these tools are inefficient for making a decision regarding the combined fault diagnosis and prediction. This paper presents a contribution to fault prediction in centrifugal compressor utilizing multi-source information fusion by a Bayesian network. The data fusion does not come from the same source, but rather from vibration analysis, oil analysis, and operating parameters. In addition, the accuracy and ability of fault prediction can be improved compared with the use of data obtained from vibration analysis only or oil analysis. The proposed method accuracy is validated on a BCL 406 type centrifugal compressor. Furthermore, the obtained results showed the effectiveness of the multi-source information fusion by Bayesian network approach gives more accuracy to decision-making in fault prediction and the developed method has an effect in predicting the combined faults.